What‘s the difference?Diversity constructs as separation, variety, or disparity in organizations.
Harrison, D.A., & Klein K.J. (2007)
Academy of Management Review
How to measure diversity in organizations.
How to measure diversity in organizations (Harrison, Klein, 2007)
1. Harrison, D.A., & Klein K.J. (2007)
Academy of Management Review
Elena Tecchiati
What‘s the difference?
Diversity constructs as separation,
variety, or disparity in organizations
2. The challenges
¤ Diversity in organizations
¤ Diversity in researches and investigation
¤ New theories
¤ New studies
¤ 19 „diversity“ investigations in 1988 vs. 134 in 2003
3. Efforts from studies
¤ Cumulative findings about consequences of within-unit
differences have been weak and/or inconsistent
¤ According to different authors the results were „mixed“
¤ Theories and analyses must be refined
¤ Elaborating mediators & moderators
4. Problem
¤ The very construct of diversity requires closer examination
and refinement
¤ Diversity ≠ heterogenity, dissimilarity or dispersion
5. Diversity: a definition
¤ The distribution of differences among the members of a
unit with respect to a common attribute (X), such as
tenure, ethnicity, conscientiousness, task attitude or pay.
¤ Unit-level compositional construct
¤ Describing diversity of a given attibute within a unit, one
describes the unit as a whole
6. What do we talk about when we
talk about diversity?
9. Diversity of attributes
¤ Such as age, values, and personality
¤ limits within-unit behavioral and social integration
¤ fosters conflict and turnover
¤ Diminishes morale, cohesion, and performance
(Williams & O‘Reilly, 1998)
14. Separation: overview
MEANING AND
SYNONYMS
ATTRIBUTE SHAPE
AT MAXIMUM
DIVERSITY
ATTRIBUTE
EXAMPLES
PREDICTED
OUTCOMES
FOUNDATIONAL
THEORIES
Composition of
differences in
(lateral) position
or opinion among
unit members,
primarily of value,
belief, or attitude;
Disagreement or
opposition
Bimodal
distribution, with
half of unit
members at
highest and
lowest endpoints
of the attribute S
continuum
Opinion, beliefs,
values and
attitudes,
especially
regarding team
goals and
processes
Reduced
cohesiveness,
more
interpersonal
conflict, distrust,
decreased task
performance
Similarity
attraction;
social
categorization;
attraction,
selection and
attribution (ASA)
15. Investigations in separation
¤ Reduced separation = greater similiarity = higher levels of
cooperation, trust and social integration.
¤ Increase of separation = low cohesion, high conflict, high
rates of withdrawal, and poor performance
¤ Minimum separation = unit members‘ position is
equivalent (perfect agreement within the unit)
¤ Homogenity is beneficial for the group
17. Variety: overview
MEANING AND
SYNONYMS
ATTRIBUTE SHAPE
AT MAXIMUM
DIVERSITY
ATTRIBUTE
EXAMPLES
PREDICTED
OUTCOMES
FOUNDATIONAL
THEORIES
Composition of
differences in
kind, source, or
category of
relevant
knowledge or
experience
among unit
members; unique
or distinctive
information
Uniform
distribution, with
even spread of
members across
all possible
categories of the
attribute V (no
continuum)
Content
expertise,
functional
background,
nonreduntant
network ties,
industry
experience
Greater
creativity,
innovation, higher
decision quality,
more task
conflict,
increased unit
flexibility
Information
processing; law of
requisite variety;
variation,
selection, and
retention (VSR)
18. Variety summarized
¤ Members differ from another qualitatively
¤ The attribute has no high or low
¤ „categorical variability“ (Miner, Haunschild, & Schwab,
2003: 790)
19. Maximum and mininum variety
¤ Minimum variety: all members belong to the same
category
¤ Maximum variety: the richest possible distribution of
information > perfect homogenity (maximum
heterogenity within a unit)
¤ When two categories, moderate variety is not possible
20. Variety in investigations and research
¤ Units where members draw from different pools of
information, knowledge, background, etc. will make more
effective decisions and deliver more creative products than
units whose members draw from the same pool of resources.
¤ Variety as „sociocognitive horsepower“ (Carpenter, 2002:
280)
¤ Units whose members bridge structural holes in an interunit
network are likely to be more crative and productive (Burt,
2002)
¤ Heterogeneous teams can match complex competitive
challenges and uncertain contexts with a requisite level of
cognitive and experiential variety (Ferrier, 2001)
21. Moderate variety
¤ Can lead to problems of „unshared information“
because members may fail to discuss information not
shared by all or the majority of the group (Gruenfeld,
Mannix, Williams, & Neale, 1996)
23. Disparity: overview
MEANING AND
SYNONYMS
ATTRIBUTE SHAPE
AT MAXIMUM
DIVERSITY
ATTRIBUTE
EXAMPLES
PREDICTED
OUTCOMES
FOUNDATIONAL
THEORIES
Composition of
(vertical)
differences in
proportion of
socially valued
assets or
resources held
among unit
members;
inequality or
relative
concentration
Positively skewed
distribution, with
one member at
highest endpoint
of the attribute D
continuum and
others at lowest
Pay, income,
prestige, status,
decision-making
authority,
social power
More within-unit
competition,
resentful
deviance,
reduced member
input, withdrawal
Distributive
(in)justice and
(in)equity;
status hierarchy;
Tournament;
Social
stratification
24. Disparity summarized
¤ Empirical treatments of disparity in the literature not usual
¤ More common in sociology
¤ Disparity is asymmetric
¤ Disparity is high if 10% of the unit holds a great deal of D
¤ Disparity is low if 90% of the unit holds a great dela of D
25. Disparity in the research
¤ Variable D as resource, such as pay, power, prestige, status.
¤ Disparity-based research assumes that
1. Within units, members can differ in the extent to which they
hold or receive a share, amount, or proportion of D
2. Units differ in the extent to which D is distributed among or
possessed by their members
3. Differences among units in the extent to which their D is
distributed equally among unit members lead to
predictable and important consequences (e.g., fewer
member expressions of voice)
26. Disparity in theory and research
¤ Theories and investigations in disparity are rare
¤ Researches commonly predict that status, power, or pay
disparity incites competition, differentiation, and
deviance among some unit members (e.g., Bloom, 1999)
¤ Disparity may foster conformity, silence, suppression of
creativity and withdrawal (Hollander, 1958)
¤ Marked disparity in team member power diminished
team performance by distracting team members from
key tasks and interrupting the flow of information
27. Social capital
¤ Centralization may be viewed as a measure of how
unequal the individual actor values are (Wassermann &
Faust, 1994):
¤ If the valued resource D is social capital, the structure of
a network might illustrate disparity
¤ Social capital is accessed and conveyed through
interpersonal ties (Adler & Kwon, 2002)
¤ In a highly centralized unit‘s network structure, network
ties are unevenly distributed. Only one or few members
are highly central and, thus, highly influential.
29. Meanings (general)
¤ Separation reflects stand point of position
¤ Variety reflects information
¤ Disparity reflects possession
30. More precision
¤ For future research, using the correct definition of diversity
¤ Theory building regarding diversity is enhanced by
authors‘ explicit specification and justification of the
diversity type of interest
¤ Stronger difference when maximum diversity
31. Why precision?
¤ The same distribution may mean different types of
diversity, may need different explanations and
interpretations
¤ E.g. differences in growing from minimum to maximum
¤ Maximum separation: unit members polarize
¤ Maximum variety: there is one of a kind
¤ Maximum disparity: unbalance within the unit members
32. New theoretical approaches
¤ Maybe max. separation leads to subunits identification
and weak unit identification
¤ Maybe max. variety does not lead to conflict but to
openness toward other ideas, as no one shares the same
idea
33. Guideline
¤ Theory building should careful visualize the shape and
consequences of maximum separation, maximum variety
and maximum disparity
¤ Focus from differences within dyads to the pattern of
differences withing the unit as a whole
35. Most frequent demographic variables
¤ Age
¤ Sex
¤ Race/ethnicity
¤ Organization and team tenure
¤ Education level
¤ Educational content
¤ Functional background
36. Educational content &
Functional background
¤ Qualitative differences in the kinds of information held by
unit members
¤ They could be conceptualized as variety, as separation
or as disparity
37. Tenure
¤ Researchers may conceptualize it a s separation
¤ „Similarity in time of entrance into the group may facilitate both
attraction and interaction“ (O‘Reilly, Caldwell, & Barnett, 1989: 33)
¤ It can be conceptualized as variety
¤ Differences in experiences, information bases, internal and
external network ties
¤ It can be conceptualized as disparity
¤ Individual tenure can be positively associated with status or
authority within a team
¤ Tenure diversity within a team may result in empowerment or
disempowerment
38. Gender diversity
¤ It may be conceptualized as separation
¤ When it reflects a distribution of opposing beliefs about the
appropriateness of critical team processes or outcomes
¤ When it is negatively related to cohesion and identification
within a unit
¤ It may be conceptualized as variety
¤ When it may spark creativity and innovation
¤ It may be conceptualized as disparity
¤ „power“ differences
¤ discrimination
39. Other variables
¤ Such as age, educational level, race, or ethnicity
¤ can be conceptualized as separation, as variety and as
disparity
40. How to define the type of diversity?
¤ It depends on unit members‘ context-dependent
interpretation of the variable in question
¤ Organization
¤ Purpose of the study/investigation
¤ Unit members‘ perception
41. Guideline
¤ A precise specification of diversity type is essential
¤ It allows theorists to defferentiate and compare
conceptual models, facilitating understanding and cross-
fertilization and paving the way for empirical tests of
contrasting conceptions
43. Operationalizing Separation
Index Formula Min to max Assumed Scale
of measurement
Standard
deviation
(SD)
√ [Σ(Si – Smean)2/n] 0 to [(u – 1)/2] Interval
Mean Euclidean
Distance
Σ√ [Σ(Si – Sj)2/n]/n 0 to [(u – 1)/√(2)] Interval
44. Standard deviation
¤ √ [Σ(Si – Smean)2/n]
¤ Describes within-unit diversity as a sample value
¤ Does not estimate a population parameter
¤ Its denominator contais n (not n-1)
45. Minimum and maximum SD
¤ Minimum is „0“
¤ Maximum is (u-1)/2
¤ u is the upper bound of the continuum
¤ E.g.: continuum from 1 to 7, max. SD is (7-1)/2 = 3
¤ Maximum SD does not increase with size of unit
¤ Disadvantage: it cannot be compared across different
separation variables
46. Mean Euclidean distance
¤ Σ√ [Σ(Si – Sj)2/n]/n
¤ Distance of one member i from all the other members j
¤ Minimum is „0“
¤ Maximum measure (u – 1)/√(2) increases with team size
¤ u is the upper bound of the continuum
¤ Also cannot compared across variables with different
metrics
47. Operationalizing Variety
Index Formula Min to max Assumed Scale
of measurement
Blau 1-Σpk
2 0 to (K-1)/K Categorical
Teachman
(entropy)
-Σ[pk x 1n(pk)2] 0 to – 1 x ln(1/K) Categorical
48. Blau‘s index
¤ Qualitative distinctions for the conceptualization of
variety (not distances)
¤ Blau as the most common index for variety
¤ 1-Σpk
2
¤ To understand the formula: p is the proportion
of unit members in K categories
Peter Blau (1918-2002)
49. Maximum and mimimum Blau
¤ 0 to (K-1)/K
¤ Maximum: members of a team are spread equally
(evenness) over the K categories („richness“ of species)
¤ The Blau‘s index reflects the chance that two randomly
selected group members belong to different categories
¤ Blau‘s indexes are not directly comparable when the
number of categories is not the same across diversity
variables (less potential diversity when dyads, e.g., gender)
¤ Maximum possible variety increases with unit size
50. Teachman‘s (entropy) index
¤ -Σ[pk x 1n(pk)2]
¤ Originally developed by Shannon (1948)
¤ p as the proportion of unit members in the K categories
¤ Index rises as unit members are spread more evenly and
across a richer number of V categories
Jay Teachman (1950-)
51. Maximum and minimum of the
Teachman index
¤ 0 to – 1 x ln(1/K)
¤ Not comparable across different V-type variables when
different number of categories (as Blau index)
¤ Blau more used as it occupies a tidier range of 0 to (a
value close to) 1
52. Operationalizing Separation
Index Formula Min to max Assumed
Scale of
measurement
Coefficient of
variation
√ [Σ(Di – Dmean)2/n] Dmean 0 to √(n – 1) Ratio
Gini
coefficient
(Σ Di – Dj )/(2 x n2 x Dmean) 0 to 1 – (1/n) Ratio
53. Coefficient of variation
¤ Formula √ [Σ(Di – Dmean)2/n] Dmean captures the
asymmetry that is fundamental to the conceptualization
of diversity as disparity:
¤ SD divided by the mean
¤ Disparity reflects distance between unit members and
the dominance of those who have higher amounts of
attribute D (resource)
¤ Greater disparity when minority holds great amounts of D
¤ In sociology CV used as a measure of income inequality
54. Maximum and minimum of CV
¤ 0 to √(n – 1)
¤ CV at its maximum when n-1 individuals are at the lower
bound l of a ratio-level continuum
¤ Max CV sensitive to sample size, teams with less members
have less disparity than smaller teams
55. Lower bound l
¤ If l is „zero“ nobody holds any amount of D and the
absolute value of D held by the nth person does not
matter, as the person has all of D in the unit
¤ E.g., number of articles published in a team of 8 persons: if 7
had no articles, CV does not change if the 8th person has 10
or 100 articles published
¤ CV: √(8-1)= 2.65
¤ If l > 0, the upper bound u does matter
¤ The further the distance on D that one or a few elite persons
are above the rest of the persons within a unit, the higher the
CV
56. Gini coefficient
¤ Used in finance and economy
¤ Formula(Σ Di – Dj )/(2 x n2 x Dmean)
¤ as the sum of all pairwise absolute differences between unit
members on variable D divided by (2 x n2 x Dmean)
¤ Only appropriate for attributes that have ratio-level
properties (ratio: relationship between two numbers of
the same kind)
57. Maximum Gini coefficient
¤ 0 to 1 – (1/n)
¤ Depends on n as well
¤ But it should be less of a limiting factor than for CV when
it is used in larger vs. smaller groups
58. On the folly of conceptualizing
S or V While operationalizing D
59. Specifications
¤ Researchers already made mistakes
¤ CV most widely applied diversity index
¤ CV to assess separation or variety is misleading!!!
