Content uploaded by Frank L. Schmidt
Author content
All content in this area was uploaded by Frank L. Schmidt on Aug 14, 2017
Content may be subject to copyright.
Part I
SELECTION
c01.indd 1c01.indd 1 6/17/09 2:26:27 PM6/17/09 2:26:27 PM
COPYRIGHTED MATERIAL
c01.indd 2c01.indd 2 6/17/09 2:26:28 PM6/17/09 2:26:28 PM
1
Select on Intelligence
F RANK L. S CHMIDT
Other things equal, higher intelligence leads to better job performance on all jobs. Intelli-
gence is the major determinant of job performance, and therefore hiring people based on
intelligence leads to marked improvements in job performance – improvements that have
high economic value to the fi rm. This principle is the subject of this chapter.
This principle is very broad: it applies to all types of jobs at all job levels. Until a couple
of decades ago, most people believed that general principles of this sort were impossible
in personnel selection and other social science areas. It was believed that each organization,
work setting, and job was unique and that it was not possible to know which selection
methods would work on any job without conducting a validation study on that job in that
organization. This belief, called the theory of situational specifi city, was based on the fact
that different validity studies of the same selection procedure(s) in different jobs in the same
organization and/or different organizations appeared to give different results. However, we
now know that these “ confl icting fi ndings ” were mostly due to statistical and measurement
artifacts and that some selection procedures have high validity for predicting perfor mance
on all jobs (e.g. intelligence) and others do a poor job of predicting performance on any job
(e.g. graphology) (Schmidt and Hunter, 1981 , 1998 ). This discovery was made possible by
new methods, called meta - analysis or validity generalization methods, that allow researchers
to statistically combine results across many studies.
Meta - analysis has also made possible the development of general principles in many other
areas beyond personnel selection (Hunter and Schmidt, 2004 ; Schmidt, 1992 ). For example,
it has been used to calibrate the relationships between job satisfaction and job perfor mance
with precision (Judge, Thoresen, Bono, and Patton, 2001 ) and between organiza tional
commitment and work - related outcomes including job performance (Cooper - Hakim and
Viswesvaran, 2005 ).
W HAT IS INTELLIGENCE?
Intelligence is not the ability to adapt to one ’ s environment; insects, mosses, and bacteria
are well adapted to their environments, but they are not intelligent. There are many ways
c01.indd 3c01.indd 3 6/17/09 2:26:28 PM6/17/09 2:26:28 PM
4 FRANK L. SCHMIDT
in which organisms can adapt well to their environments; use of intelligence is only one
possible way. Intelligence is the ability to grasp and reason correctly with abstractions (con-
cepts) and solve problems. However, perhaps a more useful defi nition is that intelligence is
the ability to learn. Higher intelligence leads to more rapid learning, and the more com-
plex the material to be learned, the more this is true. Intelligence is often referred to as
general mental ability (GMA) and general cognitive ability, and we use all these terms inter-
changeably in this chapter.
Intelligence is the broadest of all human mental abilities. Narrower abilities include
verbal ability, quantitative ability, and spatial ability. These narrower abilities are often
referred to as special aptitudes. These special aptitudes do predict job performance
(although less well than GMA), but only because special aptitude tests measure general
intelligence as well as specifi c aptitudes (Brown, Le, and Schmidt, 2006 ; Schmidt, Ones,
and Hunter, 1992 ). It is the GMA component in these specifi c aptitude tests that predicts
job performance. For example, when a test of verbal ability predicts job or training per-
formance, it is the GMA part of that test – not the specifi cally verbal part – that does the
predicting (Brown et al., 2006 ).
Intelligence predicts many important life outcomes in addition to job performance: per-
formance in school, amount of education obtained, rate of promotion on the job, ulti-
mate job level attained, income, and many other things (Brody, 1992 ; Herrnstein and
Murray, 1994 ; Gottfredson, 1996 ; Jensen, 1998 ). It is even involved in everyday activities
such as shopping, driving, and paying bills (Gottfredson, 1996 ). No other trait – not even
conscientiousness – predicts so many important real world outcomes so well. In this sense,
intelligence is the most important trait or construct in all of psychology, and the most
“ successful ” trait in applied psychology.
The thousands of studies showing the link between intelligence (GMA) and job perform-
ance have been combined into many different meta - analyses. Ree and co - workers have shown
this for military jobs (Olea and Ree, 1994 ; Ree and Earles, 1991 , 1992 ; Ree, Earles, and
Teachout, 1994 ), as have McHenry, Hough, Toquam, Hanson, and Ashworth (1990) in the
famous Project A military study. ( With a budget of 24 million dollars, Project A is the larg-
est test validity study ever conducted.) Hunter and Hunter ( 1984 ) have shown this link for a
wide variety of civilian jobs, using the US Employment Service database of studies. Schmidt,
Hunter, and Pearlman ( 1980 ) have shown it for both civilian and military jobs. Other large
meta - analytic studies are described in Hunter and Schmidt ( 1996 ), Schmidt ( 2002 ), and
Schmidt and Hunter ( 2004 ). Salgado and his colleagues (Salgado, Anderson, Moscoso,
Bertua, and de Fruyt, 2003a , 2003b ) demonstrated the link between GMA and job perform-
ance across settings in the European countries. The amount of empirical evidence supporting
this principle is today so massive that it is hard to fi nd anyone who questions the principle.
