Controversies over integrity testing: Two viewpoints

4 downloads 17702 Views 970KB Size Report
University of Houston. Frank L. ... both overall job performance and counterproductive behaviors on the job (e.g., ..... r=.48, job tryout r ffi.44; biographical data r=.39, assessment center ... We could easily call these tests for "Rule breaking incli-.
JOURNAL OF BUSINESS AND PSYCHOLOGY

Volume 10, No, 4, Summer 1996

CONTROVERSIES OVER INTEGRITY TESTING: TWO VIEWPOINTS Deniz S. Ones University of Houston

Frank L. Schmidt University of Iowa

Chokalingam Viswesvaran Florida International University

David T. Lykken University of Minnesota

ABSTRACt. In the August 1993 issue of Journal of Applied Psychology, Deniz Ones, Vish V3swesvaran, and Frank Schmidt published a monograph on the validity of integrity tests as revealed by meta-analysis. They concluded that integrity tests have substantial and generalizable predictive validity for criteria of both overall job performance and counterproductive behaviors on the job (e.g., disciplinary problems, absenteeism, rule breaking, violence on the job, theft, etc.). During the Fall of 1992, the authors sent a preprint of this study to David Lykken of the University of Minnesota. stimulating the exchange of letters that is reproduced here. The doubts, concerns, and feelin~ about integrity tests expressed by Professor Lykken are likelyto be shared by others. The responses by Professors Schmidt, Ones, and Viswesvaran attempt to show that these worries are unfounded. Recently Ones, Viswesvaran, and Schmldt (1993) published a comprehensive meta-analysis of integrity test validities for predicting supervisory ratings of overall job performance and counterproductive behaviors on the job. Their results indicate t h a t integrity have generalizable substantial predictive validities for both types of criteria. The following

Address correspondenceto Fr..l¢ L. Schmldt,DepartmentofManagementand Organizatious, College of Business Administration,University of Iowa, 651 phillips Hall, Iowa City, Iowa 52242-1323. 487

©1996HumsnSciencesPre~,Inc.

488

JOURNALOF BUSINESSAND PSYCHOLOGY

letters were exchanged between David Lykken of University of 1VIinnesota and the authors, aider FrAn~ Schmldt sent him a preprint of the article that appeared and the August 1993 issue of the Journal of Applied Psychology. We feel that the questions and the issues t h a t Dr. Lykken raises in his letters are likely to be similar to those raised by others who may read our article. Our responses to his letters m a y provide answers to the most frequently asked questions about integrity tests. We have Dr. Lykken's permission to publish his letters to us along with our responses to blm. Here are the four letters. Dear Dr. Schmidt, ThAn]~g for sending me the preprint of your meta-analysis of integrity tests and also your excellent recent article in the American Psychologist. I think the former is, in many ways, a good illustration of the important points you make in the latter. However, I would contend that the meta-analysis of integriW tests also illustrates a couple of disadvantages of formalized meta-analysis t h a t are not mentioned in your article. By explicitly avoiding a "qualitative ~ or evaluative set, you failed to convey to your readers t h a t at least a part of that surprisingly large literature surveyed consists of bad or at least exceedingly dubious in-house studies, studies t h a t have not met the preliminary test of impartial peer review. Some of t h e "validity studies ~ of the Reid Report and the Staunton Survey, for example, employ polygraph screening tests as criteria. They assume either that these polygraph tests are themselves valid, a most dubious assumption, or at least that ~admissions" made during a polygraph test are a reasonable basis on which to separate the honest from the dishonest. But, of course, it m a y well be that the people who make these admissions are the relatively honest ones who wish to be open and forthcoming and that the real villains are the ones who are accustomed to denying wrongdoing. Any alleged in-house validity study that has not at least been written up in standard form and impartially screened for scientific merit is suspect. A non-evaluative meta-analysis of this type applied to the question of polygraph validity would, in my opinion, be worse t h a n worthless because t h a t literature, which I know quite well, varies from execrable to quite good. Combining the lot into one great bouillabaisse can only lead to scientific indigestion. A second problem seems to me to be more serious. Suppose we accept for the sake of argument that at least some subset of integrity tests have a predictive validity on job applicants of.39. According to m y rough calculations, if the true incidence of counterproductive behavior is 50% and the cutting score is chosen to reject 50% of applicants, such a test

DENIZS. ONES,ET AL.

