You are on page 1of 3

7.

Fairness

Unfair discrimination = when persons with equal probabilities of success on the job have
unequal probabilities of being hired for the job.
Although fairness is a socially constructed concept and is defined in different ways, test bias is a
psychometric concept and it has been defined quite clearly.

Concept of differential validity if distinct from differential prediction.


Equal correlations do not necessarily imply equal standard errors of estimate, nor do they
necessarily imply equal slopes or intercepts of group regression equations.

Differential validity = when the scores of a test have significantly different degrees of validity
for different groups of people (e.g., males vs. females).
Figure 8-1 and 8-2, p. 169: no differential validity.

An important consideration in assessing differential validity is whether the test in question


produces adverse impact.
Adverse impact = members of one group are selected at substantially greater rates than
members of another group. To understand whether that is the case, compare selection ratios
across groups.

SR1 (selection ratio 1)/ SR 2 = Adverse impact ratio. 80% rule: if adverse impact ratio < .80 à
adverse impact. 

Figure 8-3, p. 170: Adverse Impact is justifiable, since minority groups score lower.
Figure 8-4, p. 170: Valid predictor for whole group, invalid for each group separately.
Figure 8-5, p. 171: Equal validity, unequal predictor means. As long as the expectancy of
success on the job is equal for the two groups, the use of separate cut off scores is justified.
Figure 8-6, p. 171: Similar validity, dissimilar criterion means.
Figure 8-7, p. 172: Only valid for majority.
Figure 8-8, p. 172: Dissimilar criterion means, valid only for majority.
à There are numerous possibilities when heterogeneous groups are combined in making
predictions. When there is differential validity, the use of a single regression line, cut score, or
decision rule can lead to serious errors in prediction.

Evidence of differential validity provides information only on whether a selection device should
be used to make comparisons within groups. A selection procedure may be fair and yet predict
performance inaccurately and vice versa.

Differential Validity exists when:


1). There is a significant difference between the validity coefficients obtained for two subgroups;
2). The correlation found in one or both of these groups are significantly different from zero.

Single-group validity = in which a given predictor shows validity significantly different from
zero for one group only, and there is no significant difference between the two validity
coefficients. 
Using tests with adverse impact can lead to negative organizational and societal consequences
and perceptions of test unfairness. It is recommended that, when adverse impact is found, HR
specialists strive to use alternative tests with similar levels of validity, but less adverse impact. It
would be more efficient to reduce adverse impact by using available testing procedures.
Countering/ reducing adverse impact:

 Improve recruiting strategy for minorities (reduce also the overall validity);
 Use cognitive abilities in combination with non-cognitive predictors;
 Use multiple regression and other methods for combining predictors into a composite;
 Use more specific than general cognitive predictors (differences are smaller for specific
abilities);
 Use differential weighting for the various criterion facets;
 Use alternate modes of presenting test stimuli (using formats that do not have heavy
reading and verbal requirements, such as video-based tests);
 Enhance face validity (increasing motivation can help reduce adverse impact);
 Implement test-score banding* (to select among the applicants).

 Adverse impact may occur even when there is no differential validity across groups.

Tests are never perfectly reliable, and the relationship between test scores and criteria is never
perfect. Test-score banding* = is decision-making process that is based on these two grounds.

Sliding-band method = an attempt to bring together economic and social objectives within the
framework of generally accepted procedures for testing hypotheses about differences in
individual test scores.

Criterion-reference banding model = Procedure that incorporates not only reliability


information for the predictor but also reliability information for the criterion and the explicit
relationship between the predictor and criterion scores.

Differences between comparison of predictor-referenced bands and criterion-referenced bands:

 Use of validity evidence;


 Bandwidth;
 Inclusion of criterion information;
 Use of reliability information.

The interaction of examiner with examinee should be professional, courteous, caring and
respectful. Attention to these aspects of test use and interpretation is no less important than more
technical concerns. To understand the fairness and impact of the selections systems in place, it is
necessary not only to conduct technical analyses on the data but also to take into account the
perceptions of people who are subjected to the system.
From the perspective of applicants and test takers, there are 2 dimensions of fairness:
1). Distributive (perceptions of fairness of the outcomes);
2). Procedural (perceptions of fairness of the procedures used to reach a hiring decision).
Applicants’ personality profiles are related to their perceptions of fairness:
Higher on neuroticism à more negative perceptions;
Higher on agreeableness à more positive perceptions.
 

Interpersonal Context
Employers do have control of the procedures implemented and can improve the perceived
fairness of the testing process. Although tests may be technically fair and lack predictive bias,
the process of implementing testing and making decisions can be such that applicants perceive
unfairness.

Public Policy
Public interest in measurement include three functions:
1). Diagnosing needs;
2). Assessing qualifications to do;
3). Protecting against false credentials.
We have moved from naive acceptance of tests, through a period of intense aggression to tests, to
a higher acceptance of tests. 
Tests serve as instruments of public policy, and public policy must be evaluated periodically.

Good policy is not for or against tests; what really matters is how tests are used.

The use of cognitive abilities tests in combination with other predictors provides the highest level
of accuracy in predicting future performance.

You might also like