Chapter 12 — Selection Methods

Chapter 12 opens with eight learning goals, numbered 12.1 through 12.8. They map directly onto the chapter's own structure: personal history data first, then recommendations and references, honesty and polygraph testing, training-and-experience evaluation, drug and polygraph screening, and finally the employment interview in depth, closing with a look at social media and other emerging technologies.

1. 12.1 Gather personal history data from job applicants in a manner that minimizes distortions and embellishments
2. 12.2 Assess letters of recommendation and reference checks in terms of factors that affect their validity (e.g., degree of writer familiarity with the candidate and job in question)
3. 12.3 Choose an appropriate honesty test (e.g., overt vs. personality oriented)
4. 12.4 Use valid and reliable measures of past training and experience
5. 12.5 Implement drug screening and polygraph testing using appropriate legal guidelines
6. 12.6 Design and implement employment interviews taking into account possible response distortion and considering social/interpersonal, cognitive, and individual differences that affect the process and outcomes of interviews
7. 12.7 Administer structured employment interviews that maximize validity and reliability
8. 12.8 Use caution in relying on social media and other big data and technological advancements (e.g., mobile and Web-based technology, computer scoring of text, remote interviewing, and virtual reality technology) for selection purposes

Selection often begins with personal history data — biodata — found in application forms, biographical inventories, and résumés. Application forms sample past or present behavior briefly but reliably, much like a test. Studies of application forms at 200 organizations found the questions were generally job related, yet over 95% of the applications included one or more legally indefensible questions. To avoid problems, an item should be omitted if it might lead to adverse impact on protected groups, does not appear job related or tied to a bona fide occupational qualification, or might constitute an invasion of privacy (Miller, 1980).

Applicants confronted with an irrelevant or invasive question sometimes choose not to respond — but research shows employers tend to read a nonresponse as an attempt to conceal something unflattering, so applicants (especially those with nothing to hide) are ill advised to skip the question (Stone & Stone, 1987).

Biodata items can be classified along several dimensions — verifiable/unverifiable, historical/futuristic, actual/hypothetical behavior, firsthand/secondhand, external/internal, specific/general, invasive/noninvasive (Table 12.1) — complicated further by the fact that "contemporary biodata questions are now often indistinguishable from personality items in content, response format, and scoring" (Schmitt & Kunce, 2002, p. 570). Still, the core attribute of a biodata item is that it pertains to a historical event that may have shaped a person's behavior and identity (Mael, 1991). Some argue only historical, verifiable items should count as biographical; under that stricter view, a question like "Did you ever build a model airplane that flew?" would not be asked — yet Cureton found this single unverifiable item nearly as good a predictor of World War II flight-training success as the entire Air Force Battery (cited in Henry, 1965, p. 113).

Weighted Application Blanks (WAB)

The weighted application blank (WAB) technique identifies which background aspects (years of education, prior experience) reliably distinguish effective from ineffective employees, assigning weights by predictive power so each applicant gets a total score and a cutoff can be set. It works best as a rapid screening device in organizations with many employees doing similar work and adequate records, especially for jobs with costly training, high turnover, or many applicants chasing few openings (England, 1971). Weighting procedures are simple (Owens, 1976), but because WABs are raw empiricism taken to the extreme, weights must be cross-validated — otherwise observed differences may reflect chance fluctuation rather than real signal.

Biographical Information Blanks (BIB)

The biographical information blank (BIB) is closely related to the WAB: a self-report instrument, exclusively multiple-choice, usually with a larger item sample that reaches into content a WAB would not normally cover — early life experiences, hobbies, health, social relations, plus present values, attitudes, interests, and preferences (Glennon, Albright, & Owens, 1966; Mitchell, 1994). BIBs are usually built to predict success in one specific type of work, and much of their power comes from including all elements of consequence to the criterion (Asher, 1972). Development and item weighting mirror WAB methods (Mumford & Owens, 1987; Mumford & Stokes, 1992).

Résumés

Résumés are the most common source of personal history data — an estimated 1 billion paper résumés were screened annually as far back as 1975 (Brown & Campion, 1994). Because information is compressed into one or two pages, examiners reading résumés are especially prone to cognitive biases: applicants get sorted into stereotype-based categories almost automatically, and group-typical attributes get assigned to individuals even when factually wrong. "Paper people"/vignette studies test this by having judges rate hypothetical applicants' résumés (Aguinis & Bradley, 2014; Derous, Ryan, & Serlie, 2015).

Derous et al. (2015) had 60 Dutch recruiters rate résumés varying ethnicity (Dutch, Arab) and gender (female, male): Arab applicants and male applicants were both rated more negatively, an effect amplified by raters' own prejudice and by jobs with more client contact. Video résumés — recorded video/audio self-presentations — let applicants express themselves beyond the paper format, and can include animation and text as multimedia résumés (Hiemstra, Derous, Serlie, & Born, 2012). In a study of 445 unemployed job seekers who had completed a two-day job-application training in the Netherlands, video résumés were perceived as fairer than paper résumés regardless of applicant ethnicity. The chapter's overall advice: train raters to focus on job-related factors and assess interrater reliability (Brown & Campion, 1994).

Credit History

Big data has given organizations personal history information unthinkable a few years ago — about 50% of employers now run credit background checks on at least some applicants (Bernerth, 2012). Credit scores can look like an objective proxy for conscientiousness and integrity: failing to keep a promise to a lender may predict failing to keep promises at work, and financial duress may correlate with counterproductive behavior like theft.

Using credit checks for employment is legal under the federal Fair Credit Reporting Act with written applicant authorization, but California, Colorado, Connecticut, Delaware, Hawaii, Illinois, Maryland, Nevada, Oregon, Vermont, Washington, and Washington, D.C. restrict the practice. Colorado's Employment Opportunity Act (SB13-018) bars using consumer credit information unrelated to the job, requires disclosure when credit information drives an adverse action, and lets an aggrieved employee sue for an injunction, damages, or both.

The restrictions are evidence-based: Bernerth (2012) regressed FICO scores for 112 university employees/alumni on minority status, gender, marital status, educational attainment, and age. The five predictors together explained 34% of the variance in credit scores, with minority status (lower scores), less education (lower scores), and younger age (lower scores) showing the strongest effects. Since ethnicity correlates strongly with credit scores, using credit history will produce adverse impact — critics argue it traps low-score individuals in a "vicious downward spiral" where unemployment damages credit, which then further damages job prospects (Bernerth, 2012, p. 245). As with any predictor, validity evidence is required, and is especially important wherever adverse impact appears.

Job applicants can and do intentionally distort personal history data. "Sweetening" résumés is common: one study found 20–25% of résumés and applications contain at least one major fabrication (LoPresto, Mitcham, & Ripley, 1986), and self-reported distortion runs even higher when measured with the randomized-response technique, which fully guarantees anonymity (Donovan, Dwight, & Hurtz, 2003). When McFarland and Ryan (2000) instructed participants to "fake good" versus answer honestly, scores rose nearly two standard deviations under the fake-good instruction — a bigger gap for biodata than for personality traits like extraversion, openness, or agreeableness. People also differ in how well they can fake.