489

would yield about 38% false positives. If the incidence of CB is only 25%, and the cutting score is set accordingly, the false positive rate will be about 88%. Now if most of these tests are really measuring the same thing, and they are used widely for entry-level jobs, then a very large fraction of your and my grandchildren are never going to get one of those jobs. Doesn't that give you pause? I understand the argument that "it is always better to select with some validity than with none~ but I reject it in this case at least. Your friendly local sheriff might argue the same way for preventive detention of ~crime-prone~ adolescents, using the Reid Report as his basis for selection. My answer to the sheriff would be not to select at all; wait until the kid has done something wrong. My answer to McDonalds would be the some, combined with some suggestions about improving security, employee morale, and the like. I enclose a paper on this topic that might interest you. If I publish it somewhere, I shall have to redo it to take account of your meta-analysis. I'd be most interested to get your reaction to these notions.

David T. Lykken Professor Department of Psychology University of Minnesota Minneapolis, Minnesota

Dear Dr. Lykken: ThAnK you for your letter dated October 29th regarding your metaanalysis of integrity test validities.You seem to have misunderstood it or overlooked several things reading our paper. In this letterwe respond to your comments, both in your letter and in your manuscript titled ~Honesty Testing: A n Environmental Impact Assessment ~. . You contend that the database we gathered largely consists of studies employing polygraph screening tests or admissions as criteria.

We explicitly state in our study that we excluded all studies using polygraph admissions as criteria from our database. Such studies cannot be taken to indicate criterion related validity. Of the 665 validity coeffidents that were included in our meta-analyses, only 255 were based on the criterion of admissions. We analyzed 222 validities (across 68,772 data points) where the criterion was overall job performance (either su-

490

JOURNALOF BUSINESSAND PSYCHOLOGY

pervisory ratings or production records) and found substantial validity for integrity tests. Although it is not clear that external measures of counterproductive behaviors are more valid than admissions of such behaviors (the correlation between the two criteria is about .50), we analyzed the 187 validities (across 197,717 individuals) from studies using the criterion of externally measured counterproductivity separately from those using admissions. (See our Table 11). Validities were again substantial. Thus your criticism is not germaine. You liken our meta-analysis to ~one great bouiUabaisse~. Actually, our study is a legitimate distillation of knowledge from many seemingly conflicting study findings, and was accepted as such by the authors of the APA report on integrity tests and the reviewers of our manuscript at JA~. Your comment appears to be an example of "critique through gratuitous insult, ~ which should have no place in science. Consider how you yourself respond to criticisms of a similar nature made of your research on the role of heredity in the determlr~ation of personality and ability. In this connection, consider this: virtually all the personality, interest, and values measures studied by the Minnesota group have turned out to have substantial heritabilities. It is highly unlikely, therefore, that this will not also be true for integrity tests scores. If this is the case, then how can it be so shocking that they are found to have substantial real world validity? 2.

You are concerned that if integrity tests are used in personnel selection a large portion of job applicants may be denied jobs.

All selection instruments are intended to ensure that the most promising people from the applicant pool are hired. Hence, depending on the selection ratio, all selection instruments have the potential for excluding a large fraction of job applicants. That is the point in selection. Remember that use of any valid selection predictor will result in a lower false positive (and false negative) rate than its nonuse. Integrity tests have higher validity than many other available selection instruments and (more importantly) can add incremental validity over and above other procedures, as indicated in our paper. We would feel more comfortable if our grandchildren were denied a position because their scores on an integrity test were low than if they were denied the same position randomly or based on a selection procedure with no validity such as graphology. .

In your manuscript on the impact of honesty testing you assert that ~attitude~ items supplement "admissions ~ items on integrity tests. You also state that integritytests were developed as a by product of polygraph screening.

D E N I Z S. O N E S , E T AL.

