Which comes first, organizational culture or performance? A longitudinal study of causal priority with automobile dealerships

Abstract

Prior research supports a link between organizational culture and performance but generally falls short of establishing causality or determining the direction of a culture–performance (C-P) relationship. Using data collected from 95 franchise automobile dealerships over 6 years, we studied longitudinal culture–performance relationships to determine whether culture or performance has causal priority, or alternatively, whether a reciprocal relationship exists. Results from cross-lagged panel analyses indicate that culture “comes first,” consistently predicting subsequent ratings of customer satisfaction and vehicle sales. Furthermore, the positive effect of culture on vehicle sales is fully mediated by customer satisfaction ratings. Copyright © 2015 John Wiley & Sons, Ltd.

Full text

Which comes first, organizational culture or
performance? A longitudinal study of causal
priority with automobile dealerships
ANTHONY S. BOYCE
1
*, LEVI R. G. NIEMINEN
2
, MICHAEL A. GILLESPIE
3
,
ANN MARIE RYAN
4
AND DANIEL R. DENISON
5
1
Aon Hewitt, Port Washington, New York, U.S.A.
2
Denison Consulting, Ann Arbor, Michigan, U.S.A.
3
University of South Florida Sarasota–Manatee, Sarasota, Florida, U.S.A.
4
Michigan State University, East Lansing, Michigan, U.S.A.
5
International Institute for Management Development, Lausanne, Switzerland
Summary Prior research supports a link between organizational culture and performance but generally falls short of
establishing causality or determining the direction of a culture–performance (C-P) relationship. Using data
collected from 95 franchise automobile dealerships over 6 years, we studied longitudinal culture–
performance relationships to determine whether culture or performance has causal priority, or alternatively,
whether a reciprocal relationship exists. Results from cross-lagged panel analyses indicate that culture “comes
first,”consistently predicting subsequent ratings of customer satisfaction and vehicle sales. Furthermore, the
positive effect of culture on vehicle sales is fully mediated by customer satisfaction ratings. Copyright © 2015
John Wiley & Sons, Ltd.
Keywords: organizational culture; employee surveys; organizational performance
Organizational culture has long been described as a driver of firm performance, from its earliest mentions in the
research literature (Jaques, 1951; Pettigrew, 1979) to a number of popular press books (e.g., Ouchi, 1981; Peters
& Waterman, 1982). Despite early reviews that were mainly critical of empirical culture–performance (C-P) research
(Rousseau, 1990; Siehl & Martin, 1990), evidence has accumulated that an association does exist and that
certain features of organizational culture are routinely correlated with a range of organizational performance
outcomes (e.g., Hartnell, Ou & Kinicki, 2011; Sackmann, 2011). However, research to date has generally
fallen short of establishing a causal culture-to-performance effect. Alternatively, it is possible that organiza-
tional performance causes organizational culture, culture and performance are reciprocally related, or both
are caused by a third variable.
Given the unique theoretical and practical implications of these alternative forms of a C-P relationship, the ques-
tion of causal priority is of fundamental importance. In lieu of closely controlled experimentation, longitudinal C-P
studies are necessary to tease apart causal priority and rule out alternative explanations, as well as to more fully
understand the timing of C-P relationships. In her recent review of the C-P domain, Sackmann (2011) identified only
six longitudinal studies among the 55 total studies published. Importantly, none of these studies provided an explicit
test of reverse or reciprocal causality.
To address these shortcomings, the present study uses a longitudinal design that spans 6 years and includes
multiple waves of culture and performance data for sales and service departments within 95 automobile dealerships.
By studying the pattern of C-P relationships over successive years, this study provides two important contributions.
The first is causal priority: Is department culture an antecedent of performance outcomes over time? This is a
*Correspondence to: Anthony S. Boyce, Aon Hewitt, 10 Charles Avenue, Port Washington, New York 11050, U.S.A. E-mail: anthonyboyce@gmail.com
Copyright © 2015 John Wiley & Sons, Ltd.
Received 12 August 2013
Revised 10 May 2014, Accepted 26 October 2014
Journal of Organizational Behavior, J. Organiz. Behav. (2015)
Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/job.1985
Research Article

fundamental, yet untested, assumption of existing culture theory and practice. If culture is not an antecedent of
performance, then existing theoretical assumptions require revision and firms would be well advised to direct their
limited resources away from explicitly building a positive culture and toward interventions directly impacting firm
performance. To date, this question has not been appropriately examined longitudinally.
The second contribution involves the timing of C-P relationships: Over what time period (or delay) do predictive
relationships emerge? Despite general agreement that organizational cultures tend to evolve slowly over time, cur-
rent theory and research are largely silent on the temporal lags of C-P relationships (Schein, 1985). This gap in
understanding has unfortunate practical and scholarly implications. Practitioners may not fully understand or appre-
ciate the time it will take for culture improvements to demonstrate a return on investment for various performance
outcomes. With limited insights into time horizons, it will remain difficult for practitioners to compare culture inter-
ventions with other management tools and strategies. In empirical research, inferences about the substantive nature
of C-P relationships must be considered in the context of the time lags investigated. Absent a well-reasoned and
tested theory (of which we are not aware), it will be particularly difficult to evaluate null findings in future C-P
studies, particularly those studies set within one specific time frame or lag.
Prior to discussing the current study in greater detail, we first introduce several key theoretical perspectives
bearing on the C-P directionality issue, followed by a brief review of the longitudinal studies from this domain.
The C-P Link: Theory and Empirical Observations
In describing the basis for a C-P link, one must first acknowledge a certain degree of definitional ambiguity sur-
rounding the organizational culture construct (Verbeke, Volgering & Hessels, 1998). Despite this ambiguity, various
definitions generally reflect a focus on the values, beliefs, and assumptions that are held by the employees of an
organization and which facilitate shared meaning and guide behavior (e.g., Hofstede, Neuijen, Ohayv & Sanders,
1990; Schein, 1985). Consistent with our purpose, this study focuses primarily on the norms, values, and practices
that are observable and, therefore, measurable manifestations of an organization’s culture (Ashkanasy, Broadfoot &
Falkus, 2000).
Wilderom, Glunk, and Maslowski (2000) indicated that a major challenge facing researchers is the “establishment
of a theoretical basis for explaining the assumed [C-P] relation”(p. 205). Although no comprehensive theory has
emerged, two perspectives particularly relevant to the causal directionality issue are the process-oriented models that
describe how organizational cultures come to be (e.g., Schein, 1985) and the resource-based view of the firm
(Barney, 1986, 1991).
Process-oriented models. Process-oriented models follow from culture’s anthropological origin by focusing on
cultural dynamics and interpretation (Hatch, 1993; Martin, 1992). Several researchers have described culture as
adaptive and laid the groundwork for a reciprocal C-P relationship. Perhaps most notable is Schein’s (1985)
classic description of the evolution of culture across the organizational life cycle. Schein proposed that early in
the life cycle, culture mainly reflects the values of the founding leader. Subsequently, these values are challenged
as the organization struggles to adapt and respond to internal and external problems. Over time, the values that
enhance the organization’s responsiveness are passed on to “new members as the correct way to perceive, think,
andfeelinrelationtothoseproblems”(Schein, 1985, p. 9), and a set of fundamental beliefs and assumptions
become deeply engrained, reflecting the cumulative lessons learned.
Thus, the life cycle process described by Schein (1985) supports the idea of a reciprocal C-P relationship, with
firm effectiveness and survival serving as a feedback loop (Sackmann, 2011; Wilderom et al., 2000). However, there
are a few noteworthy factors that may limit the influence of the feedback loop. First, attribution theory and research
suggest that individuals are motivated to attribute their failures, and those of the groups to which they belong, to
external rather than internal factors (Kelley, 1971). This suggests that culture change is unlikely to occur unless
the people in the organization view the culture as part of the problem. Second, Schein describes how the ability
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

to change the culture becomes constrained as the organization reaches maturity. And third, characteristics of the
culture itself can lead to change resistance (Hofstede, 2001). Together, these factors suggest ways in which an orga-
nization’s momentum may constrain a performance-to-culture feedback loop and, by extension, weaken the strength
of a reciprocal relationship. Nonetheless, the potential for the dynamic interplay of culture and organizational effec-
tiveness is foundational to a process-oriented perspective.
The resource-based view. Barney (1986, 1991) described three conditions that are required for organizational culture
to have a causal impact on firm profitability. First, the culture must be valuable to the firm by allowing it to behave in
ways that facilitate a higher degree of effectiveness than competitors. Second, it must be rare. No advantage is con-
ferred if the culture is commonplace among the firm’s competitors. Third, it must be imperfectly imitable, such that
competitors cannot readily re-create the culture in their own organization. In other words, cultures that are valuable,
rare, and not easily replicated can facilitate superior performance. Thus, a resource-based view clearly positions cul-
ture as a causal antecedent of performance outcomes, provided that certain conditions are met.
Culture content. One approach to articulating how culture contributes to organizational effectiveness has been
to focus on the content of cultural values and norms (Saffold, 1988). Although a number of scholars have
developed influential trait-based frameworks comprising various dimensions of culture (e.g., Hofstede et al.,
1990; Quinn & Rohrbaugh, 1983), only a subset focus specifically on the C-P link, and even fewer have a
strong theoretical basis (Ashkanasy et al., 2000). One of the exceptions is Denison’s model of cultural effec-
tiveness, which provides the framework for the current investigation.
Denison and colleagues’theory proposes that the most effective organizations are characterized by a strong
mission and high levels of employee involvement,internal consistency, and adaptability (Denison & Mishra,
1995). Employee involvement is the extent to which the organization encourages empowerment, team-based coop-
eration, and individual learning and development; internal consistency is the degree to which there exists a clear set
of espoused values, agreement about values, and inter-departmental coordination that arises from this common and
agreed upon set of values; adaptability represents the degree to which the organization is focused on learning from
its competitors and customers and promotes flexible and adaptive responses at both the organizational and employee
levels; and finally, mission regards the extent to which the organization has a clearly articulated strategic direction
that provides context for action and goals against which progress can be tracked.
According to Denison’s theory, effective organizations have all of these cultural traits, and the balancing and
simultaneous pursuit of the competing demands these values represent is critical to organizational effectiveness.
Indeed, the most effective organizations have high levels of each trait, or a “full profile”(Denison, Nieminen &
Kotrba, 2014). In support of the theory, positive correlations between the traits and a range of effectiveness criteria
have been demonstrated in a variety of organizational, industry, and national contexts (e.g., Fey & Denison, 2003;
Gillespie, Denison, Haaland, Smerek & Neale, 2008; Yilmaz & Ergun, 2008).
Indirect longitudinal evidence
Considering the dynamic interplay of culture and organizational effectiveness from the process-oriented perspective
as well as the direct causal role of culture ascribed by the resource-based view, there is adequate rationale for
proposing both direct and reciprocal C-P relationships over time. Unfortunately, no studies that we are aware of
have assessed organizational culture and performance repeatedly over time with an explicit test of their relationship.
However, indirect evidence regarding the timing and nature of C-P relationships can be gleaned from two sources.
One form of indirect evidence can be drawn from the handful of studies that have quantitatively tested C-P rela-
tionships using time-lagged effectiveness criteria. Most research suggests that the magnitude of C-P relationships
remains stable or increases slightly within a 1- to 6-year period following culture assessment (Denison & Mishra,
1995; Gordon & DiTomaso, 1992; Zahra & Covin, 1995). However, Petty, Beadles, Lowery, Chapman and Connell
(1995) reported a slight decrease in the positive effects of cultural variables on performance after a 1-year delay.
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Because these studies assessed culture only once, their findings provide relatively limited insights into the dynamic
nature of C-P relationships over time and provide no test of reverse or reciprocal causality.
A second form of indirect evidence comes from longitudinal case studies, which follow organizations as they
undertake a culture transformation and document subsequent improvements in objective and subjective performance
indicators (e.g., Fairfield-Sonn, 1993; Frame, Nielsen & Pate, 1989; O’Regan & Lehmann, 2008). The generalizabil-
ity of these findings is difficult to establish given the focus on one particular organizational context. Also, because
the majority of case studies focus on organizations undergoing effortful culture transformation, it is further unsur-
prising that these studies have focused little attention on the possibility of reverse or reciprocal C-P relationships.
Two exceptions include case studies by Sackmann, Eggenhofer-Rehart, and Friesl (2009) and Bititci, Mendibil,
Nudurupati, Garengo, and Turner (2006), both of which describe the dynamic nature of C-P relationships over time.
In summary, the existence of a C-P relationship is relatively well established after 30 years of theoretical and
empirical developments, but with a still tenuous basis for staking a causal claim or asserting directionality. No study
that we are aware of has conducted an explicit test of the process-oriented view of culture and performance as recip-
rocally related, nor has any study adequately tested the alternative possibility that performance causes culture, not
vice versa (Siehl & Martin, 1990). Failure to consider directionality and the timing of C-P relationships is an impor-
tant deficit in the literature, with substantial theoretical and practical implications.
The present study
This study investigates C-P relationships over a multi-year period within the sales and service departments of franchise
automobile dealerships. The dealerships carried the same products and used the same performance metrics but never-
theless were owned and operated independently. Consistent with an integration perspective, culture was conceptualized
as a shared phenomenon at the department level (Martin, 1992). Although this may overlook important cultural dyna-
mics at higher or lower levels of the dealerships, we chose to focus on departments because they operate fairly indepen-
dently within the dealerships and because the employees within these groups work together as a team and share
common performance objectives. Moreover, statistical tests (see succeeding discussions) supported aggregation of
individual perceptions of culture to the department level. C-P relationships were tested using a series of cross-lagged
panel analyses. The performance variables investigated were vehicle sales (sales departments) and customer satisfaction
(sales and service departments).
Hypotheses
On the basis of our review of the C-P literature, we expected that department culture would have an effect on both
customer satisfaction and sales. Denison’s research has shown that organizations with higher mean levels or intensity
(Chatman, 1989) of cultural values and norms in each of the four traits—involvement, consistency, adaptability, and
mission—tend to be more profitable and have higher sales growth (e.g., Denison, 1984; Denison & Mishra, 1995), more
favorable customer satisfaction ratings (Gillespie et al., 2008), and higher subjective ratings of firm effectiveness (Fey &
Denison, 2003; Denison et al., 2003). Moreover, these same themes are reflected in a number of additional studies
characterizing the cultural context of high performing organizations in the service and sales sector (e.g., Lee, Yoon,
Kim & Kang, 2006; Petty et al., 1995; Ryan, Schmit & Johnson, 1996; Schneider, White & Paul, 1998).
This research suggests broadly that an effective departmental culture is characterized by high overall levels of
involvement, consistency, adaptability, and mission. Accordingly, we expected the overall culture of a department,
as indexed by its mean level across traits, to be positively related to its customer satisfaction ratings and vehicle sales
over time. In this study, we define causal priority as the relative strength of the directional relationships (culture-to-
performance vs. performance-to-culture) rather than the total absence of a reverse effect (performance-to-culture).
Therefore, we hypothesized that the overall culture of a department is a stronger predictor of subsequent perfor-
mance than vice versa.
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Hypothesis 1: In sales and service departments, the department culture has causal priority over customer satisfaction.
Hypothesis 2: In sales departments, the department culture has causal priority over vehicle sales.
As described earlier, the process-oriented view of culture draws attention to the evolution of a culture in response
to the organization’s successes and failures. As highlighted by the longitudinal case studies reviewed previously, and
indirectly supported by two empirical studies from the climate for service domain (Ryan et al., 1996; Schneider
et al., 1998), it is likely that performance outcomes operate as a feedback loop for subsequent culture change. We
therefore advanced the following hypothesis, noting that the occurrence of a reverse performance-to-culture relation-
ship is not mutually exclusive with the hypothesized causal priority of culture.
Hypothesis 3: In sales and service departments, the department culture and performance (i.e., vehicle sales and
customer satisfaction) are reciprocally related over time, as evidenced by a positive performance-to-culture feed-
back loop.
In addition to the issue of causal priority, we also sought to test the role of customer satisfaction as a mediator of
the C-P relationship in sales departments. Prior studies demonstrate that customer satisfaction is a key factor in the
performance of service-oriented organizations and is closely linked with sales performance. This is because satisfied
customers are more likely to make repeat purchases (Grewal & Sharma, 1991) and spread positive word-of-mouth
about the organization (e.g., Maxham, 2001; Terblanche, 2011), both of which have crucial implications for future
sales. Repeat purchases comprise an important part of the total new automobile sales market (Verhoef, Langerak &
Donkers, 2007), and word-of-mouth likely impacts both the number of new customers entering dealerships and pur-
chasing behavior thereafter (Söderlund, 2002). Given the well-documented linkages between culture and customer
satisfaction (e.g., Ford et al., 2008; Lee et al., 2006) and the anticipated role of customer satisfaction in future vehicle
sales, we therefore proposed the following.
Hypothesis 4: In sales departments, customer satisfaction mediates the relationship between the department
culture and vehicle sales.
Method
Sample and procedure
Over a 6-year period (2000–2005), complete data were collected from 95 franchise automobile dealerships selling
and servicing vehicles by the same manufacturer. All dealerships were located in the United States. These dealer-
ships met the following inclusion criteria: they (i) participated in all four rounds of culture surveys, (ii) had a
minimum of three survey respondents from each department, and (iii) had customer satisfaction and vehicle sales
data across all 6 years. Both culture and performance were operationalized at the level of departments—sales or
service—within each dealership.
For analytic purposes, the variables were aligned to six time points separated by approximate 1-year intervals.
Culture surveys were collected four times, starting in the first quarter (Q1) of 2000 and subsequently in Q4-2001,
Q4-2002, and Q4-2004. On the basis of the timing and availability of the performance variables, these surveys
corresponded most closely to time points T1, T3, T4, and T6, respectively. Whenever possible, the customer satis-
faction ratings and vehicle sales for each time point were derived as the mean across the two quarters preceding and
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

following each survey administration. All years were represented, with the exception that no vehicle sales informa-
tion was available for 2002.
Culture surveys were administered in paper-and-pencil in 2000 and 2001 and online in 2002 and 2004. All
employees in the sales and service departments received an invitation to complete the survey during normal working
hours. Participation was voluntary, and respondents were ensured that their responses would be kept anonymous.
For sales departments, the average number of employees completing the survey was 13 (range 3 to 42). For service
departments, the average number was 32 (range 4 to 131). No information on the demographics of respondents or
response rates within the dealerships or departments was available.
Customer satisfaction surveys were mailed to all customers purchasing or servicing a vehicle. Each quarter, an
average of 41 and 85 surveys were returned to sales and service departments, respectively, yielding response rates
of approximately 50 and 35 percent.
Measures
Culture survey
Department culture was assessed using the Denison Organizational Culture Survey (DOCS). The DOCS as-
sesses four primary cultural traits described earlier: involvement, consistency, adaptability, and mission. In
total, the survey consists of 60 items (15 items per trait), each of which uses a 5-point scale ranging from
1=strongly disagree to 5 = strongly agree. Prior studies have supported the DOCS factor structure with the
four cultural traits as second-order factors and demonstrated acceptable levels of internal consistency for the
15-item sub-scales (e.g., Denison et al., 2014). In the present study, Cronbach’sαacross time and departments
was .94 for involvement, .94 for consistency, .92 for adaptability, and .96 for mission.
In order to justify aggregating individual ratings of culture to the department level, it is first necessary to demon-
strate adequate levels of within-group agreement (r
wg(j)
), inter-rater reliability (ICC1), and group mean reliability
(ICC2; Bliese, 2000). Because the ratings of culture traits demonstrated a negative skew, r
wg(j)
was computed on
the basis of a uniform null distribution and again on the basis of a slightly negatively skewed null distribution
(LeBreton & Senter, 2008). Across years, departments, and traits, the median r
wg(j)
ranged from .78 to .96, exceed-
ing the .70 criterion recommended as adequate evidence for aggregation (Lance, Butts & Michels, 2006). Inter-rater
reliability (ICC1) corresponds to the percentage of total variance in cultural ratings that can be attributed to group
membership (Bliese, 2000). James (1982) reported a median ICC1 of .12 across studies from the organizational
literature, and Bliese indicated that values between .05 and .20 have typically been deemed acceptable. In the present
study, ICC1 ranged from .11 to .19. The reliability of group means (ICC2) also provided adequate support for
aggregation, with values ranging from .67 to .85 across years and departments. These results provide sufficient
justification for aggregating individual ratings of culture to the department level. In addition to statistical justifica-
tion, we note briefly that our focus on separate departments as the unit of analysis is justified by two additional
considerations: (i) the dependent variables measured are specific to department type (see succeeding discussions);
and (ii) sales and service departments are intact subgroups with different people (sales staff vs. mechanics) completing
different types of work (selling vs. fixing automobiles).
Consistent with our focus on investigating department culture as an antecedent of performance outcomes, we
derived an index of overall department culture by taking the mean across all four culture traits. Although this ap-
proach is not sensitive to potential differences at the trait level, this decision was made in light of several key con-
siderations. First, prior research indicates that the traits tend to be moderately-to-strongly positively correlated. In the
present study, trait correlations were generally high (i.e., .89 to .96 with a mean of .93). Second, the statistical com-
plexity of the tested models precluded use of a multidimensional approach in which the effects of the four cultural
traits could be simultaneously modeled. Finally, we note that separate analyses for each culture trait (following the
same analytic strategy reported later) revealed an overall pattern of findings that was both consistent across traits and
consistent with the overall culture results reported here. As a set, the trait-level analyses resulted in exactly 300
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

model tests, for which only eight minor model-fitting differences were observed across the four traits. Further, a
comparison of path coefficients from the final models did not identify a pattern of unique C-P relationships by trait
or in comparison with the overall culture results. Full details on the trait-level findings are available from the authors
upon request.
Customer satisfaction
Mean ratings on a single customer satisfaction survey item were available on a quarterly basis for each sales
department: “Based on your overall purchase/lease and delivery experience, how satisfied are you with XYZ
Dealership.”Similar mean customer responses to a single item were available for each service department:
“Based on this service visit overall, how satisfied are you with XYZ Dealership?”Customers rated their satis-
faction using a 4-point scale, ranging from 1 = not at all satisfied to 4 = completely satisfied. Although the use
of a single-item customer satisfaction indicator is not ideal, there is some evidence in support of the reliability
of single-item measures (Wanous, Reichers & Hudy, 1997).
New vehicle sales
The number of new vehicle sales for each dealership was available on a quarterly basis. On average, the dealerships
included in this study sold approximately 93 vehicles (median = 77) each quarter, with sales increasing slightly over
time. In order to account for a strong negative skew, sales data were logarithmically transformed prior to analyses.
Control variables
Department size and economic conditions were investigated as potential control variables. The average number of
respondents from each department was used as a proxy for department size because all employees were strongly
encouraged to complete the culture surveys and the actual number of employees was not available. Department size
was consistently correlated with the number of new vehicle sales (ranging from r= .18 to r= .27), so we controlled
for it in all analyses involving vehicle sales. Although it was plausible that department size could also affect
customer satisfaction, this was not the case in the present study (Table 1). Thus, we did not control for department
size in analyses involving satisfaction. We also investigated the role of local economic conditions (i.e., unemployment
rates) and found they were unrelated to sales or satisfaction, so they were not included as controls.
Analytic strategy
The analytic strategy compared models with increasingly strict assumptions about the underlying relationships
among variables, with the best-fitting model from each stage serving as the initial comparison model for the next
stage. Following Hu and Bentler’s (1999) recommendation, model fit was assessed using the following criteria:
SRMR ≤0.08 (and RMSEA ≤0.06 or CFI ≥0.95). Comparisons between nested models were assessed with the
chi-square difference test. All models were estimated using AMOS 7.0. Separate analyses were conducted for service
departments-satisfaction outcomes, sales departments-satisfaction outcomes, and sales departments-sales outcomes.
The timing of C-P relationships was explored by repeating analyses based on 1-, 2-, and 3-year lags. This was per-
formed because no a priori basis existed for specifying the optimal time lag of C-P relationships.
The analysis progressed through four stages (See Appendix). Valid inferences about the similarity or differ-
ences of structural relationships are contingent upon evidence that the constructs of interest have the same
underlying meaning and are measured consistently across time and between groups. Thus, stage 1 of the an-
alytic process tested the invariance of the DOCS across departments. Stage 2 began with the full cross-lagged
reciprocal model. The key features of this model are as follows: (i) cross-lagged paths from culture to out-
comes, (ii) cross-lagged paths from outcomes to culture, (iii) autoregressive paths within constructs, and (iv) residual
correlations between culture and outcomes within each measurement occasion. The cross-lagged paths estimate
the hypothesized relationships. The autoregressive lags control for prior levels of the variables when
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Table 1. Descriptive statistics and correlations for major study variables in sales departments (lower left) and service departments (upper right).
Variables
Sales Service
12345678910111213141516MSDM SD
1. Culture T1 3.67 0.33 3.41 0.26 —.49 .45 .39 .16 .23 .27 .31 .19 .17 .04
2. Culture T3 3.70 0.41 3.44 0.28 .49 —.48 .52 .03 .13 .21 .39 .30 .27 .09
3. Culture T4 3.81 0.37 3.50 0.30 .41 .36 —.40 .11 .16 .31 .39 .41 .25 .03
4. Culture T6 3.84 0.41 3.54 0.33 .16 .40 .31 —.04 .13 .16 .26 .28 .13 .17
5. Sat T1 3.71 0.15 3.49 0.16 .17 .18 .13 .14 —.72 .56 .37 .36 .25 .10
6. Sat T2 3.74 0.14 3.49 0.14 .13 .26 .02 .19 .66 —.60 .49 .28 .23 .06
7. Sat T3 3.75 0.10 3.50 0.15 .22 .22 .01 .02 .38 .47 —.68 .45 .40 .01
8. Sat T4 3.76 0.11 3.55 0.13 .20 .10 .25 .06 .49 .51 .54 —.67 .58 .02
9. Sat T5 3.80 0.08 3.60 0.12 .24 .28 .15 .18 .41 .32 .36 .50 —.62 .01
10. Sat T6 3.81 0.09 3.61 0.13 .18 .27 .21 .19 .43 .27 .37 .45 .47 —.08
11. Dept. size 12.69 6.10 31.48 13.93 .10 .03 .08 .05 .08 .04 .05 .10 .15 .14 —
12. Sales T1 1.87 0.32 —— .05 .16 .01 .10 .25 .34 .14 .22 .14 .22 .27 —
13. Sales T2 1.81 0.31 —— .05 .15 .01 .06 .24 .31 .12 .21 .10 .18 .27 .98 —
14. Sales T3 1.87 0.30 —— .07 .14 .01 .07 .30 .30 .13 .24 .14 .27 .22 .95 .96 —
15. Sales T4 1.85 0.31 —— .06 .13 .02 .09 .21 .26 .06 .18 .16 .18 .20 .92 .93 .93 —
16. Sales T5 1.91 0.30 —— .06 .19 .00 .09 .20 .28 .10 .16 .17 .21 .22 .03 .94 .93 .98 —
17. Sales T6 1.95 0.29 —— .11 .21 .01 .10 .21 .26 .11 .20 .17 .27 .18 .91 .92 .92 .96 .98
Note: Correlations in bold are significant at p<.05. Correlations in bold italics are significant at p<.01.
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

estimating the impact of the cross-lagged variable. The residual correlations were included on the basis of
evidence, provided by Anderson and Williams (1992) that failure to account for these correlations can lead
to biased estimates of the cross-lagged effects. The impact of wave-skipping autoregressive lags (e.g., T1 to
T3) was also investigated on the basis of prior research demonstrating that inclusion of these paths improved
model fit (e.g., Madon, Willard, Guyll, Trudeau & Spoth, 2006).
Stage 3 models compared three more parsimonious models to the full cross-lagged reciprocal model by removing
selected paths. The first model examined the possibility that culture influences outcomes over time, but not vice
versa, by constraining the cross-lagged paths from outcomes to culture to be zero. The second model examined
the converse where outcomes influence culture over time, but not vice versa. The third model, an autoregressive null
model, tested the alternative that no direct causal relationships exist by constraining all cross-lagged paths to zero.
Finally, stage 4 models tested the consistency of the autoregressive relationships (e.g., C-C and P-P) and C-P rela-
tionships over time. The first set of models constrained the autoregressive lags to be equal, and the second
constrained the cross-lagged paths to be equal.
Results
Measurement invariance
Measurement invariance of the DOCS across time and departments was examined at the department level
using item parcels defined by the four cultural traits. We followed the procedure outlined by Vandenberg
and Lance (2000) with one exception. The omnibus test suggested by Vandenberg and Lance can lead to
erroneous conclusions (e.g., Raju, Laffitte & Byrne, 2002) and, therefore, was excluded from interpretation.
Additionally, within-variable measurement error correlations were allowed because repeated measures of the
same variable generally results in correlated measurement errors (Bollen, 1989).
Examination of equivalence of factor patterns (i.e., configural invariance) with factor loadings freely estimated
across time and departments indicated that the model fit the data well (model 1 SRMR = 0.02, CFI = 0.99,
RMSEA = 0.04, χ
2
(120) = 232.62). Next, metric equivalence was assessed by constraining the factor loadings to
be equal across departments (Model 2a), time (Model 2b), and both time and departments (Model 2c). As demon-
strated by non-significant changes in chi-square tests, neither Model 2a nor 2b fitsthedatasignificantly worse than
the configural model. Similarly, Model 2c also did not fit the data significantly worse than 2a or 2b, supporting
metric invariance across both time and departments (Model 2c SRMR = 0.02, CFI = 0.98, RMSEA = 0.05,
χ
2
(169) = 257.23).
Next, scalar invariance was examined by constraining intercept terms to be equal across departments (Model
3a) and time (Model 3b). Scalar invariance analyses indicated a significant reduction in fit(Model3a)and
failed convergence (Model 3b). These results indicated that scalar invariance is not present and that testing
more restrictive models was unnecessary. The existence of metric invariance provides sufficient justification
for proceeding with the structural analyses (Bollen, 1989; Cheung & Rensvold, 1998). As the pattern of factor
loadings observed for the DOCS was consistent across time, these parameters were constrained to be equal.
Culture and customer satisfaction
Hypothesis 1, proposing that department culture has causal priority over customer satisfaction, was tested separately
for service and sales departments. The reciprocal effect predicted by Hypothesis 3 was also tested separately by
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

department. Hypothesis-testing models were run for all time lags, but our interpretation focuses only on the best-
fitting lag periods.
Service departments
To illustrate our model-testing process, fit statistics for all 1-, 2-, and 3-year lag models are presented in
Table 2. Overall, the 1-year lag results provided the best fit. In contrast, an autoregressive null model did
not fit significantly worse for the 2- and 3-year lag analyses, indicating that C-P effects did not emerge at
these time intervals.
On the basis of the 1-year lag, the initial full cross-lagged reciprocal model (model 1) did not satisfy all of the
criteria for acceptable fit (e.g., SRMR >0.08). However, an alternative model (Model 1a) including wave-skipping
autoregressive lags (e.g., T1 to T3, and T2 to T4) resulted in significant improvement of fit, had acceptable fit over-
all, and was retained as the comparison model for subsequent stages (SRMR = 0.04, CFI = 0.99, RMSEA = 0.05,
χ
2
(176) = 218.50). Model 2, which constrained the 1-year cross-lagged effects of customer satisfaction on culture
to be zero, did not result in significantly worse fit. In contrast, Model 3, which constrained the 1-year cross-lagged
effects of culture on customer satisfaction to be zero, did result in a significant decline in fit, as did Model 4, the
autoregressive null model. These results support a direct effect of culture on customer satisfaction but not reverse
or reciprocal effects.
Models 5 and 6 were tested to examine the consistency of the 1-year lag model over time. Model 5, which
constrained the autoregressive lags to be equal over corresponding time periods, did not result in significantly
worse fit. Similarly, imposing the constraint of equal cross-lags over time periods (Model 6) also did not result
in a significant decline in fit; thus, model 6 was retained as the final 1-year lag model (SRMR = 0.08,
CFI = 0.99, RMSEA = 0.06, χ
2
(190) = 239.38). This indicates that the effect of culture on customer satisfaction
was consistent over time. Figure 1a shows the parameter estimates for this model. The cross-lagged paths from
culture to customer satisfaction were stable over time and significant (p<.01) with standardized estimates
ranging from .16 to .20 (note that although the coefficients were constrained to be equal in the unstandardized
solution, standardization led to slightly different estimates). For service departments, the results supported the
hypothesis that culture has causal priority over customer satisfaction (Hypothesis 1) but failed to support the
hypothesis of reciprocal relationships over time (Hypothesis 3).
Sales departments
Overall, the 2-year lag, results provided the best fit to the data. As with the service department results, the
model-testing sequence resulted in retaining model 6, with equal culture-to-performance cross-lags, as the final
model (SRMR = 0.06, CFI = 0.98, RMSEA = 0.05, χ
2
(181) = 225.09). Table 3 contains fit statistics, and
Figure 1b shows the parameter estimates for this model (fit statistics for all time periods and models are
available upon request from the authors). The cross-lagged paths from culture to customer satisfaction were
stable over time and significant (p<.01) with standardized estimates ranging from .13 to .20. Therefore, re-
sults for sales departments supported the hypothesis that culture has causal priority over customer satisfaction
(Hypothesis 1) but failed to support the hypothesis of reciprocal relationships over time (Hypothesis 3).
Culture and vehicle sales
Hypothesis 2 proposed that culture would have causal priority over vehicle sales. Overall, the 2-year lag results pro-
vided the best fit. Again, Model 6, including equal 2-year culture-to-performance cross-lags, was retained as the final
model for the effects of culture on vehicle sales (SRMR = 0.06, CFI = 0.98, RMSEA = 0.07, χ
2
(196) = 296.41;
Table 3). Figure 1c shows the parameter estimates for this model. The cross-lagged paths from culture to vehicle
sales were stable over time and statistically significant (p<.05)—standardized estimates were approximately .03.
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Table 2. Summary of Fit indices for service department models (all lag periods).
Model SRMR CFI RMSEA Chi-square df
Comparison
model
ΔChi-
square Δdf
1-year cross-lagged models: culture and customer satisfaction*
1. Full cross-lag 0.12 0.98 0.07 256.79 182 Model 1a 38.29 6
1a. Full cross-lag with
wave-skipping
a
0.04 0.99 0.05 218.50 176 —
2. Cross-lagged: culture
to sat. only
0.05 0.99 0.05 222.73 179 Model 1a 4.23 3
3. Cross-lagged: sat.
to culture only
0.09 0.98 0.06 238.04 179 Model 1a 19.54 3
4. Autoregressive null
(no cross-lags)
0.11 0.98 0.06 242.60 182 Model 2 19.87 3
5. Equal
autoregressive lags
0.07 0.99 0.05 238.14 188 Model 2 15.41 9
6. Equal cross-lags
(within construct)
0.08 0.99 0.06 239.38 190 Model 5 1.23 2
2-year cross-lagged models: culture and customer satisfaction
1. Full cross-lag 0.15 0.97 0.08 277.54 182 Model 1a 39.10 6
1a. Full cross-lag with
wave-skipping
a
0.09 0.98 0.06 238.44 176 —
2. Cross-lagged:
culture to sat. only
0.09 0.98 0.06 239.08 179 Model 1a 0.64 3
3. Cross-lagged: sat.
to culture only
0.11 0.98 0.06 241.98 179 Model 1a 3.54 3
4. Autoregressive null
(no cross-lags)
0.11 0.98 0.06 242.60 182 Model 2 3.52 3
Model 3 0.62 3
5. Equal
autoregressive lags
0.12 0.98 0.06 256.18 191 Model 4 13.58 9
6. Equal cross-lags
(within construct)
b
—— — — — —
3-year cross-lagged models: culture and customer satisfaction
1. Full cross-lag 0.15 0.97 0.07 277.65 184 Model 1a 38.86 6
1a. Full cross-lag with
wave-skipping
a
0.09 0.98 0.06 238.79 178 —
2. Cross-lagged:
culture to sat. only
0.09 0.98 0.06 239.38 180 Model 1a 0.59 2
3. Cross-lagged: sat.
to culture only
0.11 0.98 0.06 241.95 180 Model 1a 3.16 2
4. Autoregressive null
(no cross-lags)
0.11 0.98 0.06 242.60 182 Model 1a 3.81 4
5. Equal
autoregressive lags
0.12 0.98 0.06 256.18 191 Model 4 13.58 9
6. Equal cross-lags
(within construct)
b
—— — — — —
Note: Table shows model results corresponding to stage 2 (labeled Models 1 and 1a), stage 3 (labeled Models 2, 3, and 4), and stage 4 (labeled
Models 5 and 6) of the described analytic strategy. This analytic strategy was repeated for 1-, 2-, and 3-year cross-lags. Chi-square values in bold
are significant at p<.05. Chi-square values in bold italics are significant at p<.01.
a
Model with autoregressive wave-skipping lags.
b
Model not applicable because the autoregressive null model was not rejected.
* The 1-year cross-lagged model with equal cross-lags from culture to customer satisfaction provided the most acceptable fit and was retained as
the final model.
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Figure 1. Final cross-lagged models
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Table 3. Summary of fit indices for sales departments models (best-fitting lag periods*).
Model SRMR CFI RMSEA
Chi-
square df
Comparison
model
ΔChi-
square Δdf
2-year cross-lagged models: culture and customer satisfaction
1. Full cross-lag 0.12 0.97 0.07 257.67 182 Model 1a 34.94 6
1a. Full cross-lag with wave-
skipping
a
0.06 0.98 0.05 222.73 176 —
2. Cross-lagged: culture to sat. only 0.06 0.98 0.05 224.45 179 Model 1a 1.72 3
3. Cross-lagged: sat. to culture only 0.08 0.98 0.06 232.81 179 Model 1a 10.08 3
4. Autoregressive null (no cross-lags) 0.09 0.98 0.06 234.53 182 Model 2 10.08 3
5. Equal autoregressive lags 0.09 0.98 0.06 256.66 188 Model 2 32.21 9
6. Equal cross-lags (within
construct)
0.06 0.98 0.05 225.09 181 Model 2 0.64 2
2-year cross-lagged models: culture and vehicle sales
1. Full cross-lag 0.09 0.96 0.09 340.40 197 Model 1a 49.77 6
1a. Full cross-lag with wave-
skipping
a
0.04 0.98 0.07 290.63 191 —
2. Cross-lagged: culture to sales only 0.05 0.98 0.07 293.55 194 Model 1a 2.92 3
3. Cross-lagged: sales to culture only 0.04 0.97 0.08 298.73 194 Model 1a 8.10 3
4. Autoregressive null (no cross-
lags)
0.06 0.97 0.08 301.65 197 Model 2 8.10 3
5. Equal autoregressive lags 0.07 0.97 0.08 330.73 203 Model 2 37.19 9
6. Equal cross-lags (within
construct)
0.06 0.98 0.07 296.41 196 Model 2 2.86 2
2-year cross-lagged models: customer satisfaction and vehicle sales
1. Full cross-lag 0.10 0.94 0.14 138.01 47 Model 1a 53.03 8
1a. Full cross-lag with wave-
skipping
a
0.07 0.92 0.11 84.98 39 —
2. Cross-lagged: sat. to sales only 0.08 0.97 0.10 86.46 43 Model 1a 1.48 4
3. Cross-lagged: sales to sat. only 0.07 0.97 0.11 95.12 43 Model 1a 10.14 4
4. Autoregressive null (no cross-
lags)
0.09 0.97 0.11 96.80 47 Model 2 10.34 4
5. Equal autoregressive lags 0.10 0.95 0.13 141.56 57 Model 2 55.10 14
6. Equal cross-lags (within
construct)
0.08 0.97 0.10 91.27 46 Model 2 4.81 3
Mediation models: culture to customer satisfaction to vehicle sales
1. Partial mediation 0.10 0.95 0.08 536.15 330 Model 1a 71.51 10
1a. Partial mediation with wave-
skipping
a
0.07 0.97 0.07 464.65 320 —
2. Full mediation 0.07 0.97 0.07 471.33 323 Model 1a 6.68 3
3a. Cross-lagged: culture to sat. only 0.07 0.96 0.07 481.79 327 Model 2 10.47 4
3b. Cross-lagged: sat. to sales only 0.09 0.96 0.07 480.45 326 Model 2 9.12 3
4. Autoregressive null (no cross-
lags)
0.09 0.96 0.07 490.89 330 Model 2 19.56 7
5. Equal autoregressive lags 0.09 0.95 0.08 536.98 339 Model 2 65.65 16
6. Equal cross-lags (within
construct)
0.07 0.97 0.07 476.79 328 Model 2 5.47 5
Note: Chi-square values in bold are significant at p<.05. Chi-square values in bold italics are significant at p<.01.
a
Model with autoregressive wave-skipping lags.
* Fit statistics for all lag periods are available upon request from the authors.
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

There was no evidence of reciprocal or reverse relationships. Therefore, results support the hypothesis that culture
has causal priority over vehicle sales (Hypothesis 2).
Mediation of the culture–sales relationships by customer satisfaction
Hypothesis 4 proposed that customer satisfaction mediates the relationship between culture and vehicle sales.
Consistent with the results described above for sales departments, the cross-lagged effects of customer satisfaction
and vehicle sales on culture were excluded, and the cross-lagged effects of culture on customer satisfaction and
vehicle sales were set to 2-year lags. Prior to testing mediation, an additional set of models was needed examining
the longitudinal relationship between customer satisfaction and vehicle sales. This corresponds to Baron and
Kenny’s (1986) third condition of mediation: a direct effect of the mediator on the dependent variable. Results supported
a 2-year lag model with wave-skipping autoregressive lags and equal cross-lagged effects from customer satisfaction to
vehicles sales (Model 6 SRMR = 0.08, CFI = 0.97, RMSEA = 0.10, χ
2
(46) = 91.27; Table 3). The cross-lagged effects
were stable over time and significant (p<.05), with standardized estimates ranging from .02 to .04.
The results of mediation tests are shown in the lower portion of Table 3. A partial mediation model included 2-year
cross-lagged effects of culture on customer satisfaction, culture on vehicle sales, and customer satisfaction on vehicle
sales (Model 1). Consistent with prior results, the inclusion of wave-skipping autoregressive lags (Model 1a) resulted
in a significant improvement in fit. A full mediation model (Model 2), removing the direct paths from culture to vehicle
sales, did not result in a significant decline in fit. In contrast, Models 3a and 3b, which removed the cross-lagged effects
of customer satisfaction on sales and the cross-lagged effects of culture on customer satisfaction, respectively, resulted
in a significant decline in fit. Finally, constraining corresponding autoregressive lags to be equal (Model 5) resulted in
significantly worse fit, whereas constraining the cross-lagged effects to be equal (Model 6) did not.
A full mediation model with equal cross-lagged effects from culture to customer satisfaction and customer satis-
faction to vehicle sales (Model 6) was retained as the final model (SRMR = 0.07, CFI = 0.97, RMSEA = 0.07,
χ
2
(328) = 476.79). Figure 2 shows the parameter estimates for this model. The cross-lagged paths from culture to
customer satisfaction were stable over time and statistically significant (p<.01), with standardized estimates ranging
from .12 to .19. The cross-lagged paths from customer satisfaction to sales were stable over time and significant
(p<.05), with standardized estimates ranging from .02 to .04. Therefore, results supported customer satisfaction
as fully mediating the culture-to-vehicle sales relationship.
Discussion
The results of our study suggest that organizational culture has causal priority over performance outcomes.
Overall, department culture was found to consistently predict higher subsequent levels of customer satisfaction
ratings and vehicle sales, with no evidence obtained for a reciprocal performance-to-culture feedback loop. In
addition, the positive effect of culture on vehicle sales was mediated by customer satisfaction. We discuss how
these findings contribute important insights regarding the nature and timing of C-P relationships, both within
specific service and sales organizational contexts and more broadly within the culture and performance
research domain.
Causal priority of C-P relationships
The causal priority of culture was evidenced by a consistent pattern in which department culture predicted subse-
quent levels of performance more strongly than vice versa, controlling for prior levels of culture and performance.
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Indeed, the cross-lagged effects of performance on department culture could be removed from all tested models
without significantly reducing fit. Support for an antecedent role is consistent with the resource-based view of
culture as a source of competitive advantage (Barney, 1986, 1991) and extends prior research linking organizational
culture to performance (e.g., Pelham & Wilson, 1996; Zahra & Covin 1995).
Timing of C-P relationships
In the present study, the immediacy and magnitude of a culture-to-performance effect depended on department and
type of performance indicator. In service departments, culture’s effect on customer satisfaction emerged consistently
for 1-year lags. In contrast, the culture-to-customer satisfaction relationship emerged at 2-year lags in sales depart-
ments, as did the effect of culture on vehicle sales. Several possibilities may account for the longer delay of C-P
relationships in sales departments.
The customer service and sales experience in automobile dealerships consists of a mix of tangible and intangible ser-
vices (Shostack, 1977). Tangible services include vehicles, parts, and maintenance, and intangible services include the
behaviors of the service provider that guide customers’impressions of service quality (Schneider & Bowen, 1985). It is
interesting to speculate that culture may have a more direct and immediate effect on customer satisfaction outcomes in
the provision of intangible services, such as in service departments where employees have a greater opportunity to be
responsive to problems and demonstrate competency and fairness in a continuous interaction with the customer. Alter-
natively, customer satisfaction with the sales department might be due in larger part to tangible factors less directly
under the influence of culture (e.g., sales price) and which take time to manifest (e.g., the car’s long-term performance).
Future research could test these observations by systematically comparing the timing of C-P relationships for tangible
and intangible service providers.
Figure 2. Final mediation model for sales department culture, customer satisfaction, and vehicle sales
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Although prior longitudinal studies have not resulted in a coherent pattern of cross-lagged C-P effects—and
more fundamentally, such studies are lacking (Sackmann, 2011)—a few observations are nevertheless warranted
regarding what is known so far and the implications for practitioners and researchers. In combination with prior
studies, our findings suggest that the magnitude of C-P relationships tends to peak somewhere between 1 and
3 years (e.g., Kotrba et al., 2012; Pelham & Wilson, 1996; Petty et al., 1995), although a relationship may be
observable over a longer time horizon. Only two published studies to date seem to stretch beyond the notion of
the 1-to-3-year peak. Zahra and Covin (1995) found that the correlations between entrepreneurial culture and
financial performance increased up to 6 years following culture assessment, whereas Gordon and DiTomaso
(1992) documented slightly increasing relationships between the cultural stability adaptability of 11 insurance
firms and their annual growth in assets and premiums up to 4 years following culture assessment. Other studies
have reinforced the stability of C-P relationships by demonstrating positive effects with performance operationa-
lized as a 3-year average (e.g., Denison & Mishra, 1995).
For practitioners, the notion of a 1-to-3-year peak could prove useful to set general guidelines about the return on
investment for culture-focused work and interventions. If replicated, the pattern identified here regarding tangible
versus intangible service outcomes could inform thinking about where to look for these returns first and where
additional patience will be necessary. More generally, we suspect this finding will confirm popular perspectives
of culture-focused work as a relatively long-term management strategy when juxtaposed with other nearer term
solutions and tools.
For researchers, the 1-to-3-year peak informs the characteristics of adequate study design within the C-P domain.
The most direct implication is that findings from cross-sectional designs should not be expected to generalize
to other time lags and may misrepresent the magnitude of C-P relationships. Obviously, as our study has
pointed out, the context and nature of the culture and performance variables under examination are likely to influ-
ence the peak and temporal stability of the relationships observed (Zahra & Covin, 1995). Perhaps more importantly,
this research and subsequent studies will lead to the emergence of new theoretical perspectives and models to help
explain and predict the nuanced pattern of C-P relationships, taking time and several of these context factors into
account.
Satisfaction versus sales outcomes
Our results also highlighted a difference in magnitude between culture-to-satisfaction and culture-to-sales relation-
ships. From one perspective, the small effect of department culture on vehicle sales might be considered too weak to
be practically meaningful. However, when considering the size of these effects, it is important to remember that prior
levels of vehicle sales and culture, as well as department size, were statistically controlled for when estimating the
cross-lagged relationships. Together, these factors accounted for 90 percent of the total variance in vehicle sales,
leaving little variance left to be accounted for by year-to-year changes in culture. These findings are consistent with
prior research that has reported large reductions in predictive relationships with financial outcomes once prior levels
are taken into account (e.g., Koys, 2001; Wright, Gardner, Moynihan & Allen, 2005). Last, it is important to point
out that even small effects can have large practical implications, particularly when the relationships are distal with a
number of intervening processes, competing causes, and random factors (Abelson, 1985). Future research could
shed additional light on the proximal-distal nature of C-P relationships by attempting to clarify potential mediating
processes.
Mediation by customer satisfaction
Our results suggest a causal chain, whereby the culture of sales departments affects customer satisfaction, which, in
turn, influences the number of vehicles sold. At each point in the sequence, relationships emerged at a 2-year lag,
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

underscoring that the impact of culture on vehicle sales is not immediate. Although the mechanisms responsible for
these linkages deserve further study, our rationale focuses on the importance of positive word of mouth and cus-
tomer loyalty. These factors seem particularly important owing to the regional markets that are typically serviced
by automobile dealerships. In other words, because it may be difficult for regionally based organizations to attract
buyers from other geographical regions, maintaining a positive reputation in their community and building a loyal
customer base seem paramount to long-term success. Our results suggest that neither is built overnight, but that a
strong culture at the foundation can be a unique point of leverage for winning and retaining customers over time.
Future research should explore other potential mediators of C-P relationships in other contexts.
Limitations
Four main limitations should be considered when interpreting the results of our study. First, there was a degree of
asymmetry and missing data that had to be overcome when aligning the culture and performance variables to
time points for analytic purposes. As a result, it is possible that results could have differed with more closely
matched timing or complete data. However, we believe that any impact of these asymmetries on the present results
was minimal, given that many of our results demonstrated clear and consistent patterns over time (e.g., all models
could be estimated with equal cross-lagged effects). Second, the relatively small number of departments included
might be considered a limitation. Although the present sample size did not result in being underpowered to
detect small relationships, it can nevertheless result in over- or under-estimated standard errors in structural equation
modeling (Muthén & Muthén, 2002). This latter point underscores the need for future studies that replicate our
findings, particularly with respect to the lack of reciprocal C-P relationships observed. Third, it is important to
highlight that the causal argument presented here is based on non-experimental data, which to some degree
reduces the strength of causal inference. Although the collection of longitudinal data and the use of cross-lagged
panel analyses strengthen our claims of causal priority (Kenny & Harackiewicz, 1979), other longitudinal designs
including quasi-experimentation should be conducted, if feasible, to further test the pattern of relationships reported
here.
Fourth, our decision to derive an overall index of department culture by taking the mean of culture traits did not
allow for a multidimensional exploration of longitudinal C-P relationships. Prior research has generally supported a
multidimensional (or configurational) rather than global culture perspective and accumulated a complex pattern of
C-P relationships across various dimensions of culture and performance criteria (e.g., Denison et al., 2014; Hartnell
et al., 2011; Sackmann, 2011). However, in our pragmatic view, the decision to model and report results based on
the overall culture (as an aggregate of measured culture traits) was justified, first, in order to achieve a parsimonious
test of temporal precedence and, second, given that a trait-level analysis failed to identify unique patterns of C-P re-
lationships or modify the substantive conclusions drawn from the overall culture results. Having demonstrated the
temporal precedence of culture generally in this study, subsequent research could build upon these findings by
exploring the combination and interaction of culture dimensions and performance criteria within a multivariate
framework (Kotrba et al., 2012).
Conclusion
Our study adds important evidence of a longitudinal culture-to-performance relationship to the research literature
(Sackmann, 2011; Wilderom et al., 2000), indicating that higher intensity of positive organizational culture traits
among automobile dealerships leads to greater customer satisfaction in later years. Moreover, culturally based gains
in customer satisfaction led to small but reliable improvements in subsequent vehicle sales. On the other hand, no
evidence of a performance-to-culture feedback loop was found, thus failing to support a reciprocal or reverse
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

causality model. Therefore, in response to our initial causal priority question, we conclude from our study that it is
culture that comes first, with performance levels to follow.
Acknowledgements
This work is based on the first author’s doctoral dissertation. He would like to extend his sincere gratitude to
Denison Consulting for providing access to the data and support throughout the process, as well as to his committee
members Ann Marie Ryan, Neal Schmitt, Steve W. J. Kozlowski, and Frederick P. Morgeson for their support and
invaluable input.
Author biographies
Anthony S. Boyce, PhD, is a Associate Partner at Aon Hewitt where he leads research and development efforts for
the Assessment and Selection practice. Tony earned his MA and PhD from Michigan State University.
Levi Nieminen, PhD, is the Director of Research & Development at Denison Consulting. His research and consult-
ing is focused on organizational culture and leadership as interrelated drivers of organizational effectiveness. Levi
holds a doctoral degree in Industrial and Organizational Psychology from Wayne State University in Detroit,
Michigan.
Michael A Gillespie, PhD, is an Assistant Professor of Psychology at University of South Florida, Sarasota. His re-
search focuses on organizational culture, critical thinking, psychological measures, and job attitudes. He earned his
BA from USF, Tampa, his MA from Michigan State University, and his PhD from Bowling Green State University.
Ann Marie Ryan, PhD, is a professor of organizational psychology at Michigan State University. Her major re-
search interests involve improving the quality and fairness of employee selection methods, and topics related to di-
versity and justice in the workplace.
Daniel Denison, PhD, is a Professor of Organization and Management at the International Institute for Management
Development (IMD) in Lausanne, Switzerland, and the Founder and Chairman of Denison Consulting. Since receiv-
ing his PhD in Organizational Psychology from the University of Michigan, he has authored numerous books and
journal articles describing his research linking organizational culture to bottom-line business performance.
References
Abelson, R. P. (1985). A variance explanation paradox: When a little is a lot. Psychological Bulletin,97(1), 129–133.
DOI: 10.1037/0033-2909.97.1.1129
Anderson, S. E., & Williams, S. J. (1992). Assumptions about unmeasured variables with studies of reciprocal relationships: The
case of employee attitudes. Journal of Applied Psychology,77, 638–650. DOI: 10.1037/0021-9010.77.5.638
Ashkanasy, N. M., Broadfoot, L.E., & Falkus, S. (2000). Questionnaire measures of organizational culture. In N.M. Ashkanasy, C.P.
M. Wilderom, & M.F. Peterson (Eds.), Handbook of organizational culture & climate (pp. 131–146). Thousand Oaks, CA: Sage.
Barney, J. B. (1986). Organizational culture: Can it be a source of sustained competitive advantage? Academy of Management
Review,11, 656–665. DOI: 10.2307/258317
Barney, J. B. (1991). Firm resources and sustained competitive advantage. Journal of Management,17,99–120. DOI: 10.1177/
014920639101700108
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Concep-
tual, strategic, and statistical considerations. Journal of Personality and Social Psychology,51, 1173–1182. DOI: 10.1037//
0022-3514.51.6.1173
Bititci, U. S., Mendibil, K., Nudurupati, S., Garengo, P., & Turner, T. (2006). Dynamics of performance measurement and
organisational culture. International Journal of Operations & Production Management,26, 1325–1350. DOI: 10.1108/
01443570610710579
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Bliese, P. D. (2000). Within-group agreement, non-independence, and reliability: Implications for data aggregation and analysis.
In K. J. Klein, & S. W. J. Kozlowski (Eds.), Multilevel theory, research, and methods in organizations: Foundations,
extensions, and new directions (pp. 349–381). San Francisco: Jossey-Bass.
Bollen, K. (1989). Structural equations with latent variables. Oxford: John Wiley & Sons.
Chatman, J. A. (1989). Improving interactional organizational research: A model of person-organization fit. Academy of Management
Review,14,333–349.
Cheung, G. W., & Rensvold, R. B. (1998). Cross-cultural comparison using non-invariant measurement items. Applied Behavioral
Science Review,7,93–110. DOI: 10.1016/S1068-8595(99)80006-3
Denison, D. R. (1984). Bringing corporate culture to the bottom line. Organizational Dynamics,13,4–22. DOI: 10.1016/
0090-2616(84)90015-9
Denison, D. R., Haaland, S., & Goelzer, P. (2003). Corporate culture and organizational effectiveness: Is there a similar pattern
around the world? Advances in Global Leadership,3, 205–227. DOI: 10.1016/S1535-1203(02)03011-3
Denison, D.R., & Mishra, A. K. (1995). Toward a theory of organizational culture and effectiveness. Organization Science,6(2),
204–223. DOI: 10.1287/orsc.6.2.204
Denison, D. R., Nieminen, L. R., & Kotrba, L. (2014). Diagnosing organizational cultures: A conceptual and empirical review
of culture effectiveness surveys. European Journal of Work and Organizational Psychology. DOI: 10.1080/
1359432X.2012.713173
Fairfield-Sonn, J. W. (1993). Moving beyond vision: Fostering cultural change in a bureaucracy. Journal of Organizational
Change Management,6,43–55. DOI: 10.1108/EUM0000000001211
Fey, C., & Denison, D. R. (2003). Organizational culture and effectiveness: Can an American theory be applied in Russia?
Organization Science,14, 686–706. DOI: 10.1287/orsc.14.6.686.24868
Ford, R. C., Wilderom, C. P., & Caparella, J. (2008). Strategically crafting a customer-focused culture: An inductive case study.
Journal of Strategy and Management,1, 143–167. DOI: 10.1108/17554250810926348
Frame, R. M., Nielsen, W. R., & Pate, L. E. (1989). Creating excellence out of crisis: Organizational transformation at the
Chicago Tribune. Journal of Applied Behavioral Science,25, 109–122. DOI: 10.1177/0021886389252001
Gillespie, M., Denison, D., Haaland, S., Smerek, R., & Neale, W. (2008). Linking organizational culture and customer
satisfaction: Results from two companies in different industries. European Journal of Work and Organizational Psychology,
17, 112–132. DOI: 10.1080/13594320701560820
Gordon, G., & DiTomaso, N. (1992). Predicting corporate performance from organizational culture. Journal of Management
Studies,29, 783–798. DOI: 10.1111/j.1467-6486.1992.tb00689.x
Grewal, D., & Sharma, A. (1991). The effect of salesforce behavior on customer satisfaction: An interactive framework. The
Journal of Personal Selling & Sales Management,11,13–23.
Hartnell, C. A., Ou, A. Y., & Kinicki, A. (2011). Organizational culture and organizational effectiveness: A meta-analytic
investigation of the competing values framework’s theoretical suppositions. Journal of Applied Psychology,96, 677–694.
DOI: 10.1037/a0021987
Hatch, M. J. (1993). The dynamics of organizational culture. Academy of Management Review,18,657–693. DOI: 10.2307/258594
Hofstede, G. (2001). Culture’s consequences,(2
nd
ed.). Thousand Oaks, CA: Sage.
Hofstede, G., Neuijen, B., Ohayv, D. D., & Sanders, G. (1990). Measuring organizational cultures: A qualitative and quantitative
study across twenty cases. Administrative Science Quarterly,35, 286–316. DOI: 10.2307/2393392
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new
alternatives. Structural Equation Modeling,6,1–55. DOI: 10.1080/10705519909540118
Jaques, E. (1951). The changing culture of a factory. London, UK: Tavistock.