Four situational levers reduce distortion: (1) more objective, verifiable items are harder to fake, since fear of being caught deters distortion (Kluger & Colella, 1993); (2) option-keyed items — where each response alternative is scored only if it correlates with the criterion — are less fakeable (Kluger, Reilly, & Russell, 1991); (3) warning applicants of a lie scale, and using biodata in a non-evaluative classification context, both reduce distortion (Kluger & Colella, 1993; Fleishman, 1988); and (4) requiring applicants to elaborate on their answers forces more accurate recall and reduces impression management (Schmitt & Kunce, 2002). Elaboration reduced scores by about .6 SD in one federal civil-service pilot study of 311 examinees, and a 600-plus-student study replicated the effect (Schmitt, Oswald, Kim, Gillespie, & Ramsay, 2003).

Validity evidence for biodata is strong across many occupations — life insurance agents, law enforcement, service station managers, sales clerks, engineers, architects, research scientists, Army officers — against criteria like turnover (the most common), absenteeism, salary growth, performance ratings, publication counts, training success, creativity, sales volume, and theft. Reilly and Chao (1982) reviewed 58 biodata validity studies and found an average validity of .35; Hunter and Hunter's (1984) meta-analysis of 44 studies found .37; a later meta-analysis of eight salesperson studies using supervisory ratings found a corrected mean of .33 (Vinchur, Schippmann, Switzer, & Roth, 1998).

In a concurrent-validity study of 300-plus clerical employees, a rationally selected, empirically keyed, cross-validated biodata inventory added incremental variance over personality and cognitive-ability measures: about 6% for quantity/quality of work, 7% for interpersonal relationships, and 9% for retention (Mount, Witt, & Barrick, 2000) — empirical support for Owens's (1976) claim that personal history data illuminate what does and doesn't contribute to effective performance.

Bias, Adverse Impact, and What Biodata Mean

Since Title VII of the 1964 Civil Rights Act, personal history items face intense legal scrutiny. Cascio (1976b) found cross-validated validities of .58 (minority) and .56 (nonminority) for female clerical employees against a tenure criterion, with no significant scoring differences between groups — supporting a single scoring key for both. Bobko and Roth's (2013) meta-analysis found biodata inventories are relatively free of adverse impact when items avoid cognitive-ability content, but most evidence comes from concurrent designs using incumbent samples, which likely understate ethnicity-based differences; they estimated a black–white standardized mean difference of d = .31 for biodata heavy in KSAs.

Beyond criterion-related validity, courts and regulators care about job relatedness: an item with no rational job connection (e.g., "applicant does not wear eyeglasses" predicting theft) is unlikely to survive scrutiny if it produces adverse impact. External/empirical keying — scoring items purely against an external criterion — remains the most popular method, but it does not tell us what construct is actually being measured (Stokes & Searcy, 1999). The rational approach instead starts from job analysis, forms hypotheses about what success requires, and searches existing research for items or factors addressing those hypotheses (Stokes & Cooper, 2001). Brown and Campion (1994) found recruiters naturally infer language/math ability from education items, physical ability from sports items, and leadership from prior-authority or social-activity items — nearly every item is read as saying something about motivation.

The rational approach extends beyond employment testing: Douthitt, Eby, and Simon (1999) built a biodata scale measuring openness to dissimilar others, reasoning that travel and integrated schooling expose people to more difference (e.g., "How extensively have you traveled?", "How racially/ethnically integrated was your high school?"). Even with a rational approach, an item's validity can shift depending on the life stage it's anchored to (childhood versus recent years) — specific developmental framing helps applicants answer more accurately (Dean & Russell, 2005).

Recommendations and reference checks supply four types of information: (1) employment and educational history, including degree/GPA confirmation; (2) evaluation of character, personality, and interpersonal competence; (3) evaluation of job performance ability; and (4) willingness to rehire. For a recommendation to be meaningful, four preconditions must hold: the recommender had adequate opportunity to observe the applicant in job-relevant situations, is competent to evaluate, is willing to be open and candid, and expresses evaluations so the employer can interpret them as intended (McCormick & Ilgen, 1985). Unwillingness to be candid is probably the most damaging failure — though unverifiable, harmful unfavorable information can also expose the writer to defamation liability (libel in writing, slander orally) (Ryan & Lasek, 1991).

Written recommendations are widely seen as low-value: in a survey of about 600 HR professionals, 80% agreed "letter inflation is a problem that will never be entirely alleviated" (Nicklin & Roch, 2009). Evidence backs the skepticism — Reilly and Chao (1982) found average recommendation validity of just .14. A meta-analysis of academic settings found correlations of .13 with medical-school GPA (N = 916), .12 with clinical/internship performance (N = 1,120), and a higher .28 with college GPA (N = 5,155) (Kuncel, Kochevar, & Ones, 2014). Meta-regression showed letters added only .003 of variance to graduate-school GPA prediction and .011 to faculty performance ratings beyond undergraduate GPA and GRE scores, though degree-attainment prediction improved a bit more (.024) (Gonzalez-Mulé & Aguinis, in press).

The core problem: recommendations rarely include unfavorable information, so they fail to discriminate among candidates. Letter writers' affective disposition also shapes letter length, which in turn shapes favorability (Judge & Higgins, 1998) — so a letter often reveals more about its author than its subject. Still, decisions get made on letters, especially in academia (Nicklin & Roch, 2009); useful letters should include (Knouse, 1987): the writer's degree of familiarity with the candidate (time known, time observed weekly), familiarity with the job (aided by the requester supplying a job description), specific performance examples (goals, task difficulty, environment, coworker cooperation), and the comparison group used to judge the candidate.

Reference checks are legally permissible despite common employer misconceptions — employers may seek applicant information, use it in selection, and share results with other employers (Sewell, 1981; Hedricks, Robie, & Oswald, 2013). Employers can even be liable for negligent hiring if they should have discovered an applicant's unfitness (e.g., prior job-related convictions, violence propensity) that later causes harm (Gregory, 1988; Ryan & Lasek, 1991) — failing to check closely enough is itself a legal risk.

Hunter and Hunter's (1984) meta-analysis found reference checks average .26 validity. To be most useful, reference checks should be consistent (the same standard applied to every applicant), relevant (focused on items that truly distinguish effective from ineffective employees), written (documented to support the hiring decision), and grounded in public records (court records, workers' compensation, bankruptcy filings) (Ryan & Lasek, 1991; Sewell, 1981).

Structured Telephone Reference Checks (STRC)

Taylor, Pajo, Cheung, and Stringfield (2004) tested a structured telephone reference check across 448 calls covering 244 customer-contact applicants (about two referees each), run by recruiters at six firms over eight months, averaging 13 minutes per call. Referees rated the applicant against others in similar roles on a 1 (below average) to 5 (outstanding) relative scale — chosen specifically to reduce leniency — and were asked to elaborate, a second leniency safeguard. Of 244 applicants, 191 were hired, and performance data existed for 109 who survived to the first appraisal cycle. Regressing supervisory performance ratings on the three STRC constructs (conscientiousness, agreeableness, customer focus) produced R² = .28, with customer focus the only significant predictor (standardized β = .28).

Polygraph instruments infer truthfulness from physiological measures (e.g., heart rate) in response to questions. They remain common for event-specific investigations (post-crime) but are used only on a limited basis for employment screening, sharply restricted by the 1988 Employee Polygraph Protection Act — a federal law that bars private employers (except security-service firms and controlled-substance manufacturers) from requiring or requesting preemployment polygraph exams, and restricts testing of current employees to narrow circumstances. Some federal agencies (e.g., the Department of Energy) still use polygraphs given international terrorism concerns.

Beyond ethics (Aguinis & Handelsman, 1997a, 1997b), the central controversy is validity: can physiological measures actually detect deception (Saxe, Dougherty, & Cross, 1985)? The National Research Council's Committee to Review the Scientific Evidence on the Polygraph (2003) quantitatively analyzed 57 independent studies and concluded: polygraph accuracy for screening is almost certainly lower than for specific-incident testing; the physiological indicators can be consciously altered; and using the polygraph for security screening forces an unacceptable trade-off between falsely flagging loyal employees and missing real security threats. The committee found the polygraph's accuracy "insufficient to justify reliance on its use in employee security screening in federal agencies" (p. 6) — a conclusion echoed by a survey of Society for Psychophysiological Research members and APA Division 1 fellows (Iacono & Lykken, 1997), who added that polygraph tests can be beaten with countermeasures.

No alternative — including brain-activity measures via electrical/imaging studies — has yet been shown to outperform the polygraph, so it will likely remain in use for security screening until better alternatives emerge (Committee to Review the Scientific Evidence on the Polygraph, 2003).

Honesty testing is a multimillion-dollar industry, especially since polygraph use has been curtailed and "ban-the-box" laws restrict asking about prior convictions early in hiring. Written honesty (integrity) tests split into two types. Overt integrity tests ask about attitudes toward theft/dishonesty (endorsing rationalizations, beliefs about theft frequency, punitiveness, perceived ease of theft) and direct admissions (dollar amounts stolen, drug use, gambling). Personality-based measures instead predict a wide range of counterproductive behavior — substance abuse, insubordination, absenteeism, bogus workers' comp claims, passive aggression — via broader traits like socialization and conscientiousness. Despite different content, both types share a common latent structure reflecting conscientiousness, agreeableness, and emotional stability (Berry, Sackett, & Wiemann, 2007).

Do honesty tests work? Broadly yes. Ones, Viswesvaran, and Schmidt (1993) meta-analyzed 665 validity coefficients from 576,460 test-takers and found average validity of .41 for supervisory performance ratings (similar for overt and personality-based tests), but only .13 for predicting theft itself. Van Iddekinge, Roth, Raymark, and Odle-Dusseau (2012a) reran the analysis on a smaller, more methodologically vetted set (104 studies, 134 samples) and found much lower corrected validities: .15 for job performance, .16 for training performance, .32 for counterproductive work behaviors, and .09 for turnover. Test vendors pushed back hard (Harris et al., 2012), arguing more weight should go to the Ones et al. figures; Van Iddekinge et al. (2012b) countered that vendor gatekeeping had blocked their access to additional technical reports. The chapter treats this as an open, unresolved methodological dispute (Sackett & Schmitt, 2012).

Four unresolved issues limit confidence in honesty tests. First, construct validity is unclear — integrity tests are not interchangeable with each other, and more work is needed linking them to specific traits like object beliefs, negative life themes, and power motives (Mumford, Connelly, Helton, Strange, & Osburn, 2001) rather than only broad personality dimensions. Second, women score about .16 SD higher than men, and applicants 40+ score about .08 SD higher than those under 40 (Ones & Viswesvaran, 1998), with no clear explanation. Third, unlike ability (where someone who tests low rarely suddenly becomes smart), a person with a genuinely reformed moral history may still be honestly "locked into" a low integrity score — a classification-error risk for people with a criminal past (Lilienfeld, Alliger, & Mitchell, 1995). Fourth, intentional distortion is a real threat (Alliger, Lilienfeld, & Mitchell, 1996) — ironically, applicants are likely to be dishonest on an honesty test: McFarland and Ryan (2000) found fake-good instructions raised scores 1.78 SD over honest instructions. Test publishers also carry an inherent conflict of interest in validating their own products, echoing the test-fairness discussion in Chapter 8.

Alternatives — Conditional Reasoning and Situational Judgment

Conditional reasoning testing (Frost, Chia-Huei, & James, 2007; James et al., 2005) presents what look like ordinary inductive-reasoning problems, but scores respondents' implicit biases and preferences revealed through their chosen "solutions" — biases that typically operate below conscious awareness. A second alternative folds integrity into a situational judgment test (detailed in Chapter 13), where applicants choose the response closest to what they would actually do (Becker, 2005).

Judgmental evaluation of prior work experience and training — from résumés and applications — is a standard part of initial screening, ranging from purely subjective judgment to a formal, standardized method. Evaluating experience is genuinely hard because it has both qualitative and quantitative components that interact and accumulate over time (Aguinis, O'Boyle, Gonzalez-Mulé, & Joo, 2016); work experience is multidimensional and temporally dynamic (Tesluk & Jacobs, 1998). Even so, it pays off as a predictor: a study of 800-plus U.S. Air Force enlisted personnel found ability and experience have linear, noninteractive effects on performance (Lance & Bennett, 2000), and other military research showed experience items predict performance beyond cognitive ability and personality (Jerry & Borman, 2002). A National Association of Colleges and Employers survey of 200-plus staffing professionals confirmed that experienced hires are rated more favorably than new graduates on most characteristics (Rynes, Orlitzky, & Bretz, 1997).

McDaniel, Schmidt, and Hunter (1988) compared four methods for evaluating work experience and found the behavioral consistency method had the highest mean validity, .45. It requires applicants to describe major achievements in job-related dimensions that supervisors have identified as maximally distinguishing superior from minimally acceptable performers; achievement statements are then scored against anchored rating scales built by subject matter experts.

The accomplishment record (AR) method (Hough, 1984) is a related approach best suited to professionals — formalizing the common claim "my record speaks for itself." It is a biodata/maximum-performance/self-report instrument tapping a part of an individual's history not captured by ordinary biographical inventories; it correlates essentially zero with aptitude scores, honors, grades, or prior activities/interests. Development starts from critical incidents that define important performance dimensions, from which raters build principles and scales. The method yields: complete definitions of key job dimensions; summary principles highlighting what to look for at each achievement level; job-expert-agreed example accomplishments at each level; and numerical equivalents translating accomplishments into a quantitative index. Applied to 329 attorneys, the AR produced a respectable .82 reliability for overall performance ratings and a .25 validity coefficient, and appeared fair across gender and race groups.

Academic qualifications, by contrast, carry surprisingly little weight — and can even hurt candidates with weak work experience, whose chances of being hired fall as academic credentials rise (Singer & Bruhns, 1991). A national U.S. Census Bureau survey of 3,000 employers found the most important hiring characteristics were attitude, communication skills, and previous work experience; the least important were grades, school reputation, and teacher recommendations (Applebome, 1995). Worse, when grades are used, they tend to produce adverse impact on ethnic minority applicants (Roth & Bobko, 2000).

Drug screening began in the military, spread to sports, and is now common in employment — about 50% of U.S. employers screen all job applicants (Lieberman, 2017). Critics call it a privacy violation and note inaccuracy risk from cheating (Box 12.2), though even critics concede that safety-critical jobs (nuclear plant operators, commercial pilots) warrant testing; perceived danger to the worker, coworkers, or the public predicts acceptability of testing (Murphy, Thornton, & Prue, 1991).

The largest study of its kind: the U.S. Postal Service urine-tested 5,465 applicants without using results in hiring decisions or informing local managers. Six months to a year later, applicants who had tested positive were absent 41% more often and fired 38% more often, with no difference in voluntary turnover — results holding after adjusting for age, gender, and race. The Postal Service subsequently implemented nationwide preemployment drug testing (Wessel, 1989). Legally, two 1989 Supreme Court rulings upheld mandatory urinalysis/blood tests for railroad crews in accidents and urine tests for U.S. Customs drug-enforcement applicants; an employer has a legal right to ensure competent, safe job performance, giving adequate legal grounds for drug testing when off-the-job use could impair performance or endanger coworkers.

To avoid legal challenge: inform employees/applicants in writing of the drug policy, include the policy and testing possibility in employment contracts, and frame the program around medical/safety benefits. If testing current employees too, give advance notice that it will be routine (Angarola, 1985). Fairness perceptions improve with advance notice, a right to appeal, a safety framing, minimized invasiveness, and supervisor training (Konovsky & Cropanzano, 1991; Tepper, 1994) — and are also shaped by employee characteristics, such as having a friend who failed a test (Aguinis & Henle, 2005).

Computer-Based Screening (CBS)

Faster processors and richer software now let organizations run computer-based screening (CBS) — job applications, structured interviews, and other tests globally, around the clock (Jones & Dages, 2003). At the low-interactivity end, CBS is just an electronic page turner (Olson-Buchanan, 2002); Nike uses interactive voice response by phone, the Air Force uses computer-adaptive testing (CAT) routinely (Ree & Carretta, 1998), and Home Depot and JCPenney use varied screening technologies (Chapman & Webster, 2003; Overton, Harms, Taylor, & Zickar, 1997). CAT presents average-difficulty items first, then raises difficulty after correct answers and lowers it after incorrect ones, using item response theory (Chapter 6) to estimate the applicant's trait level — something impossible to implement with paper-and-pencil testing.

CBS advantages (Kantrowitz et al., 2011; Olson-Buchanan, 2002): easier, more standardized administration (no inconsistent human proctors), automatic response recording that reduces data-entry error, remote access that widens the applicant pool, accommodations for applicants with disabilities (captioned audio, modified input devices), and preliminary evidence that Web-based assessment does not worsen adverse impact. Concerns include cost, potential cheating, and skepticism that high-stakes tests belong in unproctored Internet settings (Tippins et al., 2006); CAT partly addresses cheating since each applicant gets a different item set. The "digital divide" — unequal low-income access to the Internet — is a further implementation barrier (Stanton & Rogelberg, 2001).

Olson-Buchanan (2002) found CBS innovation lagging behind computer technology generally, due to development cost, a scientific-guidance lag on CBS reliability/validity, and doubt about tangible payoff. Encouragingly, Ployhart, Weekley, Holtz, and Kemp (2003) tested nearly 5,000 telephone-service-representative applicants and found proctored Web-based biodata testing produced similar or better psychometric properties (lower means, more variance, higher internal-consistency reliability) than paper-and-pencil. Applicant reactions to CAT track perceived performance (Tonidandel, Quiñones, & Adams, 2002), so item-selection tweaks that raise the share of correctly answered items can also improve perceptions.

Interview use in selection is almost universal (Moscoso, 2000), partly because it does far more than screen: it is a communication process where applicants learn about the job and organization and form realistic expectations. When an applicant is accepted, employment terms are typically negotiated in the interview; when rejected, the interviewer performs a public-relations function, since a rejected applicant should still leave with a favorable impression of the organization. Studies confirm that interviewer interpersonal skill, listening, recruiting ability, and information-conveying skill shape applicants' evaluations of the interviewer and the company (Kohn & Dipboye, 1998; Schmitt & Coyle, 1979) — though the interviewer's behavior barely affects whether an applicant accepts an offer if one comes (Powell, 1991).

As a selection device, the interview does two things well: it fills information gaps left by other tools (e.g., unclear application responses; Tucker & Rowe, 1977), and it assesses factors only observable face-to-face — appearance, speech, poise, interpersonal competence, likely person–organization fit (Cable & Judge, 1997). Well-designed interviews can even capture constructs other tools miss, such as empathy (Cliffordson, 2002) and personal initiative (Fay & Frese, 2001). A review of 388 rated characteristics across 47 interview studies found personality traits tied to conscientiousness (responsibility, dependability, persistence) and applied social skills (interpersonal relations, team focus, working with people) are rated more often than any other construct (Huffcutt, Conway, Roth, & Stone, 2001). Interviews also add incremental prediction of job performance beyond cognitive ability and conscientiousness (Cortina, Goldstein, Payne, Davison, & Gilliland, 2000) and beyond experience (Day & Carroll, 2002).

Response Distortion in the Interview

Interview distortion is probable and tends to run toward upgrading, not downgrading, prior experience (Weiss & Dawis, 1960) — driven by social desirability bias, the tendency to answer in the direction that looks good to the interviewer. Applicants also use influence tactics, especially self-promotion (Stevens & Kristof, 1995).

Does a computer interviewer reduce distortion? Martin and Nagao (1989) found candidates report GPA and test scores more accurately to computers than to human interviewers — possibly a "big brother" effect, since computer-recorded answers feel more subject to instant cross-checking. But high-status candidates resented computer interviews far more than low-status candidates did. Richman, Kiesler, Weisband, and Drasgow's (1999) meta-analysis of 61 studies (673 effect sizes) found computer-based interviews reduce social-desirability distortion versus face-to-face, especially for highly sensitive behavior like illegal drug use — likely because a computer interview feels more impersonal and removes evaluation-arousing social cues.

A subtler distortion channel is impression management — ingratiation and self-promotion are the two tactics most effective at swaying interviewer ratings favorably (Higgins & Judge, 2004; Lievens & Peeters, 2008; Roulin, Bangerter, & Levashina, 2015). Across five real-time video-coding experiments, interviewers could not reliably detect impression management, though they were somewhat better at spotting honest impression management (truthfully describing real accomplishments) than deceptive impression management (embellished or fabricated credentials) — and experienced interviewers were no better than novices at telling the difference. Training helps when it focuses on story-related cues (vagueness, contradictions) rather than nonverbal cues (gaze aversion, posture shifts, fidgeting) (Roulin et al., 2015).

Reliability and Validity

An early, uncorrected meta-analysis of just 10 studies found interview validity of .14 for supervisory ratings (Hunter & Hunter, 1984). Later meta-analyses correcting for range restriction told a more encouraging story: Wiersner and Cronshaw (1988) found a mean corrected validity of .47 across 150 studies; McDaniel, Whetzel, Schmidt, and Maurer (1994) found .37 across 245 coefficients from 86,311 people (higher for research criteria, .47, than administrative decisions, .36); Marchese and Muchinsky (1993) found .38 across 31 studies; Huffcutt and Arthur (1994) found .37 across 114 entry-level studies; and Schmidt and Rader (1999) found .40 for structured telephone interviews across 40 studies. The numbers converge tightly around the high .30s to high .40s.

On reliability, Huffcutt, Culbertson, and Weyhrauch (2013) found a mean interrater reliability of .68 across 125 coefficients (N = 32,428), with an 80% credibility interval from .42 to .94. Structure drove the spread: reliability was .36 at low structure and .76 at high structure — since reliability caps validity (Chapter 7), the clearest lever for improving interview validity is improving its structure.

Posthuma, Morgeson, and Campion (2002) reviewed 278 studies of interview decision making and organized findings into four categories: (a) social/interpersonal factors, (b) cognitive factors, (c) individual differences, and (d) structure. The chapter follows this taxonomy.

Social/Interpersonal Factors

Similarity leads to attraction, attraction to positive affect, and positive affect to higher ratings (Schmitt, Pulakos, Nason, & Whitney, 1996); similarity also raises expectations of future performance (García, Posthuma, & Colella, 2008). Lin, Dobbins, and Farh (1992) found African American and Latino (but not white) interviewees rated higher when the interviewer shared their race — an effect eliminated by including at least one different-race panelist, with no similar effect for age. When an interviewer perceives shared attitudes with an interviewee, competence and affect ratings rise (Howard & Ferris, 1996). These similarity effects are modest and are reduced or eliminated by structured interviews and diverse interviewer panels.

On verbal cues, applicants were more likely to be hired when the interviewer talked more and silence was scarcer (Anderson, 1960); interview length tracks applicant quality (higher-quality applicants get more deliberation time) and the interview's expected length (Tullar, Mullins, & Caldwell, 1979). Nonverbal cues have a small but real impact (DeGroot & Motowidlo, 1999) — smiling, attentive posture, and closer interpersonal distance produce more favorable ratings (Imada & Hakel, 1977). Nonverbal cues interact with gender in surprising ways: Aguinis, Simonsen, and Pierce (1998) found a man making direct eye contact was rated more credible, but an exact replication with a woman displaying identical eye contact found her rated as coercive (Aguinis & Henle, 2001a). Overall, concise, complete, on-topic answers matter more than nonverbal behavior alone (Parsons & Liden, 1984; Rasmussen, 1984); strong nonverbal behavior only helps when verbal content is also good, and can hurt ratings when verbal content is weak.

Cognitive Factors

Interviewers are limited information processors with predictable biases (Kraiger & Aguinis, 2001). Dipboye (1982, 1992) modeled a self-fulfilling prophecy explaining how preinterview impressions (from reference letters or applications) distort evaluation through both behavioral bias (treating applicants in ways that confirm prior impressions) and cognitive bias (selectively attending to or recalling confirming information). Content-coded interviews showed favorable first impressions triggered confirmatory interviewer behavior — positive regard, "selling" the company, giving more job information — while gathering less information from the applicant, who in turn behaved more confidently and built better rapport (Dougherty, Turban, & Callender, 1994).

Preinterview test/biodata scores also bias interviewers: in a study of 577 life-insurance-agent candidates, interview ratings best predicted hiring and job survival for applicants with low biodata scores, and worst for applicants with high biodata scores — interviewers deferred to a high test score and gave the interview little weight, but leaned harder on the interview (making finer distinctions) when the test score was poor (Dalessio & Silverhart, 1994).

A decade-long McGill University research program (Webster, 1964, 1982) found early impressions dominate final decisions and rarely reverse — crystallizing after a mean of just four minutes. The interview is also primarily a search for disqualifying information: one unfavorable impression triggered a reject decision 90% of the time, while positive information carried much less weight (Bolster & Springbett, 1961). Even a handshake matters: Stewart, Dustin, Barrick, and Darnold (2008) found handshake quality shaped hiring recommendations by signaling extraversion, holding appearance and dress constant — though women received lower handshake ratings than men without receiving lower overall suitability ratings.

The McGill studies' most important finding may be that interviewers develop a personal prototype of a "good applicant" and favor candidates who match it (Rowe, 1963; Webster, 1964). When interviewers hold negative stereotypes that diverge from actual job requirements, evaluations of equally qualified minority candidates can suffer (Arvey, 1979); gender stereotypes work similarly, since traits linked to managerial success culturally resemble masculine more than feminine role traits (Aguinis & Adams, 1998), which may explain lower scores for female applicants (Arvey & Campion, 1982).

Contrast effects are well documented: an average candidate rated after several unfavorable candidates in a row tends to be rated favorably, because interviewers use other candidates as an implicit standard (Hakel, Ohnesorge, & Dunnette, 1970; Heneman, Schwab, Huett, & Ford, 1975; Landy & Bates, 1973). These effects are stubborn — warnings and anchoring procedures alone didn't eliminate them (Wexley, Sanders, & Yukl, 1973); only an intensive workshop combining observation, rating practice, and immediate feedback produced lasting change, a result replicated by Latham, Wexley, and Pursell (1975), whose workshop participants alone avoided contrast, halo, similarity, and first-impression errors six months later.

On information recall, Carlson, Thayer, Mayfield, and Peterson (1971) had 40 managers watch a 20-minute videotaped interview and then take a 20-question factual test: the average manager missed 10 of 20 items, and half missed enough to fail badly — except those who had taken notes during the interview, who scored accurately. The least accurate recallers defaulted to a halo strategy (rating the candidate uniformly higher with less variability), while accurate recallers used an individual-differences strategy, correctly recognizing intraindividual variation. Note-taking alone does not automatically improve interview validity, however — interviewers must be trained on what behaviors to note (Burnett, Fan, Motowidlo, & DeGroot, 1998; Middendorf & Macan, 2002); other memory aids include mentally reconstructing the interview's context and retrieving information from multiple starting points (Mantwill, Kohnken, & Aschermann, 1995).

Applicant Appearance and Other Personal Characteristics

Physical attractiveness helps only in jobs where attractiveness is itself relevant, but being unattractive is never an advantage (Beehr & Gilmore, 1982). Perceived obesity produced a small but statistically significant negative effect in one study (Finkelstein, Frautschy Demuth, & Sweeney, 2007), though another found overweight applicants were no less likely to be hired for high-public-contact jobs than low-contact ones (Pingitore, Dugoni, Tindale, & Spring, 1994).

Ethnicity alone shows limited direct bias evidence (Arvey, 1979; McDonald & Hakel, 1985) — the real effect runs through interviewer–applicant race similarity, not applicant race per se. But accent and name interact: applicants with an ethnic name who also spoke with an accent were rated less positively than ethnic-named applicants without an accent or nonethnic-named applicants either way (Segrest Purkiss, Perrewé, Gillespie, Mayes, & Ferris, 2006). A study of 1,334-plus police officers found a three-way interaction: African American interviewers rated African American interviewees more favorably than white applicants only when seated on a predominantly African American panel (McFarland, Ryan, Sacco, & Kriska, 2004).

Disability-status evidence is mixed — some studies show no effect (Rose & Brief, 1979), some show more negative ratings (Arvey & Campion, 1982), and some show more positive ratings (Hayes & Macan, 1997), likely because moderators like rater empathy shift the direction (Cesare, Tannenbaum, & Dalessio, 1990).

Applicant personality predicts interview success. In a study of 85 graduating seniors, conscientiousness (r = .38) and extraversion (r = .27) predicted follow-up interview invitations, while extraversion (r = .34), agreeableness (r = .27), openness (r = .23), and neuroticism (r = –.21) predicted job offers (Caldwell & Burger, 1998); once self-reported preparation was controlled, conscientiousness alone predicted interview invitations and extraversion/neuroticism (negative) alone predicted offers. A second study found trait negative affectivity affects interview success indirectly, through job-search self-efficacy and job-search intensity (Crossley & Stanton, 2005), and interview anxiety levels also relate to interview performance (McCarthy & Goffin, 2004).

Even artificial scents matter: in a controlled study, women rated applicants higher when they wore perfume/cologne, but men rated them lower — possibly reflecting gender differences in filtering out irrelevant grooming/appearance cues (Baron, 1983).

Coaching

Coaching — modeling, behavioral rehearsal, role playing, lecture (Maurer & Solamon, 2007; Tross & Maurer, 2008) — measurably improves interview performance. Maurer, Solamon, and Troxtel (1998) coached police and firefighter candidates on interview logistics, structured-vs-unstructured formats, needed KSAs, role-play practice, and interview tips; coached participants outscored noncoached peers across four job types (police sergeant/lieutenant, fire lieutenant/captain), holding up for three of four jobs after controlling for precoaching knowledge and motivation. A follow-up (Maurer, Solamon, Andrews, & Troxtel, 2001) replicated the effect.

Interviewer Experience, Cognitive Complexity, and Mood

Contrary to the intuitive hypothesis that similarly experienced interviewers should agree with each other (Rowe, 1960), Carlson (1967) found experience level had no effect on interrater agreement — interviewers get little benefit from routine experience because their day-to-day job lacks the training and feedback needed for real learning (Jacobs & Baratta, 1989). Experience improves decisions only when paired with higher cognitive complexity, which is really doing the work (Dipboye & Jackson, 1999). Ferguson and Fletcher (1989) found cognitive complexity linked to greater accuracy for female raters but not male raters — an unresolved, understudied finding.

Baron (1993) induced positive, negative, or neutral affect in 92 undergraduates before a simulated interview with applicants described as highly qualified, ambiguous, or unqualified. Mood shaped ratings only when qualifications were ambiguous or poor: with ambiguous qualifications, positive-affect raters scored the applicant higher than negative-affect raters; with unqualified applicants, positive-affect raters actually rated them lower than negative-affect raters did. Mood had no effect at all when the applicant was clearly highly qualified.

Structure has four dimensions (Chapman & Zweig, 2005): questioning consistency, evaluation standardization, question sophistication, and rapport building. Structure increases when questions are grounded in job analysis, the same questions are asked of every candidate, follow-up prompting/elaboration is limited, better question types are used (e.g., situational questions), interviews run longer with more questions, ancillary information (applications, résumés, scores, recommendations) is controlled, applicants can't ask their own questions until after the interview, each answer is rated on multiple scales with detailed anchors, detailed notes are taken, multiple and consistent interviewers are used across all applicants, interviewers receive extensive training, and statistical (rather than clinical) prediction is used (Campion, Palmer, & Campion, 1997).

The payoff is large and consistent. First, structured interviews are simply more valid: corrected validities range .35–.62 for structured interviews versus .14–.33 for unstructured ones (Campion et al., 1997). Second, structure shrinks racial subgroup differences — a meta-analysis found d = .32 (white–Black) across 10 low-structure studies versus d = .23 across 21 high-structure studies (Huffcutt & Roth, 1998), though range restriction makes both figures understated (Roth, Van Iddekinge, Huffcutt, Eidson, & Bobko, 2002). Third, structured interviews are far less likely to be successfully challenged in court for illegal discrimination (Williamson, Campion, Malos, Roehling, & Campion, 1997).

Terpstra, Mohamed, and Kethley (1999) reviewed 158 U.S. federal hiring-discrimination cases from 1978–1997: 57% of cases challenged unstructured interviews, versus only 6% challenging structured interviews. More striking, unstructured interviews were ruled non-discriminatory in only 59% of challenges, while structured interviews were ruled non-discriminatory in 100% of challenges — a compelling legal case for structure despite HR managers' persistent reluctance to adopt it (van der Zee, Bakker, & Bakker, 2002).

Why does structure work so well? Higher reliability alone doesn't fully explain it (Schmidt & Zimmerman, 2004); the more likely answer is that structured and unstructured interviews measure different constructs (Huffcutt, Conway et al., 2001). Structured interviews, built from job analysis, systematically assess job knowledge/skills, organizational fit, interpersonal/social skills, and applied mental skills like problem solving — giving them greater job relatedness. When structured, interviewers know exactly what to ask (a more consistent behavior sample across applicants) and what to do with the answers (better-informed ratings).

Experience-Based vs. Situational Questions

Structured interviews split into two question types. Experience-based (past-oriented) questions ask what the applicant actually did in comparable past situations (Janz, 1982; Motowidlo et al., 1992) — "Can you tell me about a time when...?" — resting on the assumption that past behavior best predicts future behavior. Situational (future-oriented) questions instead ask applicants to imagine a scenario and describe how they'd respond (Latham, Saari, Pursell, & Campion, 1980; Maurer, 2002) — "What would you do if...?" Situational interviews are highly valid and notably resistant to contrast error and race/gender bias, largely because of behaviorally anchored rating scales: Maurer (2002) found untrained business students rating videotaped situational-interview answers showed more accuracy and agreement than experienced police officers using a non-situational structured format.

Both formats derive from job analysis using the critical-incidents method (Chapter 9); incidents become interview questions, each answer is independently rated by two-plus interviewers on a five-point scale, and job experts write behavioral anchors for scores of 1, 3, and 5.

Taylor and Small's (2002) meta-analysis compared 30 situational validities and 19 experience-based validities: mean corrected validities of .45 (situational) and .56 (experience-based). Restricting to studies using behaviorally anchored rating scales raised both: .47 (situational, 29 coefficients) and .63 (experience-based, 11 coefficients). Mean interrater reliabilities were similar — .79 (situational) and .77 (experience-based). Some studies suggest situational interviews may be less valid for higher-level or more complex jobs (Pulakos & Schmitt, 1995; Huffcutt, Weekley, Wiesner, DeGroot, & Jones, 2001), but the meta-analytic evidence found no differential validity by job complexity for either format.

Social Media

By 2016, Facebook exceeded 1.9 billion users, Twitter 330 million, and LinkedIn 500 million (statista.com) — a massive source of big data on people's history, attitudes, preferences, and behavior (Harlow & Oswald, 2016; McFarland & Ployhart, 2015). An EEOC official noted roughly 75% of recruiters research applicants online, and 70% of those recruiters report rejecting candidates as a result (Roth, Bobko, Van Iddekinge, & Thatcher, 2016) — despite thin evidence that the practice is valid or fair.

Kluemper, Rosen, and Mossholder (2012) had raters judge 56 employed students' hireability from their Facebook pages and correlated scores with supervisor performance ratings: validity of .28 — but on a small sample, a hypothetical job, and a performance measure from the students' current (not necessarily supervisory) job. A larger, more troubling study had 86 recruiters rate graduating students' Facebook pages on hireability items (e.g., "I can see how this person would be an attractive applicant"); about 14 months later, supervisor performance data were collected for 142 hired students (Van Iddekinge, Lanivich, Roth, & Junco, 2016). Facebook ratings were essentially unrelated to supervisor performance ratings (r = –.13 to –.04), turnover intentions (r = –.05 to .00), or actual turnover (r = –.01 to .01), and added nothing beyond cognitive ability, the Big Five, core self-evaluation, self-efficacy, and GPA.

The chapter's verdict: social media data being abundant does not make it useful. It may still help recruitment broadly (wider, more global applicant pools) and help applicants get realistic previews (LinkedIn connections, Glassdoor ratings unfiltered by leadership), but its use specifically for selection decisions remains empirically unproven — "the jury is still out" (McFarland & Ployhart, 2015).

Mobile and Web-Based Selection

Per Tippins's (2015) review, computer/mobile delivery enables richer assessment formats — audio/video content, avatars, animation, a single assessment portal, drag-and-drop or thermometer-scale responses that avoid the scale-coarseness problem (Chapter 6). Large electronic test-taker databases let vendors build norms by job/industry/region and generate detailed applicant reports, useful for tracking promotion-relevant data over a career.

Pitfalls remain. Fast Internet access varies sharply by country — 73% in the U.S., 82% Canada, 94% South Korea, 86% Japan, 91% Switzerland, 88% Netherlands, but just 12% Panama, 4.9% India, 4.2% Philippines — so Web-based testing doesn't guarantee equal opportunity, especially if incidental KSAs like computer familiarity or typing speed aren't actually job related. Other challenges: uncontrolled test-environment distractions and cheating risk (addressable via proctors or webcams), plus applicant privacy concerns that employers should proactively address by explaining data storage and use safeguards.

Computer Scoring of Text

Campion, Campion, Campion, and Reider (2016) demonstrated automated essay scoring (AES)/computer-automated scoring (CAS), part of the computer-aided text analysis (CATA) family (Chapter 6), to score applicants' narrative responses. The core challenge is the information retrieval problem — matching a query's words to a document's words when applicants may describe "leadership" using the word manager instead. Using SPSS-IBM's Premium Modeler, Campion et al. built a program scoring six competencies (communication, critical thinking, people skill, leadership, managerial skill, factual knowledge) by identifying key terms and modeling relations among them. Computer-based scores matched human-rater reliability, showed construct validity, and delivered substantial cost savings versus human scoring.

Remote Interviewing

Videoconferencing and telephone interviewing let employers reach distant applicants cheaply (Chapman & Rowe, 2002; Schmidt & Rader, 1999), but differ meaningfully from face-to-face interviews. Phone interviews strip out visual cues (Silvester & Anderson, 2003) — which can reduce nonverbal-based interviewer bias, but also changes the process. Videoconferencing without a duplex system (both parties talking simultaneously) can alter interview dynamics.

A hybrid approach records a face-to-face interview for review by raters who weren't present (Van Iddekinge, Raymark, Roth, & Payne, 2006); a 113-student simulation found face-to-face ratings scored significantly higher than videotaped ratings of the same interviews, suggesting the two are not interchangeable. In a field comparison, 70 applicants for a multinational oil corporation job were randomly split into face-to-face-then-phone and phone-then-face-to-face groups (Silvester, Anderson, Haddleton, Cunningham-Snell, & Gibb, 2000): phone ratings (M = 4.30) were consistently lower than face-to-face ratings (M = 5.52) regardless of order — possibly because phone interviewers focus more on content than extraneous cues (which could make phone interviews more valid), or because applicants took the phone interview less seriously or had less practice with the format. A separate experimental comparison of face-to-face versus videoconferencing found applicants more satisfied with both the interviewer's and their own performance in the face-to-face condition (Chapman & Rowe, 2002).

Virtual Reality Technology (VRT)

VRT can place applicants in simulated job environments — a truck-driving simulator, a simulated chemistry lab — to demonstrate competence without real trucks or real hazardous chemicals, gathering information relevant to future on-the-job performance (Aguinis, Henle, & Beaty, 2001). Challenges include sopite syndrome (eyestrain, blurred vision, headache, balance disturbance, drowsiness; Pierce & Aguinis, 1997), cost and limited commercial availability (though falling — a Google Daydream View headset cost under $100), and technical limits like movement lag and cartoonish graphics. The chapter expects these limitations to keep shrinking as technology advances.

Cascio and Aguinis close every chapter with an "Evidence-Based Implications for Practice" list — worth reading as a final checklist before any quiz or discussion post.

• Several methods are available for early-stage selection screening. None offers a "silver bullet" — use them in combination rather than in isolation.
• Personal history data (application forms, BIBs, résumés) are most useful when built on a rational approach: questions and data collection grounded in job analysis and explicit hypotheses linking item constructs to job-performance constructs.
• Recommendations and reference checks are most useful when applied consistently to every applicant and when the information gathered is genuinely relevant to the position.
• Polygraph testing is likely to produce errors; administrators should remember that physiological indicators can be consciously altered by applicants.
• Honesty/integrity tests are either overt or personality-oriented. Given unresolved challenges, consider alternatives to the traditional paper-and-pencil format, including situational-judgment and conditional-reasoning tests.
• Training-and-experience evaluations are most useful when directly relevant to specific job-related areas.
• Drug screening is most effective and least legally vulnerable when framed within a safety-and-health context, as part of a comprehensive drug-use policy.
• Employment interviews are used almost universally. Validity is shaped by social/interpersonal issues, cognitive biases, individual differences in both interviewer and interviewee, interview structure, and format (face-to-face vs. videotaped).
• Big data and technological advances (social media, mobile/Web-based selection, virtual reality) create new opportunities, but much remains unknown about their validity and reliability before wider recommendation.

Chapter 12 ends with thirteen discussion questions. Each is paired below with a concise model answer grounded in the chapter's content.

1. How can the usefulness of recommendations and reference checks be improved?

Require the four preconditions the chapter names: adequate opportunity to observe the candidate in job-relevant situations, competence to evaluate, willingness to be candid, and clear communication of the evaluation. Structure the process (consistent, relevant, written, and grounded in public records), consider structured telephone reference checks with relative rating scales and elaboration prompts to curb leniency, and remember written letters average only .14 validity — reference checks (.26) are the stronger tool.

2. Is the use of video résumés and credit history scores equally fair for all job applicants? Why or why not?

No. Video résumés appear to reduce perceived unfairness across ethnicities relative to paper résumés (Hiemstra et al., 2012), but credit history scores correlate strongly with minority status, education, and age (Bernerth, 2012) — meaning their use predictably produces adverse impact regardless of intent. Several states now restrict credit checks for exactly this reason; video résumés carry no comparable documented adverse-impact pattern in the chapter.

3. As CEO of a large retailer considering drug testing to screen new hires, what elements should the policy include?

Put the policy in writing and communicate it to all employees and applicants; include it in employment contracts; frame it around health and safety benefits rather than surveillance; give advance notice before testing; preserve a right to appeal; minimize invasiveness; and train supervisors on fair administration. This mirrors the Postal Service's approach and the chapter's list of fairness-enhancing procedures (Konovsky & Cropanzano, 1991; Tepper, 1994; Angarola, 1985).

4. What instructions minimize response distortion on a biodata instrument?

Tell applicants items will be verified where possible, warn them a lie scale is embedded in the instrument, and frame the instrument as non-evaluative/classification-oriented rather than a pass/fail gate. Where feasible, require elaboration on answers — describing the specific situation behind a response — since this consistently reduced distortion by about .6 SD in the chapter's cited studies (Schmitt & Kunce, 2002; Schmitt, Oswald, Kim, Gillespie, & Ramsay, 2003).

5. What is the difference between personality-based and overt honesty tests, and what does each measure?

Overt integrity tests directly ask about attitudes toward theft/dishonesty and about admissions of past theft or illegal activity. Personality-based measures instead assess broader dispositional traits — socialization, conscientiousness — that predict a wide range of counterproductive behavior (substance abuse, insubordination, absenteeism, passive aggression) without asking about theft directly. Despite the different surface content, both share a common latent structure of conscientiousness, agreeableness, and emotional stability (Berry, Sackett, & Wiemann, 2007).

6. Are you in favor of or against polygraph testing for security screening at airports? Why?

The chapter's evidence leans against it: the National Research Council concluded polygraph accuracy for screening is insufficient to justify its use, since it forces a choice between falsely flagging loyal employees and missing genuine security threats, and results can be beaten through conscious countermeasures. A defensible position argues against polygraph-based airport screening on these grounds while acknowledging no alternative technology has yet proven superior.

7. An interviewer asks a question you believe invades your privacy. What do you do?

The chapter's own research cautions against simply refusing to answer: employers tend to interpret nonresponses as concealment (Stone & Stone, 1987). A better approach is to answer factually but briefly, or to ask the interviewer to clarify the question's job relevance — documenting the exchange afterward if it seems like a pattern, since legally indefensible questions (not job related, potential adverse impact, or invasive) remain a documented problem on over 95% of studied application forms.

8. Employers generally weight experience over academic qualifications. Why, and should it be so?

Evidence shows experience-based methods like behavioral consistency evaluation reach validities as high as .45, and employer surveys rank attitude, communication skills, and work experience above grades, school reputation, and recommendations (Applebome, 1995). Academic credentials can even backfire for candidates with weak experience (Singer & Bruhns, 1991), and grade-based screening tends to produce adverse impact on minority applicants (Roth & Bobko, 2000). This weighting is reasonably justified where jobs require demonstrated behavior over credentialed potential, though it should still rest on job analysis rather than blanket policy.

9. Discuss the advantages of computer-based screening (CBS). Given these advantages, why isn't it more popular?

CBS standardizes administration, automates response recording, widens the applicant pool via remote access, and can accommodate disabilities. Adoption lags because of development cost, a persistent lag in scientific guidance on CBS reliability/validity, doubt about tangible payoff (Olson-Buchanan, 2002), and the digital divide that limits access for low-income applicants (Stanton & Rogelberg, 2001).

10. Your boss asks you to build a training program for employment interviewers. How do you proceed?

Base the program on the chapter's seven suggestions: ground questions in job analysis; standardize the six-step structured interview format; train interviewers to anchor rating scales with concrete examples; include supervised practice interviewing diverse and disabled applicants; teach interviewers to focus on story-related deception cues rather than nonverbal cues; and document the whole system for legal defensibility. To measure whether it's working, track interrater reliability before and after training, and, per Latham, Wexley, and Pursell (1975), test whether contrast/halo/similarity/first-impression errors persist six months later — only intensive, feedback-based workshops produced durable change in the chapter's cited research.

11. Discuss the advantages of structured over unstructured interviews. Why are HR managers still reluctant to use them?

Structured interviews show higher validity (.35–.62 vs. .14–.33), smaller racial subgroup differences, and dramatically better legal outcomes (100% vs. 59% ruled non-discriminatory when challenged). Reluctance persists because structure feels rigid, reduces interviewer autonomy and rapport-building flexibility, and requires more upfront investment in job analysis, question design, and interviewer training than a casual conversation-style interview (van der Zee, Bakker, & Bakker, 2002).

12. Discuss potential pitfalls in using social media for selection purposes.

The strongest available study found Facebook-based hireability ratings essentially uncorrelated with actual supervisor performance ratings, turnover intentions, or turnover (Van Iddekinge, Lanivich, Roth, & Junco, 2016), and added nothing over standard predictors like cognitive ability and the Big Five. Pitfalls include unclear job relatedness of what's being judged, risk of adverse impact from indirectly signaled demographic information, inconsistent application across candidates, and legal exposure from using unverified, non-job-related content in a hiring decision.

13. What empirical evidence would clarify when and how social media can be used for selection?

Large-sample, predictive (not just concurrent) validity studies linking specific social-media-derived ratings to actual job performance and retention, replicated across job types and demographic groups; adverse-impact analyses comparable to those required for biodata and tests; and construct-validity work clarifying exactly what a "Facebook rating" measures, the same rational-approach standard the chapter applies to biodata and honesty tests.

Every bolded or explicitly defined term in Chapter 12, in one line each, in the order the chapter introduces them.

A single table capturing every selection method's headline validity figure and defining feature — everything needed to answer a cold-call question about Chapter 12 without re-reading it.