491

Both of your claims are misrepresentations. Most integrity tests do not use admissions items. None of the personality-based tests have admissions items, and only a few overt tests have them. And only 2 integrity tests out of 43 were developed as derivatives of polygraph administrations. . On p. 11 of your manuscript, you ask if any selection methods could be less open, scientific and controllable than preemployment integrity screening. There are several methods that we can think of. How about handwriting analysis or astrology for starters? Can you argue that such methods, despite no validity evidence, are superior to integrity tests, which we have shown to have substantial predictive validity? More importantly, this statement by you, like many of your other statements, appears to spring from a highly emotional dislike of integrity tests. How else can one explain the fact that you label as fatal criticisms of integrity tests properties you apparently consider harmless in the case of ability tests? What is the basis of your strong negative emotional reaction? .

In your manuscript, on p. 15, you state that screening by integrity tests would mean that millions of people would be excluded from the work force.

This is a straw man argument. If integrity tests were widely used in the U.S. economy, they would help allocate more honest people to jobs requiring high levels of security and less honest people to jobs where security is not as crucial (or is counterbalanced by close supervision). Therefore, using integrity tests would not cause any unemployment, but the efficient allocation of human resources to jobs. The number of jobs that must be filled would remain unchanged, and so therefore would be the number of people obtaining jobs. The key point you are missing is that in an economy with low unemployment it is not possible for most people not to get a job. 6.

On page 17 of your manuscript you state that integrity test validities are supported by publisher's clnims rather than data.

This is erroneous. Our meta-analysis is based on voluminous hard data. Data this extensive do not exist for most other selection instruments, with the exception of ability tests. Even in the domain of personality tests, validity data is based on less than 100,000 individuals. Val-

JOURNAL OF BUSINESS AND PSYCHOLOGY

492

idity data on integrity tests is based on more than 600 studies and half a million datapoints. . Finally, you conclude that honesty testing ~is likely to do considerable harm, and.., should be closely monitored and regulated. ~ (p.

18) There is no basis for this statement. In your arguments against integrity tests you are concerned that the innocent should not be presumed guilty until proven so. Yet you presume worthlessness for integrity tests despite the massive evidence that we present in our meta-analysis to the contrary. As the APA Task force report on integrity tests points out integrity tests should be evaluated on the same basis and using the same criteria as other psychological tests. You seem to apply different s t a n d a r d s to integrity tests than you would to ability test and more mainstream personaUty tests. Most of your arg~iments can be made against any psychological test used in personnel selection. In evaluating the merits of various tests used in employment settings, the most important factors involve reliabilityand validity.This is what we focused on in our metaanalysis. W e hope that you will go back and reread our paper more carefully and will reconsider some of the statements you advanced in your letter and your article.

Frank L. Schmidt R. L. Sheets Professor of H u m a n Resources

Deniz Ones Doctoral Candidate

C. Viswesvaran Doctoral Candidate

Dear Dr. Schmldt et al., Th~n]~ for your detailed response to m y letter of 29 October. I have to say I r e r e a d m y letter after getting yours because I hadn't intended it to be as critical as you seemed to have found it. For example, your Point 1 attributes to me something t h a t I did not in fact "contend." I said t h a t "some ~ of the Reid Report and Staunton Survey studies use polygraph or admissions criteria--not t h a t ~ahe database [you] gathered largely consists of' such studies. Similarly, m y comment about "one great bouillabaisse" was m e a n t to refer to a hypothetical "non-evaluative metaanalysis.., applied to polygraph validit3f', not to your paper. .The point t h a t I did intend to m a k e seems to me still to be a fair

DENIZ S. ONES, ET AL.

493

one. You explain that you excluded studies based on polygraph and admissions criteria but were all studies based on job performance included? What do the data look like if you analyze separately only the studies t h a t have been published in edited journals?mif you look at only the studies done by disinterested investigators? You must yourselves have felt t h a t the studies you varied in scientific merit, in the quality of the criteria or of the analysis, and in the clarity of presentation. What if you three authors had independently rated these studies, in order to estimate the reliability of your ratings, and then plotted the obtained validities as a function of the consensus rating? Would the best studies have achieved as high validity estimates as the worst ones? My point is that a meta-analysis requires the reader to trust the authors' judgment and perspicacity since, unlike the standard Psychological Bulletin review, he csnnot compare the author's evaluation of specific studies with the reader's own judgment about those studies. That being so, it seems to me that meta-analyses should strive to provide the reader with as much reassurance as possible as to the way in which these qualitative issues were handled. Your letter makes much of the thousands of data points and hundreds of studies, all of which is certainly important, but a vast amount of garbage is still garbage. (No, I am not saying that you reviewed garbage but merely that I should not have to depend on your personal reputation in mRk!ng this judgment.) On the question of the social impact of low-validity screening: If your grandchildren are denied a job on the basis of a coin toss, they have a 50:50 chance of getting the next job that uses random selection. But if they are denied the job because they are consistent false-positives on some weak but widely used test, they are likely to remain jobless for some time. Surely that is a valid concern? In your Point 5, you say that, since the ratios jobs to people is relatively constant, the effect of a false-positive screen will not be to keep all the victims unemployed. This is a clever debating point but isn't it a bit superficial? Here we have all the entry level jobs that involve hand]iug money, using the universal ~honestf' test which doesn't in fact measure honesty (as far as we know) but has correlates which make it we~k]y predictive of supervisor ratings of dependability and so on. The cutoff scores exclude about half the applicants, many of whom are persistent false positives. No doubt they can find jobs at last, working for the sewer department, in the army, in the slaughter house, but not in the b~nk or the retail store or, indeed, any white collar job with a firm large enough to employ testing. False-positives on some aptitude test might at least get a lower-level job on which they can demonstrate their ability and win promotion but there is less chance for appeal from a false verdict on an honesty test.

494

JOURNALOF BUSINESSAND PSYCHOLOGY

Is there not a kind of Gresham's Law in personnel psychology which says t h a t a cheap and weakly valid screening procedure tends to dis= courage good management, surveillance, and well-planned provision of opportunities for individuals to prove their qualities through job perforo mance? I once talked with the chief security m a n at Farr~hR in Las Vegas about pre-employment screening. He pointed out that they use elaborate monitoring of their employees so that they know whom they can trust; there are few false=positives or false-natives working in Vegas casinos. It seems too bad that my honorable grandson, who aspires to a career in retailing or bsnklng, might have difficulty getting in on the ground floor in these industries but could be successful as a blackjack dealer. I haven't seen all the honesty tests you surveyed (nor can a reader tell which ones you surveyed) but I have seen the Reid and Staunten and Holland House tests which were the earliest on the scene a n d still are widely used, are they not? My description of these tests in my paper is not a ~misrepresentation. ~ (A rather strong word for you to have used, don't you th;nl~?) My paper was written as a response to the OTA and APA Task Force reports, both of which lamented the lack of data other t h a n publisher's claims. As I indicated in my letter, my paper will have to be revised in light of your meta°analysis but it seems unreasonable for you to be so antagonistic about its failure to take account of your broader survey in advance of my seeing it. It is true that failing an honesty test has rather more disconcerting and damaging implications than failing an Aptitude for Butter Packing test. But it is also true that many of my concerns about honesty testing apply with nearly equal force to all psychological selection devices t h a t (1) have only weak predictive validity and (2) have weak or 1]nl~uown validity as assessments of meaningful trait unities such as IQ. T h a t is, an IQ test that is valid but only weakly predictive of performance in some job seems to me less objectionable than a criterion-keyed test that is equally weakly predictive but also doesn't validly measure any recognized psychological dimension. And both are less objectionable t h a n a test t h a t claims to measure some important attribute, such as honesty, but cs n n o t be shown to do so with high validity. Perhaps what I should do is to publish a p~mphlet on how to pass integrity tests. Does it not bother you that this would be quite easy to do and t h a t most job applicants who can read would make it past the cut a l ~ r only brief instruction? Well, enough. I intended to make a constructive suggestion about meta°analysis. Your paper on this topic suggests t h a t there m a y come to be a division between the workers in the vineyards and the meta-an= alysts, analogous to the distinction between theoretical and experimental physicists. I don't see that happening until the meta-analysts find a

DENIZ S. ONES,ET AL.

495

way of giving the consumers a better sense of the variations in character and quality of studies being fed into the hopper.

David Lykken Professor Department of Psychology University of Minnesota Minneapolis, Minnesota

Dear Dr. Lykken: Thank you for your response of February 2 to our first letter. You raised several new issues and reiterated some that were in your previous letter. Here are our answers to your questions. 1. Were all studies utilizing job performance criteria included in our analyses? Yes, all studies reporting validities for job performance were included. We were able to find 222 validity coefficients using the job performance criteria and all of them were included in our meta-analysis. .

Would the results of the meta-analysis be different had we included only the studies published in peer reviewed journals? Or had we included only studies done by disinterested investigators?

Of the 665 validities contributing to our database, 247 validity coefficients came from the published literature or the published reviews of integrity tests; of these 67 were published. To address the concern that there could be some systematic difference in validities from the published source compared to unpublished sources, we computed the correlation between the validity coefficients reported and the dichotomous variable of validities reported in published vs. unpublished studies. This correlation was -.02. The negative sign of the correlation indicates that published studies reported higher validities, but the low absolute value of the correlation indicates that any differences in validities are negligible. Hence, in our database, the published vs. unpublished distinction for the validities is inconsequential. This analysis was presented in the study we sent you earlier. In response to your question of whether our results would have been different if we had included only Uunsponsored" research, two practices issues to consider are the definition of Usponsored ~ research and

496

JOURNALOF BUSINESSAND PSYCHOLOGY

adequacy of sponsorship acknowledgments in research reports. It is unclear just what sponsorship refers to. Research conducted by in house psychologists? Financial support to researchers? Provision of test booklets free of charge? Or simply permission to use the integrity test? In addition to these definitional concerns, the research reports are often lacking in unambiguous acknowledgments of the status of the research relative to sponsorship questions. Therefore, it is not feasible to determine with accuracy how many studies are publisher sponsored vs. independent and correlate this dichotomous variable with the validity coefficients reported. For example, we know several cases where the authors are neither affiliatedwith a particular test publisher nor acknowledge any sponsorship, but have conducted research exclusivelyusing one particular test,the results of which were published by the test publisher as technical reports. W e do not feel that sponsored research is a threat to the integrity of our database. Note that about one fifthof the validities were actually calculated by us. In addition, the enlargement of our database from 77 to 665 validities over a period of two years did not substantiaUy change the conclusions of our paper. Finally,we would like to reiterate the guiding principle enunciated by the A P A Task Force report on integritytesting:Absent a strong arg~iment to the contrary, the evaluation of honesty tests should not be carried out using standards higher than or different from those that would apply to other types of tests used for making decisions about people. Some psychologists continue to be skeptical about integritytest research, in large part because most of it has been conducted by the test publishers. However, m a n y of these researchers are I/O psychologists in good standing and no evidence has been produced to indicate that negative findings have been withheld. To point up the nature of this double standard, consider the Differential Aptitude Test Battery (DAT). M a n y of the D A T validities studies in the D A T Manual were conducted by the test authors or publishers.Yet you would pres~imably not consider them suspect simply for that reason; most psychologistswouldn't. W e could give m a n y other similar examples. M a n y of our colleagues with good reputations are tied to various integrity tests.For example, Richard Arvey (of the University of Minnesota, your university)is the co-author of the revised Personnel Reaction Blsn!~; Harrison Gough is the original author of Personnel Reaction Blank, Robert Hogan is the author of the ReliabilityScale, Philip Ash has ties to both the Reid Report and the PSI of London House, George Pajaanen, a researcher with the Personnel Decisions, Inc.,developed the PDI-EI as a part of his dissertationat the University of Minnesota. This listcan go on and on. W h a t basis do we have to discount all the research they or their colleagues or students have done? Further, you seem to ignore the fact that you are endorsing a catch

DENIZS. ONES,ET AL.

497

22: APA test standards and APA ethical standards require test authors to conduct research on the validity of the instruments they offer. Yet when they do so, you reject their studies as being of no scientific value because of their connections to the instrument. So they are damned if they conduct validity studies and dRmned if they don't: catch 22! .

If we had made judgments about the quality of the studies included in our meta-analysis, would the better studies have had validities siml]ar to the worse studies?

E|imlnating studies from any meta-analysis based on "methodological weAl~nesses ~ is not as reasonable or desirable as it m a y seem. N o research study can be defended against all possible counter hypotheses; hence, no study can be without alleged ~methodological inadequacy. ~ In our meta-analysis database we saw no evidence of methodological problems that would systematically bias study outcomes, other than the statistical artifacts for which we corrected. These clearly demonstrable methodological inadequacies (e.g.,unreliability of measures, range restriction, small sample sizes, dichotomization of continuous variables) cause studies to underestimate validity. The hypothesis of methodological inadequacy should be tested only after two hypotheses have first been rejected. First, one should determine if the variation across all studies can be accounted for by statistical artifacts (including sampling error). If all the variation is due to artifacts, then there can be no variance due to methodological inadequacy. Second, if there is substantial variation across studies, then theoretically plausible moderator variables should be tested. If these moderator variables account for all the remaining variance, then there can be no variance due to methodological inadequacy. This was very close to being the case in our meta-analysis. In most of the cells in Tables 8 and 11 of the paper we sent you, there was very little remaining variance--no more than has been found previously for studies of ability and aptitude. . Does our meta-analysis require the reader to trust our judgment rather than allowing the reader to compare Ms/her evaluation of specific studies with our evaluation? For most meta-analyses data are drawn mostly from studies in the published literature. Ours was a special case of meta-analysis where the readers could not immediately check the original studies or the analysis because most of the data came from unpublished sources. But anyone interested in obt~inlng the primary studies done on various tests can write to the publishers for the studies. This is how we obtained the

498

JOURNAL OF BUSINESS AND PSYCHOLOGY

studies we meta-analyzed. All test publishers make their studies and technical reports available to researchers. 5.

Do integrity tests have "weak predictive validity ~ for job performance?

No, integrity tests have fairly high validities when compared to the other predictors used in personnel selection. Integrity tests have a m e a n predictive validity of .41 for supervisory ratings of job performance. Mean validities of other widely used personnel selection devices for the same criterion are as follows: Ability/Aptitude r ffi.53, job knowledge r=.48, job tryout r ffi.44; biographical data r=.39, assessment center ratings r = . 3 7 , unstructured interviews rffi.40, reference checks r-~.26, amount of job experience rffi.18; and education r=.10. Given a validity of .41 for integrity tests, using integrity tests in personnel selection would improve prediction and decrease both the false positive and false negative rates. 6. What do honesty tests measure? What is the construct validity evidence for integrity tests? In your letter, you indicate that honesty tests do not in fact measure honesty. For prediction purposes, it does not matter what "integrity tests ~ measure. We have shown that they have substantial predictive validity. But maybe for you, construct validity is as important as (maybe more important than) criterion-related validity. We agree with you that construct validity is important from a theoretical and a practical perspective. On the last page of your letter you imply that honesty tests do not ~validly measure any psychological dimension.~ The value we see in integrity tests is due to the fact that they are closely linked to the personality domain. Many integrity tests were derived from the California Psychological Inventory, the Hogan Personality Inventory and even the MMPI. Furthermore, we have research underway investigating which of the Big Five dimensions of personality integrity tests tap into. So far our results reveal that the most consistent correlates of integrity tests are the dependability and conscientiousness scales of various personality inventories. Our theory is that integrity tests tap into a broadly defined conscientiousness (trustworthiness) dimension of personality. We will share with you a copy of our paper on the construct validity of integrity tests when it is completed. 7. Which integrity tests did we include in our study? (You complain that a reader cannot tell which ones we surveyed).

DENIZS. ONES,ET AL.

499

We report in our Table 1 all the integrity tests contributing data to our analyses. (Again, we have to ask ourselves whether you actually read our paper). 8. Does failing an integrity test have more severe implications than failing an aptitude test? We don't t]~ink so, and we do not understand why you do. Consider failing a general mental ability test. It is more damaging to one's ego to be told he/she has low intelligence than to be told one has undesirable attitudes toward thei~ of counterproductive job behaviors. And one's ability scores are probably even more stable from test to test, employer to employer, etc. So where is the difference? But regardless of the selection device used by a company, applicants are not usually given the precise reason why they are not hired. Further, it is reasonable to suspect that any company would share honesty (or ability) test score information with any outsiders. Finally, there are requirements of confidentiality for all psychological test scores (including integrity tests). Integrity tests should be held exactly to the same standards as all other psychological tests, and failing an integrity test is no worse than failing an aptitude test. The label "honestf' seems to be troubling to you. We don't use the phrase ~honssty testing", but perhaps the word ~utegritf' is equally troubling to you. We could easily call these tests for "Rule breaking inclinations", or "Conscientiousness~ or even ~work ethics or ~sociallzation"?. Would other test names be more acceptable to you than integrity? Is the question of labeling the key to understanding your strongly emotional reaction? 9.