James, L. R. (1982). Aggregation bias in estimates of perceptual agreement. Journal of Applied Psychology,67, 219–229.
DOI: 10.1037/0021-9010.67.2.219
Kelley, H. H. (1971). Attribution in social interaction. New York: General Learning Press.
Kenny, D. A., & Harackiewicz, J. M. (1979). Cross-lagged panel correlation: Practice and promise. Journal of Applied Psychology,
64(4), 372–379. DOI: 10.1037//0021-9010.64.4.372
Kotrba, L. M., Gillespie, M. A., Schmidt, A. M., Smerek, R. E., Ritchie, S. A., & Denison, D. R. (2012). Do consistent corporate
cultures have better business performance? Human Relations,65, 241–262.
Koys, D. J. (2001). The effects of employee satisfaction, organizational citizenship behavior, and turnover on organizational
effectiveness: A unit-level, longitudinal study. Personnel Psychology,54,101–114. DOI: 10.1111/j.1744-6570.2001.tb00087.x
Lance, C., Butts, M., & Michels, L. (2006). The sources of four commonly reported cutoff criteria: What did they really say?
Organizational Research Methods,9, 202–220. DOI: 10.1177/1094428105284919
LeBreton, J. M., & Senter, J. L. (2008). Answers to 20 questions about interrater reliability and interrater agreement. Organiza-
tional Research Methods,11, 815–852. DOI: 10.1177/1094428106296642
Lee, S., Yoon, S. J., Kim, S., & Kang, J. W. (2006). The integrated effects of market-oriented culture and marketing strategy on
firm performance. Journal of Strategic Marketing,14, 245–261. DOI: 10.1080/09652540600825548
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Madon, S., Willard, J., Guyll, M., Trudeau, L., & Spoth, R. (2006). Self-fulfilling prophecy effects of mothers’beliefs on
children’s alcohol use: Accumulation, dissipation, and stability over time. Journal of Personality and Social Psychology,
90, 911–926. DOI: 10.1037/0022-3514.90.6.911
Martin, J. (1992). The culture of organizations: Three perspectives. New York: Oxford University Press.
Maxham, J. G. (2001). Service recovery’sinfluence on consumer satisfaction, positive word-of-mouth, and purchase intentions.
Journal of Business Research,54(1), 11–24. DOI: 10.1016/S0148-2963(00)00114-4
Muthén, L. K., & Muthén, B. O. (2002). How to use a Monte Carlo study to decide on sample size and determine power.
Structural Equation Modeling,4, 599–620. DOI: 10.1207/S15328007SEM0904_8
O’Regan, N., & Lehmann, W. (2008). The impact of strategy, leadership, and culture on organisational performance: A case
study of an SME. International Journal of Process Management and Benchmarking,2, 303–322. DOI: 10.1504/
IJPMB.2008.021790
Ouchi, W. G. (1981). Theory Z: How American business can meet the Japanese challenge. Reading, MA: Addison-Wesley.
Pelham, A. M., & Wilson, D. T. (1996). A longitudinal study of the impact of market structure, firm structure, and market orientation
culture on dimensions of small-firm performance. Journal of the Academy of Marketing Science,24,27–43. DOI: 10.1007/
BF02893935
Peters, T. J., & Waterman, R. (1982). In search of excellence. New York: Harper and Row.
Pettigrew, A. M. (1979). On studying organizational cultures. Administrative Science Quarterly,24,570–581. DOI: 10.2307/
2392363
Petty, M. M., Beadles, N. A., Lowery, C. M., Chapman, D. F., & Connell, D. W. (1995). Relationships between organizational
culture and organizational performance. Psychological Reports,76, 483–492. DOI: 10.2466/pr0.1995.76.2.483
Quinn, R. E., & Rohrbaugh, J. (1983). A spatial model of effectiveness criteria: Towards a competing values approach to orga-
nizational analysis. Management Science,49, 917–947. DOI: 10.1287/mnsc.29.3.363
Raju, N., Laffitte, L., & Byrne, B. (2002). Measurement equivalence: A comparison of methods based on confirmatory factor
analysis and item response theory. Journal of Applied Psychology,87, 517–529. DOI: 10.1037//0021-9010.87.3.517
Rousseau, D. M. (1990). Assessing organizational culture: The case for multiple methods. In B. Schneider (Ed.), Organizational
culture and climate (pp. 153–192). San Francisco: Jossey-Bass.
Ryan, A. M., Schmit, M. J., & Johnson, R. (1996). Attitudes and effectiveness: Examining relations at an organizational level.
Personnel Psychology,49, 853–882. DOI: 10.1111/j.1744-6570.1996.tb02452.x
Sackmann, S. A. (2011). Culture and performance. In N. Ashkanasy, C. Wilderom, & M. Peterson (Eds.), The handbook of
organizational culture and climate (2nd edn., pp. 188–224). Thousand Oaks, CA: Sage Publications.
Sackmann, S. A., Eggenhofer-Rehart, P. M., & Friesl, M. (2009). Sustainable change: Long-term efforts toward developing a
learning organization. Journal of Applied Behavioral Science,45, 521–549. DOI: 10.1177/0021886309346001
Saffold, G. (1988). Culture traits, strength, and organizational performance: Moving beyond ’strong’culture. Academy of
Management Review,13, 546–558. DOI: 10.2307/258374
Schein, E.H. (1985). Organizational culture and leadership. San Francisco: Jossey-Bass.
Schneider, B., & Bowen, D. E. (1985). Employee and customer perceptions of service in banks: Replication and extension.
Journal of Applied Psychology,70(3), 423–433. DOI: 10.1037/0021-9010.70.3.423
Schneider, B., White, S. S., & Paul, M. C. (1998). Linking service climate and customer perceptions of service quality: Tests of a
causal model. Journal of Applied Psychology,83(2), 150–163. DOI: 10.1037/0021-9010.83.2.150
Shostack, G. (1977). Breaking free from product marketing. Journal of Marketing,41,73–80. DOI: 10.2307/1250637
Siehl, C., & Martin, J. (1990). Organizational culture: A key to financial performance? In B. Schneider (Ed.), Organizational
climate and culture (241–281). San Francisco: Jossey-Bass.
Söderlund, M. (2002). Customer familiarity and its effects on satisfaction and behavioral intentions. Psychology & Marketing
Scandinavian experiences,19(10), 861–880. DOI: 10.1002/mar.10041
Terblanche, N. S. (2011). The mediating role of customer satisfaction and loyalty in predicting the word-of-mouth of supermarket
customers. Management Dynamics: Journal of the Southern African Institute for Management Scientists,20(4), 23–37.
Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, prac-
tices, and recommendations for organizational research. Organizational Research Methods,3,4–69. DOI: 10.1177/
109442810031002
Verbeke, W., Volgering, M., & Hessels, M. (1998). Exploring the conceptual expansion within the field of organizational
behaviour: Organizational climate and organizational culture. Journal of Management Studies,35,303–329. DOI: 10.1111/
1467-6486.00095
Verhoef, P. C., Langerak, F., & Donkers, B. (2007). Understanding brand and dealer retention in the new car market: The
moderating role of brand tier. Journal of Retailing,83,97–113. DOI: 10.1016/j.jretai.2006.10.007
Wanous, J. P., Reichers, A. E., Hudy, M. J. (1997). Overall job satisfaction: How good are single-item measures? Journal of
Applied Psychology,82, 247–252. DOI: 10.1037/0021-9010.82.2.247
A. S. BOYCE ET AL.
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job

Wilderom, C. P., Glunk, U., & Maslowski, R. (2000). Organizational culture as a predictor of organizational performance. In N.
Ashkanasy, C. Wilderom, & M. Peterson (Eds.), The handbook of organizational culture and climate (pp. 193–209). Thousand
Oaks, CA: Sage.
Wright, P. M., Gardner, T. M., Moynihan, L. M., & Allen, M. R. (2005). The relationship between HR practices and firm
performance: Examining causal order. Personnel Psychology,58(2), 409–446. DOI: 10.1111/j.1744-6570.2005.00487.x
Yilmaz, C., & Ergun, E. (2008). Organizational culture and firm effectiveness: An examination of relative effects of culture traits
and the balanced culture hypothesis in an emerging economy. Journal of World Business,43, 290–306. DOI: 10.1016/j.
jwb.2008.03.019
Zahra, S. A., & Covin, J. G. (1995). Contextual influences on the corporate entrepreneurship-performance relationship: A
longitudinal analysis. Journal of Business Venturing,10,43–58. DOI: 10.1016/0883-9026(94)00004-E
APPENDIX ANALYSIS FLOWCHART
CULTURE–PERFORMANCE
Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. (2015)
DOI: 10.1002/job