There has been an important development since the fi rst edition of this book appeared
in 2000: a new and more accurate method for correcting for the biases created by range
restriction has been developed and applied (Hunter, Schmidt, and Le, 2006 ; Schmidt, Oh,
and Le, 2006 ; Schmidt, Shaffer, and Oh, 2008 ). (Range restriction is the condition in which
variability of the predictor (here intelligence) in one ’ s sample of people ( job incumbents) is
artifi cially lower than in the population of people ( job applicants) one wants to get esti-
mates for.) Application of this procedure to existing data shows that previous estimates of
the validity of GMA – including those in the 2000 version of this chapter – were under-
estimated by 25% to 30%. In this chapter, I present the updated, more accurate validity
estimates. When performance is measured objectively using carefully constructed work
c01.indd 4c01.indd 4 6/17/09 2:26:28 PM6/17/09 2:26:28 PM
SELECT ON INTELLIGENCE 5
sample tests (samples of actual job tasks), the correlation (validity) with intelligence measures
is about .84 – 84% as large as the maximum possible value of 1.00, which represents perfect
prediction. When performance is measured using ratings of job performance by super-
visors, the correlation with intelligence measures is .66 for medium complexity jobs (over
60% of all jobs). For more complex jobs, this value is larger (e.g. .74 for profes sional and
managerial jobs), and for simpler jobs this value is not as high (e.g. .56 for semi - skilled jobs).
Another performance measure that is important is amount learned in job training pro-
grams (Hunter et al., 2006 ). Regardless of job level, intelligence measures predict amount
learned in training with validity of about .74 (Schmidt, Shaffer, and Oh, 2008 ).
W HY DOES INTELLIGENCE PREDICT JOB PERFORMANCE?
It is one thing to have overwhelming empirical evidence showing a principle is true and
quite another to explain why the principle is true. Why does GMA predict job perform-
ance? The primary reason is that people who are more intelligent learn more job knowl-
edge and learn it faster. The major direct determinant of job performance is not GMA
but job knowledge. People who do not know how to do a job cannot perform that job well.
Research has shown that considerable job knowledge is required to perform even jobs
most college students would think of as “ simple jobs, ” such as truck driver or machine
operator. More complex jobs require even more job knowledge. The simplest model of
job performance is this: GMA causes job knowledge, which in turn causes job perform-
ance. But this model is a little too simple: there is also a causal path directly from GMA
to job performance, independent of job knowledge. That is, even when workers have
equal job knowledge, the more intelligent workers have higher job performance. This is
because there are problems that come up on the job that are not covered by previous job
knowledge, and GMA is used directly on the job to solve these problems. Many studies
have tested and supported this causal model (Hunter, 1986 ; Ree, Earles, and Teachout,
1994; Schmidt, Hunter, and Outerbridge, 1986 ). This research is reviewed by Schmidt
and Hunter ( 1992 ), Hunter and Schmidt ( 1996 ), and Schmidt and Hunter ( 2004 ). It has
also been shown that over their careers people gradually move into jobs that are consistent
with their level of GMA (Wilk, Desmariais, and Sackett, 1995 ; Wilk and Sackett, 1996 ).
That is, a process that sorts people on GMA takes place gradually over time in everyday
life. People whose GMA exceeds their job level tend to move up to more complex jobs;
and people whose GMA is below their job level tend to move down.
There is a broader theory that explains these research results: the traditional psycho-
logical theory of human learning (Hunter and Schmidt, 1996 ; Schmidt and Hunter, 2004 ).
This theory correctly predicted that the effect of GMA would be on the learning of job
knowledge. The false theory of situational specifi city became widely accepted during the
fi rst eight decades of the 20th century in considerable part because personnel psychol-
ogists mistakenly ignored the research on human learning.
Many lay people fi nd it hard to believe that GMA is the dominant determinant of job
performance. Often they have known people who were very intelligent but who were dis-
mal failures on the job because of “ bad behaviors ” such as repeated absences from work,
carelessness at work, hostility toward the supervisor, unwillingness to work overtime to
meet a deadline, or stealing from the company. These are examples of so - called “ counter-
productive work behaviors ” (CWBs). Integrity tests predict CWBs with a validity of about
c01.indd 5c01.indd 5 6/17/09 2:26:28 PM6/17/09 2:26:28 PM
6 FRANK L. SCHMIDT
.35 (Ones, Viswesvaran and Schmidt, 1993 ). People with lower scores on integrity tests
show more CWBs. The personality trait of conscientiousness also predicts CWBs (again,
negatively). However, a recent large - scale study ( N ⬎ 800) found that GMA predicted
CWBs with a validity of .47; when the more accurate correction for range restriction is
applied, this fi gure becomes .57. So it is possible that the best predictor of CWBs is GMA.
People who are more intelligent show fewer CWBs.
There is also a facet of job performance called “ contextual performance ” (CP). CP
is just good citizenship behaviors, while CWB is bad citizenship behaviors as discussed
above. CP behaviors include willingness to help train new employees, willingness to work
late in an emergency or on a holiday, supporting the community relations and reputation
of the company, and many other such behaviors. CP behaviors and CWBs are differ-
ent from core job performance but are often confused with core job performance by lay
observers. CP and CWB behaviors are predicted by measures of the personality traits of
conscientiousness and to a lesser extent agreeableness (Dalal, 2005 ). We do not yet know
whether GMA predicts CP behaviors; these studies have yet to be done. Low ability leads
to an inability to perform well; low conscientiousness and low agreeableness lead, not pri-
marily to low performance on core job tasks but to lack of CP and/or more displays of
organizationally disruptive behaviors (CWBs). These disruptive behaviors are more visible
to lay observers (and to many supervisors) than differences between employees in core job
performance, probably because they appear so willful. On the other hand, a low ability
employee has diffi culty learning how to perform the job, but if he/she has a “ good atti-
tude, ” this employee often seems like less of a problem than one showing CWBs. This
makes it diffi cult for some to clearly see the GMA – performance link in the real world
(Hunter and Schmidt, 1996 ).
