presentation

advertisement
Stability of Job Analysis Findings
and Test Plans over Time
Calvin C. Hoffman, PhD
Los Angeles County Sheriff’s Department
Presented to PTCSC on April 14th, 2010
Coauthors: Carlos Valle, Gabrielle Orozco-Atienza, &
Chy Tashima.
INTRODUCTION
• Job analysis (JA) provides foundation for many
human resources activities [recruitment,
placement, training, compensation,
classification, and selection (Gatewood &
Feild, 2001)].
• In content validation research, JA is used to
minimize the “inferential leaps” in selecting or
developing selection instruments.
INTRODUCTION
• There is little guidance on how often to revalidate or revisit the validity of selection
systems (Bobko, Roth, & Buster, 2005).
– Uniform Guidelines (1978) - “There are no
absolutes in the area of determining the
currency of a validity study.”
– SIOP Principles – “…organizations should
develop policies requiring periodic review of
validity of selection materials and methods.”
INTRODUCTION
• Our position is that if “revalidation” is
needed, researchers must pay attention
to job analysis.
• For example:
– Changes in duties?
– Changes in technology?
– HR systems changes?
– Changes in required KSAs?
SETTING
• Extensive litigation regarding sergeant
promotional process.
• Consent decree governed all aspects of JA,
selection system design, selection system
operation, and actual promotions for over 25
years.
• Organizational policy requires updating job
analysis every five years.
– Policy does not consider important factors
such as costs and legal context.
SETTING
• 2006 Sergeant Exam - Conducted extensive multimethod JA.
• 2009 Sergeant Exam – Unsure about need for
additional JA given recency of JA data.
• SIOP Principles - “The level of detail required of
an analysis of work is directly related to its
intended use and the availability of information
about the work. A less detailed analysis may be
sufficient when there is already information
descriptive of the work” (p.11).
CURRENT STUDY
• Few changes in sergeant job were
expected during the three-year span.
– Could conclude that new JA is not needed,
and reuse the existing 2006 test plan.
– Given history of litigation surrounding this
exam, the Principles would support
conducting an additional JA.
• Choosing to err on the side of caution, we
performed a slightly abbreviated JA to
support 2009 exam.
CURRENT STUDY
• Study examined the stability of the JA
data over a two-year span. Focuses on
the similarity of:
– task and KSA ratings by two independent
groups of incumbents
– the test plans for the written job knowledge
test.
METHOD - 2006 JA
• Structured JA interviews were conducted onsite with incumbents, along with job
observation, facility tours, and “desk
observation”.
• From these data sources, a work-oriented job
analysis questionnaire (JAQ) was drafted
consisting of major tasks and KSAs.
• JAQ survey (incumbents), SME linkage
ratings.
METHOD - 2009 JA
• Did not conduct additional JA interviews.
• Relied on the existing 2006 JAQ as a
starting point for the 2009 JA effort.
• Otherwise, followed same process.
METHOD
Participants
Invite
d
91
Participated
69*
Response
Rate
Sampling Method
76%
Incumbents
(Sergeants) chosen by
Personnel Captain.
75%
Incumbents
(Stratified random
sample) chosen by
researchers.
2006
Exam
65
2009
Exam
49*
*Both JAQs were administered online via a web survey
METHOD
• SMEs (2006 N = 13, 2009 N = 10) in both
studies performed linkage ratings to establish
relationship between task domains and KSA
domains using a 4-point relevance scale.
– JAQ x linkage ratings data were further
reviewed and fine-tuned by SMEs.
• JAQ data helped determine relative weight and
content of test plans (written test, appraisal of
promotability, and structured interview).
RESULTS - TASKS
Correlation
2006
Test
2009
Test
Mean Task
Importance
Rating
t-test
3.7
Dependant
t (28) = 7.65; p < .001 two-tailed;
d = 1.74
3.4
Independent
t (56) = 3.23; p < .01 two-tailed;
d = .85
r =.83
RESULTS - KSAs
Correlation
2006
Exam
2009
Exam
Mean KSA
Importance
Rating
t-test
3.8
Dependant
t (29) = 3.94; p < .001 two-tailed;
d = 1.54
3.7
Independent
t (56) = 0.90; p > .05 two-tailed;
d = .23
r =.96
RESULTS – TEST PLAN
2006
Test
Total
Items
Knowledge
Domains
From JA
Domains
Omitted
from Test
Plan
102
31
6
Recall v.
Correlation
Reference
84%
104
30
3*
Importance
Rating
3.8
Agreement
2009
Test
M KSA
r =.85
3.7
*Of six knowledge domains omitted in 2006, three knowledge domains
were retained in 2009, for a total of 13 items. All were included as
Reference items in 2009.
DISCUSSION
• JA data were highly stable over time, despite
significant M differences observed.
• Mean task importance ratings (r = .83)
– About 1.0 standard deviation larger than meta-analytic findings
of intrarater reliability (rate-rerate) of JA ratings data reported by
Dierdorff and Wilson (2003), r = .68 (n = 7,392; k = 49) over an
average of 6 months.
• Mean KSA importance ratings (r = .96)
– No comparison could be made because NO estimate of KSA
stability over time could be located in literature.
• Test plans (r =.85)
– Number of items allocated to specific knowledge domains was
highly similar.
DISCUSSION
• Differences between mean task
ratings might be attributable to:
– differences in the selection of JAQ
respondents
– decreased sensitivity in organization
regarding sergeant promotional exam
(i.e., no new lawsuits!).
DISCUSSION
• Findings did not translate into major
differences in the test plans resulting
from the JA efforts even with:
– different SMEs,
– different survey respondent selection
methods,
– significant differences in mean task and
KSA ratings.
DISCUSSION
• Although five new domains were
included in the 2009 test, they were
incorporated as Reference
questions wherein candidates are
provided resource material to
answer questions.
CONCLUSION
• We considered costs and risks in determining
whether to revalidate our selection system.
– The greater the number of intervening years
between validation studies, the higher the risk the
organization assumes.
– The shorter the intervals of time between
revalidation efforts, the costlier and more
burdensome they are for the organization.
• We were conservative due to the legal context. Might
have followed a different path if guidelines regarding
revalidation efforts were clearer.
CONCLUSION
• We encourage researchers and practitioners to
conduct and share any research findings that
might help in creating detailed practice
guidelines on revalidation efforts.
• More information is needed to close the
disconnect between the requirement to
maintain currency of validity information and
the lack of clear guidance regarding how often
is “often enough.”
Questions?
• Copies of slides and the conference
paper are available:
– Email request to: choffma@lasd.org
Download