TBC: Job productivity and IQ

Arun Gupta (gupta@mrspock.mt.att.com)
Sun, 2 Apr 1995 14:18:42 GMT

[From : The Bell Curve Debate : History, Documents, Opinions ]

Leon Kamin, in his essay "Lies, Damned Lies and Statistics"
addresses the issue of IQ and on-job performance, with the following:

Begin Quote :

Tests of cognitive ability ..... are claimed by Hernnstein and Murray
to be excellent predictors of "job productivity".

Thus an employer concerned with the bottom line would do well to hire,
no matter what the job, those applicants with high IQ test scores:
"the smart busboy will be more productive than the less-smart busboy...."

But how do we measure the "productivity" of an employee ? The vast
majority of studies "validate" the predictive power of IQ tests by
demonstrating that supervisors assign higher ratings to workers with
high test scores.

That fact, of course, tells us that supervisors think highly of workers
with high test scores -- most of whom share various traits (whiteness
is one of them) with most supervisors. It does not necessarily tell
us that high-IQ workers are more productive.

There is also an extensive research literature which demonstrates that
workers with high IQs possess more "job knowledge", as assessed by
written multiple-choice tests. High IQ workers are also more likely
to pass written qualifying examinations given at the end of training
courses for particular jobs. But again, these facts do not demonstrate
that -- once on the job -- high-IQ workers are really more productive.

There have been some studies, many conducted by the military, in
which the criterion for job productivity has involved actual work
samples, or "hands-on" tests. Maier and Hatt, in a technical report
cited by Hernnstein and Murray, explain that "hands-on job performance
tests have intrinsic validity because of their high fidelity to
the skills required to perform job tasks ... [they] are the benchmark
measure for evaluating the job relatedness of surrogate measures of
job performance, such as written tests, ratings and grades".

With an understanding of how psychologists measure job productivity,
we can now follow Herrnstein and Murray as they grapple with the
problem of whether experience on the job can "make up for less

They conclude that "the difference in productivity associated with
differences in intelligence diminishes only slowly and partially.
Often it does not diminish at all. The cost of hiring less
intelligent workers may last as long as they stay on the job."

To arrive at this bleak conclusion, they cite only two studies (both
in the military) which used work samples or hands-on tests. Their
description of one study is false; their description of the other
study is accurate, but incomplete.

Herrnstein and Murray assert that Schmidt et. al. studied armor
repairmen, armor crewmen, supply specialists and cooks "extending
out to five years of experience and using three different measures
of job performance". They indicate that the researchers found
high-IQ workers to begin at higher levels, and to continue to
outstrip low-IQ workers by the same amount, in all jobs, for all
measures, for five years.

That much is basically true, but it obscures an important fact. In
all measures -- work samples, job knowledge tests, and supervisory
ratings -- both high- and low- IQ workers improved steadily with
experience. Thus in work sample scores, a low IQ worker after two
years was about as productive as a high IQ worker after one year of
experience. Facts of that sort are not irrelevant to the productive
utilization of "human capital".

But more : despite Herrnstein and Murray's claim that the study
extended out to five years, 194 of the 1457 workers had had more than
five years experience. The work sample scores of such highly
experienced low-IQ workers had completely caught upt to those of
equally experienced high-IQ workers ! The supervisory ratings of
the experienced low-IQ workers were actually higher than those of
high-IQ workers, although a substantial gap remained in the "job
knowledge" tests. These embarrassments were explained away by the
study's authors with an appeal to "a fluke of sampling error" and
an assertion that "findings in the highest experience group are

The second military study cited by Herrnstein and Murray is that of
Maier and Hiatt. That study was described, accurately enough, as
finding that a difference favoring high-IQ workers persisted over time
when "job knowledge" was the criterion, but disappeared when a
work sample was the measure.

The data in fact indicated that, for both ground radio repairers and
automotive mechanics, high-IQ workers initially outscored low-IQ
workers on both hands-on and written tests. But after four or five
years of experience, the low-IQ workers actually did better on the
hands-on test than those with the high IQs !

On the written test of "job knowledge", low-IQ workers showed no signs
whatever of catching up to the superior multiple-choice testing skills
of their high-IQ betters.

Maier and Hiatt concluded that the military's IQ test was "a valid
predictor of job performance as measured by hands-on tests", but the
content validity of hands-on tests "is sensitive to job experience".
That is a psychometrician's way of saying that after a few years on
the job the correlation between IQ and worker productivity was actually
slightly negative.

This military research, I think, has a genuine and deep meaning.
The kinds of people who don't do well on standardized tests have
some trouble catching on to job requirements in the early going;
but with experience their actual work performance catches up to
that of their more academically talented peers. Their problem appears
to be that even when they are doing the job excellently, they have
no "job knowledge". They don't "know" how to do the job, they just
"do" it; or at least they can't write down what they do know. That,
in the view of Hernnstein and Murray, is sufficient reason to consign
them to unemployment.

End Quote.

-arun gupta