An Examination Of Biodata Theory-Based Constructs In A Field Context

An Examination of Biodata Theory-Based
Constructs in a Field Context
Michelle A. Dean
San Diego State University
Craig J. Russell
University of Oklahoma
This study examined whether items from a biodata inventory were consistent with theory-
based constructs as suggested by the ecology model. The model posits five constructs
explain biodata predictive ability. Items were sorted using the theory-based construct
domains. The resulting scale scores were assessed for evidence of content, criterion-
related, and construct validity in a sample of 6032 automated systems controllers. Results
suggested moderate support for the theory-based constructs. A notable finding was that
items capturing a theoretical construct were impacted by the life stage in which they were
anchored. Results further suggest the need to examine if certain life stages are more
predictive than others or perhaps if certain constructs are more predictive when related to
a particular life stage.
B iographical information (biodata) personnel selection
inventories demonstrate high predictive validity
across employment venues and criteria (Hunter & Hunter,
1984; Reilly & Chao, 1982; Reilly & Warech, 1990;
Russell & Dean, 1994; Schmitt, Gooding, Noe, & Kirsch,
1984). Despite cumulative evidence of successful perfor-
mance prediction, biodata research has received little
praise for innovative theoretical explanations of its ability
to predict criterion performance (cf. Fleishman, 1988).
Schmidt, Ones, and Hunter (1992) suggested the need
for greater attention to be paid to constructs underlying
biodata items to advance biodata theory and practice. They
argued that lack of attention to latent construct domains
prevents realization of biodata’s full potential. Although
evidence of criterion validity is important, Stokes and
Cooper (2001) suggested that approaches in addition to
criterion validation (such as construct and content valida-
tion approaches) should be used to understand biodata
within the context of a theoretical framework. Fleishman
(1988) suggested that the greatest opportunities for under-
standing human performance in organizations will come
from examining relationships between life experiences and
job performance. The ability to examine life experience-
performance links makes biodata unique among selection
devices. However, a much greater understanding of these
linkages is needed in order to proactively guide life
experience interventions to achieve desired levels of
knowledge, skills, and abilities (KSAs).
The present study builds on past theoretical efforts by
examining a biodata inventory to determine whether
theory-based latent construct domains were present in a
set of rationally developed items and whether scales
derived from items loading onto these constructs exhibited
construct, content, and criterion validity in a large-scale
biodata selection application. A brief review of existing
biodata theory is presented followed by a description of the
current study.
Biodata Theory
Efforts undertaken to develop explanatory models for
biodata’s predictive ability started with Wernimont and
Campbell’s (1968) capstone description of the behavioral
consistency principle (Owens & Schoenfeldt, 1979).
Wernimont and Campbell described the notion of beha-
vioral consistency as ‘‘that familiar piece of conventional
wisdom, ‘The best predictor of future performance is past
performance’’’ (p. 372). Unfortunately, this is often where
the discussion begins and ends even though this principle is
not a feasible explanation for biodata (Dean, Russell, &
Muchinsky, 1999). Biodata items generally do not predict
future performance (or behavior) from past measures of the
*Address for correspondence: Michelle A. Dean, Department of
Management, College of Business Administration, San Diego State
University, 5500 Campanile Drive, San Diego, CA 92182-8238, USA.
E-mail: mdean@mail.sdsu.edu
INTERNATIONAL JOURNAL OF SELECTION AND ASSESSMENT VOLUME 13 NUMBER 2 JUNE 2005
139
r Blackwell Publishing Ltd 2005, 9600 Garsington Road, Oxford OX4 2DQ, UK and
350 Main Street, Malden, MA 02148, USA.

same. Instead, biodata is hypothesized to capture causal
events or correlates of causal events influencing job
candidates’ future behavior. The consistency principle can
only serve to explain predictive power for the small subset
of biodata items that ask about past behavior identical to
criterion behavior of interest (Dean et al., 1999).
Biodata theory has moved forward incrementally from
the consistency principle to address why past behaviors/
performance, or ‘‘life events,’’ predict non-identical future
performance. A number of hypotheses and partial explana-
tions have been developed. For example, Mael (1991) used
social identity theory to suggest life events shared by
different social groups meaningfully and causally contri-
bute to KSA development and future performance. Alter-
natively, Dean et al. (1999) hypothesized sequences of
negative life events combine with individual differences in
resilience to causally influence KSA development and
future performance.
Ecology Model
Perhaps the most comprehensive and well-developed
explanation of biodata criterion-related validity was
provided by Mumford, Stokes, and Owens’ (1990) ecology
model, derived from Owens’ developmental-integrative
model (Owens, 1968, 1971, 1976; Owens & Schoenfeldt,
1979). The ecology model describes causally recursive
sequences of life events as a learning process. Specifically,
Mumford et al. (1990) described a sequence of KSA as well
as values and expectancies as shaping individuals’ actions
consistent with their environment at time 1 (t1). Individual
actions at t1 yield a new environment at time 2 (t2)
including performance outcomes resulting from t1 activ-
ities and possibly a new profile of KSAs and modified
preferences or desires for outcomes, resulting from
application of KSAs. This rationale also acknowledged
that individuals have hereditary and environmental ‘‘bag-
gage’’ that shapes their choices. Each choice situation
encountered requires adaptation and is viewed as a
potential developmental experience. Hence, the model
describes an iterative process of choice, development, and
adaptation. Over time, individuals are hypothesized to
develop characteristic patterns of choices and behaviors.
In an attempt to pinpoint possible dimensions under-
lying the model, Nickels (1990) developed a framework of
characteristics and individual differences posited to influ-
ence performance by reviewing the individual differences
literature and research on known predictive relationships
between past behavior/experience and later performance.
Three general categories of individual differences emerged
and were labeled: ‘‘Personality Resources,’’ ‘‘Social Re-
sources,’’ and ‘‘Intellectual Resources.’’ Biodata items
labeled ‘‘Personality Resources’’ glean information from
past experiences regarding the respondents’ emotional
attributes. Items labeled ‘‘Social Resources’’ capture past
experiences that suggest an ability to engage in self-
monitoring behavior and be socially sensitive. Biodata
items labeled ‘‘Intellectual Resources’’ tap past experiences
demonstrating the ability to assimilate and retain knowl-
edge. Two additional categories related to motivational
processes (e.g., goals, needs) and beliefs/attitudes thought
to influence subsequent decisions and actions (e.g., locus of
control, self-esteem, self-efficacy) were labeled ‘‘Choice
Processes’’ and ‘‘Filter Processes,’’ respectively.
The ecology model suggested that these five latent
construct domains ‘‘facilitate the attainment of desired
outcomes while conditioning future situational choice by
increasing the likelihood of reward in certain kinds of
situations’’ (Mumford & Stokes, 1992, p. 81). The ecology
model is graphically portrayed in Figure 1.
Prior to the first formal descriptions of the ecology
model, Mumford, Wesley, and Shaffer (1987) hypothesized
that, over time, individuals ‘‘manifest an internally
consistent pattern of environmental transactions resulting
in systematic activity selections and formation of a stable
developmental trajectory’’ (p. 294). Formation of a pre-
dictable pattern of environmental interaction was said to
lead to a ‘‘crystallized,’’ or stable, adaptive style. Mumford
et al. (1987) tested their hypothesis by administering four
biodata inventories to over 750 undergraduate college
students between the ages of 18 and 30. Results indicated
stable developmental trajectories could be identified by late
adolescence and that certain later experiences inconsistent
with the individual’s adaptive style could derail or alter an
individual’s previously stable adaptive style. Factors or
dimensions used to operationalize adaptive styles included
the five ecology model domains. Although Mumford et al.
(1987) did not examine whether these life experiences
exhibited criterion validity, they did find evidence suggest-
ing that latent constructs in the ecology model evolve
systematically over time.
Mumford, Costanza, Connelly, and Johnson (1996)
described and tested a method of developing construct-
valid biodata items originally suggested by Russell (1994).
Like Mumford et al. (1987), Mumford et al.’s (1996)
purpose was not to directly test the ecology model.
However, Mumford et al. (1996) did develop biodata
items tapping aspects of Filter Processes, Social Resources,
Personality Resources, and Choice Processes and found
evidence for the construct and criterion validity of the scale
scores.
The ecology model posits that the five latent construct
domains influence one another as well as subsequent
outcomes of interest. Further, the theory suggests that
events and behaviors related to the latent ecology model
construct domains are dependent on environmental ‘‘op-
portunities’’ conditioned by life tasks and developmental
stages (Mumford, Snell, & Reiter-Palmon, 1994).
Although multiple theories of developmental life stages
exist (e.g., Erickson, 1963; Kegan, 1982; Magnusson,
1988), none have been explicitly linked (theoretically or
empirically) to ecology model construct domains or their
140 MICHELLE A. DEAN AND CRAIG J. RUSSELL
International Journal of Selection and Assessment r Blackwell Publishing Ltd 2005

criterion validity. Theories that may be useful for explain-
ing include constructive/developmental theory (Kuhnert &
Lewis, 1987; Kuhnert & Russell, 1990; Russell & Kuhnert,
1992) and developmental time windows (Rovee-Collier,
1995).
Kuhnert and colleagues proposed a constructive/devel-
opmental theory of leadership development in which
individuals go through six stages. Kuhnert and Russell
(1990) borrowed Kegan’s (1982) metaphor of an object
being viewed through a lens to describe the key character-
istics of successive developmental stages. The objects being
viewed through the lens were hypothesized to change as
individuals grow and develop through different life stages.
The values, beliefs, and needs constituting an early life
‘‘lens’’ were not available to be viewed initially by the
individual. As individuals grow into later developmental
stages, the values, beliefs, and needs constituting the earlier
‘‘lens’’ would become viewable.
Rovee-Collier (1995) proposed the concept of develop-
mental time windows as critical periods where information
about a current event is integrated with information and
knowledge acquired from previous events. The same
information encountered outside of the time window is
less likely to be integrated. Findings cited by Rovee-Collier
(1995) suggested information gleaned from life events
encountered outside of key developmental time windows
may never approach the level of influence achieved when
the events are encountered within the window, are not
assimilated into the reservoir of collective memory, and are
less likely to impact future behavior. For example, the same
event ‘‘A’’ that might be highly developmental if it occurs in
high school (e.g., ‘‘mostly A’s’’ in response to the question
‘‘what kinds of grades did you receive in science classes?’’)
may not have the same effect if encountered in college.
Meaning ascribed to event A is dependent on brief
alignments and configurations of events, the confluence
of which generates the requisite motivation, attention,
abilities, and environmental circumstances for personal
development and change. Proper alignment of all of these
episode characteristics may tend to occur at only certain life
or career stages. The absence of other early context events
when event A occurs may preclude any developmental
influence (i.e., timing is everything). Rovee-Collier (1995)
speculated that developmental time-windows might be the
cornerstone of individual differences in cognitive domains
used when integrating successive experiences.
Given that development is dependent on successive life
events and stages, it becomes important to consider the role
that life stages may play in explaining biodata predictive
ability. Recent biodata research has taken a construct-
oriented approach for measuring constructs tied to a specific
job (Allworth & Hesketh, 1999, 2000; Karas & West, 1999;
Stokes & Cooper, 2001; Stokes & Searcy, 1999). Ecology
model constructs may be useful for explaining biodata
predictive ability across a wide range of jobs.
Current Study
The purpose of the current study is to determine whether
items from a biodata inventory designed for personnel
selection display psychometric characteristics (e.g., factor
structure, internal consistency reliability) and associations
(e.g., convergent, discriminant, and criterion validity)
consistent with ecology model-based construct domains.
The degree to which evidence shows that biodata item
content is related to ecology model domains and these
Intellectual Resources:
Cognitive ability
Memory recall
Knowledge assimilation
Choice Processes:
Goals and needs
Values and beliefs
Opportunity assessment
Personality Resources:
Conscientiousness
Extraversion
Emotional stability
Social Resources:
Interpersonal skills
Self-monitoring
Empathy
Filter Processes:
Self-esteem
Self-efficacy
Locus of control
Situational Choices
and Behaviors
Environmental
Outcomes
Adapted from Mumford, M. D. & Stokes, G. S. (1992). Developmental determinants of
individual action: Theory and practice in applying background measures. In M. D.
Dunnette and L. Hough (Eds.), Handbook of industrial and organizational psychology
(2nd ed., vol. 3, pp. 61−138). Palo Alto, CA: Consulting Psychologists Press.
Figure 1. The ecology model.
BIODATA THEORY 141
r Blackwell Publishing Ltd 2005 Volume 13 Number 2 June 2005