Of course, low conscientiousness can lead to less effective performance if it results in
reduced effort (see Chapter 2 , this volume). For objective measures of job performance,
empirical evidence indicates that on typical jobs this effect is limited, probably because
most jobs are fairly structured, reducing the scope for individual differences in effort to
operate (Hunter, Schmidt, Rauschenberger and Jayne, 2000 ; Hunter and Schmidt,
1996 ). However, it is important to remember that when supervisors rate job perform-
ance, they incorporate into their ratings both CP behaviors and CWBs, in addition to
core job performance (Orr, Sackett, and Mercer, 1989 ; Rotundo and Sackett, 2002 ).
Hence supervisory ratings refl ect a combination of core job performance and citizen-
ship behaviors, both good and bad. In the case of ratings, low conscientiousness and low
agreeableness lead to poorer citizenship behaviors, which lead to lower ratings of overall
performance. For the typical job, the weight on conscientiousness in predicting objectively
measured core job performance is only 20% as large as the weight on GMA. In predict-
ing supervisory ratings of job performance, it is 40% as large (Schmidt, Shaffer, and
Oh, 2008 ).
W HAT IS REQUIRED TO MAKE THIS PRINCIPLE WORK?
There are three conditions that are required to make this principle work. That is, there
are three conditions that are required for companies to improve job performance levels by
using GMA in hiring and to reap the resulting economic benefi ts.
c01.indd 6c01.indd 6 6/17/09 2:26:28 PM6/17/09 2:26:28 PM
SELECT ON INTELLIGENCE 7
Selectivity
First, the company must be able to be selective in who it hires. For example, if the labor
market is so tight that all who apply for jobs must be hired, then there can be no selection
and hence no gain. The gain in job performance per person hired is greatest with low
selection ratios. For example, if one company can afford to hire only the top scoring 10%,
while another must hire the top scoring 90% of all applicants, then with other things
equal the fi rst company will have a much larger gain in job performance.
There is another way to look at this: companies must provide conditions of employ-
ment that are good enough to attract more applicants than they have jobs to fi ll. It is
even better when they can go beyond that and attract not only a lot of applicants, but
the higher ability ones that are in that applicant pool. In addition, to realize maximum
value from GMA - based selection, employers must be able to retain the high performing
employees they hire.
Measuring general mental ability
Second, the company must have some way of measuring GMA. The usual and best
procedure is a standardized employment test of general intelligence, such as the
Wonderlic Personnel Test. Such tests are readily available at modest cost. Less valid are
proxy measures such as grade point average (GPA) or class rank. Such proxy measures
are partial measures of intelligence. Also, intelligence can be assessed to some extent
during the employment interview (Huffcutt, Roth, and McDaniel, 1996 ), although this
is a much less valid measure of GMA than a standardized written test.
Variability in job performance
Third, the variability in job performance must be greater than zero. That is, if all appli-
cants after being hired would have the same level of job performance anyway, then noth-
ing can be gained by hiring “ the best. ” This condition is always met. That is, on all jobs
studied there have been large differences between different workers in quality and quan-
tity of output. Hunter, Schmidt, and Judiesch ( 1990 ) meta - analyzed all available studies
and found large differences between employees. In unskilled and semi - skilled jobs, they
found workers in the top 1% of performance produced over three times as much output
as those in the bottom 1%. In skilled jobs, top workers produced 15 times as much as bot-
tom workers. In professional and managerial jobs, the differences were even larger. These
are very large differences, and they are the reason it pays off so handsomely to hire the
best workers.
There is another advantage to hiring the best workers: the pool of talent available for
future promotion is greatly increased. This is of great value to employers, because it helps
ensure high performance all the way up through the ranks of managers. When the right
people are promoted, their value to the fi rm in their new jobs is even greater than in their
original jobs. Thus selection of high ability people has implications not only for the job
they are hired onto, but for other jobs in the organization, too.
c01.indd 7c01.indd 7 6/17/09 2:26:29 PM6/17/09 2:26:29 PM
8 FRANK L. SCHMIDT
A RE THERE EXCEPTIONS TO THIS PRINCIPLE?
As long as the three conditions described above are met, there are no known exceptions to
this principle. That is, there are no known cases or situations in which it is inadvisable
to select employees for general intelligence.
However, there are some people, particularly labor leaders, who believe there is an
exception. These people believe that companies should not select on mental ability if they
can select on job experience instead. That is, they believe that job experience is a bet-
ter predictor of job performance than general intelligence. What does research show?
For applicants with job experience of between none and fi ve years, experience is a good
predictor of job performance. But in the range of higher levels of experience, say from
fi ve to 30 years of job experience, job experience does not predict performance very well
(Schmidt, Hunter, Outerbridge, and Goff, 1988 ; Hunter and Schmidt, 1996 ). On most
jobs, once people have about fi ve years of experience, further experience does not contrib-
ute much to higher performance. This is probably because experience beyond fi ve years
does not lead to further increases in job knowledge. This, in turn, may be due to the fact
that after fi ve years of on - the - job learning, people in the typical job are forgetting job
knowledge about as fast as they are learning new job knowledge.
Another important fact is this: even for new hires in the one to fi ve year range of job experi-
ence, where experience is a valid predictor of job performance, the validity declines over time.
That is, experience predicts performance quite well for the fi rst three years or so on the job and
then starts to decline. By 12 years on the job, experience has low validity. But GMA continues
to predict job performance quite well even after people have been on the job 12 years or more.
