You are on page 1of 12

The Reliability and Validity of

The OCAI

Human Resources Develo


pment Consortium

Reliability Study 1

by Quinn and Spreitzer (1991) in which 796 executives from 86


different public utility firms rated their own organization's culture.
They included top managers (13 percent of the sample), uppermiddle managers (45 percent), middle managers (39 percent), and
line and staff workers (2 percent).
Cronbach alpha coefficients (a reliability statistic) were computed for
each of the culture types being assessed by the instrument. Each
coefficient was statistically significant and very satisfactory
compared to normal standards of reliability.
Coefficients were .74 for the clan culture, .79 for the adhocracy
culture, .73 for the hierarchy culture, and .71 for the market culture.
In other words, respondents tended to rate their organization's
culture consistently across the various questions on the instrument.

Human Resources Develo


pment Consortium

Reliability Study 2

Yeung, Brockbank, and Ulrich (1991) also provided evidence of reliability in their
study of 10,300 executives in 1064 businesses. These businesses included many of
the corporations on the list of Fortune 500 companies. The key respondents were
human resource executives and various associates that these executives selected to
complete the assessment instrument.
The number of respondents averaged nine per business. The OCAI was used to
gather data on the culture of each of these organizations. The various question
alternatives were grouped together into the appropriate culture types, and reliability
coefficients were computed. The results showed that the clan culture reliability
was .79, the adhocracy culture reliability was .80, the hierarchy culture
reliability was .76, and the market culture reliability was .77. In each case,
reliability coefficients exceeded satisfactory levels. Parenthetically, Yeung et al. found
that the largest percentage of firms was dominated by the hierarchy culture (44
percent), clan and adhocracy cultures were next (15 and 14 percent, respectively),
and, surprisingly, no firms were dominated by the market quadrant. All had moderate
emphasis on the market culture type. Six percent of the firms had all the cultures
equally dominant, and 22 percent had no culture emerge as dominant.

Human Resources Develo


pment Consortium

Reliability Study 3

In still another study, Zammuto and Krakower (1991) used this


instrument to investigate the culture of higher education institutions.
More than 1300 respondents, including administrators (39 percent
of the sample), department chairpersons (34 percent), and trustees
(27 percent) rated the culture of their organizations, resulting in
relia-bility coefficients for each of the culture types as follows: clan
relia- bility = .82, adhocracy reliability = .83, hierarchy reliability = .
67, and market reliability = .78.
Numerous additional studies can be cited (see, for example,
Peterson, et aI., 1991), but in every case that we know of the reliability of these culture types has shown patterns consistent with
those reported above. In other words, sufficient evidence has been
produced regarding the reliability of the OCAI to create confidence
that it matches or exceeds the reliability of the most commonly used
instruments in the social and organizational sciences.

Human Resources Develo


pment Consortium

Validity : OCAI

The multi trait-multi method analysis was produced by using two different
instruments to assess organizational culture. One instrument was the OCAI
being explained here. The other instrument assessed the same cultural
dimensions using a different response scale, namely, a Likert-type scale
where each alternative scenario was rated from 1 to 5.
The scales of the four culture quadrants represented the four traits and the
two different instruments represented the two methods. The goal of the
analysis was to determine if the variance explained between the four traits
(the four cultures) exceeded the variance accounted for by the method used
(the two different instruments).
To produce evidence of validity, correlation coefficients in the same culture
quadrant should be significantly different from zero and of moderate
magnitude (Campbell & Fisk, 1959). Convergent validity was supported, as
it turned out, when the multi trait multi method correlation matrix was
examined. As required, all diagonal correlation coefficients were statistically
different from zero (p < .001), and they ranged between .212 and .515, a
moderate level of correlation.

Human Resources Develo


pment Consortium

Validity : OCAI

Discriminant validity was tested in three ways. In the first test, scales in the
same culture quadrant were tested to see if they correlated higher with each
other than they did with scales of different culture quadrants measured by
separate instruments (Campbell & Fisk, 1959).
Twenty-three out of the twenty-four comparisons were consistent with
expectations, providing solid support for discriminant validity. In the second
test, scales in the same culture were expected to correlate higher with each
other than with scales in a different culture quadrant measured by the same
method (Campbell & Fisk, 1959). In sixteen of the twenty-four scales this
was the case, providing moderate support for discriminant validity. In the
third test, the same pattern of interrelationships was expected to exist within
and between each of the independent methods (Campbell & Fisk, 1959).
Kendall's co- efficient of concordance was computed, which produced a
coefficient of .764 (p < .001), indicating strong support for discriminant
validity. In other words, these three tests using the multi trait-multi method
procedure provided support for both convergent and discriminant validity of
the model and the instrument.

Human Resources Develo


pment Consortium

Validity : OCAI

The multidimensional scaling procedure also produced strong


support for convergent and discriminant validity. shows the
positioning of each culture type when measured with two different
types of instruments.
Guttman and Lingoes' coefficient of alienation
r = .076) and Shepherd and Kruskal's stress coefficient (stress = .
056) indicate a satisfactory fit of the data to the model (see Kruska1
& Wish, 1978). Moreover, it can be seen from the figure that each
culture type appears in the appropriate quadrant, that like cultures
are positioned closer together than to other unlike culture types, and
that each culture type is positioned in a different quadrant. Strong
support is provided, in other words, for convergent and discriminant
validity using this multidimensional scaling technique.

Human Resources Develo


pment Consortium

Human Resources Develo


pment Consortium

Psychometric Analyses of the


Managerial Skills Assessment
Instrument (MSAI)
QuadrantQuadrant
Correlation
Dimension-Dimension
Correlation
Item Dimension Correlation

Human Resources Develo


pment Consortium

Human Resources Develo


pment Consortium

10

Human Resources Develo


pment Consortium

11

Item-Dimension Correlations

An examination of the correlations with each item and the other items in its theorized
dimension (within-dimension correlations) compared to the correlations between each
item and the other three dimensions (outside-dimension correlations) reveals that
every competency dimension has strong reliabilities (well above .50, a very strong
reliability using ipsative measures). Two items on the survey-no. 31 and no. 60appear to be rather weak measures of their theorized dimensions, and eliminating
them from the MSAI would strengthen the psychometric power of the questionnaire.
On the other hand, they are items that assess important aspects of a manager's
behavior, and, although not strongly correlated with other items on the survey, they
are important aspects of a successful manager's behavior.
In sum, these analyses provide strong support for the MSAI as an instrument that can
assist the culture change process. It maps very well the relationships among
quadrants and competency dimensions theorized by the Competing Values
Framework. The critical management skills being assessed by the instrument
possess the same theorized relationships to one another as do the culture quadrants.
It may be used with some confidence, then, in assisting managers to develop
competencies that will foster culture change in desired directions.

Human Resources Develo


pment Consortium

12

You might also like