domains explain variance in criterion performance would
jointly constitute support for the ecology model and the
constructs it proposes as a sound theoretical framework for
future biodata personnel selection efforts.
Method
Sample
Data were collected from a sample of 6032 newly hired
automated systems controllers for a governmental agency
who completed a biodata inventory as part of the selection
process over a period of 6 years. This particular job
requires the ability to multitask several duties including
monitoring radar screens, making decisions to manage air
space, and directing those in the air space. The sample was
90% white and 83% male. The size of the sample was
beneficial for generating stable empirical weights and
cross-validating the inventory.
Measures
Biodata. Biographical information was collected using
a 142-item biodata questionnaire. Each item had five
response options, and respondents were instructed to
choose only one response option per item. Where applic-
able, an escape response option was included to allow
applicants to opt out of the question if they had no prior
experience on the topic the item addressed. Only applicants
who responded to all 142 items (N 5 4767) were included
for keying and cross-validating the inventory. Each
applicant’s overall biodata score was a weighted sum of
his/her chosen response options. The biodata inventory
was scored using response option empirical keying (Kluger,
Reilly, & Russell, 1991), which treats each individual item
response option as a single dichotomous ‘‘item’’ coded ‘‘0’’
or ‘‘1.’’ Each individual response option is weighted by its
point biserial correlation (rpb) with the criterion as
estimated in the 80% (N 5 3787) of the sample randomly
assigned for key development. Each applicant’s biodata
score was the simple sum of point biserial correlations
associated with the response options s/he chose added to a
base score of 100. The response option scoring weights
developed from the key development sample were sub-
sequently applied to the holdout sample (N 5 980) to
estimate cross validities.
The rpb is a special case of the Pearson product–moment
correlation (r) applicable when correlating a truly dichot-
omous variable (e.g., response options either chosen or not)
with a continuous variable (e.g., a performance measure)
and is the most efficient means (i.e., best linear unbiased
estimator) of capturing the strength of the relationship
between an item response option and the criterion of
interest. In the only published direct comparison of
alternative biodata keying procedures (Devlin, Abrahams,
& Edwards, 1992), response option-based scoring yielded
the highest criterion validities. A similar scoring method
that generates weights approximating point biserial corre-
lations, but requiring substantially less computational
resources, is known as the ‘‘vertical percent difference’’
method. This method is described in detail by Kluger et al.
(1991) and Russell, Mattson, Devlin, and Atwater (1990).
The biodata inventory was administered to applicants as
part of research undertaken to develop new procedures for
possible future use in competitive selection examinations.
Applicants completed the inventory voluntarily and were
aware that their biodata responses were not being used for
selection purposes. This, in addition to the use of response
option empirical keying, lessened the possibility of
response inflation. Regardless, Kluger et al. (1991) found
response inflation does not influence criterion validities
exhibited by response option-scored biodata.
The biodata items were rationally developed using
approaches outlined by Mumford and Owens (1987).
These included a review of qualification standards for the
job, job analyses, previous biodata efforts at this govern-
mental agency, interviews with training personnel to
determine characteristics of successful candidates, and
interviews with supervisors to ascertain characteristics
differentiating good and poor employees.
General Mental Ability Measure. Data were available
on a 110-item Office of Personnel Management (OPM) test
that was also administered to these applicants as part of the
selection process. This test measured traditional cognitive
aptitudes such as arithmetic reasoning, data interpretation,
table reading, and spatial relations.
Criterion. The criterion used in this research was a
composite performance measure gathered from applicants
during a 9-week training/screening program. Successful
program completion was a precursor to placement in this
position. The program focused on basic job rules and
procedures and then tested candidate knowledge through
written exams and laboratory simulations. The criterion
was a weighted combination of three performance assess-
ments: (1) paper and pencil multiple-choice tests adminis-
tered throughout the program that assessed candidate
ability to acquire and retain basic job knowledge, (2) job
simulation ratings on six 30 min laboratory simulations
scored using instructor technical assessments of number of
errors and instructor assessments of trainee performance,
and (3) a final paper and pencil examination that assessed
trainees’ ability to apply job rules and procedures. These
categories were weighted 20%, 60%, and 20%, respec-
tively, and summed to form an overall composite score.
Analyses and Results
Content Validity
The biodata inventory’s content validity was examined
relative to the hypothesized ecology model construct
domains using a Q-sort procedure. Five individuals with

