Gordon: Department of Busincss Administration, Rutgers Universily
Gordon: Department of Busincss Administration, Rutgers Universily
Gordon: Department of Busincss Administration, Rutgers Universily
0022-2380 $3.50
G.
GEORGE GORDON
Department of Busincss Administration, Rutgers Universily
NANCYDITOMASO
Graduate School of Managnncnt, Rutgers University
ABSTRACT
This article investigates the relationships of culture strength and two substan-
tive cultural values with corporate performance. Culture strength is measured
by the consistency of responses to survey items across people and the two
cultural values are measured by items on the survey that relate to either
adaptability or stability. The data, from management surveys of 11 US
insurance companies in 1981, were correlated with asset and premium grow,th
rates from 1982 to 1987. Results indicate that both a strong culture regardless
of content and a substantive value placed on adaptability are associated with
better performance for two to three subsequent years on both criterion
measures. The results support the findings of Denison (1990) that strength
of culture is predictive of short-term performance. The present results,
however, suggest a more complex contingency model than that proposed by
Denison.
INTRODUCTION
ment to the firm (Koberg and Chusmir, 1987; OReilly, 1983; Posner et al.,
1985).
Very few empirical studies have related cultural characteristics to some
measure of corporate financial performance. Gordon ( 1985) contrasted com-
panies in dynamic industries, where technologies, participants and products
changed frequently, with companies in the more static utilities industry,
where few such changes occurred. He found that companies in highly
dynamic industries were characterized by cultural values that enhanced
adaptability, whereas utilities were characterized by cultural values that
enhanced stability. He further found that the same values differentiated the
fastest growing and most profitable companies within each type of industry
from the less successful ones. O n the other hand, Reynolds (1986) found that
employee responses to a culture questionnaire in a company identified as
excellent by Peters and Waterman (1982) did not differ from those in two
other companies with less impressive performance.
Two studies of culture and performance have used data from the Survey
of Organizations (Taylor and Bowers, 1972) to determine relationships
between employee perceptions and attitudes and firm success. Hansen and
Wernerfelt ( 1989) refer to these as organizational variables and Denison
(1984) as cultural, but the differences are more semantic than substantive.
Hansen and Wernerfelt classified the emphasis on human resources and
emphasis on goal accomplishment scales from the Survey of Organizations
as organizational variables and classified industry profitability, relative mar-
ket share and company size as economic variables. Their study tested the
relative importance of organizational and economic factors in predicting five-
year return on assets. They found both the emphasis on human resources
and on goal accomplishment to be significant predictors, with the organiza-
tional data accounting for about twice as much variance as the economic.
Denison (1984) related two characteristics from the Survey of Organiza-
tions, organization of work and decision-making practices to subsequent
returns on sales and investment. He found higher returns for companies
above the average on each measure than for companies below, with differ-
ences tending to widen across the five years following the survey. This study
is also the only one which, in addition to examining the impact of cultural
traits, attempted to determine the impact of cultural strength (conceptualized
as consistency) on organizational performance.
Denison defined consistency as the inverse of the variance in questionnaire
responses across work groups within companies. Defined this way, Denisons
concept of consistency is an amalgam of Louis ( 1985) concepts of sociological
and psychological penetration, since low variance implies both pervasiveness
and homogeneity of perceptions. Denison (1984; 1990) found low variances
on four different traits - organization of work, emphasis on human resources,
decision-making processes and co-ordination - significantly correlated with
companies standardized Return on Investment (ROI) for the subsequent
two years. This latter study gives tentative support to the notion that a firms
culture strength, as defined by the degree of agreement on cultural character-
istics across respondents, relates to subsequent financial performance.
PREDICTING CORPORATE PERFORMANCE 787
HYPOTHESES
METHOD
The version of the instrument used in this study asked managers to describe
their companies on 61 items using seven-point scales presented in a semantic
differential type format. Earlier analyses of these items yielded eight factors
(Gordon and Cummins, 1979) ranging in reliability (coefficient alpha) from
0.80 to 0.87. The eight factors are labelled:
These factors are consistent with many of the dimensions discussed in the
culture literature. Examples include action orientation (Peters and Water-
man, 1982), integration/communication, and innovation/ risk-taking (Kan-
ter, 1983; Reynolds, 1986) and fairness of rewards and development and
promotion from within (Alston, 1986). As noted earlier, however, no set of
cultural traits is widely recognized as critical or accepted as comprehensive.
We do not have measures for several values that have become important in
the culture literature, namely, a focus on quality, service and so on. We,
nevertheless, believe the factors provide a good overall representation of the
content alluded to in much of the literature on corporate culture. For the
purposes of this study, average scores on the highest loading items on each
factor were computed for each company to produce the eight culture content
scales.
