The Use of Biodata for Employee Selection

Past research and future directionsObjectives:1. Provide a selective but representative

review of the research that has been conducted on the use of biodata for employee selection

2. To constructively critique this research to highlight deficiencies that may limit the conclusions that should be drawn

3. To stimulate important future research on biodata that avoids the limitations of past research

BiodataArticle Overview

1. Biodata Research: A selective review of the research

1.1. A Study by Goldsmith (1922)

1.2. Defining and operationalizing biodata: Differences in definitions and the types of items used

1.3. Methods of gathering biodata

1.4. Strategies used for developing biodata scales

1.5. The Reliability of Biodata Scales

1.6 The Validity of Biodata Scales

1.7 Adverse Impact 1.8 Applicant Reactions to

Biodata 1.9 Incremental Validity 1.10. The Accuracy of Biodata 1.11. Computing a Biodata

Scale Score: Unit Weighting versus differential weighting

1.12. The Generalizability of Biodata Scales

2. Past Biodata Research: Three potential concerns

2.1 The Heavy Reliance of Past Studies on a Concurrent Validity Design

2.2 The Type of Biodata Scale Used

2.3 The Lack of Information Provided on Biodata Items

3. What is biodata? And why does it predict employee behavior?

3.1 What is biodata? 3.2 Why Does Biodata

Predict Employee Behavior?

4. Biodata: Future research directions 4.1 What is biodata? 4.2 Do results for concurrent

validity studies generalize to a selection context?

4.3 Increased research with an item-focus

4.4 Greater focus on the use of technology

4.5 Ways to increase the accuracy of biodata information

4.6 The value of a biodata clearinghouse

Rethinking the use of a factorial biodata development strategy

5. Concluding Remarks

1.2. Defining and Operationalizing BiodataDifferences in definitions and the types of items

used

Factual information about life and work experiences, as well as items involving opinions, values, beliefs, and attitudes that reflect a historical perspective.

Narrowly definedBehaviors and events that occurred earlier in life

How many jobs have you had in the past 5 years?How long have you been in your previous job?

Broadly DefinedTemperament, assessment of working conditions,

values, preferences, skills, aptitudes, and abilities.I like doing things with other people.My teachers regarded me as a sociable boy/girl.

1.2. Defining and Operationalizing BiodataDifferences in definitions and the types of items used

Mael’s (1991) definition:Does include items pertaining to historical

events that may have shaped the person's behavior and identity

Does not include items that address such variables as behavioral intentions, self-descriptions of personality traits, personal interests, and ability

Advantages of Mael’s historical nature definition of biodata:

Accuracy in reporting of discrete verifiable events

More favorable view of questions that applicant views as job related and reflect experience under applicants control

BiodataIntroduction

“One of the best selection devices for predicting turnover”

Organizations rarely use biodata (<17%)Less than one page devoted to biodata in

Evers, Anderson, and Voskuijl's Handbook of Personnel Selection (2005)

A PsychINFO database search in 2008 for the term 'biodata' turned up one article

*Survey of 255 HR professionals ranked biodata as lacking in terms of validity, practicality, and legality

1.1. A Study by Goldsmith (1922)Examined the ability of 9 “personal history” items to

predict the first-year sales of insurance agents Marital status Education Belonging to clubs

Found that using a person’s biodata score would improve hiring decisions made

58 of the 259 individuals receiving a score of 4 or above were considered successful (22%)

11 of the 243 individuals receiving a score less than 4 were considered successful (4%)

Then and Now:Used few biodata itemsA number of items used would not be used today (age)Did not report data on the relationship of a given item and salesProvided an explanation for using each item

1.3. Methods of Gathering Biodata

Web-basedTelephonePaper-and-pencil

Study by Ployhart, Weekley, Holtz, and Kemp (2003) compared scores from paper-and-pencil measure to those obtained from a web-based version of the measure

o Web based group had lower mean score (may suggest less faking)

o Lower scores in terms of skew and kurtosis

Mumford (1999) suggested there may be benefits from using a greater variety of data gathering methods.