knowledge of biodata applications (one Ph.D. and four
advanced doctoral students in industrial/organizational
psychology and human resource management) read the
latent construct definitions, examined biodata item con-
tent, and Q-sorted the inventory’s 142 items onto the five
construct domains. The Q-sort instructions are presented
in the Appendix A. Agreement on the construct domain
categorization of an item among all Q-sort judges was
achieved on 116 items (82% agreement). Eleven of the 142
items were not sorted onto any of the five ecology model
construct domains. Group discussion among the Q-sort
judges was used to resolve initial disagreement on the
categorization of 26 items. Discussion continued until all
judges agreed on the appropriate categorization of each
these disputed items. One hundred thirty-one items were
eventually sorted into one of the five ecology construct
domains. The high level of agreement of the Q-sort
procedure provided initial evidence of the content validity
of this biodata instrument in terms of its ability to capture
ecology model domains.
Of the 131 items sorted onto the construct domains, 19
had categorical response options and were thus excluded
from subsequent factor analysis given that scales scores on
categorical responses are uninterpretable. Hence, 112
items were deemed appropriate for factor analysis. In
terms of how these items were sorted into ecology model
categories, 21 were Q-sorted onto the Social Resources
domain, 10 onto the Personality Resources domain, 37
onto the Intellectual Resources domain, 42 onto the Choice
Processes domain, and two onto the Filter Processes
domain, respectively.
Criterion-Related Validity
Means, standard deviations, and correlations for the study
variables generated from cross-validation sample are
presented in Table 1. The overall cross-validity obtained
for the biodata instrument was r 5.36. Response options
from items Q-sorted onto Intellectual Resources and
Choice Processes ecology model construct domains yielded
the highest subscale cross-validities (r 5.31 and r 5.23,
respectively).
Construct Validity
The 112 items that were Q-sorted and appropriate for
factor analysis were subjected to common factor analysis
using oblique rotation to create item parcels within each
Q-sort item grouping. Notably, factor analysis results
suggested the items Q-sorted onto ecology model construct
domains tended to be reflected by a number of homo-
geneous factors that were interpretable and reflected
different ‘‘eras’’ or ‘‘life stages.’’ This suggested that
perhaps meaningfully different chronological developmen-
tal opportunities existed within ecology model construct
domains. For example, factor analysis results suggested
responses to items tapping life events related to intellectual
resources were clearly delineated into high school, college,
and early career developmental time periods. A total of 11-
item parcels were identified via common factor analysis
across all ecology model domains and were used as
indicators in subsequent confirmatory factor analyses
(CFA). Six parcels emerged from ecology model domains
tapping early career life events, two parcels emerged from
ecology model domains tapping college era events, and
three parcels emerged from ecology model domains
tapping high school era events. Factor analysis results are
available from the first author upon request.
Correlational analyses provided some evidence for the
ecology model construct validity. General cognitive ability
scores correlated more highly with the intellectual
resources items (r 5.13) than with any other ecology
Table 1. Means, standard deviations, and correlations for the cross-validation samplea
Variable Mean SD 1 2 3 4 5 6 7 8 9 10 11
1. Criterion 70.86 11.48
2. Entire Biodata Inventory 100.51 1.06 .36/.43b
3. Social Resources Items 100.01 .21 .18 .07
4. Personality resources items 100.01 .09 .16 .44 .11
5. Intellectual Resources Items 100.07 .87 .31 .92 .11 .27
6. Choice Processes Items 100.31 .27 .23 .54 .10 .44 .27
7. Filter Processes Items 99.99 .03 .01 .38 .16 .21 .30 .27
8. Cognitive Ability Test 71.10 11.50 .16/.41 .14 .08 .04 .13 .01 .05
9. Mid-term Examination 93.14 6.98 .54 .22 .05 .08 .21 .16 .03 .11
10. Skills Test 90.32 7.30 .83 .31 .22 .09 .29 .18 .00 .21 .44
11. Behavioral Ratings 62.66 14.26 .98 .29 .16 .14 .28 .22 .02 .14 .47 .69
12. Final Examination 74.89 14.41 .52 .24 .04 .09 .23 .16 .02 .08 .42 .44 .43
a
All correlations above .075 are significant at po.05, N 5 741.
b
Correlations reported after forward slash are corrected for range restriction on the cognitive ability test.
BIODATA THEORY 143

