The validity of the Reid Report for selection of corrections staff.

Although estimates for industry losses from employee theft and counter-productivity vary across reports, employee honesty and integrity is a serious concern for many employers. The solution for most organizations concerned with integrity issues has been honesty testing. Sackett, Burris, and Callahan (1) have distinguished between two types of honesty tests: disguised purpose tests and overt integrity tests. Disguised purpose tests use personality-based measures to predict future counterproductivity. Overt tests, also known as clear purpose tests, ask direct questions about applicants' attitudes regarding dishonest behavior. Most, including the Reid Report, request admissions to past transgressions. (2) Some overt integrity tests are also designed to measure drug use, job satisfaction, burnout, potential for violence, etc. Still, overt tests are generally designed to measure a narrow construct--namely, theft potential.

Honesty tests eliminate applicants by considering responses in relation to a "profile" of employee thieves. These potential criminals think about stealing more often than others, are more tolerant and less punitive towards others who steal, and accept rationalizations for theft. Overt integrity tests make direct inquiries relating to these attitudes such as "I often think about stealing something even though I do not actually take it."

Factor analysis of the Reid Report, one of the three most popular overt integrity tests, shows that the test contains two major factors that Cunningham, Wong, and Barbee (3) have labeled "punitiveness" and "projectiveness." The "punitiveness factor" contends that honest people maintain high personal standards and expect relatively severe punishment for those who are dishonest. The "projectiveness factor" has items that indicate honest individuals project their honesty and believe that most others are honest as well.

Integrity Test Validity

Early validity research of integrity tests compared scaled scores to admissions in polygraph interviews and to polygraph outcomes (i.e., pass-fail). Sackett and Harris (4) found validity coefficients ranging from .29 to .86 using polygraphs as the criterion, with a mean validity coefficient of .49. Serious questions exist, however, about the usefulness of polygraph tests as criterion. Another popular but questionable criterion for evaluating overt tests has been anonymous admissions of past theft. Sackett and Harris found validity coefficients ranging from .29 to .86 when theft admissions were the criterion. However, they warn that numerous factors are likely to influence correlations using this criterion. For example, some studies include admissions in both the predictor and the criterion. Thus, correlations with self-reports are potentially enhanced by common method variance. (5)

Guastello and Rieke (6) warn that use of the theft admission criterion is especially susceptible to the effects of "faking good." Since overt integrity tests do very little to disguise their intent, it is possible for applicants to lie or assume an honest "role" when completing the form. A study by Lobello and Sims (7) found that convicted felons instructed to "fake good" on an honesty test significantly outscored both those instructed to be honest and those given no instructions. Also, correlations of .43 have been found between the Reid Report and the 16PF faking good scale. (8) Therefore, Guastello and Rieke argue that correlations between subjects' responses to attitude questions and their admissions to past transgressions may be a better measure of candor or ingenuousness than dishonesty. They contend that scoring any overt integrity test without correcting for faking is useless for predicting behavior.

A better criteria to examine integrity test validities has been external, nonpolygraph criteria. Such criteria can include production records, supervisory ratings, absences, turnover, disciplinary action, and detected theft on the job. (9) As would be expected, validities for correlations using external criteria are not nearly as high as those that use admissions of past behaviors. In their meta-analysis of integrity test validities, Ones, Viswesvaran, and Schmidt (10) report that overt integrity tests predict supervisory ratings with a true validity of .30. They also found a mean true validity of .47 for prediction of a broad range of counterproductive behaviors. Naturally, higher correlations can be expected when a broad range of criteria are considered. (11)

Lack of criterion variance is the major limiting factor in what most psychologists would agree is the best validation technique for these tests: correlation between predicted theft and detected theft. (12) Without exception, this validation technique finds significant differences between terminated employees and the average test score. (13) However, the low percentage of terminated employees requires the use of very large sample sizes to make compelling arguments about the test's validity. Often, predictive studies must use more accessible criteria such as average cash register shortage to indirectly measure the effects of employee dishonesty. (14) Time series designs have also contributed to the literature on integrity test validities. In these studies, shrinkage of store inventory is measured before and after the implementation of an honesty test on new applicants. (15) A problem with these studies is that they generally lack a control group; therefore, it is difficult to say with any certainty that the introduction of the integrity test caused the reduction in shrinkage. (16) The problem is that employee theft may not be the major contributor to shrinkage; other explanations could include poor bookkeeping, customer theft, spoilage, or misplacement. (17) Still, studies often find a significant reduction in shrinkage within six months to one year of implementing honesty tests. (18)

Construct validity of integrity tests has been examined in a number of ways. Lilienfeld, Andrews, Stone-Romero, and Stone (19) used convicted felons and college students to correlate scores on the Reid Report with scores on personality measures. High scores on the integrity test negatively correlated with negative emotionality (r = -0.41) and positively correlated with constraint (r = 0.40). A similar study conducted by Murphy and Lee (20) found that a correlation existed between the London House PSI and the Hogan Personality Inventory measure of conscientiousness (r = 0.41). Both of these studies make solid arguments for the construct validity of overt integrity tests.

Although there are drawbacks and shortcomings to each of these validation strategies, the vast majority of research on these tests has been consistently positive. (21) Still, one must remember that most of the research has been conducted by the test publishers. (22) This problem is often exacerbated by the proprietary nature of scoring keys and the lack of cooperation by publishers towards independent researchers. (23) Sackett et al. agree that, without more independent research, it cannot be ascertained whether only successful research by the industry gets published.

Validation of a narrowly-based overt integrity test such as the Reid Report poses a unique difficulty in a corrections setting. For corrections officers, potential theft is rarely a concern; officers in most positions have no access to cash or valuable property. Integrity issues center around inappropriate conduct with the inmate population, introduction of contraband, and general counterproductivity. Obviously, these criteria are normally associated with broad-based measures such as disguised purpose tests. Yet, the Public Safety Reid Report is often used for screening corrections officers. Considering this incongruity, the following research question is suggested:

Research Question 1: What is the validity coefficient when using the Reid Report--a narrowly-based integrity test--for the selection of corrections officers?

False Positive Debate

Outside of validation strategies, one of the most controversial topics in integrity testing is the high false positive rate. Recommended cut-off scores for integrity tests eliminate between 25 and 75 percent of applicants, yet the percentage of theft-related terminations is very low. (24) Proponents of honesty testing argue that a significant number of those who failed also engaged in theft but were not caught.

Martin (25) argues that all psychological tests have a non-zero false positive rate, so the important consideration is a comparison of the paper-and-pencil tests' false positive rate with the rates that would result from other measures. Provided integrity testing has some use in a particular field and the employer has enough applicants to fill positions despite the high rejection rates, these tests have utility for employers and should be used. (26)

Opponents of honesty tests such as Guastello and Rieke (27) contend that, unlike traditional employment tests that measure likelihood of good performance, these tests purport to measure the likelihood of bad performance and falsely accuse applicants of being thieves. By comparing applicants' attitudes on theft to a standard, organizations are denying jobs because the applicants do not agree with the test publisher's belief system. Also, false positive rates are inflated by factors such as applicants' social desirability, level of education and other factors unrelated to integrity. Furthermore, the high rejection rate of these tests implies that one out of every three working Americans is a thief, which borders on paranoia.