Is it a concern that some individuals may be false-positives on integrity tests?

Any selection device, short of having perfect validity, will result in some false positives and false negatives. False positives, the decision error you are concerned about, occur on ability tests as well. You claim ~false-positives on some aptitude test might at least get a lower-level job on which they can demonstrate their ability and win promotion but there is less chance for appeal from a false verdict on an honesty test." What makes you think that false-positives on an integrity test will not get lower level jobs where they can demonstrate their honesty and get promoted? Furthermore, integrity tests are almost never used alone but in conjunction with other selection devices such as ability tests and application b]ank q.

500

JOURNAL OF BUSINESSAND PSYCHOLOGY

10. Do screening procedures discourage good m a n a g e m e n t and surveillance at the workplace? We do not deny the importance of good management techniques in orgAni~.ations. But given better workers (brighter, more honest and hard-working workers), good management can produce even better results. It is likely that close surveillance at work can produce aversion, frustration, and low morale in w o r k e r s w t h e syndrome of ~ h y don't my superiors t r u s t me? Why do they watch me all the t i m e ~ So we see using integrity tests before hiring as a good investment that can reduce unnecessary surveillance costs later on (and the added psychological costs). 11.

Can applicants be taught to appear honest on integrity tests? Can test takers alter their scores on integrity tests?

Yes, when instructed some people can alter their scores on integrity tests. However, there is evidence that faking by applicants is infrequent. Furthermore, response distortion, if it exists at all, does not destroy the criterion-related validities of integrity tests. Our study showed this. Given all this, it would be a waste of their time for anyone to publish a pamphlet on how to pass specific integrity tests. Such a pamphlet would not affect prediction or validity but would violate test security requirements of the APA Test Standards and APA Ethical Principles. Ethical standard 2.10 states "Psychologists make reasonable efforts to m~intain the integrity and security of tests and other assessment techniques consistent with the law, contractual obligations, and in a m a n n e r t h a t permits compliance with the requirements of this Ethics Codex. (American Psychologist, 1992, p. 1604). We do not see how individuals m a y be instructed to pass integrity tests without revealing to them the exact nature of the questions used in the inventories. We hope that we were able to ~nRwer most of your questions on preemployment integrity testing. We can see that most of your objections arise from your drawing a parallel between the polygraph and integrity tests and from your emotional aversion to the phrases ~honesty testing ~ and "integrity testing ~. Integrity testing is closer to personality assessm e n t t h a n it is to the polygraph. The criticisms that apply to t h e polygraph do not apply to the integrity testing domain and we hope t h a t our responses to your questions illustrate this point.

Frank L. Schmidt R.L. Sheets Professor of Human Resources

Deniz S. Ones Doctoral Candidate

C. ~swesvaran Doctoral Candidate

DENIZ S. ONES, ET AL.

501

REFERENCES American Psychological Association. (1992). Ethical principles of psychologists and code of conduct. American Psychologist, 47, 1597-1611. Goldberg, L. R., Grenier, J. 1~, Guion, R. M., Sechrest, L. B., & Wing, H. (1991). Questionnaires used in the prediction of trustworthiness and preemployment selection decisions. Washington, DC: American Psychological Association. Lykken, D. T. (Unpublished Manuscript). Honesty Testing: An Environmental Impact Assessmant. Department of Psychology, University of Minnesota, Minneapolis: MN. Ones, D. S.; Viswesvaran, C.; & Schmidt, F. L. (1993). Comprehensive Meta-Analysis of Integrity Test Validities: Findings and Implications for Personnel Selection and Theeties of Job Performance. Monograph. Journal of Applied Psychology, 78, 679-703. U.S. Congressional Office of TechnologyAssessment. (1990). The use of integrity tests for pre-employment screening (Report No. OTA-SET-442).Washington, DC: U.S. Government Printing Office.