model construct domain item group. Although the intel-
lectual resources items are not substitutes for standardized
cognitive ability tests, these correlations provide some
evidence of construct validity for this subgroup of items.
Table 2 reports correlations among unit-weighted parcel
scale scores. Correlations within and across item parcels
suggested that biodata items sorted as being consistent with
a specific construct but anchored in different life stages
captured something meaningfully about the particular
construct over time. Correlations among parcel scale scores
tapping a particular life stage (e.g., high school, college,
and early career) were relatively high (average r 5.41),
constituting evidence of biodata item convergent validity
around life stages. The average correlation between parcel
scale scores tapping a particular ecology model construct
domain was r 5.17. An average across-life stage parcel
correlation of .15 suggested item responses exhibited
discriminant validity across life stages. The average
correlation among ecology model construct domains was
r 5.23. These results provide initial discriminant and
convergent validity evidence supporting a model integrat-
ing life stage construct domains within the ecology model.
Additionally, items were grouped by life stage and
correlated with the criterion. A priori we suspected that
items tapping earlier life stages (e.g., high school) would
have a lower correlation with the job performance measure
compared with items tapping later stages closer to the
current job (e.g., college or early career). Instead, we found
that the correlations were not significantly different from
each other. Specifically, the high school, college, and job
scales correlated r 5.23, .20, and .19, respectively, with the
criterion. Correlations among the life stage item groupings
are presented in Table 3.
Finally, CFA using LISREL were performed in order to
test the ecology model’s ability to explain relationships
within these data. Typical CFA practice uses averages of
item groupings, or parcels, as indicators of latent construct
domains (Schau, Stevens, Dauphine, Del Vecchio,
1995). Hall, Snell, and Foust (1999) suggested item parcels
are more reliable and likely to be normally distributed
relative to individual items and hence preferred as CFA
indicators. Therefore, items were grouped into the 11
parcels identified in exploratory factor analyses for input as
indicators in CFA.
CFA results suggested a model in which item parcels
were loaded onto their respective ecology model domains
did not fit the data well (Model 1: w2
[34, N 5 6036] 5
8145.78, po.001; Goodness of Fit Index (GFI) 5 .81,
Comparative Fit Index (CFI) 5 .19, Normed Fit Index
(NFI) 5 .56, Parsimonious Goodness of Fit Index
(PGFI) 5 .56, and Root Mean Square Error of Approxima-
tion (RMSEA) 5 .35). The resulting fit indices were not
acceptable using commonly used guidelines (e.g., Medsker,
Williams, Holahan, 1994). Therefore, the ecology
model as conceived via the model’s construct domains did
not provide a good fit to these data. Given the literature on
life stages and their important role in individual develop-
ment, and because of the fact that exploratory analyses
suggested different stages of development existed within
the ecology model construct domain items, this led us to
examine the viability of a post hoc, alternative ‘‘life stages’’
ecology model.
Based on these results and the theoretical rationale
provided by Kuhnert and Russell (1990) and Russell and
Kuhnert (1992), the latent measurement model was
reconfigured around developmental periods associated
Table 2. Item parcel correlationsa
Parcel content 1 2 3 4 5 6 7 8 9 10 11 12
Early Career/Job
1. Criterion –
2. Social Resources/job .05 .80b
3. Personality Resources/job .03 .49 .55
4. Intellectual Resources/job .18 .59 .55 .86
5. Choice Processes/job .13 .44 .33 .43 .78
6. Filter Processes1/job .03 .26 .20 .30 .14 –
7. Filter Processes2/job .03 .32 .20 .22 .17 .24 –
College
8. Intellectual Resources/college .19 .08 .15 .14 .10 .16 .14 .94
9. Choice Processes/college .03 .07 .09 .09 .07 .14 .16 .82 .83
High School
10. Social Resources .05 .30 .19 .16 .03 .12 .16 .38 .39 .67
11. Intellectual Resources .21 .06 .16 .28 .06 .22 .15 .20 .08 .01 .86
12. Personality Resources .08 .50 .15 .30 .28 .17 .26 .06 .04 .19 .08 .56
a
All correlations above .075 are significant at po.05, N 5 741.
b
Parcel coefficient as are presented on the diagonal.

with high school, college, and early career experiences.
This proposed modification is also consistent suggestions
by with Mumford et al. (1987) and Mumford and Stokes
(1992) that the ecology model framework is dependent on
environmental opportunities that are conditioned by
developmental life stages, and Rovee-Collier’s (1995) work
on developmental time windows, which suggested that
there are critical periods that influence individuals’ abilities
to integrate information from successive life experiences.
The proposed life stages ecology model was tested by
arranging item parcels as indicators of the life stage
captured. This model fit the data better than the original
ecology model formulation (Model 2: w2
[41, N 5 6036] 5
2148.22, po.001; GFI 5 .94, CFI 5 .89, NFI 5 .88,
PNFI 5 .58, and RMSEA 5 .09). Modification indices
generated in the test of Model 2 were examined in an
attempt to achieve better fit. One of the three Personality
Resources parcels initially loaded onto the ‘‘high school’’
factor, although some but not all items in this parcel
addressed a singular point in time. This parcel had a very
low CFA loading (.05) on the high school factor and a low
l–w modification index (consistent with the fact that no
single developmental time period was tapped). Addition-
ally, the two Filter Processes parcels had high error terms
relative to other indicators (which was likely because of the
fact that each parcel contained only one item). Therefore,
the three above-mentioned parcels were subsequently
removed from the model. Model 3, as illustrated in Figure
2, generated slightly better model fit indices: (Model 3: w2
[17, N 5 6036] 5 1347.62, po.001; GFI 5 .95, CFI 5 .92,
NFI 5 .92, PGFI 5 .45, and RMSEA 5 .11). Examination
of fit indices across the models suggested the life stages
ecology model yielded a better fit to the data compared
with the ecology model as originally proposed. Further
inspection of the life stages model showed that the high
school and college factors were highly correlated (.73).
Given they both tapped educational experiences, these two
factors were combined to form an ‘‘educational experi-
ences’’ factor to see if a more parsimonious model would
provide a better fit to the data. However, this modifica-
tion did not improve model fit: (Model 4: w2
[19,
N 5 6036] 5 2104.32, po.001; GFI 5 .92, CFI 5 .88,
NFI 5 .87, PNFI 5 .49, and RMSEA 5 .13.
Discussion
Four sets of analyses provided evidence of the ecology
model’s content, criterion-related, and construct validity.
First, the Q-sort procedure in which items were sorted into
ecology model construct domains provided a test of pro-
totypic conceptions and content relevance. Q-sort judges
agreed on sorting the majority of items onto the five
ecology model construct domains. This high level of inter-
judge agreement supports inferences of content validity.
Second, biodata scores and the construct domain scores
were regressed onto the criterion, thereby providing a
demonstration of the biodata inventory’s and the scale
scores’ criterion-related validity. Meaningful criterion
validities were found for items falling in all but one of the
five ecology model construct domains. Criterion validities
associated with construct domains are consistent with the
ecology model prediction that life experiences falling in
these domains should causally influence subsequent KSA
acquisition and job performance.
Third, results reported in Tables 2 and 3 support biodata
scale convergent and discriminant validity of life stage-
oriented ecology model construct domains. Specifically,
factor analysis results suggested that biodata items load
more highly within one life stage-specific ecology model
construct domain than within the same construct at a
different life stage or with other construct domains.
Estimates of internal consistency reliability reached gen-
erally accepted standards for items loading within indivi-
dual stages when at least six items loaded dominantly on
them. Correlations among parcel scale scores tapping
a particular ecology model construct domain were rela-
tively low. However, an ecology model based around life
stages suggested parcels within a life stage should exhibit
Table 3. Correlations between high school, college, and
early career domainsa
Variable 1 2 3
1. Criterion –
2. High school domain .23
3. College domain .20 .18
4. Early career domain .19 .08 .17
a
All correlations above are significant at po.05, N 5 741.
Interpersonal
Resources - Job
Interpersonal Resources
- High School
Personality Resources
- Job
Intellectual Resources
- College
High School
- Job
- High school
Choice Processes -
Job
Choice Processes -
College
College
Job
Figure 2. Relationship of ecology model-based item par-
cel indicators with life stages latent constructs (Model 3).
BIODATA THEORY 145