Sample
The 11 companies in the sample were all in the life/health sector of the
insurance business and varied in assets from $691 million to $18.7 billion,
with a median of $1.9 billion. Participants in each company consisted of
everyone in the top four to five levels of management. Questionnaires were
distributed to participants (with a cover letter from the company chairperson)
through each companys internal mail and were returned directly to the
investigator in pre-addressed envelopes. All questionnaires were administered
and returned during a two-month period in mid-1981. Returns exceeded 90
per cent of those surveyed and ranged from 34 to 132 participants per
company, with a mean of 77, for a total of 850.
The fact that respondents were limited to managers represents both
advantages and disadvantages. On the one hand, the management, especially
the middle to upper level management represented in this study, is clearly
not a representative sample of the employees in the companies studied.
Previous research has shown that management is considerably more positive
about their companies than people at lower levels (Hay Group, 1986).
Previous work also has shown that different levels in a company may
represent different sub-cultures (Davis, 1985; Martin and Siehl, 1983; Riley,
PREDICTING CORPORATE PERFORMANCE 789
1983). On the other hand, it is the management, especially senior manage-
ment, that must support, if not initiate, any major efforts on the part of their
companies. As Schein (1990, p. 11 1) claims; . . . cultural origins and dyna-
mics can sometimes be observed only in the power centers where elements
of the culture are created and changed by founders, leaders, and powerful
persons. Following the same logic, culture measured at this level will be most
predictive of future behaviour and performance of the firm. Also, because
each survey included only middle and upper management, the samples are
much more comparable across companies than if, as is often done in compara-
tive studies, different employee groups had been sampled.
Culture Measures
For purposes of this study, three separate measures were used. The first,
culture strength, was designed as an attempt to replicate Denisons (1984; 1990)
findings that consistency of company values predicts short-term profit per-
formance. Culture strength was measured by the inverse of each of the eight-
scale standard deviations averaged across all eight scales for each company.
While Denison ( 1990) operationalized the culture strength construct by
computing the variance across groups within companies, we computed stan-
dard deviations across individuals within companies. The standard deviation,
of course, represents a measure of dispersion around the mean, and thus is
also consistent with Saffolds (1988) recommendation that one measure
culture strength by dispersion. Since smaller standard deviations imply
greater agreement among managers on the types of values driving the
company, we assume that such agreement - or consistency - is an indication
of the strength of the organizational culture.
The other measures, adaptability and stability, were based on Gordons
(1985) work indicating that companies tend to develop cultures that match
their environments. The adaptability measure is a combination of two of the
scales, action orientation and innovation/risk-taking, because in Gordons
work the better-performing companies in dynamic or fast-changing industries
(high tech manufacturers) scored high on those measures. Stability is a
combination of three factors, integration/communication, development and
promotion from within and the fairness of reward. In Gordons study, these
scores were high in the better-performing companies in stable industry
environments (utilities). Gordon found none of the other three factors related
to financial performance in either stable or dynamic environments.
Because the use of survey data to study cultural phenomena is still
relatively unexplored, the literature offers little guidance regarding the best
form of measure to use, and arguments can be made for several measures.
We employed two different variations of the adaptability and stability mea-
sures. First, we used an unweighted average of the raw scores on the scales
comprising each measure. These will be referred to subsequently as adapta-
bility (Adap) and stability (Stab). Second, we used standardized deviation
scores, which set all means equal to zero. That is, for each company, the
mean across the eight factor scores was subtracted from each individual factor
score and divided by the standard deviation across the eight factors. This
procedure produces standardized deviation scores and is similar to that
790 GEORGE G . GORDON AND NANCY DITOMASO
Measures of Pezfonnance
Measuring corporate performance in the life insurance industry presents a
unique problem because the majority of companies in the industry (and seven
of the eleven in the sample) are mutual companies, owned by policyholders
and not by stockholders. Because of this unique ownership structure, such
companies do not make profits per se and do not use generally accepted
accounting principles (GAAP) in reporting company results. Thus, the
profitability measures used by Denison (1990), such as the Return on Sales
(ROS) or Return on Investment (ROI), are not available for many life
insurance firms.
One measure commonly employed in the life insurance industry is the gain
or loss in admitted assets, or growth in assets. This measure reflects the
income contribution of many business functions, including sales, investment,
actuarial and underwriting. Growth in assets thus represents the contribution
of a variety of profitable and efficient activities to the companys growth.
Growth in assets, however, includes a large component of very stable invest-
ment earnings, and thus tends to cloud performance changes resulting from
short-term management actions.