What this means is that job experience is not a substitute for GMA. In the long run,
hiring on intelligence pays off much more than hiring on job experience (Hunter and
Schmidt, 1996 ). So if you had to choose, you should choose GMA. However, typically,
you do not have to choose; more than one procedure can be used. It may be desirable to
use both experience and GMA in hiring; as discussed later, it is usually best to use multiple
hiring methods. But in this case, the weighting given to GMA should be higher than the
weighting given to job experience.
I SSUES IN IMPLEMENTING AN ABILITY - BASED HIRING SYSTEM
Can intelligence be too high?
One issue is whether an applicant can have too much intelligence for a job. Recently,
an applicant was rejected for a job as a police offi cer in a New Jersey city on grounds that
his intelligence test score was too high! This city believed something that many people
believe: that intelligence leads to better job performance but only up to a point . After that, more
intelligence leads to lower job performance. Hundreds of studies have shown that this is
false. Higher intelligence leads to better job performance up to the highest levels of intel-
ligence (Coward and Sackett, 1990 ). There is a straight line (linear) relationship between intelli-
gence and job performance. Why do so many people believe otherwise? Probably because
they imagine a university professor or a medical doctor working as a janitor, and they think
“ This person would be so bored with this job that he would do a poor job. ” They forget
that the university professor or doctor would never apply for the janitor ’ s job to begin with.
c01.indd 8c01.indd 8 6/17/09 2:26:29 PM6/17/09 2:26:29 PM
SELECT ON INTELLIGENCE 9
Among people who actually apply to get real jobs, there is a straight line relationship between
intelligence and performance; the higher the intelligence, the better the job performance.
Hence, we do not have to worry about hiring people who are too intelligent for the job.
Does only intelligence matter in jobs?
A second issue is the one alluded to earlier: Although intelligence is the best predictor
of job performance, it does not follow that use of intelligence alone in hiring is the best
way to select people. In fact, it is well known that other predictors can be used along with
intelligence to produce better predictions of job performance than intelligence alone. For
example, for most jobs an intelligence test combined with an integrity test (a composite
personality of conscientiousness, emotional stability, and agreeableness) is 20% more valid
than an intelligence test alone. Adding a structured employment interview to an intelli-
gence test increases validity by 14% (Oh, Schmidt, and Shaffer, 2008 ). It is almost always
possible to add supplementary measures that increase validity. Some of these measures
are discussed in other chapters in this book (e.g. Barrick and Mount ’ s chapter on selection
of conscientiousness and emotional stability).
Are there legal risks in selecting for intelligence?
A third issue is the potential for legal risks. Members of some minority groups, particularly
blacks and Hispanics, typically have lower average scores on GMA tests, leading to lower
hiring rates. Government agencies such as the Equal Employment Opportunity Commission
refer to these lower hiring rates as “ adverse impact. ” The term adverse impact is deceptive,
because it implies that the GMA tests create the difference in test scores, when in fact the
tests only measure real pre - existing differences in mental skills. This is shown by the fact
that minorities and non - minorities with the same test scores have the same level of later
job performance. That is, the test scores predict equally accurately for all groups; they are
predictively fair or unbiased (Schmidt, 1988 ; Wigdor and Garner, 1982 ).
Despite this fact, a lower hiring rate for minorities does sometimes lead to lawsuits.
Employers can win these suits by demonstrating that the tests are valid predictors of job
performance. Today, such demonstrations rely increasingly on summaries of the kinds
of research fi ndings discussed in this chapter, rather than on studies conducted by the
employer. (This is part of the move away from the theory of situational specifi city, dis-
cussed earlier.) Since around the mid 1980s, employers have been winning more and more
such suits, and today they prevail in 80% or more of such suits. Research shows that the
value of the increases in job performance from good selection overshadows any potential
legal costs stemming from defending against such suits. But a key fact is that today there
are far fewer such suits to begin with. Currently, less than 1% of employment - related law-
suits are challenges to selection tests or other hiring procedures. This is almost certainly
due to the greatly reduced chances of winning such suits.
Political risks
However, this does not mean that all employers are willing to use intelligence tests in hir-
ing. Although the percentage of employers using GMA tests has been increasing, some
c01.indd 9c01.indd 9 6/17/09 2:26:29 PM6/17/09 2:26:29 PM
10 FRANK L. SCHMIDT
fi rms view even the possibility of a lawsuit as a public relations disaster. They feel that
even if they win, they still lose on the public relations front. And they believe that public
relations problems can reduce sales and profi ts. These fi rms – mostly larger companies
that sell directly to consumers – are willing to tolerate lower levels of job performance
to avoid even the possibility of such a problem. Unfortunately for such fi rms, not using
GMA tests does not remove the possibility of lawsuits. Other selection procedures also
produce “ adverse impact. ” Employers have tried to reduce adverse impact by introducing
various forms of minority preferences in hiring, but courts have recently begun to strike
down many forms of minority preferences. For example, under the 1991 Civil Rights Act,
it is illegal to adjust test scores or other scores to equalize minority and non - minority hir-
ing rates. This issue is one that will probably remain unsettled for some time.
Many fi rms that rarely use written GMA tests build oral GMA tests into the interview
process. For example, in many employment interviews at Microsoft, job applicants are asked
to solve complex mental puzzles that require high GMA to answer correctly. In fact, even
ordinary job interviews have been found to be correlated with GMA scores (Huffcutt et al.,
1996 ). And, as would be expected from this fact, it has recently been found that even ordi-
nary job interviews show larger minority – majority differences (and thus “ adverse impact ” )
than was previously believed to be the case (Roth, Bobko, Switzer, and Dean, 2001 ).
The effect of testing for intelligence on employee attitudes
A fi fth issue is whether the use of mental ability tests turns off applicants. Some have
argued that applicants do not like to take ability tests. However, surveys of applicant
attitudes reveal that they view mental ability and GMA tests as generally relevant to job
performance (more so than they do personality, bio - data, and integrity tests, for example),
and that they do not have a negative attitude toward such tests (Hausknecht, Day, and
Thomas, 2004 ). It also appears to be the case that when GMA or other ability tests are
used, applicants view the selection requirements as being higher and this increases the
status of the job and hence its attractiveness. That is, something that is harder to attain is
viewed as being more valuable.
The economic value of hiring on intelligence
A fi nal issue is whether the economic value of the job performance gains from GMA -
based hiring is cancelled out by higher wages and salaries. The argument is that if a fi rm
hires more intelligent people, they will have to pay them more and this will cancel out the
gains from the increased job performance. However, in most cases it appears that there is
no increase in compensation costs, at least initially. This is especially likely to be the case
when few of the fi rm ’ s competitors use GMA measures in their hiring. Typically, there is a
pool of available applicants in the area for a particular type of job, and the higher GMA
applicants have no immediate effective way to command higher initial wages.
However, after some time on the job, when higher GMA employees have developed
high levels of performance, the employer can afford to share some of these gains with such
employees in the form of higher wages or salaries. In some cases, this might be necessary
to retain high performing employees. In any event, the payoff to the employer in terms of
enhanced job performance is much greater than any increase in compensation cost.
c01.indd 10c01.indd 10 6/17/09 2:26:29 PM6/17/09 2:26:29 PM
SELECT ON INTELLIGENCE 11
Although most employers, for most jobs, do not pay different people in the same job at
different rates, they do typically promote the top workers to higher level jobs, and this does
result in higher pay. But at promotion the value of the worker ’ s performance to the fi rm
increases much more than the worker ’ s pay, creating another large net benefi t to the fi rm of
good selection. On the other hand, employers that hire only mediocre or poor workers at
entry level fi nd that their higher level jobs also become fi lled with mediocre or poor per-
formers. Again, as noted earlier, selection based on GMA improves performance not only
in the job in question, but also later in higher level jobs in the fi rm.
C ASE EXAMPLES
We will fi rst look at two negative examples and then examine two positive examples of
real world applications of GMA - based hiring.
US Steel plant at Fairless Hill, PA
Up until 1978, the US Steel plant at Fairless Hills, PA, selected applicants into their skilled
trades apprentice programs based on the applicants ’ total scores of a battery of ability
tests. These total scores were a good measure of GMA, and selection was from the top
down. The plant maintained apprentice programs in the wide variety of skilled trades
needed to run a steel mill: machinists, tool and die makers, electricians, sheet metal workers,
etc. The local unit of the United Steelworkers Union, however, did not like this selection
method. In negotiations with the union, the company agreed to modify the selection sys-
tem. In the new system, all applicants who scored above a low cut - off on each test, set at
about the 7th grade level, were considered equally qualifi ed and eligible for hire. Only a
few applicants were screened out by this procedure. Applicants in the passing group were
selected based on plant seniority only. Hence, this plant went from a GMA - based hiring
system to one in which GMA played only a very minor role.
The apprentice training center at Fairless Hills was a well - run facility that kept excel-
lent records of apprentice performance from both before and after the change in the
selection system. These records showed that after the new selection system was intro-
duced, performance plummeted. Scores on the mastery tests of amount learned in train-
ing declined markedly. The fl unk - out and drop - out rates increased dramatically. The
training time and training costs of those who did make it through the program increased
substantially – because many apprentices had to retake multiple units in the training. And
fi nally, the ratings of later performance on the job out in the plant declined.
This was a well - controlled natural quasi - experiment. The only change made was the
lowering of mental ability standards in selection. The training program and the tests
given in the program remained the same. The decline in performance was clearly due to
the lower intelligence of the new apprentices.
The Washington, DC police force
Up until the mid 1980s the Washington, DC police force was one of the best in the USA.
Applicants were selected for Police Academy training based on a general intelligence test
c01.indd 11c01.indd 11 6/17/09 2:26:30 PM6/17/09 2:26:30 PM
12 FRANK L. SCHMIDT
constructed for the District of Columbia by the US Offi ce of Personnel Management
(OPM), as required by then existing Congressional regulations. This test had been chal-
lenged legally and the case had gone all the way to the US Supreme Court, where it had
been upheld. A background investigation was also part of the selection process. The mayor
of Washington, Marion Barry, repeatedly voiced opposition to both the test and the back-
ground check on grounds that the failure rate on both was higher for blacks. In 1987, when
Congress relinquished control over the selection process to the Mayor ’ s offi ce, Barry took
responsibility for the selection process out of OPM ’ s hands. He then eliminated both the
GMA test and the background test. The replacement selection process was somewhat
unclear, but reputedly involved fairly perfunctory interviews.
The fi rst consequence was that the fl unk - out rate in the Police Academy soared, with
over 80% of the new hires being incapable of completing the required training. Failure
rates that high were viewed as unacceptable, and so the content of academy training was
“ dumbed down. ” When this reduced the failure rate only slightly, the content was further
dumbed down, and then dumbed down again. This process of successive adjustments ulti-
mately “ solved ” the fl unk - out problem.
However, the police offi cers being produced were incompetent. Large numbers of mur-
der indictments had to be dismissed because the reports written by the offi cers on the scene
were unintelligible, due to the low literacy levels. The solution rate for murder cases, for-
merly one of the highest in the USA, declined precipitously to one of the lowest. Firearms
accidents soared because offi cers did not know how to use their sidearms properly.
Complaints of police abuse and incompetence from citizens soared. In addition, crime on
the police force became quite common. For example, a group of police offi cers was found
to be selling handguns previously confi scated from criminals back to criminals ! These changes
and others are described by Carlson ( 1993a , 1993b ).
In this example, unlike the US Steel example, two things are happening. First, people
low in intelligence are being hired, resulting in plummeting job performance. Second,
criminals are being hired because there was no background investigation to ensure that
they were not, and the result was crime on the police force.
Employment in the federal government
We now turn to a more positive example – or at least a less negative one. For many jobs in
the federal government, people can either be hired from the outside using a GMA test or
they can be promoted from within. When they are promoted from within, GMA tests are
usually not used – although they sometimes are. Instead, people are evaluated based on
records of their education and training and on appraisals by their supervisors of their
performance in their present jobs. These procedures do have some validity but would not
be expected to be as valid as GMA - based hiring.
So we can ask the following question. After people have been on the job some time,
is the job performance higher for those initially selected using a GMA test? Government
researchers at OPM addressed this question in a detailed study of three representative mid -
level government jobs: IRS auditor, social security claims examiner, and customs inspector.
In each of these jobs, people hired both ways had been on the job from fi ve to eight years.
c01.indd 12c01.indd 12 6/17/09 2:26:30 PM6/17/09 2:26:30 PM
SELECT ON INTELLIGENCE 13
The measure of job performance was unusually good: it was the sum of a hands - on work
sample test, a job knowledge test, and supervisory ratings of job performance.
In all three jobs, those selected years earlier using GMA tests had higher job perform-
ance. The average job performance of the non - GMA - selected employees was at the 50th
percentile, while that of the GMA - selected employees was at the 70th percentile. This is a
large difference. If this difference is projected over the federal workforce as a whole, it
amounts to billions of dollars per year in increased output (Schmidt, Hunter, Outerbridge,
and Trattner, 1986 ). We can also look at this another way. Americans expect their federal
government to perform a wide variety of socially important tasks (e.g. administer the
social security program, protect homeland security, run the federal tax system fairly and
accurately, catch people who commit federal crimes, etc.). To the extent that the federal
government hires less competent people, these jobs are done less well. As shown in this
research, failure to select on GMA results in the hiring of less competent people and
produces lower job performance.
This study was a reasonably controlled quasi - experiment. During the study, the rese archers
did not know which employees had initially been selected using a GMA measure and which
had not. The only relevant difference between the two groups of workers was the method by
which they had been hired. This study provides strong evidence that GMA - based hiring pays
off in higher job performance.
The Philip Morris plant in Cabarrus County, North Carolina
The US Employment Service began a new nationwide program of employment testing,
operated through state employment offi ces, in the early 1980s. Like its earlier program,
it was based on the General Aptitude Test Battery (GATB). One of the three abilities
measured in that program was GMA (the other two were general perceptual ability and
general psychomotor ability). This new program was based on the methods of meta - analysis
or validity generalization that were mentioned at the beginning of this chapter.
The large Philip Morris plant in Cabarrus County, North Carolina, was one of the fi rst
employers to subscribe to this testing program. They signed an agreement under which
the state employment service tested and referred the higher scoring applicants to Philip
Morris for possible hire. For the jobs at Philip Morris, most of the weight was placed on
GMA in determining who was hired.
The human resources department at Philip Morris decided to conduct a study to com-
pare the performance of GATB – GMA - selected workers and workers hired without use
of the test. They found that the GMA - selected workers were superior across a variety of
performance measures. For example, there was a 35% gain in output. The GMA - selected
workers learned 8% more skills during job training, had 25% fewer operator failures and
58% fewer disciplinary actions. The incidence of unsafe job behaviors was 35% less and the
reduction in work days lost to accidents was 82% .
These are large differences. The Philip Morris personnel researchers, Dennis Warmke
and William Van Arnam, noted the employment interview used might have contributed
somewhat to the performance superiority of these workers. However, they stated that
because it was the GMA test that screened out most of the applicants who were not hired,
c01.indd 13c01.indd 13 6/17/09 2:26:30 PM6/17/09 2:26:30 PM
14 FRANK L. SCHMIDT
the GMA test was the dominant infl uence producing the performance improvements.
This research is described in McKinney ( 1984 ).
C ONCLUSION
Higher intelligence leads to better job performance on all jobs, and the increases
in job performance resulting from hiring on GMA have high economic value for
organizations. Higher intelligence causes higher job performance primarily because
it causes people to learn job knowledge faster and to learn more of it. However, intel-
ligence is also used directly on the job to solve performance - related problems, inde-
pendent of prior job knowledge. The primary requirement that an organization must
meet to make GMA - based hiring work well is the ability to attract job applicants and
to retain them once they are hired. Despite beliefs to the contrary, hiring on job experi-
ence is inferior to hiring on GMA. Although GMA is the most important determinant
of job performance, it is not the only determinant. Therefore, fi rms should use other
valid procedures along with GMA. Finally, we have seen four concrete, graphic, real
world examples of the impact of GMA on job performance.
R EFERENCES
Brody , N. ( 1992 ). Intelligence ( 2nd edition ). San Diego, CA : Academic Press Inc .
Brown , K. , Le , H. , and Schmidt , F. L. ( 2006 ). Specifi c aptitude theory revisited: Is there incre-
mental validity for training performance? International Journal of Selection and Assessment, 1 4 ,
87 – 100 .
Carlson , T. ( 1993a ). D.C. blues: The rap sheet on the Washington police . Policy Review ,
Winter, 27 – 33 .
Carlson , T. ( 1993b ). Washington ’ s inept police force . The Wall Street Journal , November 3.
Coward , W. M. , and Sackett , P. R. ( 1990 ). Linearity of ability – performance relationships:
A re - confi rmation . Journal of Applied Psychology , 75 , 295 – 300 .