1.4. Strategies Used for Developing Biodata Scales

Researchers use a combination of strategies:1. Empirical“Dust-bowl empiricism”

2. Behavioral Consistency“Best predictor of future behavior is past behavior”

3. Rational/DeductiveJob Analysis/Theories

4. FactorialAttempting to explain ‘why’ there is a correlation

5. SubgroupingDifferent groups use different constructs when answering

Scale Development1. The Empirical Approach

“Dust Bowl Empiricism” No theory is involved behind the study. Solely refers to instances arising from entirely inductive processes. We just want to know which items are significantly correlated to form the scale.

Large pool of items are used, those that are predictive are chosen for use in the scale

Ideally a cross-validation study would be conducted

Example: Finding a high correlation between two variables, job turnover and amount of jobs held in past five years, and including ‘amount of jobs held in past five years’ as part of your biodata scale.

Scale Development2. Behavioral Consistency Approach

Past behavior predicts future behaviorSelects items that are consistent with the

criterion of interestCausal variables are usually not

investigatedExample: When interested in predicting

turnover ask, “how long have you been at your most recent job?”Stable work ethic?

Scale Development3. Rational/Deductive Approach

Conduct a job analysis to determine KSAs relevant for the criterion of interest or

Use recent research/theories in development of questions

Criterion of interest = voluntary turnoverKnowledge of the job (realistic

expectations) reduces voluntary turnoverExample: “Do you know someone who works for the organization?”

Scale Development 4. Factorial Approach

Principal Axis Factor Analysis/ Principal Components Analysis

Explain why biodata scales predict the criterion of interest

Extracts underlying 'factors' that cause the statistical relationship to existEmpirical example: Job turnover and amount of jobs held in

past five years? Age

Behavioral Consistency example: turnover and amount of time at most recent job?

Work ethic

Rational/Deductive example: Knowing someone working for the organization and voluntary turnover?

Realistic expectations

Scale Development5. Subgrouping

Different groups may have different patterns of constructs that underlie their responses to biodata itemsTypes of biodata items that best predicted military suitability for high school graduates differed from those that predicted suitability for non-graduates

1.5. The Reliability of Biodata ScalesOne construct

e.g. past experience interacting with peopleCoefficient alpha appropriate Estimates range from .50-.80

A variety of constructse.g. marital status, age, schooling completed, and number of jobs held in the past 5 yearsCoefficient alpha not appropriateTest-Retest reliability may be appropriateEstimates range from .60-.90

1.6 The Validity of Biodata ScalesCriterion-related Validity

Research shows that biodata is a good predictor of:Job performanceVoluntary turnover

1.9 Incremental ValidityMount, M. K., Witt, L. A., & Barrick, M. R. (2000)

Biodata added unique variance in predicting supervisory ratings of performance beyond that accounted for by tenure, general mental ability, and the Big Five personality traits

Allworth and Hesketh (2000)Biodata scale accounted for unique variance in

performance ratings when added after a cognitive ability test

1.7 Adverse ImpactCauses for concern occur when

biodata items are used regarding:Educational levelCognitive ability (GPA)

Use careful item screeningCompared to other selection devices, biodata

has modest adverse impact

1.8 Applicant Reactions to Biodata

Poor face validityApplicants are likely to react negatively to items

that are perceived as lacking job relatedness, fakable, and overly personal in nature

Studies usually do not involve applicantsStudents or current employees

1. How did you typically prepare for final exams in college? A. Studied a few hours every day across several weeks B. Studied many hours over a few days C. Studied the entire night before each exam D. Did not study

2. How often are your library books overdue? A. Always B. Often C. Rarely D. Never E. I never take books out of the library

3. To what extent have you enjoyed being given a surprise party? A. Not at all B. To a slight extent C. To a moderate extent D. To a great extent E. I have never been given a surprise party

4. In the past year, how many times have you thrown something when you were angry? A. 0 times B. 1 - 2 times C. 3 - 4 times D. 5 - 6 times E. 7 or more

BiodataSample FBI Inventory

This inventory contains 40 questions about yourself. You are to read each question and select the answer that best describes

you from the choices provided. Answer the questions honestly; doing otherwise will negatively affect your score.

1.10. The Accuracy of Biodata

StudentsFairly AccurateExternal verification from parents supports the

self-reported student dataApplicants

Accuracy was mixed when studies were conducted in a selection context

Faking 'good' answers

1.11. Computing a Biodata Scale Score

Unit Weighting versus differential weighting

1. Correlational (Unit) MethodCompute a simple correlation between an

item and the criterion, then use this value to weight the item More highly correlated items receive higher weights

2. Differential Regression MethodSelect all biodata items that are significantly

correlated to the criterion and unit weight them. Differential regression method is most beneficial

when the correlations among the items are low, there are relatively few items, and there is a large sample

Both methods tend to provide comparable results.

1.12. The Generalizability of Biodata Scales

Will a biodata scale developed in one organization be valid if applied in another organization?

In the U.S., research shows that biodata scales have predicted:Brown (1981)

Sales volume for insurance agents across 12 companies

Rothstein, Schmidt, Erwin, Owens, and Sparks (1990)Performance of supervisors across

organizationsCarlson, Scullen, Schmidt, Rothstein,

and Erwin (1999)Rate of promotions across 24 organizations

1.12. The Generalizability of Biodata Scales(International)

Laurent (1970)Valid scale for managers in the US was

also valid in predicting management success in Denmark, Norway, and the Netherlands

Dalessio, Crosby, and McManus (1996)Scale used to select insurance agents in

the US used with equal effectiveness in the United Kingdom and Ireland

1.12. The Generalizability of Biodata Scales Overtime

Brown (1978)Scale developed in 1933 for selecting insurance agents

predicted survival and performance of agents in 1969-1971

Rothstein, H. R., Schmidt, F. L., Erwin, F. W., Owens, W. A., & Sparks, C. P. (1990)Validity coefficients of studies done in 1974 and 1985

were similarCarlson, K. D., Scullen, S. E., Schmidt, F. L.,

Rothsteing, H., & Erwin, F. (1999)Scoring key for the Manager Profile Record yielded valid

scores up to 11 years after the key was developed*Stability likely due to researchers using items that were generic/attributes of the jobs tapped by the biodata items have not changed greatly

2.Past Biodata ResearchThree potential concerns

2.1 Heavy reliance on concurrent validity designs

2.2 Type of biodata scale used2.3 Lack of information provided on

biodata items

2.1. The Heavy Reliance of Past Studies on a Concurrent Validity Design

Stokes, Hogan, and Snell (1993)Studied sample of incumbents

working in a sales position and applicants who had applied for the position

o Developed two scales to predict turnover (i.e. job applicant scale and job incumbent scale)

o Validities of scale were similar• Job Incumbent .22• Job Applicant .23

Switched the scales, i.e. gave job applicant the job incumbent scale

● Validity Coefficient = .08● Biodata scales developed for

each group had no items in common

Harold, McFarland, & Weekley (2006)425 call center employees

and 410 applicants respond to 20 biodata item Validity coefficients higher for job

incumbents (.27) than job applicants (.18)

Are results taken from current employees comparable to job applicants?

2.2 The Type of Biodata Scale Used

Generic vs. Situation-specific ScalesDeveloping situation-specific biodata scales may result in higher validity than a more generic scaleoSituation-specific validity coefficient = .33

oGeneral validity coefficient = .22Expensive to develop

• Writing items• Pilot testing

*Generic scale is better than no scale

2.3 The Lack of Information Provided on Biodata Items

Researchers often do NOT report the actual items they used due to:

Lengthy biodata measures, < 100 items, journal space issue

Used biodata items sold by vendors who do not allow publication of their items

Therefore, most studies have not reported:1. Correlation between each biodata item and the

criterion used in the study2. How each item was weighted in creating the scale3. Whether an item provided unique variance in

predicting a criterion variable4. Whether an item had adverse impact5. Correlations among biodata items

2.3 The Lack of Information Provided on Biodata Items

Imagine you are developing a new biodata scale. How would this omitted information be beneficial?

Valid predictors in past studiesAdverse impactNon-significant findings

Allow selection of biodata items that are of maximum value while limiting the number of items that are used

3. What is Biodata? And Why Does It Predict Employee Behavior?

3.1 What is Biodata? 3.2 Why does it predict employee

behavior?

3.1 What is Biodata?

Article’s Position: Biodata consists of applicant’s past behavior and experiencesPast behaviors and experiences can reflect events that occurred

in a work context (quit a job without giving notice), an educational setting (graduated from college), a family environment (traveled considerably growing up), community activities (led a cub scout troop), or other domains (active in local politics)

Does not mean that past experiences are unrelated to such variables as interests, personality, values, knowledge, and skills

Schmidt et al. (1999) It is likely that an individual who possesses certain interests,

personality traits, values, and/or KSAs will be more likely to seek out certain situations that are captured by historical biodata

In summary: Many of the variables (personality traits) that have commonly been confounded with biodata are actually antecedents of consequences of the personal experiences that biodata taps

3.2 Why Does Biodata Predict Employee Behavior?

Most studies focus on criterion-related validity and few models offer an explanation to ‘why’

Mumford, Owens, and Stokes (1987, 1990) developed the (Interactive) Ecological Model to help determine the “why”

3.2 Why Does Biodata Predict Employee Behavior?

Person's life begins with certain environmental and hereditary resources...

• A nurturing mother• excellent eyesight

...and certain limitations...• Substandard nutrition• Poor coordination

...which determine individual differences early in life. • High cognitive ability• Poor health• Self-confidence

Given these individual differences, an individual attempts to maximize adaptation to the environment.

• The ecological model presumes that an individual makes decisions about what situations to enter.. • What college to attend• Whether to accept a job offer

…based upon the perceived value of the outcomes

• Social status• Financial rewards• Intrinsic satisfaction

…that are likely to be derived from the situations.• Interactive: An individuals choice at a given point in time about

what situation to enter affects his subsequent development, which influences his future choices of situations, which affect future development, etc. Thus, over time, an individual may develop new skills, satisfy existing needs, increase academic goals, or decrease his work ethic.

3.2 Why Does Biodata Predict Employee Behavior?

Time 1An individual possesses several

attributes, based on these attributes the model suggests the individual will

actively choose to enter a new situation/environment that is perceived

to aid in development.

Time 2The experience of the new environment

will lead to changes in the person’s attributes.

Time 3The individual is now different on one or

more attributes than at Time 1.

Environmental Experience at Time 2

Environmental Experience at Time 2

.

Non-choice/Uncontrollable events: e.g. unemployment, health problems

3.2 Why Does Biodata Predict Employee Behavior?

By limiting the definition of biodata to past behaviors and experiences, the data collection is on events reflected in the boxes labeled 'Environmental Experience at Time 2'.

Did you graduate from college?How long were you at your most recent job?What percentage of you college expenses have you paid?

This model explains how defining biodata by past behavior and experiences does NOT mean a biodata score is unrelated to such variables as conscientiousness, ability, interests, knowledge, etc. Rather, it shows how such variables are likely antecedents and/or consequences of an individual's behaviors and experiences.

Better way to get information that could be fakedAsking how long someone was in a prior position in sales helps

gauge the persons dependability, knowledge of a sales position (realistic job expectations), communication skills, etc. without asking 'Are you a dependable person'

4. Biodata: Future Research Directions

4.1 What is biodata?Doubtful that rigorously designed empirical biodata studies will result

in a consensus, more likely that cogent arguments by experts will need to persuade the research community

4.2 Do results for concurrent validity studies generalize to a selection context?

Increase predictive validity research designs Concurrent validity coefficients may overestimate coefficients for

applicants

4.3 Increased research with an item-focus

Increase attention to item-level issues Work vs. education, amount vs. time

4.4 Greater focus on the use of technology Individual differences different questions different scoring keys?

4. Biodata: Future Research Directions

4.5 Ways to increase the accuracy of biodata information

Elaboration lowers scores (decrease faking), but does this increase validity?

4.6 The value of a biodata clearinghouseEasier scale development but risk of compromising ‘answers’

4.7 Rethinking the use of a factorial biodata development strategy

Varimax rotation-constructs found may be correlated (explain little variance)

PCA-large number of biodata items used warrants large sample size

May result if valid biodata items being dropped from final scale

Confirmatory Factor Analysis may be more appropriate

5. Concluding RemarksConcerns are unfounded or only true for certain biodata scalesValidity

Research has shown biodata to be an excellent predictor

LegalityAdverse impact and a lack of face validity may be minimized by careful

selection of items

PracticalityBiodata scales do not need to involve a large number of items

Barrick and Zimmerman (2005) and O’Connell et al. (2002) used < 10 items

To further use of biodata and research 3 issues need to be addressed

1. Agreement on what biodata is2. Greater reliance on predictive validity designs3. Greater attention given to the specific biodata items used in

studies