High frequency attitude measurement in longitudinal designs

skillfulbuyerUrban and Civil

Nov 16, 2013 (3 years and 10 months ago)

80 views

High frequency attitude measurement in longitudinal designs


On capturing alpha, beta, and gamma changes in self
-
report measures.


Omar N. Solinger


Supervisor: Dr. Woody van Olffen

Promotor: Prof. Dr. Robert A. Roe


Maastricht University

Department of Organization & Strategy

The Netherlands


3
rd

METEOR PhD Colloquium

Maastricht, June 13, 2007



Outline


Measuring attitudes over time


Why high frequency longitudinal designs?


The standard
one shot

instrumentarium


Presenting LOCUST: Longitudinal
OcCUpational States Tracker.


Critiques to the new methodology


Research in progress





Why high frequency designs?


How do we include ‘time’ in
conceptualization as well as in research
methodology?


Instead of YES/NO, move towards
HOW?/WHEN?


Instead of levels (“how high?”), study
dynamics (“development”)


Metaphor: snapshot or video recording (= high
frequency snapshots!)

Why the
one shot

instrumentarium
falls short


Impractical


Unnecessarily intrusive & demanding


Massive attrition


Leniency


Insensitive
to micro changes


Likert scales (1
-
5)


Imperfect memory of past scores => unreliability of observed
change.


Static
assumptions


Cronbach’s Alpha is based on single
-
snapshot thinking


Item wordings (“in general”) render a ‘summary cognition’

The alternative…


Losing reliability?

A 1
-
item measure of job satisfaction


All things considered, how satisfied are you with your job
?”

Highly satisfied
-


-

highly dissatisfied.


Or the more sophisticated summation of job
-
factors, s.a.
work, supervision, present pay, promotion opportunities,
relations with coworkers.


Simplicity is just as valid as complexity!!

Wanous et al., (1997), JAP

Anchoring: curse or blessing?



Anchoring
-
bias:
A tendency to fixate on initial
information, from which whe then fail to
adequately adjust for subsequent information



However, LOCUST
makes use of anchoring

to reduce false variation and make small
variations meaningful.

=> anchoring is NOT a problem, but a blessing.

Pilot
-
study experiment (in progress)


Students assigning their perceived development
of language proficiency are randomized over
conditions:


Condition A: Assigning their language proficiency by
tweaking the bar
without
the graph.


Condition B: tweaking the bar
with
a graph of past
scores.


Possibility to compare perceived proficiency
development in the three conditions to exam
scores.

Measurement inequivalence over time


(Golembiewski, Billingsley, & Yeager, 1976; Horn & McArdle, 1992; Vandenberg & Self, 1993; Lance,
Vandenberg, & Self, 2000; Vandenberg & Lance, 2000)




Alpha change
: “real change”. The kind
of change we are interested in.


Beta change
: a change in the
respondent’s metric over time (scale
recalibration)


Gamma change
: a reconceptualization
of the concept in the minds of the
respondents over time (concept
redefinition)

Retrospectively adjusting scores

Pilot
-
study experiment (in progress)


Condition A: Assigning their language
proficiency by tweaking the bar
without
the
graph.


Condition B: tweaking the bar
with
a graph of
past scores.


Condition C: Tweaking the bar with the
possibility to retrospectively adjust past
scores.

Critiques to retrospective
adjustment of scores


Recollection bias
: measuring beta change or
systematic memory distortions? (Armenakis &
Bedeian, 1982)


Regression to the mean
: eliminating extremities
in retrospect.


Embellishment of recollections
:
past is
remembered as more positive than it actually felt
at the time (Armenakis & Bedeian, 1982)


The experience of the past is ‘
colored’ by current
affective states





A solution to gamma change?

A subjective definition of
organizational commitment

Concluding remarks


OB research to date lacks the appropriate
instrumentarium to conduct temporal research


Whatever critique and limitation there may be,
we believe this is a step forward in designing
self
-
report measurement instruments that are
not confounded by other types of change, and
hence allow us to make meaningful inferences
on longitudinal change.










Questions, Comments?





o.solinger@os.unimaas.nl