Cooper - Hakim , A. , and Viswesvaran , C. ( 2005 ). The construct of work commitment: Testing
an integrative framework . Psychological Bulletin , 131 , 241 – 259 .
Dalal , R. S. ( 2005 ). A meta - analysis of the relationship between organizational citizen-
ship behavior and counterproductive work behavior . Journal of Applied Psychology, 90 ,
1241 – 1235 .
Gottfredson , L. S. ( 1996 ). Why g matters: The complexity of everyday life . Intelligence , 24 ,
79 – 132 .
Hausknecht , J. P. , Day , D. V. , and Thomas , S. C. ( 2004 ). Applicant reactions to selection
procedures: An updated model and meta - analysis . Personnel Psychology , 57 , 639 – 683 .
Herrnstein , R. J. , and Murray , C. ( 1994 ). The Bell Curve: Intelligence and Class Structure in
American Life. New York : The Free Press .
Huffcutt , A. I. , Roth , P. L. , and McDaniel , M. A. ( 1996 ). A meta - analytic investigation of
cognitive ability in employment interview evaluations: Moderating characteristics and
implications for incremental validity . Journal of Applied Psychology , 81 , 459 – 473 .
Hunter , J. E. ( 1986 ). Cognitive ability, cognitive aptitudes, job knowledge, and job per-
formance . Journal of Vocational Behavior , 29 , 340 – 362 .
c01.indd 14c01.indd 14 6/17/09 2:26:30 PM6/17/09 2:26:30 PM
SELECT ON INTELLIGENCE 15
Hunter , J. E. , and Hunter , R. F. ( 1984 ). Validity and utility of alternate predictors of job
performance . Psychological Bulletin , 96 , 72 – 98 .
Hunter , J. E. , and Schmidt , F. L. ( 1996 ). Intelligence and job performance: Economic and
social implications . Psychology, Public Policy, and Law , 2 , 447 – 472 .
Hunter , J. E. , and Schmidt , F. L. ( 2004 ). Methods of Meta - analysis: Correcting Error and Bias in
Research Findings. Thousand Oaks, CA : Sage .
Hunter , J. E. , Schmidt , F. L. , and Judiesch , M. K. ( 1990 ). Individual differences in output
variability as a function of job complexity . Journal of Applied Psychology , 75 , 28 – 42 .
Hunter , J. E. , Schmidt , F. L. , and Le , H. ( 2006 ). Implications of direct and indirect range
restriction for meta - analysis methods and fi ndings . Journal of Applied Psychology, 91 ,
594 – 612 .
Hunter , J. E. , Schmidt , F. L. , Rauschenberger , J. M. , and Jayne , M. ( 2000 ). Intelligence,
motivation, and job performance . Chapter in C. L. Cooper and E. A. Locke (eds), I/O
Psychology: What We Know about Theory and Practice . Blackwell Publishers , pp. 278 – 303 .
Jensen , A. R. ( 1998 ). The g Factor: The Science of Mental Ability . Westport, CT : Praeger .
J u d g e , T. A. , Thoresen , C. J. , Bono , J. E. , and Patton , G. K. ( 2001 ). Another look at the rela-
tionship between job satisfaction and job performance . Psychological Bulletin , 127 , 376 – 407 .
McHenry , J. J. , Hough , L. M. , Toquam , J. L. , Hanson , M. L. , and Ashworth , S. ( 1990 ).
Project A validity results: The relationship between predictor and criterion domains .
Personnel Psychology , 43 , 335 – 354 .
McKinney , M. W. ( 1984 ). Final report: Validity generalization pilot study . Southern Test
Development Field Center, US Employment Service, Raleigh , North Carolina .
Oh , I. , Schmidt , F. L. , and Shaffer , J. ( 2008 ). Comprehensive updating of the validity and
incremental validity of selection procedures based on 95 years of research . Working
paper, Department of Management, Tippie College of Business, University of Iowa,
Iowa City, Iowa .
Olea , M. M. , and Ree , M. J. ( 1994 ). Predicting pilot and navigator criteria: Not much
more than g . Journal of Applied Psychology , 79 , 845 – 851 .
Ones , D. S. , Viswesvaran , C. , and Schmidt , F. L. ( 1993 ). Comprehensive meta - analysis of
integrity test validities: Findings and implications for personnel selection and theories
of job performance . Journal of Applied Psychology , 78 , 679 – 703 .
Orr , J. M. , Sackett , P. R. , and Mercer , M. ( 1989 ). The role of prescribed and nonpre-
scribed behaviors in estimating the dollar value of performance . Journal of Applied
Psychology , 74 , 34 – 40 .
Ree , M. J. , and Earles , J. A. ( 1991 ). Predicting training success: Not much more than g .
Personnel Psychology , 44 , 321 – 332 .
Ree , M. J. , and Earles , J. A. ( 1992 ). Intelligence is the best predictor of job performance .
Current Directions in Psychological Science , 1 , 86 – 89 .
Ree , M. J. , Earles , J. A. , and Teachout , M. ( 1994 ). Predicting job performance: Not much
more than g . Journal of Applied Psychology , 79 , 518 – 524 .
Roth , P. I. , Bobko , P. , Switzer , F. S. , and Dean , M. A. ( 2001 ). Prior selection causes biased
estimates of standardized ethnic group differences: Simulation and analysis . Personnel
Psychology , 54 , 591 – 617 .