¤ Researchers sought to assess separation or variety, but
assessed disparity using CV
¤ Interpretations can only be wrong
60. Example
¤ „Commitment to meeting project deadlines“
conceptualized as separation
¤ Teams have equal separation
¤ SD of commitment in Team S1 and S2 = 10
¤ But in Team S1 level of commitment is 40 and Team 2 is 20
¤ If used CV
¤ Team 1 CV= .25
¤ Team 2 CV = .5
61. Guideline
¤ Conceptualization must be aligned with
operationalization
¤ CV is not a universal diversity index
¤ Researchers must specify diversity of research and use
the index already mentioned
63. SD and CV
¤ Difference: CV is ratio between SD and mean
¤ Attention: mean and SD are two independent
(incorrelated) measures
Mean 1 Mean 2
Same SD
Team A Team B
64. Guideline
¤ Mean is necessary in tests of diversity as separation and
disparity
¤ Researchers in the fields of separation and disparity
should first statistically control the within-group mean of
the attribute
66. Previous studies
¤ Some researchers have often conceptualized one form
of diversity and operationalized another one
¤ There is no way of correction after studies
¤ Validities may be called into question
67. „Overall“ Diversity
¤ Researchers seek to assess overall within-unit diversity
averaging diversity indexes arriving to a single index
¤ But: a unit or team itself does not have diversity
¤ An ATTRIBUTE of individuals within units has diversity
¤ It is not possible to sum different indexes
¤ Interpretation is not possible
¤ Thinking of the possibility to add indexes of the same type of
diversity (e.g., gender and ethnicity if diversity)
¤ But the two variables should be positively correlated
¤ In the literature no sign of positive correlation
68. Critical methodological drawbacks
to this approach
¤ Overall diversity treats the causal force of each
component variable as equal masking effects that might
be due mainly to one variable rather than another
¤ Overall diversity measures masks substantive differences
among units that have the same overall composite score
(e.g., team with 2 whites and 6 African-Americans, or six
whites and 2 African-Americans, Blau index is the same)
69. Guideline
¤ Simple operationalization of overall diversity should be
avoided unless
¤ Theoretical motivations for their aggregate effects are clear,
or
¤ Evidence of their convergent validity can be shown.
70. Perceived Diversity
¤ In some studies units were asked about their perception
of diversity
¤ This approach is ok if test theories address perception of
differences
¤ Perceived diversity may have more proximal explanatory
power than actual diversity
71. Construct validity
¤ Measures of perceived diversity are not likely to be
construct-valid measures of „actual“ diversity
¤ Lack of accurate assessment of the unit members
¤ Perceived diversity ratings are likely to be biased, relative
to measures of actual diversity
¤ Ingroup members may overestimate their own unit‘s
diversity
72. Guideline
¤ Measures of perceived diversity should not substitute for
measures of actual diversity
¤ Measures of perceived diversity can provide an
operationalization of a useful, substantive construct -
members‘ perception of unit diversity - that may be
related to but is different from actual unit diversity
73. Instruments for perceived diversity
¤ Researchers may need to develop distinct instruments for
perceived separation, perceived variety and perceived
disparity
¤ If so, their questions, response formats, and anchors
should reflect the diversity type under consideration
74. Sampling the Full Range: Between-
Unit Variance of Within-Unit Diversity
¤ To avoid range restriction, researchers need samples where
sufficient between-unit variability in diversity
¤ to allow effects to be revealed
¤ A significant interaction indicates asymmetric separation effects
(i.e., the impact of separation depends on the level of S)
¤ Researchers need samples that have different amounts of
category richness or unevenness across units
¤ Samples must include
¤ units in which variety is very low and
¤ units in which variety is very high
76. Suggestions
¤ The diversity typology presented suggests new fields of
research:
¤ Unit-level consequences of within-unit inequality in power,
status, and other valued resources
¤ Diversity and social network
¤ Antecedents of separation, variety, and disparity in
organizations
¤ Also research about the relationship among the three
diversity types
77. Relationship among the three types
of diversity
¤ Considering
1. Strategies for disentangling assumptions of
demographic separation, variety, and
disparity
2. Relationships among and interactive
impacts
3. Multilevel diversity effects
78. Disentangling separation
¤ How to proceed when researchers are interested in
demographic diversity?