convergent validity. Correlations among parcel scale scores
tapping common life stages exhibited strong evidence of
convergent validity. Correlations between parcels from
different life stages also exhibited evidence of discriminant
validity. These results suggest that construct domains
hypothesized by the ecology model do explain applicant
responses to biodata items when the model is modified to
incorporate developmental life stages.
Finally, CFA provided construct validity evidence for a
life stages configuration of the ecology model constructs.
Given results of the exploratory factor analysis and
convergent and discriminant validity evidence, it was not
surprising that initial CFA results failed to support the
a priori five-factor ecology model construct domain. Sub-
sequent interpretations of item parcel content and CFA
modification indices suggested an interpretable life stages
ecology model fit the data well. Specifically, results
suggested meaningful differences in ecology model construct
domains occurred across life experiences. We chose not to
portray the ecology and life stages models as competing, as
we feel that the authors of the ecology model would not
consider the notion of life stages as being inconsistent with
the ecology model in general. Nonetheless, the ecology
model does not specify when relevant life stages might occur
or the developmental content of these stages.
These results suggest future ecology model research, and
perhaps biodata research in general that incorporates a life
stage perspective may yield a better understanding of
biodata predictive ability. This suggestion is also consistent
with constructive/developmental theory (Kuhnert Rus-
sell, 1990) and Rovee-Collier’s (1995) research on develop-
mental time windows. High school, college, and early career
experiences captured by the biodata items in this study may
reflect key events that shape and change how individuals
view themselves and their environment over time.
A theoretical contribution of this research is the
suggestion that items capturing the same theoretical
construct may be impacted by the life stage, or era, in
which the item is anchored. For example, items tapping the
Intellectual Resources domain in the college era seemed to
capture a meaningfully different aspect of Intellectual
Resources than those anchored in early career (r 5.14
between scales). We believe these findings may be general-
izable to any construct domain that biodata intends to tap.
Whatever latent construct domain biodata items may
intend to reflect, these findings suggest future investigators
should generate items tapping multiple time periods across
one’s life history to ensure items from the key (i.e.,
predictive) ‘‘eras’’ are captured. More research is needed
to examine the influence of anchoring items in a particular
time context. It may be important to examine if certain life
stages are more predictive than others or perhaps that
certain constructs may be more predictive when anchored
in a particular life stage. Additional research is also needed
to see if the relationships here hold for other samples and
contexts.
The finding that items capturing the same theoretical
construct may be impacted by the life stage in which the
item is anchored speaks to the importance of item context
when developing items to measure a particular construct.
This is a similar notion to research on personality tests
which found that adding item context in terms of work-
related activities such as ‘‘I am always on time for work’’
rather than ‘‘I am always on time’’ tended to increase
criterion validity (Schmit, Ryan, Stierwalt, Powell,
1995). The current study’s findings suggest that similar
consideration be given to biodata items as has been
examined in the personality literature. Framing an item
around a specific, hypothesized developmental time era
may give insight into which developmental periods are
most predictive and may help applicants provide more
accurate responses by giving them a specific context to
which to relate their response.
Limitations
Limitations of the current study include the fact that
biodata items studied were from a typical biodata
instrument not specifically based on the ecology model
constructs and the non-orthogonal nature of the ecology
model construct domains in general. Despite these limita-
tions, results suggest the ecology model was useful in
explaining this particular biodata inventory’s explanatory
and predictive ability. These findings suggest that future
research is warranted examining items specifically devel-
oped to tap ecology model construct domains. Applicant
responses to biodata items specifically developed to
capture ecology model construct domains across different
developmental stages should yield even stronger findings.
The ecology model domains by their nature are not
perfectly orthogonal (e.g., social resources and personality
resources domains). One might have expected overlap
between latent ecology model construct categories to have
caused Q-sorters some frustration and possible low levels
of consensus. Fortunately, current Q-sorters did not report
this kind of frustration in their debriefing. Additionally, it
should be noted that the Personality Resource items are not
appropriately viewed as ‘‘personality’’ measures per se.
Rather, biodata items address life experiences deemed to be
reflective of personality characteristics. Biodata items
should not be considered measures of latent personality
constructs or a substitute for items specifically designed to
measure personality or its dimensions (e.g., the Big Five).
Similarly, intellectual ability-loaded biodata items should
not be viewed as a substitute for items specifically designed
to assess cognitive ability.
Future Research
Future research is warranted on biodata item development
efforts designed to tap construct domains proposed around
developmental eras, or life stages. Researchers should use