Rotundo , M. , and Sackett , P. R. ( 2002 ). The relative importance of task, citizenship, and
counterproductive performance to global ratings of job performance: A policy - capturing
approach . Journal of Applied Psychology , 87 , 66 – 80 .
c01.indd 15c01.indd 15 6/17/09 2:26:30 PM6/17/09 2:26:30 PM
16 FRANK L. SCHMIDT
Salgado , J. F. , Anderson , N. , Moscoso , S. , Bertua , C. , and de Fruyt , F. ( 2003b ). International
validity generalization of GMA and cognitive abilities: A European community meta -
analysis . Personnel Psychology , 56 , 573 – 605 .
Salgado , J. F. , Anderson , N. , Moscoso , S. , Bertua , C. , de Fruyt , F. , and Rolland , J. P.
( 2003a ). A meta - analytic study of general mental ability validity for different occu-
pations in the European community . Journal of Applied Psychology , 88 , 1068 – 1081 .
Schmidt , F. L. ( 1988 ). The problem of group differences in ability scores in employment
selection . Journal of Vocational Behavior , 33 , 272 – 292 .
Schmidt , F. L. ( 1992 ). What do data really mean? Research fi ndings, meta analysis, and
cumulative knowledge in psychology . American Psychologist , 47 , 1173 – 1181 .
Schmidt , F. L. ( 2002 ). The role of general cognitive ability in job performance: Why there
cannot be a debate . Human Performance , 15 , 187 – 210 .
Schmidt , F. L. , and Hunter , J. E. ( 1981 ). Employment testing: Old theories and new
research fi ndings . American Psychologist , 36 , 1128 – 1137 .
Schmidt , F. L. , and Hunter , J. E. ( 1992 ). Development of causal models of processes
determining job performance . Current Directions in Psychological Science , 1 , 89 – 92 .
Schmidt , F. L. , and Hunter , J. E. ( 1998 ). The validity and utility of selection methods in
personnel psychology: Practical and theoretical implications of 85 years of research
fi ndings . Psychological Bulletin , 124 ( 2 ), 262 – 274 .
Schmidt , F. L., and Hunter , J. E. ( 2004 ). General mental ability in the world of work:
Occupational attainment and job performance . Journal of Personality and Social Psychology , 86 ,
162 – 174 .
Schmidt , F. L. , Hunter , J. E. , and Outerbridge , A. N. ( 1986 ). The impact of job experi-
ence and ability on job knowledge, work sample performance, and supervisory ratings
of job performance . Journal of Applied Psychology , 71 , 432 – 439 .
Schmidt , F. L. , Hunter , J. E. , Outerbridge , A. N. , and Goff , S. ( 1988 ). The joint relation
of experience and ability with job performance: A test of three hypotheses . Journal of
Applied Psychology , 73 , 46 – 57 .
Schmidt , F. L. , Hunter , J. E. , Outerbridge , A. N. , and Trattner , M. H. ( 1986 ). The econ-
o mic impact of job selection methods on the size, productivity, and payroll costs of the
Federal work - force: An empirical demonstration . Personnel Psychology , 39 , 1 – 29 .
Schmidt , F. L. , Hunter , J. E. , and Pearlman , K. ( 1980 ). Task difference and validity of
aptitude tests in selection: A red herring . Journal of Applied Psychology , 66 , 166 – 185 .
Schmidt , F. L. , Oh , I. , and Le , H. ( 2006 ). Increasing the accuracy of corrections for range
restriction: Implications for selection procedure validity and other research fi ndings .
Personnel Psychology , 59 , 281 – 305 .
Schmidt , F. L. , Shaffer , J. A., and Oh, I. ( 2008 ). Increased accuracy for range restriction
corrections: Implications for the role of personality and general mental ability in job
and training performance . Personnel Psychology , 61 , 827 – 868 .
Schmidt , F. L. , Ones , D. S. , and
Hunter , J. E. ( 1992 ). Personnel selection . Annual Review of
Psychology , 43 , 627 – 670 .
Wigdor , A. K. , and Garner W. R. (eds) ( 1982 ). Ability Testing: Uses, Consequences, and Controversies
(Report of the National Research Council Committee on Ability Testing). Washington,
DC : National Academy of Sciences Press .
Wilk , S. L. , Desmarais , L. B. , and Sackett , P. R. ( 1995 ). Gravitation to jobs commensurate
with ability: Longitudinal and cross - sectional tests . Journal of Applied Psychology , 80 , 79 – 85 .
Wilk , S. L. , and Sackett , P. R. ( 1996 ). Longitudinal analysis of ability – job complexity fi t
and job change . Personnel Psychology , 49 , 937 – 967 .
c01.indd 16c01.indd 16 6/17/09 2:26:31 PM6/17/09 2:26:31 PM
SELECT ON INTELLIGENCE 17
EXERCISES
Hiring offi ce workers
You are the human resources director at a large fi rm and you are faced with designing a
system for hiring offi ce workers. An offi ce manager comes to you and says the fi rm should
not use written GMA tests because of the danger of law suits. He says he knows GMA is
important to job performance but maintains that you can use “ GMA - loaded ” interviews
to measure GMA and thus get the benefi t of using GMA without leaving a “ paper trail ”
of test scores that could stimulate a law suit. Respond to this manager based on what
you learned from this chapter. What would you tell him? What is the foundation for your
response?
Educating the CEO
You are the human resources director in your organization. The CEO calls you to her
offi ce for a meeting and tells that she knows from 35 years of experience in dealing with
people that the key determinant of high job performance is personal values and sense
of responsibility. She says she would like to have all hiring in the company done using
measures of values and sense of responsibility. Based on what you learned in this chap-
ter, what would you tell her? What is the basis for the position you are taking?
c01.indd 17c01.indd 17 6/17/09 2:26:31 PM6/17/09 2:26:31 PM
c01.indd 18c01.indd 18 6/17/09 2:26:31 PM6/17/09 2:26:31 PM