Printer Friendly

It's all in the name: failure-induced learning by multiunit chains.

Multiunit chains are a conspicuous feature of the modern economy. Chains are collections of service organizations that produce similar goods and services in several markets and are linked together under common ownership into larger superorganizations that make considerable effort to standardize and coordinate the behavior of their components (Ingram and Baum, 1997a). Often, geographic location is the only difference among a chain's components. Chains proliferated during the twentieth century and have now penetrated deeply into nearly every service field in which customers have some direct contact with the organization (Bradach, 1997; Greve and Baum, 2001).

Despite their significance, only recently have multiunit organizations begun to receive regular attention in organization theory and strategic management. Much of the recent work on multiunit organizations draws on ideas from organizational and interorganizational learning. This is not surprising. Chains emphasize replicating and coordinating a standard set of routines or capabilities in multiple locations. This emphasis on leveraging a set of valuable routines by spawning similar components across multiple geographical markets invites discussion of learning and knowledge transfer. Through transfer learning among components, multiunit chains not only replicate but also improve their routines, resulting in learning curves within each component and spillover learning among them (Darr, Argote, and Epple, 1995; Argote, 1999; Ingram and Baum, 2001; Mitchell et al., 2002). Multiunit organizations are also conduits for diffusion, making them prone to early adoption of new practices and strategies, both because practices and strategies spread rapidly within them and because their presence in multiple markets provides greater exposure to innovations (Greve, 1995, 1996).

In addition to generating scale economies and reducing operating costs, chains' strategic emphasis on standardization helps reduce uncertainty about their components' quality by combining them under a common corporate umbrella (Ingram, 1996). Consistency across a chain's components raises consumers' perceptions of their reliability in producing a given quality of service repeatedly. Accountability is also higher, because interdependence pressures each component to maintain and enhance its chain's standards--poor quality service in any component can damage the entire chain's reputation. This constraint increases beliefs in the trustworthiness of chain components, reducing consumers' search and monitoring costs.

Although not all chains try to develop reputations, many do. Securing such benefits requires consumers being able to identify various components as members of a chain in which they trust. Chains that give their components names that link them to each other and to the chain thus meet a necessary condition for building a reputation (Ingram, 1996). Research has shown that common naming of components is vital to the growth and survival of entire chains (Ingram, 1996; Ingram and Baum, 1997b) and their individual components (Ingram and Baum, 1997a; Baum and Ingram, 1998), with chains naming their components in common outperforming those that do not. Notably, the benefit of common naming to a chain depends on the degree to which other chains in the same field are also using the strategy (Baum, 1999). Thus, the benefit of adopting common naming in a given setting depends on the diffusion of common naming strategies in that population.

Although research has demonstrated the benefits of common naming for multiunit chains, no study has yet attempted to explain what leads chains to adopt this strategy. Nor has research examined why, despite the apparent benefits of common naming, some chains employ an alternative "local" naming strategy, in which components' names identify them with their locations, rather than with the chain (Ingram, 1996). A learning perspective seems particularly germane to an explanation of chains' naming strategies. Name choices are made repeatedly by chains and so are likely to be subject to an experiential process in which chains learn by repeating naming choices that appear successful. Naming strategy also seems a likely candidate for interorganizational learning. The uncertainty resulting from a growing chain's inexperience with naming multiple components can be reduced, for example, by observing and imitating (or avoiding) the naming strategies of other chains based on their outcomes.

A fundamental mechanism facilitating learning is experience (Huber, 1991; Cyert and March, 1992). According to learning theory, organizations adjust their behavior based on their own past performance and the performance of other organizations (Haveman, 1993; