Psychology › Section M · Industrial-organizational psychology › Topic M.2
Selecting and evaluating employees
Every hiring decision is a prediction: this person, out of all who applied, will do the job well. Industrial psychology tests which predictions work, how to train people once hired and how to judge their performance fairly and within the law.
Hiring as prediction
Industrial psychology, the first of the three areas met in the previous topic, matches people to jobs. The process has four linked steps: describe the job, select applicants who fit it, train them, and appraise their performance. Each step can be tested scientifically. Does an interview predict who performs well? Does training change behaviour? Do ratings measure performance or the rater's biases? Fairness runs through all four steps, as good practice and as law.
Job analysis
Job analysis is the systematic description of a job. A task-oriented analysis lists what the job involves, often scoring each task's frequency, difficulty and importance. A worker-oriented analysis, sometimes called a job specification, lists the knowledge, skills, abilities and other characteristics (KSAOs) a person needs. Analysts gather information by observing skilled workers, interviewing job holders and supervisors, and using questionnaires. The US Department of Labor's free O*NET database holds detailed descriptions of hundreds of occupations. Selection tests, training plans and appraisal forms should all trace back to the job analysis.
Selection tools and validity
Organisations use cognitive ability tests, job knowledge tests, work samples (doing a slice of the real job), personality questionnaires, integrity tests, interviews and assessment centres, where candidates complete exercises while trained observers rate them. The key question for each tool is predictive validity: how well do scores correlate with later job performance? A validity coefficient is a correlation, \(r\), and \(r^2\) gives the proportion of variation in performance the tool accounts for. Large reviews find that structured interviews, job knowledge tests, work samples and cognitive ability tests are among the better predictors. Among personality traits, conscientiousness predicts performance most consistently across jobs.
Worked example: comparing two selection methods
A firm compares interview ratings with job performance a year later. Its unstructured interview has a validity of \(r = 0.20\); a new structured interview has \(r = 0.40\). How much better is the new method? 1. Square each coefficient: \(0.20^2 = 0.04\) and \(0.40^2 = 0.16\). 2. Interpret: the old interview accounts for 4% of the variation in performance, the new one for 16%. 3. Compare: doubling \(r\) quadrupled the variance accounted for. 4. Check the limits: even 16% leaves most of the variation unexplained, so combining several valid tools usually predicts better than any one. 5. Conclusion: the structured interview is clearly better, but no method guarantees a good hire.
Structured interviews
In an unstructured interview the interviewer asks whatever seems relevant, so different candidates face different questions and answers are judged by impression. In a structured interview every candidate gets the same job-related questions, prepared in advance from the job analysis, and answers are scored on the same rating scale, often by a panel. Reviews, including a 1994 meta-analysis by Michael McDaniel and colleagues, find that structured interviews predict job performance better. Interviews are also open to bias: interviewers tend to favour candidates who resemble them, and first impressions, smiling, eye contact and accent can sway ratings. Structure reduces, but does not remove, these effects.
Training and mentoring
Training begins with orientation, now often called onboarding, which covers policies and, just as importantly, the organisation's culture: its values, norms and ways of making decisions. Job-specific training follows, along with retraining when tools or duties change. Mentoring pairs a new employee, the protégé, with an experienced one; reviews find small but real benefits for protégés' performance, satisfaction, pay and promotion, and benefits for mentors too. Donald Kirkpatrick's widely used model evaluates training at four levels: reaction (did trainees like it?), learning (did they gain knowledge?), behaviour (do they use it on the job?) and results (did productivity, safety or profits improve?).
Performance appraisal
Performance appraisal is the formal evaluation of how well employees do their jobs, used for feedback, pay, promotion and training decisions. Ratings are prone to errors. The halo effect is letting one strong impression colour every rating. Leniency is rating everyone too high, central tendency is rating everyone as average, and recency is weighting the last few weeks too heavily. Remedies include ratings tied to the job analysis, behaviourally anchored rating scales that describe what each score looks like, and rater training. In 360-degree feedback, ratings come from supervisors, peers, direct reports, the employee and sometimes customers; their disagreements can prompt useful discussion.
Worked example: auditing an appraisal
A manager rates all twelve team members "excellent" on every item. One member, Sam, who is charming in meetings, also gets top marks for report accuracy, although Sam's reports often need correcting. Diagnose and fix. 1. All-excellent ratings for everyone: leniency error, which makes the ratings useless for comparing people. 2. Sam's accuracy rating driven by charm: halo effect. 3. Check against the job analysis: report accuracy is a listed task and can be measured directly, for example by the number of corrections. 4. Fixes: behaviourally anchored scales for each task, notes collected through the year, rater training and a second rater. 5. Caution: feedback itself does not always help; a 1996 review found that about a third of feedback interventions made performance worse.
Fair hiring and the law
In the United States, the Equal Employment Opportunity Commission (EEOC) enforces federal laws forbidding job discrimination based on race, colour, religion, sex (including pregnancy, sexual orientation and gender identity), national origin, age (40 or over), disability or genetic information. Key laws include Title VII of the Civil Rights Act (1964), the Age Discrimination in Employment Act (1967) and the Americans with Disabilities Act (1990), which requires reasonable accommodation, such as ramps or flexible breaks, and bars disability questions before a job offer. The Supreme Court's Bostock decision (2020) applied Title VII to sexual orientation and gender identity. Federal law does not ban every question about age, but basing decisions on it is illegal, so careful employers avoid asking.
Narrow exceptions and hidden bias
A bona fide occupational qualification (BFOQ) is a rare, legally accepted exception that lets an employer select by sex, religion, national origin or age when that trait is essential to the job; race can never be a BFOQ. Courts read BFOQs narrowly, and customer preference is not enough: in the 1970s Pan Am was not allowed to hire only women as flight attendants. The mandatory retirement age for US airline pilots, 65 under federal rules, is an accepted age-based limit. Bias also works below the law's radar. Studies have found that attractive applicants are favoured in hiring decisions, and that recruiters with negative automatic associations about obesity invited fewer equally qualified applicants with obesity to interview.
Common mistakes to avoid
First, skipping the job analysis: without it, tests and appraisals have nothing valid to aim at. Second, trusting the unstructured "gut feeling" interview over structured methods. Third, reading a validity coefficient as a percentage: \(r = 0.40\) accounts for 16% of performance variation, not 40%. Fourth, judging training by trainees' enjoyment alone, the lowest Kirkpatrick level. Fifth, confusing the halo effect, one impression colouring all ratings, with leniency, rating everyone high. Sixth, assuming a BFOQ can be based on race or on customers' preferences: neither is allowed.
Key takeaways
Job analysis is task-oriented or worker-oriented (KSAOs). Selection tools are judged by predictive validity, with r² giving the share of performance variation explained; structured interviews outperform unstructured ones. Training is evaluated at Kirkpatrick's four levels, and mentoring brings small real benefits. Appraisals suffer from halo, leniency, central tendency and recency errors, reduced by behaviourally anchored scales; 360-degree feedback adds viewpoints. US law, enforced by the EEOC, forbids discrimination on protected traits; BFOQs are narrow and never based on race.
Flashcards
Flashcard
A systematic description of a job's tasks and requirements
Test your understanding
Question 1 of 5
Why do structured interviews usually predict job performance better than unstructured ones?
Question 2 of 5
Three months after a safety course, supervisors check whether trainees now follow the new procedures on the job. Which of Kirkpatrick's levels is this?
Question 3 of 5
A manager gives Lee top marks for punctuality and accuracy mainly because Lee once handled a difficult client brilliantly. This is an example of:
Question 4 of 5
A selection test has a predictive validity of r = 0.50. What share of the variation in job performance does it account for?
Question 5 of 5
Under Title VII, which characteristic can never be used as a bona fide occupational qualification?