A measure more sensitive to management actions is new premium income,
clearly a key revenue component in any life insurance company. However,
during the period of the study, the nature of the revenue stream in the
insurance industry was changing very dramatically. Following the interest
rate fluctuations of the 1970s, many companies began to offer interest-
sensitive products to their clients in addition to insurance, and the significant
changes in policy portfolios resulted in large short-term gains. These gains,
PREDICTING CORPORATE PERFORMANCE 79 1
however, were temporary and began to wane as the easy sales were
exhausted. Such spurts and subsequent settling back during this period
significantly affected the consistency of year to year performance of the
insurance industry and of the companies in our sample.
Assets and total premiums for each company were obtained from Bests
Insurance Reports (1988) for the years 1981-7 and growth rates were
calculated for each of the years from 1982 to 1987. Because of significant
acquisitions and divestitures of business by one of the companies from 1984
to 1987, only their 1982 and 1983 growth data were usable. Thus, the n for
1982 and 1983 is 11 companies while the R for 1984 to 1987 is only 10. Because
the study includes a single industry in a single time period, there is no need
to standardize performance against multi-industry, multi-period norms, as
was necessary for Denison.
RESULTS
Perfonname
Table I presents the correlations in year-to-year performance for the com-
panies in the sample. As anticipated, asset growth is more consistent than
premium growth. One can also see from the negative correlations between
1985-6 and 1986-7 that considerable changes in premium growth took place
at that time. This pattern suggests that major environmental changes affected
these companies beginning in 1985, making it unlikely that any hypothesized
relationships would hold up over the entire period.
Table I1 presents the yearly growth rates for the insurance industry as a
whole and for our sample. Again, the greater stability of asset over premium
growth rates is clear. The industry experienced an unusual drop in premiums
in 1983 and an unusual spurt in 1986. Although the sample as a whole was
not subject to fluctuations on premium growth as wide as the populations,
some of the individual companies did have extreme fluctuations on this
measure. The most dramatic was one company which experienced a swing
from 0.1 per cent in 1985 to 77.5 per cent in 1986, and then back to 5.5 per
cent in 1987. The premium growth of the sample is sometimes above and
sometimes below that of the industry, but asset growth is consistently below
the industry.
Conelalionr
Year 1WP3 1983/4 1984/5 1985/6 19W7
Table 11. Year-to-year growth in the industry and the current sample
Sample
Assets 9.6 10.3 9.5 10.8 10.8 8.8
Premiums 12.2 16.0 16.6 7.7 14.5 13.3
Industry
Assets 11.9 11.3 10.4 14.2 13.5 11.3
Premiums 14.0 -1.6 13.3 15.7 24.5 9.7
Culture Variables
Table I11 presents the intercorrelations among our culture measures. As one
can see, culture strength, as measured by the average standard deviation
across factors, is significantly related to both stability ( T = 0.88) and adapta-
bility ( r = 0.59). Adaptability and stability are themselves positively corre-
lated 0.75, and are thus not highly differentiated when measured by raw
factor scores. But when we partial out the overall average of the factor scores
from the correlation between adaptability and stability, we get a partial
correlation of -0.81, which is what one would expect from their conceptual
definition.
The more predictable -0.79 correlation of DevStab and DevAdap further
supports the claim that the deviation scores are more consistent with the
cultural values emphasized in each company than are raw scores. Adap and
DevAdap are positively correlated (0.92), but Stab and DevStab have a
negative and non-significant relation of -0.20. Even so, for the reasons
already discussed, we report all of the measures in the relationship with
performance, but we focus on the deviation scores.
Table IV. Correlations between measures of culture and average growth of assets and premiums
Assets growth
AvgSD 0.31 0.44+ 0.45+ 0.68.' 0.08 0.26
Stab 0.39 0.38 0.30 0.64. 0.10 0.18
Adap 0.21 0.23 0.12 0.56. 0.19 0.23
DevSta b -0.07 -0.04 0.15 -0.14 -0.25 -0.32
DevAdap 0.22 0.30 0.26 0.55. 0.37 0.33
Premium growth
AvgSD 0.78*+ 0.47' 0.44+ -0.13 -0.16 0.46'
Stab 0.73+* 0.25 0.54+ -0.15 -0.02 0.12
Adapt 0.65** 0.32 0.63. -0.07 -0.57. -0.03
DevS t a b -0.42+ -0.30 -0.37 -0.08 0.69+* 0.14
DevAdap 0.57* 0.44+ 0.54+ 0.10 -0.75'. 0.00
CONCLUSIONS
NOTES
*We would like to thank Emilio Venezian for his help during the preparation of this
manuscript.
[I For each organization let:
D(factor) = S(factorl-M
SD
where S(factor) is the factor score
M is the mean of 8 factor scores
SD is the standard deviation of the 8 factor
scores from their mean
Then for each organization:
DevAdap=D(Action) +D(Innovation)
DevStab =D(Integration)+D(Development) +D(Fairness)
REFERENCES