Running Head: PERSONNEL TESTING

Personnel Testing:

Predicting Occupational Personality Test Scores

Loyola Marymount University

Hugo Fernandez

Temre Green

Kira Shymanski

Jennifer Zamora

Psychology 542, Section 01

Dr. Mills

October 25, 2001

Furham, A., Drakeley, R. (2000). Predicting Occupational Personality Test Scores. Journal of Psychology, 134, 103-112.

            Predicting Occupational Personality Test Scores examined the relationship between students' actual test scores and their self-perceived test scores based on the Hogan Personality Inventory.  Even though the students were given descriptions of the content being measured, this experiment found that the students overestimated their scores.  

            The theory behind this experiment was based on the belief that students would be able to best predict the concepts they were most familiar with, sociability and intellectance (how interested they are in intellectual matters) and least able to predict prudence.  They also predicted that the students would overestimate their scores about the positive content they were being asked to measure, such as likeability and prudence, but underestimate characteristics such as poor adjustment.

            The participants were all 1-st year psychology students attending a beginning course on personality measurement.  They had all been tested 2 months before coming to the university and none of them had had any prior instruction on personality theory or assessment, nor had they seen this particular instrument (HPI) used before.  These students were a homogeneous group in terms of age, socio-economic status and educational qualifications. 

            The participants completed the HPI.  The results of the test are given in the form of 7 different scores: the Adjustment Scale, the Ambition Scale, the Sociability Scale, the Likeability Scale, the Prudence Scale, the Intellectance Scale, and the School Success Scale.  After the participants completed the instrument, they were asked a week later to estimate their own scores after being given a complete description of the scales that were measured. The results found that the closest approximations were for school success and intellectance, where the poorest correlation was for ambition and adjustment. 

            This study showed that although people overestimate their actual test scores, they are fairly good at predicting their own personality scores.  Although, overall, it seemed the scores were easy to predict, some were easier to predict than others were.  For example, adjustment, ambition and sociability were easier to predict than prudence and school success.  This might be because adjustment, ambition and sociability are more often discussed in an everyday setting than prudence and school success.

            The relevance of these studies is important because personality tests are used frequently in an occupational setting.  Job applicants who are faced with taking a personality test, especially one with high face validity, would be able to predict the qualities that the employer is looking for and therefore alter their scores to fit the employers needs. 

Predicting Occupational Personality Test Scores

I. Interest in and use of Personality Tests

1. By organizations

1.      Recruitment

2.      Selection

3.      Promotional procedures

2. By psychometrics

1.      Test validity in making job performance predictions

3. Human resource managers require face validity in tests, even at the expense of construct, concurrent, and predictive validity

1.      Disadvantages

a.       Easier for test takers to fake results

b.      Easier for test takers to predict their own scores

2.      Advantage

a.       Easier for test purchasers to explain use of test to test takers

4. Several Studies

1.      Vingoe (1966) found introverts to be more aware of their introversion than extraverts were of their extraversion

2.      Furnham (1990a) found students were best at predicting their scores on the dimensions of morningness-eveningness, extraversion and thinking, and wanting and expressed inclusion on the measures Morningness-Eveningness Questionnaire, Myers-Briggs Type Indicator, and Fundamental Interpersonal Relations Orientation-Behavior, respectively

3.      Furnham (1997) found students were best at predicting their conscientiousness, then extraversion, and then neuroticism on their Big Five personality scores

4.      Furnham and Henderson (1983) found subjects to be unable to predict their assertiveness or locus of control scores

5. Questions and theories about Personality Dimensions

1.      Which dimensions are people able or unable to predict with any accuracy?

2.      Are some people significantly better at predicting their own scores than others?

3.      People are probably able to predict scores for dimensions they understand

4.      Concepts that are part of everyday language may be significantly predicted

a.       cognitive framework accessed

5.      Concepts that are not part of everyday language are more difficult to predict

a.       poorly formed cognitive framework

6.      Is accuracy of self-prediction a question of main effects or interaction?

7.      Certain people may detect personality dimensions in others through the manifestations of their own personality

a.       Stresses individual differences

6. This Study

1.      Examines how well people predict their scores on personality tests used in occupational assessment

2.      Predicted subjects could best guess their scores on dimensions with which they were familiar

3.      Predicted subjects would overestimate positive attributes and underestimate negative attributes

II. Participants (N=88)

1. First year students in personality measurement class
2. No prior instruction on personality theory or assessment
3. Homogeneous in terms of:

1.      age

2.      educational qualifications

3.      social class

III. Questionnaire

1. Hogan Personality Inventory (7 scores)

1.      Adjustment Scale

-         Degree of calmness and self-acceptance or tenseness and self-critique a person has

2.      Ambition Scale

-         Degree of confidence, status seeking, & values achievement a person has

3.      Sociability Scale

-         Degree to which a person needs and enjoys social interaction

4.      Likeability Scale

-         Reflects social sensitivity, tact, and perceptiveness

5.      Prudence Scale

-         Concerns self-control and conscientiousness

6.      Intellectance Scale

-         Degree of interest in intellectual matters:

1.)    science

2.)    art

3.)    philosophy

7.      School Success Scale

-         Degree of enjoyment in academic activities and value in education

IV. Procedure

1. Subjects complete questionnaire
2. A week later they estimate their own scores

-         Subjects were thoroughly given and explained:

1.)    How scores on the seven scales worked

2.)    Means and standard deviations from a group similar to themselves

3.)    A similar distribution of scores on a graph

3. Subjects then guessed their own score by writing a cross on the graph where they believed they would fall in the distribution
4. Subjects then received their true scores

V. Results

1. Table 1

1.      Strongest correlations

a.       sociability

b.      adjustment

c.       ambition

2.      Weakest correlations

a.       prudence

b.      likeability

c.       school success

2. Table 2

1.      All estimated-actual correlations were positive and significant

a.       Three over .60 (or high)

1.)    adjustment

2.)    ambition

3.)    sociability

4.)    intellectance

b.      Three under .35

1.)    school success

2.)    intellectance

3.)    openness to experience

2.      Correlation between estimated scores produced only 4 of 21 significant.

3.      Correlation between actual scores were 8 of 21 significant

VI. Discussion

1. Predictability of personality subscale score depends on face validity and use of concepts in everyday interaction

1.      Some subscales seemed easier to predict than others

a.       Adjustment

b.      Ambition

c.       Sociability

2.      Others scales seemed more difficult to predict

a.       Prudence

b.      School Success

2. Well established rating issues

1.      Indicated by

a.       overestimation of scores

b.      restricted use of range

c.       numerous intercorrelations between the estimation measures

2.      Example – Halo Effect

3. Relevance for applied Psychology

1.      Job applicants may be tempted to fake results to fit the profile they believe employers are looking for

2.      Faking on an item-by-item basis may not yield the same scores as faking on a subscale score basis

a.       An ability to predict scale scores does not imply an equivalent ability to fake them

3.      Useful to figure out which tests are easier to fake scores for

4. Major problems

1.      Difficult to set an absolute criterion to which all judgments can be made.

2.      Maybe it is not that people can or can’t predict their personality scores, maybe it is that the instruments used don’t measure them accurately.

a.       Hence this study is essentially a comparison of a formal method of personality assessment and an intuitive method.

5. Two additional points

1.      Subjects predicted scores, not personality per se

a.       Possible to predict scores after observing types of questions used to assess personality

2.      If measure has good construct validity, formal personality scores are reasonable criteria of accuracy

A.     Identify up to three points made by the author that the panel found especially interesting or informative.

1. One interesting point the authors made had to do with the idea that accuracy of self-predictions may be a question of main effects and/or interaction.  “Certain individuals are highly sensitive to detecting personality dimensions (from related social behavior) in others as a function of their own personalities.”  This is interesting not only in the context of this study, but it also helps explain the occurrence of the defense mechanism known as projection.
2. A very informative point is that “if items do not have face validity but do have construct, concurrent, and predictive validity, they should be more difficult to predict and fake.  Yet, test purchasers are loath to use test with low face validity because they feel uncertain as to how to explain the use of the test to the person completing it.”  The authors are saying here that people who use these tests would rather have tests that can be explained more easily to test takers than tests that do a good job of deciphering the more appropriate applicants from the less appropriate applicants in order to save time and money.  This being the case, a lot of these tests are not serving the purpose for which they were designed.
3. Perhaps the most interesting point made by the authors is that “Instead of arguing that a person’s own personality estimations do or do not accurately reflect the assessment of psychological instruments, it may be just as meaningful to argue precisely the opposite, concluding that it is the assessment devices of psychologists that do not seem able to judge a person’s actual personality very accurately.”  The authors are reiterating one of the biggest problems that this type of research possesses – you can’t infer causation from a correlation.

B.     Identify up to three arguments made by the author that the panel either disagreed with and/or for which you think the author made a weak case.  Why?

None

C.     Identify up to three concepts that, even after reading the material, the panel still had questions about, or that the panel would have liked the author to have explained further.

1. The authors said, “Test takers could significantly under- or overestimate their scores yet obtain high (even perfect) correlations.”  What does the author mean by this?  How can someone have high correlations, but still over- and underestimate their scores?  Doesn’t having a high correlation mean that they don’t over- or underestimate their scores?
2. Something else we were unsure of was how the authors reached some of the results that they did.  In the results section, they stated, “ Although the correlation between the estimated subscale scores yielded few significant rs (4 of 21), nearly half of the correlations between the actual scores were significant (8 of 21).”  From the tables they included, it is not clear how they came up with these conclusions, not is it clear where they came up with the numbers 4 of 21 and 8 of 21.
3. The authors explained in the discussion that “The overestimation of scores and the restricted use of the range as well as the numerous intercorrelations between the estimated measures do indicate a number of well-established rating issues, such as the halo effect . . .” and that “ . . . the restriction of range would reduce correlations.”  We are unsure how the restriction of range can indicate well-established rating issues like the halo effect or how a restriction of range would reduce correlations.  It seems almost like the opposite should happen.  Perhaps our understanding of “range” is not appropriate in the context the author is using it, but he doesn’t clarify exactly what he means by it.