The routines literature argues that_ as repetitive patterns of

Document Sample
The routines literature argues that_ as repetitive patterns of Powered By Docstoc
					  A co-organizing view of the impact of variant executions and
context irregularity on service errors in the waste collection sector

                          Scott F. Turner
                 R. H. Smith School of Business
                       Van Munching Hall
                      University of Maryland
                     College Park, MD 20742
                        Tel: 301-405-7186
                        Fax: 301-314-8787

                          Violina Rindova
                  McCombs School of Business
                    University of Texas-Austin
                         Austin, TX 78712
                         Tel: 512-471-7975
                        Fax: 512-471-3937

                      Version: April 30, 2007

                  A co-organizing view of the impact of variant executions and
                context irregularity on service errors in the waste collection sector


This study examines the impact of routine functioning on error generation in a services setting.
From one view, routine functioning is said to have beneficial effects in the form of developing
skills that reduce errors, but an alternative view suggests detrimental effects as corresponding
tendencies towards mindless behavior increase the risk of errors. To better understand the effect
of routine functioning on error generation, we examine contextual conditions more closely. We
distinguish between two types of context in which routines are executed: a larger objective
context and a local enacted context. Collectively, we argue that the objective context in which
routines are executed, the enacted context created by historical executions, and the current
execution of the routine jointly determine the generation of service errors. We test our
hypotheses in the context of municipal waste collection for one of the largest cities in the United
States, finding evidence that is consistent with our arguments.


In organizational theory and strategy, routines are a central concept (Hannan and Freeman, 1984;
March and Simon, 1958; Nelson and Winter, 1982; Stene, 1940). As repetitive patterns of action
involving interdependent actors, routines provide a powerful concept for explaining the behavior
of organizations. Central to the routine-based perspective is the efficiency criterion (Karim and
Mitchell, 2000), which is manifest at multiple levels within organizations. For individuals,
routines foster the development of habits that provide cognitive and behavioral efficiency
(James, 1914/1890; Simon, 1976/1945). At group and organization levels, routines enable
efficient means of coordination and governance (Coriat and Dosi, 1998; Gersick and Hackman,
1990; Nelson and Winter, 1982), and these efficiencies stem in large part from the organizing
process, whereby routines enable the fusing together of individual-level habits (Cohen and
Bacdayan, 1994; Weick, 1979). Across this body of work, substantial attention is directed to
efficiencies in the productive and governance costs required to generate a focal product or
service (Karim and Mitchell, 2000).

However, we have less consensus regarding the impact of routine functioning on the quality of
the generated product or service. Some scholars suggest beneficial effects of routine
functioning, as it facilitates skill development which leads to higher-quality outputs (James,
1914/1890; Nelson and Winter, 1982). Further, ecology scholars suggest that consistency
generates external value in the form of conferred legitimacy from stakeholders (Hannan and
Freeman, 1984). By contrast, a habits-based stream of research suggests detrimental quality
effects from routine functioning, as its tendency towards mindless behavior results in
inappropriate actions and greater risk of errors (Cohen and Bacdayan, 1994; Gersick and
Hackman, 1990).

The purpose of this study is to examine the implications of routine functioning for error
generation in a services context. While learning scholars have placed increasing interest on the
study of organizational errors (Haunschild and Sullivan, 2002), this phenomena is important for
routines scholarship as well (Weick, 1987; Weick and Roberts, 1993). Errors can range from the
minor to the catastrophic, and we note that researchers highlight the importance of developing
our understanding of the small-scale errors (Repenning and Sterman, 2002). Such errors are
pervasive, have learning value, and are often focal outcomes in their own right (i.e. customer
satisfaction). Further, catastrophic errors research emphasizes the tendency for small-scale
errors to propagate into large-scale disasters (Perrow, 1984; Weick, 1987). In particular, we
focus on service errors to better understand the process failures that occur at the boundaries of
organizations, where organizations intersect with their customers.

In this paper, we view routines as interaction patterns that organize the behavior of employees
and co-organize the behavior of customers. We make two core arguments. First, consistent with
past research, we view the execution of routines as context dependent (Cohen, et al., 1996;
Nelson and Winter, 1982). Specifically, the current execution of a routine is shaped by the
regularity of its larger contextual conditions. Second, in departure from past research, we view
the execution of routines as context creating. This arises as the recurring nature of routine
execution enacts an interlocking structure between employees and customers (Turner and
Rindova, working paper; Weick, 1979). Moreover, the extent of variation in the recurring

executions affects the degree of "structuring" in this local, enacted context -- thereby influencing
the subsequent flexibility and effectiveness of the routine. In our view, the objective context in
which routines are executed, the enacted context created by past executions, and the current
execution of the routine jointly determine the extent of error generation. We test a corresponding
set of hypotheses in the context of municipal garbage collection for one of the largest cities in the
United States in 2005, and the results provide evidence consistent with our arguments.


In routines theory, variation and consistency are the two dialectical opposites that capture the
central tension in executing organizational processes and deriving benefits from routine
functioning. While routines are characterized by their substantive similarity in sequences of
action across executions (Cohen and Bacdayan, 1994; Cohen, et al., 1996), variation is also
present in these action sequences (Feldman and Pentland, 2003; Pentland, 2003). Further, while
routines scholars suggest that consistency provides efficiency benefits by automating the
organizing process, a mindless view of automating human action suggests a simultaneous
increase in the risk of process failure.

Service errors are an important form of process failure (Tucker and Edmondson, 2003). These
errors are instances in which an organization fails to provide service or incorrectly provides
service to customers. By contrast with traditional efficiency metrics, errors capture an
effectiveness dimension of process outcomes. Errors represent important indicators of
effectiveness for a wide range of services organizations, including health care (Tucker and
Edmondson, 2003), air transport (Haunschild and Sullivan, 2002; Weick and Roberts, 1993), and
power generation (Perrow, 1984).

In services settings, we note that greater levels of behavioral interaction between organizations
and their customers can increase the incidence of errors. Greater behavioral interaction arises for
many forms of service provision, as the customer must first provide an input. For example,
healthcare requires patients to supply medical conditions, education needs the presence of
students, and garbage collection requires customers to supply waste materials.

Service errors can be small- or large-scale. While rare and catastrophic errors understandably
attract substantial attention, it is interesting to observe that examinations of these large-scale
failures often refocus our attention on the importance of small-scale errors as originating sources
for catastrophic failure (Gersick and Hackman, 1990; Perrow, 1984; Weick, 1987). Yet, existing
research provides limited understanding of the relationship between organizational processes,
such as routines, and the generation of small-scale errors.

Variant Execution and Context Irregularity

In this study, we argue that two forms of variant executions, historical and current, and context
irregularity collectively influence the generation of service errors. Historical variant executions
represent the extent of sequential variety associated with prior executions of the routine
(Pentland, 2003). Thus, this concept captures the extent to which the action sequence of the
routine has been consistently executed in the past. Current variant executions capture the extent

to which the current execution of the routine departs from its historical executions, representing
the degree to which the current action sequence is dissimilar from preceding executions of the
routine. Context irregularity represents the extent to which the larger context in which routines
are executed is dissimilar from its regular state. While some change is always present across
contexts for execution, context irregularity refers to changes in the larger context that have more
substantive consequences for routine functioning (Gersick and Hackman, 1990; Wood, Tam and
Witt, 2005).

Context as Objective and Enacted

For routines theory, "context dependence is fundamental; the effectiveness of a routine is not
measured by what is achieved in principle but by what is achieved in practice; this generally
means that the routine might be declared effective in some specific contexts, but perhaps not in
others" (Winter in Cohen, et al., 1996: 662). Moreover, scholars emphasize that multiple facets
of context influence routine functioning and effectiveness. In particular, scholars direct attention
to the dependence of routines on local context, which complements the functioning of a specific
routine, and the larger context, which represents the objective environment within which routines
are executed (Cohen, et al., 1996; Nelson and Winter, 1982).

While scholarly attention has been directed to the determining effect of context on the execution
of routines (Becker, 2004; Cohen, et al., 1996), we argue that historical routine executions also
enact local context. Given the complementary nature of routines and their contexts, we propose
that historically-enacted context and current variant executions jointly determine service error
generation. With greater variation in historical executions, there is less enactment of the
localized context, less co-organizing constraints from customers, and more mindful employee
actors. This results in fewer process failures, as greater employee mindfulness and greater
organizational slack facilitate absorption of various disturbances encountered during routine
execution. Yet more variant current executions result in greater service errors as the current
execution departs from established contextual support mechanisms. Moreover, we argue that
there is an interactive effect. Greater consistency of historical executions results in greater
enactment of local context, such that the routine becomes more dependent on the local context,
and current variant executions are more likely to lead to service errors.

We also consider dependence on the larger objective context. We argue that given
complementary routine-context relationships, error generation is also determined by the
objective context and its alignment with current executions. As there is greater irregularity in the
objective context, employees and customers have less supporting structure to prompt and guide
their task-related actions, suggesting that greater likelihood of errors. But irregular contextual
conditions reduce the positive effect of current variant executions on error generation, as context
irregularity adjusts employee and customer expectations for variant current executions. Figure 1
illustrates routine execution as context dependent and context creating.


Our first hypothesis focuses on the relationship between context irregularity and service errors.
Context dependence is a fundamental premise in the routines and habits literatures (Cohen, et al.,

1996; Nelson and Winter, 1982; Wood, Tam and Witt, 2005). Routines and habits are
established within particular contexts, and the recurring stimuli and structure from the context
prompt and support established patterns of action (Becker, 2004; Wood, Quinn, and Kashy,

For instance, the ability to execute an operating routine in a manufacturing plant requires that
complementary equipment be available and arranged in the manner in which the actors in the
routine are accustomed. Similarly, for services like overnight delivery, relevant contextual
features include days (e.g., weekday vs. weekend) and weather conditions (e.g., clear vs.
blizzard). These factors influence transit patterns and were accounted for in routine
establishment, and they impact the extent to which actors in the routine receives supporting
structure from the execution context.

Thus, when the context changes from regular to irregular, the fit between the routine and context
is disrupted, and the likelihood for errors increases.

Hypothesis 1. The greater the context irregularity, the greater the number of service errors.

While the objective context can generate unpredictable variations in obvious ways, we suggest
that the historical executions of a routine generate a local enacted context, which is defined by
the historically-recurring exchanges between the organization and its customers. These repeated
exchanges create a local context that is routine-specific because they structure the interaction
patterns among employees and between employees and customers (Turner and Rindova, working
paper; Weick, 1979). Moreover, depending on the degree of consistency in past executions, this
context will vary in the degree of structuring, and therefore in the flexibility it affords

First, consider the role of organizational employees. With greater variation in historical
executions, employees retain a state of mindfulness in task execution (Levinthal and Rerup,
2006; Weick and Roberts, 1993). By contrast, low variation in historical executions tends to
promote a state of mindless task processing, which increases the risk of misapplied habits and
routines with corresponding error consequences (Gersick and Hackman, 1990; Zellmer-Bruhn,
2003). Thus, with greater variation in historical executions, employees are more likely to engage
in mindful information processing and less likely to make errors associated with the "blind spot"
property of routines (Cohen and Bacdayan, 1994).

Next, consider the role of customers. As noted previously, in many services settings, customers
are both suppliers and recipients in service arrangements (e.g., healthcare, air travel, postal
service). As such, organizations typically establish formal rules to govern the terms of exchange
with their customers (March, Schultz, and Zhou, 2000; Weber, 1991/1948). For instance, with
air travel, passengers must arrive at the gate at least 20-30 minutes prior to the stated boarding
time to ensure that they will be allowed to board the plane. While these formal rules may not be
the only logics that govern day-to-day service exchanges, it is important to recognize that they
represent the preferred governance logic for service organizations, sufficiently important to be
stipulated as a formal rule. Organizations establish these rules of exchange governance to build
slack into the organization-customer relationship.

With greater variation in historical executions, customers have less ability to predict the
particular characteristics (e.g., timing) of any one service exchange and are more likely to follow
the formal rule, which provides the main source of predictability in the exchange. Customers
following the formal rules provide service organizations with greater slack to perform the service
in the face of inevitable process disturbances, reducing the likelihood of errors.

However, the recurring nature of executions result in a structured enacted context, as service
engagements take the form of environmental stimuli for customers. Consistently-recurring
stimuli enable customers to develop expectations and habits for typical service exchanges. This
facilitates a co-organizing process between customers and organizations, enabling customers to
evolve from governed by the formal rule to governed by their co-organized habits (Turner and
Rindova, working paper; Weick, 1979). This reduces organizational slack for handling
inevitable process disturbances and increases the likelihood of errors due to misaligned
organization-customer exchanges.

Thus, due to its effects on both employee and customer behavior, we expect that as historical
variation in routine executions increases, errors will decrease.

Hypothesis 2. The more variant the historical executions, the fewer the service errors.

Next we propose that the effect for current variant execution is in the opposite direction of that
for historical variant executions. Here we expect that more variant current executions result in
greater service errors. From the organization side, employees develop work habits and routines
with repetitive task experience (Cohen, et al., 1996; Weick, 1979). Such habits and routines are
embedded in particular contextual structures, which facilitate smooth performance of the
execution. With a more variant current execution, employees face a reduction in the structural
support to which they have become accustomed, resulting in greater likelihood of generating
service errors.

A similar pattern unfolds on the customer side. As customers share recurring exchange
experiences with an organization, they are more likely to co-organize themselves with the routine
(Turner and Rindova, working paper). The co-organizing process is supported by coaligning
temporal structures, like time of day or week, which psychologists identify as particularly strong
guides to habitual behaviors (Wood, Neal and Quinn, working paper). For instance, if a mail
carrier arrives every day at 2pm, customers are likely to establish outgoing mail habits centered
on this time -- even if departs substantially from the formal rule for having outgoing mail ready.
As a result, with a more variant current execution, customer actions in the exchange relationship
are more prone to misalignment, increasing the risk of errors. Returning to our example, if the
mail carrier arrives at an unexpected time of 10am, the likelihood of undelivered mail increases
due to greater misalignment with co-organized customer habits.

Hypothesis 3. The more variant the current execution, the greater the service errors.

Our fourth hypothesis expects that greater context irregularity will diminish the positive effect of
current variant execution on service errors. Recall the basic argument for the current variant

execution effect as follows. For employees, the increasing likelihood of errors stems from
established employee habits firing in variant manners with limited contextual familiarity and
support. Similarly, for customers, errors are more likely as the recurring organizational pattern
of service, which provides a form of stimuli-based structure, is disturbed and misaligns with the
customers' co-organized habits.

Given that baseline, as context irregularity increases, both employees and customers reduce their
expectations for routine functioning. For employees, this increases the mindfulness of their
behavior, and for customers, it can serve to untie their co-organized habits, increase their
propensity to act in line with the formal rules, and provide greater organizational slack. Thus,
greater context irregularity will have a dampening effect on the positive effect of current variant
execution on service errors.

Hypothesis 4. The greater the context irregularity, the lower the effect of current variant
execution on service errors

Our last hypothesis predicts that greater historical variation in executions will reduce the positive
effect of current variant execution on service errors. Our argument for the current execution
effect leads to error generation due to misfit between current variant executions and established
habits for executing the routine. In other words, with greater variation in historical executions,
employees develop weaker habits, retain more mindful behavior, and are less likely to generate
errors during a current variant execution. Similarly, more variant historical executions result in
less consistent organization-provided stimuli for customers, such that customers are less likely to
establish habits that co-organize their behavior to align with the organizational routine. As such,
under conditions of more variant historical executions, current variant executions are less likely
to result in service exchange problems with customers. In sum, we expect historical variation in
executions to reduce the positive effect of current variation execution on service errors.

Hypothesis 5. The more variant the historical executions, the lower the effect of current variant
execution on service errors.



Our empirical setting is one of municipal services, specifically the collection of garbage.
Garbage collection has several characteristics that are particularly appropriate for studying
routines. First, the routines literature identifies repetition and action interdependence as key
elements behind the emergence of a routine (Becker, 2004; Stene, 1940). Waste collection is a
repetitive task, as garbage service is provided weekly for each household. Notable
interdependence of actions is present in this setting, as field staff interact with others in the
organization (e.g., peers, supervisors), with customers, and with the general environment (e.g.,
dogs, pedestrians). Second, routines scholars highlight the challenges of routine identification
(Becker, 2004; Cohen, et al., 1996). In this setting, routines are clearly delineated in terms of
geography and time. Routines are classified geographically in terms of routes, which are distinct

geographic domains for which service is to be provided each week, and each routine is bounded
temporally within the confines of a typical work-day.

Our specific empirical context is the collection of garbage in the city of San Diego. San Diego is
one of the ten largest cities in the United States, and its Environmental Services Department is at
the leading edge of capturing service delivery data through sophisticated information systems.
Data supplied by the Environmental Services Department spans three domains. Service errors
are obtained from the department's work order system. Variant executions are captured with a
technique for determining the dissimilarity between sequences; this approach is utilized in
diverse areas from molecular biology to sociology to human language to bird songs (Abbott and
Hrycak, 1990; Kruskal, 1983; Pentland, 2003). Spatial locations of service delivery events (i.e.,
waste collection pick-ups) are identified with geographical information systems software, such
that standard collections (i.e., within the route domain) can be identified as distinct from atypical
work tasks undertaken by crews (i.e., collection pickups made outside the route domain). Each
data component is initially processed independently and then integrated together for subsequent
statistical analysis.

The dataset includes seven months of data from June 2005 through December 2005. We used
the first two months, June and July, in setting a baseline for historical variant executions.
Therefore, we conducted statistical analysis on five months of data: August 2005 through
December 2005. Due to data magnitude and corresponding management challenges, our analysis
focused only on Monday garbage collection routes. Within the population of approximately 70
Monday routes for the organization, we randomly selected 22 routes for analysis. Routes are
served by one-person waste collection crews that execute routines utilizing automated equipment
technology. The level of analysis is the routine execution-day.

For one collection day, November 21, data was not available for any routes due to systemic
problems with the information monitoring systems. Excluding this day, data was missing for
18% of the sample. To gain additional insight as to the randomness of missing data, we
requested additional information for a subset of missing data from IT administrators with the
department. Their response indicated that the missing data was largely due to periodic failure
with data capture elements of the information monitoring system (e.g., failure of a GPS unit).
Therefore, for purposes of our analysis, we can treat the data as missing at random, and we
performed listwise deletion for the related observations (Roth, 1994).


Table 1 presents descriptive statistics for the variables.

Dependent variable

Our dependent variable, Service Errors, is a count of the number of customer and supervisor
reports of missed collection service. The information systems identified each incident (i.e.,
service error) by route and by reporting date. We aggregated all service errors reported between
the day of routine execution and the sixth day following the execution and treated this value as
the number of service errors for that day of routine execution. Between August 2005 and

December 2005, we observed 335 service errors reported across all the routes. The range of
service incidents per route-execution day was 0 to 19.

Explanatory variables

Context Irregularity was operationalized with a binary variable that indicated whether the typical
collection service day fell on a city-observed holiday. Between August 2005 and December
2005, there were two city-observed holidays that occurred on Mondays: Labor Day on Monday,
September 5, and Christmas on Monday, December 26. Our operationalized variable, City
Holiday, was coded as a 1 if route executions fell on one of these two days, and 0 otherwise.

Historical Variant Executions and Current Variant Execution were operationalized with an
optimal string matching technique (Abbott and Hrycak, 1990; Kruskal, 1983; Pentland, 2003).
The technique captures dissimilarity between sequences in terms of the insertions, deletions and
substitutions required to transform one sequence into another sequence. The Levenshtein
distance determines the least costly set of insertions, deletions and substitutions and represents
the cost of sequence transformation. Consistent with prior work, we set the cost of each
transforming operation to 1, and for a given set of sequences, we use the average of distances
between pairs of sequences for operationalization (Pentland, 2003). For Historical Variant
Executions, for a given route execution, we calculated the average distance among all execution
sequences for that route within the prior 30 days. For Current Variant Execution, we calculated
the average distance between the current execution sequence and all execution sequences for that
route within the prior 30 days.1

Control variables

To account for potential influences from change in task load (Gersick and Hackman, 1990), we
controlled for work load by including three measures. City Work Load represents the average
number of waste collections made by all crews in the sample on the focal day. Route Work Load
indicates the number of waste collections made by a crew on the focal routine execution day, and
Route Additional Work Load represents the number of waste collections made by the crew
outside its route for the focal execution day.

To account for employee experience (Gersick and Hackman, 1990; Wood, Neal and Quinn,
working paper), we included a Route Experience variable, which indicates whether the field
employee had executed collections on this route within the past 30 days. We also accounted for
potential entraining effects associated with start times (Bluedorn, 2002). For start times, we
controlled for variation in past task start times as well as the divergence of the current start time
relative to that of prior executions. For a given route execution, Past Start Time Divergence is
the average of the absolute values of the start time difference in minutes between all execution
days for that route within the past 30 days. Current Start Time Divergence is the average of the

  We based the 30-day window on related qualitative research by the authors in a similar solid waste collection context. In that
research, informants indicated that employees typically form related work habits and routines within 4-6 weeks following the
initiation of employment or a major organizational change. Given the computational demands of the calculations, we focused on
the 30-day window.

absolute values of the difference between the start time of the current route execution day and the
start times of all executions for that route within the past 30 days.

Models and Analysis

We employed a Poisson regression model for analyzing the data and included fixed effects for
the routes. In separate analyses, we considered two robustness examinations for the model,
focusing on the potential for overdispersion to bias standard errors. Our sensitivity analyses
employed a fixed effects Poisson model with bootstrapped standard errors as well as a negative
binomial conditional fixed effects model. The sensitivity analyses provided substantively similar
results to those presented in Table 2.


Table 2 reports our analyses using a nested approach to examine the empirical evidence for our
hypotheses. All models included fixed effects for the route. As a baseline, Model 1 included all
the control variables. Models 2 through 6 added the explanatory variables independently, and
Model 7 added the explanatory variables simultaneously.

Model 2 addressed Hypothesis 1, which predicts that service errors will be greater in irregular
contexts. We expected a positive coefficient for collections associated with city-observed
holidays, and we found strong support (p<0.01). Hypothesis 2 was examined with Model 3.
This hypothesis expects that as historical variant executions increase, there will be fewer service
errors. Consistent with our prediction, we found a negative coefficient (p<0.01). Thus, our
results support Hypotheses 1 and 2.

Hypothesis 3 expects that the more variant the current execution, the greater the service errors.
We addressed this hypothesis with Model 4, and we observed a positive and significant
coefficient (p<0.05), indicating support for Hypothesis 3.

In Models 5 and 7, we examined Hypothesis 4. Our expectation is that city-recognized holidays
will have a dampening effect on the positive effect of current variant execution on service errors.
However, we did not find statistical support for Hypothesis 4. We discuss this finding in greater
detail below. Finally, Models 6 and 7 addressed Hypothesis 5. We expected that more variant
historical executions will reduce the positive effect of current variant execution on service errors.
We observed a negative coefficient for the interactive term (p<0.01), and thus we find support
for Hypothesis 5.

Extending Analysis

As noted above, we did not find statistical support for Hypothesis 4. This hypothesis states that
the positive effect of current variant execution on service errors will be reduced in irregular
contexts. A key element of our argument is that customers have lower expectations for
consistent executions when contextual conditions are disturbed. But for this effect to be present,
customers must recognize the disturbed context. We conducted an extending analysis to examine
whether the "magnitude" of the holiday may influence our results.

Within our empirical window of August-December 2005, there are two city-recognized holidays:
Labor Day and Christmas. In the United States, customers are more likely to expect Christmas
as a city-recognized holiday relative to Labor Day. With this in mind, we separated the two
holiday event variables (i.e., Christmas as the more recognized holiday) and reanalyzed the
results. For the main effect of holiday (Hypothesis 2), we observed positive coefficients for both
holidays, but the results were only statistically significant for the Christmas holiday (p<0.01).
For the interaction between the Labor Day holiday and current process divergence, we observed
a positive coefficient that was not significant. By contrast, and consistent with our rationale for
Hypothesis 4, we observed a negative and significant coefficient for the interaction between the
Christmas holiday and current variant execution (p<0.05).


The objective of this study was to gain a better understanding of the consequences of routine
functioning for error generation in a services setting. We proposed that routine functioning is
both context dependent and context creating. Specifically, routine functioning is dependent upon
the larger, objective context that provides recurring and supporting structure for executions. At
the same time, the consistent nature of routine executions enacts a local context, which structures
an interlocking of habits between employees and customers. Moreover, we propose that greater
variation in the current execution of the routine leads to greater error generation, particularly
when the local, enacted context is firmly established by historical consistency and when
conditions are regular for the larger, objective context. Our findings are consistent with the
proposed theory, and we offer related implications for the literatures that examine routines,
organizational errors and high-reliability organizations.

Implications for the Routines Literature

Our findings suggest implications for the routines literature in terms of context enactment and
context dependence. First, while the consistencies associated with routine functioning enable
efficiencies (Cohen and Bacdayan, 1994; Weick, 1979), in the process of enactment, they
simultaneously lead to tighter coupling with exchange partners. In certain instances, tighter
coupling may be advantageous as suggested by research on the formation of inter-organizational
routines between organizations and suppliers (Kotabe, et al., 2003) and between those of alliance
partners (Zollo, et al., 2002). But at the same time, in enacting interlocked structures with
customers, organizations reduce slack for task completion, have additional challenges associated
with managing stakeholder linkages, and face greater risk of process failure. For organizations,
this implies the presence of an efficiency-robustness trade-off associated with consistency of

Second, we contribute to the idea of context dependence by distinguishing between the
historically-developed enacted context and the exogenously-determined objective context.
While contextual conditions have a constraining effect on routine execution, our findings have
implications for the conditions under which routines have greater flexibility for change and
experimentation. For the enacted context, it suggests that routines are more flexible when the
routine is nascent to the context or when there are points of substantial turnover among

stakeholders. From a planning perspective, organizations that anticipate future changes may
choose to introduce more variant executions in efforts to begin to unravel the interlocked
structure between organizations and customers. Alternatively, organizations may seek to align
changes or experiments during irregular times for the objective context, as stakeholders expect
less consistency in routine functioning.

Organizational Errors and High-Reliability Organizations

This study also contributes to our understanding of the determinants of organizational errors, a
pervasive phenomena that has received limited conceptual and empirical attention in
organizational theory (Haunschild and Sullivan, 2002). But emerging scholarly interest is
reflected in two areas. One stream of research focuses on learning theories and adopts a
backward-looking view, examining how organizations learn from past errors (Haunschild and
Sullivan, 2002; Sitkin, 1992; Tucker and Edmondson, 2003). Another stream examines high-
reliability organizations with a forward-looking view, focusing on how organizations anticipate
and prevent errors (Weick, 1987; Weick and Roberts, 1993).

The role of variety is a key theme in both streams of research. In the backward-looking area,
scholars argue that organizations reduce error rates based on the information diversity embedded
in past errors, and correspondingly find that organizations with greater heterogeneity in past error
incidents have lower rates of subsequent error occurrence (Haunschild and Sullivan, 2002).
From the forward-looking stream, Weick (1987) argues that high reliability organizations, or
those with ongoing low rates of error occurrence, benefit from greater variety in the
organizational system that monitors for signals of potential problems.

Our study contributes to this literature in the linkages among variety, slack and error generation.
In research on high-reliability organizations, two central issues are reliability and organizational
slack. Weick (1987) emphasizes that one means by which organizations achieve a high-
reliability state is by building organizational slack. Our study findings reveal an important
tension. As organizations become more reliable in their routine executions, they reduce the
organizational slack that can inhibit error generation. This arises as customers receive reliable
service executions as consistent environmental stimuli to which they evolve and co-organize
their behavior. Moreover, it draws attention to tension within the concept of high-reliability
organizations (e.g., highly-reliable, low-error organizations). Specifically, we find that the value
of variety partially stems from unreliable organizational functioning, which enhances
mindfulness, preserves organizational slack, and loosens coupling between organizations and

In sum, this study enhances our understanding of the role of context in routines theory. While
previous research emphasizes the context-dependent nature of routines, this study finds that
routines are both context dependent and context creating, with corresponding implications for the
effect of routine functioning on error generation.


Abbott A, Hrycak A. 1990. Measuring resemblance in sequence data: An optimal matching
analysis of musicians' careers. American Journal of Sociology 96(1): 144-185

Becker MC. 2004. Organizational routines: A review of the literature. Industrial and Corporate
Change 13(4): 643-677

Bluedorn AC. 2002. The Human Organization of Time: Temporal Realities and Experience.
Stanford University Press

Cohen MD, Bacdayan P. 1994. Organizational routines are stored as procedural memory:
Evidence from a laboratory study. Organization Science 5(4): 554-568

Cohen MD, Burkhart R, Dosi G, Egidi M, Marengo L, Warglien M, Winter S. 1996. Routines
and other recurring action patterns of organizations: Contemporary research issues. Industrial
and Corporate Change 5(3): 653-698

Coriat B, Dosi G. 1998. Learning how to govern and learning how to solve problems: On the co-
evolution of competences, conflicts and organizational routines. In AD Chandler, P Hagstrom, O
Solvell (Eds.), The Dynamic Firm: The role of technology, strategy, organization and regions:
103-133. Oxford University Press: Oxford

Feldman MS, Pentland BT. 2003. Reconceptualizing organizational routines as a source of
flexibility and change. Administrative Science Quarterly 48: 94-118

Gersick CJG, Hackman JR. 1990. Habitual routines in task-performing groups. Organizational
Behavior and Human Decision Processes 47: 65-97

Hannan MT, Freeman J. 1984. Structural inertia and organizational change. American
Sociological Review 49(2): 149-164

Haunschild PR, Sullivan BN. 2002. Learning from complexity: Effects of prior accidents and
incidents on airlines' learning. Administrative Science Quarterly 47: 609-643

James W. 1914/1890. Habit. Henry Holt and Company: New York

Karim S, Mitchell W. 2000. Path-dependent and path-breaking change: Reconfiguring business
resources following acquisitions in the U.S. medical sector, 1978-1995. Strategic Management
Journal 21: 1061-1081

Kotabe M, Martin X, Domoto H. 2003. Gaining from vertical partnerships: Knowledge transfer,
relationship duration, and supplier performance improvement in the U.S. and Japanese
automotive industries. Strategic Management Journal 24: 293-316

Kruskal JB. 1983. An overview of sequence comparison. In D Sankoff, JB Kruskal (Eds.), Time
Warps, String Edits and Macromolecules: The theory and practice of sequence comparison: 1-
44. Addison-Wesley Publishing Company: Reading, MA

Levinthal D, Rerup C. 2006. Crossing an apparent chasm: Bridging mindful and less-mindful
perspectives on organizational learning. Organization Science 17(4): 502-513

March J, Simon H. 1958. Organizations. John Wiley and Sons: New York

March JG, Schulz M, Zhou X. 2000. The Dynamics of Rules: Change in written organizational
codes. Stanford University Press: Stanford, CA

Nelson RR, Winter SG. 1982. An Evolutionary Theory of Economic Change. Harvard University
Press: Cambridge

Pentland BT. 2003. Conceptualizing and measuring variety in the execution of organizational
work processes. Management Science 49(7): 857-870

Perrow. 1984. Normal Accidents: Living with high-risk technologies. Basic Books: New York.

Repenning NP, Sterman JD. 2002. Capability traps and self-confirming attribution errors in the
dynamics of process improvement. Administrative Science Quarterly 47: 265-295

Roth PL. 1994. Missing data: A conceptual review for applied psychologists. Personnel
Psychology 47: 537-560

Simon HA. 1976/1945. Administrative Behavior: A study of decision-making processes in
administrative organization. Third Edition. Free Press: New York

Sitkin SB. 1992. Learning through failure: The strategy of small losses. Research in
Organizational Behavior 14: 231-266

Stene EO. 1940. An approach to a science of administration. The American Political Science
Review 34(6): 1124-1137

Tucker AL, Edmondson AC. 2003. Why hospitals don't learn from failures: Organizational and
psychological dynamics that inhibit system change. California Management Review 45(2): 55-

Turner SF, Rindova V. Working paper. Co-organizing and re-organizing: Stabilization and
adaptation in routine functioning.

Weber M. 1991/1948. Bureaucracy. In HH Gerth, CW Mills (Eds.), From Max Weber : Essays
in sociology, New ed.: 196-244. Routledge: London ; New York

Weick KE. 1979. The Social Psychology of Organizing. McGraw-Hill, Inc.: New York. Second

Weick KE. 1987. Organizational culture as a source of high reliability. California Management
Review 29(2): 112-127

Weick KE, Roberts MJ. 1993. Collective mind in organizations: Heedful interrelating on flight
decks. Administrative Science Quarterly(38): 357-381

Wood W, Neal DT, Quinn JM. 2006. Repetition in everyday life: Habit prediction, change and
regulation. Working Paper

Wood W, Quinn JM, Kashy DA. 2002. Habits in everyday life: Thought, emotion, and action.
Journal of Personality and Social Psychology 83(6): 1281-1297

Wood W, Tam L, Witt MG. 2005. Changing circumstances, disrupting habits. Journal of
Personality and Social Psychology 88(6): 918-933

Zellmer-Bruhn ME. 2003. Interruptive events and team knowledge acquisition. Management
Science 49(4): 514-528

Zollo M, Reuer JJ, Singh H. 2002. Interorganizational routines and performance in strategic
alliances. Organization Science 13(6): 701-713

Figure 1. Routine Executions as Context Dependent and Context Creating

                                    Objective context                    Objective context

                                  • context dependence
                                  • context as supporting and
                                    prompting executions

 Executions of routine
                                  • context creating
                                  • executions as structuring
                                    and co-organizing

                                     Enacted context                      Enacted context

                                  Historical Executions                   Current Execution


Table 1. Variable Summary Statistics and Product-Moment Correlations (N = 273 routine-day executions)

     Variable                                              Mean      StdDev      Min        Max         1        2         3        4         5         6        7         8         9        10       11    12
 1   ServiceErrors                                           1.16      2.39      0.00       19.00     1.00
 2   CityWorkLoad                                          536.21    109.88      0.00      696.75     -0.07    1.00
 3   RouteWorkLoad                                         591.44    397.35      0.00     2318.00     0.07     0.26      1.00
 4   RouteAdditionalWorkLoad                               122.15    253.96      0.00     1472.00     -0.04    -0.03     -0.22    1.00
 5   RouteExperience                                         0.87      0.34      0.00        1.00     -0.20    -0.03     0.12     0.01      1.00
 6   PastStartTimeDiverge                                   23.03     29.25      0.67      152.00     0.03     0.13      -0.10    0.07      -0.03     1.00
 7   CurrentStartTimeDiverge                                21.71     27.11      1.33      207.00     0.04     0.00      -0.17    0.03      -0.01     0.51      1.00
 8   CityHoliday                                             0.11      0.31      0.00        1.00     0.24     -0.05     -0.03    -0.06     -0.11     0.14      0.05     1.00
 9   HistoricalVariantExecution                              0.64      0.12      0.34        0.88     -0.05    -0.03     -0.10    0.03      0.02      0.33      0.18     0.08      1.00
10   CurrentVariantExecution                                 0.65      0.12      0.34        0.99     0.18     -0.23     -0.25    0.06      -0.43     0.19      0.27     0.18      0.55      1.00
11   CityHoliday*CurrentVariantExecution                     0.08      0.22      0.00        0.93     0.24     -0.05     -0.05    -0.06     -0.13     0.16      0.07     0.99      0.10      0.22     1.00
12   HistoricalVariantExecution*CurrentVariantExecution      0.42      0.13      0.12        0.73     0.05     -0.15     -0.19    0.04      -0.22     0.32      0.27     0.15      0.89      0.86     0.19   1.00

     Variable                                             Variable Description
 1   ServiceErrors                                        Number of reported garbage collection misses or customer call requests
 2   CityWorkLoad                                         Average number of garbage collections made across all routes in the sample
 3   RouteWorkLoad                                        Number of within-route garbage collections made
 4   RouteAdditionalWorkLoad                              Number of outside-route garbage collections made
 5   RouteExperience                                      Employee experience on the route (1 if the employee has collected in this route within past 30 days, 0 otherwise)
 6   PastStartTimeDiverge                                 Absolute Value of the average differences in route execution start times within past 30 days
 7   CurrentStartTimeDiverge                              Absolute value of the difference between the current route execution start time and that of all historical route executions within past 30 days
 8   CityHoliday                                          Context Irregularity (1 if typical collection day is a city-recognized holiday, 0 otherwise)
 9   HistoricalVariantExecution                           Average of the Levenshtein Distance between all route executions within the past 30 days
10   CurrentVariantExecution                              Average Levenshtein distance between the current route execution and that of all historical route executions within past 30 days
11   CityHoliday*CurrentVariantExecution                  Interaction between City Holiday and Current Variant Execution
12   HistoricalVariantExecution*CurrentVariantExecution   Interaction between Historical Variant Execution and Current Variant Execution

Table 2. Poisson Regression Estimates for Service Errors, with fixed effects for the routes
N = 257 routine-day executions, with one set of route observations dropped due to invariance in service errors

                                                                        1                2          3            4            5            6            7
                                       Explanatory Variables          Coeff.           Coeff.     Coeff.       Coeff.       Coeff.       Coeff.       Coeff.

                CityWorkLoad                                         -0.002**         -0.002**   -0.002**     -0.002**     -0.002**     -0.002**     -0.002**
                RouteWorkLoad                                        0.001**          0.001**     0.000*      0.001**      0.001**      0.001**      0.001**
    Control     RouteAdditionalWorkLoad                               -0.001           -0.001     -0.001       -0.001       -0.001       -0.001*      -0.001*
   Variables    RouteExperience                                      -1.081**         -0.917**   -1.039**     -0.868**     -0.749**     -0.590**      -0.442*
                PastStartTimeDiverge                                   0.003           -0.001    0.007**        0.002       -0.002      0.009**       0.005*
                CurrentStartTimeDiverge                               0.004*          0.005**     0.004*        0.003        0.004        0.002        0.003
                CityHoliday                                                           1.077**                                1.462                     0.446
                HistoricalVariantExecution                                                       -2.977**                              12.264**     15.605**
  Hypotheses    CurrentVariantExecution                                                                       1.433*       1.354*      15.821**     18.311**
                CityHoliday*CurrentVariantExecution                                                                        -0.568                      0.925
                HistoricalVariantExecution*CurrentVariantExecution                                                                     -23.195**    -27.873**

                Model Log-likelihood                                 -409.95**    -385.630**     -399.77**   -408.139**   -384.110**   -382.629**   -357.286**

** p<0.01, * p<0.05 (one-tail tests)