any one of a number of construct domain-targeted biodata
item generation procedures (Russell, 1994) to generate
items from construct domains across developmental life
stages. This study also suggests future research is needed to
examine stage characteristics, such as what events signal a
developmental stage is beginning or ending. For example,
does a life event have the same developmental influence or
criterion validity regardless of the stage or time window in
which it occurs (Kuhnert Russell’s, 1990, model suggests
it would not)? Item generation procedures described by
Mumford et al. (1996) and Russell (1994) should be
adapted to target life stage-specific events hypothesized to
influence constructs underlying the five ecology model
domains. Additional research should also shed light on
whether support exists for a model containing these five
latent construct domains during all developmental periods
or whether different domains are more pertinent during
periods of rapid change and development.
In sum, this research provided support for the ecology
model as a viable theoretical rationale for biodata
predictive ability in an applied personnel selection context.
The ecology model provides adequate explanatory power
for biodata’s predictive ability; however, this explanatory
power might be further improved by incorporating the
concept of life stages within the originally explicated
construct domains. We expect further advances in biodata
theory development will occur when future investigators
employ developmental life stages when generating biodata
items. Future research should examine whether attending
to life stages in biodata theory development and subse-
quent biodata practice will yield stronger theoretical and
empirical support for the use of biodata in personnel
selection.
Appendix A
Biodata Item Q-sort Exercise
Part I: Instructions
Thank you for participating in this research study. This
exercise should take approximately 45 min to complete and
involves sorting 142 items into the construct domain
category in which you believe a particular item most
represents. The categories you will be asked to use are
labeled:
1. Social Resources
2. Personality Resources
3. Intellectual Resources
4. Choice Processes
5. Filter Processes
6. Other
Please put the number of the category that best fits the item
in the blank next to the respective item. If an item does not
seem to fit clearly into any category, note this by choosing
option ‘‘6’’ for ‘‘Other.’’
Definitions for the categories are provided on the following
three pages. Please read through them and be certain of
your understanding of these categories prior to participat-
ing in this exercise.
Please read through each item and place the number (1–6)
of the category you think best represents that particular
item in the blank by that item. For example:
______During my last year in high school, the average
number of hours of I spent in extracurricular school
activities per week was:
A. more than 20
B. 16–20 hours
C. 10–15 hours
D. fewer than 10 hours
E. none
For example, placing a ‘‘1’’ in the space available would
suggest that this item most closely fits into the Social
Resources category.
Part II: Category Definitions
1 5 Social Resources
Place an item in this category if you believe the item has the
ability to tap a person’s interpersonal or social skills.
Interpersonal skill is related to a person’s ability to
communicate with others (oral, written, etc.). Involvement
in group activity may also signal ones’ adeptness for
interacting with others. Greater number of group member-
ships may suggest individual effectiveness in self-monitor-
ing, social sensitivity, and ability to adjust behavior to
match that of the group.
2 5 Personality Resources
Place an item in this category if you believe the item taps
aspects of a person’s personality. This category represents
consistent patterns of behavior and characteristics such as
adaptability, persistence, extraversion (e.g., sociable, asser-
tive, ambitious), emotional stability (e.g., secure, anxious,
well-adjusted), conscientiousness (e.g., dependable, effi-
cient, achievement oriented), and openness to experience
(e.g., cultured, curious, broad-minded).
3 5 Intellectual Resources
Place an item in this category if you believe the item taps
a person’s intellectual skills/resources. This category
represents attributes that enable one to assimilate and
BIODATA THEORY 147

retain knowledge affecting one’s ability to make choices
and perform efficiently and effectively. Items capturing past
experiences requiring or aiding in the development of
general cognitive ability.
4 5 Choice Processes
Place an item in this category if you believe it has the ability
to tap an individual’s motivation. This category represents
the values and choices that precede individual behavior and
performance. Items dealing with motivational issues
include goal orientation, personal performance standards,
and desirability of the reward (e.g., ‘‘valence’’ in expec-
tancy theory terms).
5 5 Filter Processes
Place an item in this category if you believe the item has the
ability to tap a person’s self-perception. This category
represents the beliefs and attitudes, which may influence
self-perception and consequently decisions an individual
makes. This category includes characteristics such as self-
esteem, self-efficacy, and locus of control.
6 5 Other
If the item does not fit into any of the above categories,
place the item in the ‘‘other’’ category.
References
Allworth, E. and Hesketh, B. (1999) Construct-oriented biodata:
Capturing change-related and contextually relevant future
performance. International Journal of Selection and Assessment,
7, 97–111.
Allworth, E. and Hesketh, B. (2000) Job requirements biodata as a
predictor of performance in customer service roles. International
Journal of Selection and Assessment, 8, 137–147.
Dean, M.A., Russell, C.J. and Muchinsky, P.M. (1999) Life
experiences and performance prediction: Toward a theory of
biodata. In G. Ferris (Ed.), Research in personnel/human
resource management. Greenwich, CT: JAI Press.
Devlin, S.E., Abrahams, N.M. and Edwards, J.E. (1992) Empirical
keying of biographical data: Cross-validity as a function of
scaling procedure and sample size. Military Psychology, 4,
119–136.
Erickson, E.H. (1963) Childhood and society. New York: Wiley.
Fleishman, E.A. (1988) Some new frontiers in personnel selection
research. Personnel Psychology, 41, 679–701.
Hall, R.J., Snell, A.F. and Foust, M.S. (1999) Item parceling
strategies in SEM: Investigating the subtle effects of unmodeled
secondary constructs. Organizational Research Methods, 2,
233–256.
Hunter, J.E. and Hunter, R.F. (1984) Validity and utility of
alternative predictors of job performance. Psychological Bulle-
tin, 96, 72–98.
Karas, M. and West, J. (1999) Construct-oriented biodata develop-
ment for selection to a differentiated performance domain.
International Journal of Selection and Assessment, 7, 86–96.
Kegan, R. (1982) The evolving self: Problem and process in human
development. Cambridge, MA: Harvard University Press.
Kluger, A.N., Reilly, R.R. and Russell, C.J. (1991) Faking biodata
tests: Are option-keyed instruments more resistant? Journal of
Applied Psychology, 76, 889–896.
Kuhnert, K.W. and Lewis, P. (1987) Transactional and transforma-
tional leadership: A constructive/developmental analysis. Acad-
emy of Management Review, 12, 648–657.
Kuhnert, K.W. and Russell, C.J. (1990) Theory and practice in the
selection and development of organizational leaders. Journal of
Management, 16, 595–607.
Mael, F.A. (1991) A conceptual rationale for the domain and
attributes of biodata items. Personnel Psychology, 44,
763–792.
Magnusson, D. (1988) Individual development from an interac-
tional perspective: A longitudinal study. Hillsdale, NJ: Erlbaum.
Medsker, G.J., Williams, L.J. and Holahan, P.J. (1994) A review of
current practices for evaluating causal models in organizational
behavior and human resources management research. Journal of
Management, 20, 439–464.
Mumford, M.D., Costanza, D.P., Connelly, M.S. and Johnson, J.F.
(1996) Item generation procedures and background data scales:
Implications for construct and criterion-related validity. Person-
nel Psychology, 49, 361–398.
Mumford, M.D. and Owens, W.A. (1987) Methodological review:
Principles, procedures, and findings in the application of
background data measures. Applied Psychological Measure-
ment, 11, 1–31.
Mumford, M.D., Snell, A.F. and Reiter-Palmon, R. (1994)
Personality and background data: Life history. In G.S. Stokes,
M.D. Mumford and W.A. Owens (Eds.), The biodata handbook:
Theory, research, and applications (pp. 583–625). Palo Alto,
CA: Consulting Psychologists Press.
Mumford, M.D. and Stokes, G.S. (1992) Developmental determi-
nants of individual action: Theory and practice in applying
background measures. In M.D. Dunnette (Ed.), Handbook of
industrial and organizational psychology (2nd Edn., Vol. 3,
pp. 61–138). Palo Alto, CA: Consulting Psychologists Press.
Mumford, M.D., Stokes, G.S. and Owens, W.A. (1990) Patterns of
life history: The ecology of human individuality. Hillsdale, NJ:
Lawrence Erlbaum Associates.
Mumford, M.D., Wesley, S.S. and Shaffer, G.S. (1987) Individuality
in a developmental context: II. The crystallization of trajectories.
Human Development, 30, 291–321.
Nickels, B.J. (1990). The construction of background data
measures: Developing procedures which optimize construct,
content, and criterion-related validities. Unpublished doctoral
dissertation, Georgia Institute of Technology, Atlanta, GA.
Owens, W.A. (1968) Toward one discipline of scientific psychology.
American Psychologist, 23, 782–785.
Owens, W.A. (1971) A quasi-actuarial basis for individual assess-
ment. American Psychologist, 26, 992–999.
Owens, W.A. (1976) Background data. In M.D. Dunnette (Ed.),
Handbook of industrial and organizational psychology (pp.
609–644). Chicago: Rand McNally.
Owens, W.A. and Schoenfeldt, L.F. (1979) Toward a classification of
persons. Journal of Applied Psychology, 64, 569–607.
Reilly, R.R. and Chao, G.T. (1982) Validity and fairness of some
alternative employee selection procedures. Personnel Psychol-
ogy, 35, 1–62.
Reilly, R.R. and Warech, M.A. (1990). The validity and fairness of
alternative predictors of occupational performance. Paper
invited by the National Commission on Testing and Public
Policy, Washington, DC.
Rovee-Collier, C. (1995) Time windows in cognitive development.
Developmental Psychology, 31, 147–169.

Russell, C.J. (1994) Generation procedures for biodata items: A
point of departure. In G.S. Stokes, M.D. Mumford and W.A.
Owens (Eds.), Biodata handbook: Theory, research, and use of
biographical information in selection and performance predic-
tion (pp. 17–38). Palo Alto: Consulting Psychologists Press.
Russell, C.J. and Dean, M.A. (1994). Exploration of a point biserial
bootstrap method of empirical biodata keying. Presented at the
ninth annual meeting of the Society of Industrial and Organiza-
tional Psychology, Orlando, FL.
Russell, C.J. and Kuhnert, K.W. (1992) New frontiers in manage-
ment selection systems: Where measurement technologies and
theory collide. Leadership Quarterly, 3, 109–135.
Russell, C.J., Mattson, J., Devlin, S.E and Atwater, D. (1990) Predictive
validity of biodata items generated from retrospective life
experience essays. Journal of Applied Psychology, 75, 569–580.
Schau, C., Stevens, J., Dauphine, T.L. and Del Vecchio, A. (1995)
The development and validation of the survey of attitudes
towards statistics. Educational and Psychological Measurement,
55, 868–875.
Schmidt, F.L., Ones, D.S. and Hunter, J.E. (1992) Personnel
selection. Annual Review of Psychology, 43, 627–670.
Schmit, M.J., Ryan, A.M., Stierwalt, S.L. and Powell, A.B. (1995)
Frame-of-reference effects on personality scale scores and
criterion-related validity. Journal of Applied Psychology, 80,
607–620.
Schmitt, N., Gooding, R.Z., Noe, R.D. and Kirsch, M. (1984) Meta-
analyses of validity studies published between 1964 and 1982
and the investigation of study characteristics. Personnel Psy-
chology, 37, 407–422.
Stokes, G.S. and Cooper, L.A. (2001) Content/construct approaches
in life history form development for selection. International
Journal of Selection and Assessment, 9, 138–151.
Stokes, G.S. and Searcy, C.A. (1999) Specification of scales in
biodata form development: Rational vs. empirical and global vs.
specific. International Journal of Selection and Assessment, 7,
72–85.
Wernimont, P.F. and Campbell, J.P. (1968) Signs, sample, and
criteria. Journal of Applied Psychology, 52, 372–376.
BIODATA THEORY 149

An Examination Of Biodata Theory-Based Constructs In A Field Context

More Related Content

Similar to An Examination Of Biodata Theory-Based Constructs In A Field Context (20)

More from Nat Rice (20)

Recently uploaded (20)

An Examination Of Biodata Theory-Based Constructs In A Field Context