¤ Answer: Test the theoretical assumptions specific to each
diversity type
¤ Examination of the individual-level relationship of the
variables within units required
79. Prior analysis
¤ Researchers can resolve uncertainty analyzing more
concretely in order
¤ to test the fundamental and distinctive assumptions
underlying diversity as separation, as variety, and as
disparity
¤ Payoff: moving forward deciphering the nature and
effects of demographic diversity
80. Relationships Among and
Interdependent Effects of Diversity
Types
¤ Separation, variety, and disparity may be causally
related
¤ They may have consequences for the unit outcomes
¤ Diversity of one type can engender divesity of a second
type
81. Separation
¤ Separation can engender variety
¤ E.g., if unit members are sharply separated, they might be
motivated to get new information to support their position,
including seeking out others to support them
82. Variety
¤ Variety can engender separation
¤ Variety in disciplinary training may lead to separationin
support for qualitative research than others
¤ Variety can lead to separation and than to conflict (but not
always)
83. Disparity
¤ Disparity can engender variety and/or separation
¤ Disparity may cause separation in beliefs, attitudes, or
values related to unit processes and outcomes that might
dislodge the current status hierarchy
¤ Disparity may cause variety
¤ Unit members that have less status, power, and influence are
likely to form coalitions with others who are similarly less
powerful
84. Diversity Moderating Diversity
¤ The three diversity types may interact to influence unit
outcomes
¤ Ideas about joint effects of within-unit demographic
differences (Lau and Murnighan, 1998): in order to
become faultlines within a unit, the diversity attributes in
question must:
1. Be apparent to unit members
2. Covary or coincide strongly within the unit
3. Create a small number of homogeneous subgroups of
factions
85. Forming a faultline
¤ The correlation of two attributes within a unit insufficient
to form a faultline
¤ E.g., task satisfaction and organizational commitment: their
joint occurrence will not necessarily divide the unit into two
or more clearly differentiated factions
¤ But: as the number of factions grows, faultlines will weaken
¤ Strong faultlines occur within a unit, when
¤ two or more variables have coincidentally maximum
separation or disparity,
¤ or both,
¤ but also when they are coupled with only modest variety
86. Variety and Faultline
¤ When variety is high, faultlines will weaken
¤ As unit members cannot be divided into two or just a few
sharply divided subsets
87. Effects of separation
¤ Low separation may allow a team to realize the benefits
of team members‘ variety of expertise and experience
¤ Team with no separation in member goals, attitudes, or
beliefs but high in variety may fail to use team‘s ideas
¤ Minimal separation of attitudes within the unit may trigger
¤ constructive debate and discussion,
¤ stimulating members to reveal to one another their
distinctive knowledge and expertise
88. Joint Impact of the Three Forms of
Diversity
¤ Example:
¤ Research team: Separation high (attitudes towards
paradigm study), no/moderate D or V(number of
publications)
¤ But if D and V were higher than S, the team would have
conflict, but able to reach across factional boundaries to
leverage its variety
89. Multilevel Influences
¤ Diversity as a multilevel construct
¤ But most studies of diversity are single-level studies (unit-level)
¤ One possible approach:
¤ move up a level of analysis and consider diversity across
units within organizations along the S,V, and D attribute in
question
¤ Effects of within-unit diversity may depend on the
composition (diversity) of the organizational context
91. Conclusion
¤ Considering the three types of diversity in research
¤ Complementary as important factor approaching diversity
measurement
¤ The three types of diversity differ in their substance, shape,
maxima, and implications
¤ Researchers should specify diversity types
¤ Key question: „What‘s the difference?“
¤ Result: cleaner, more cumulative understanding of diversity
in organizations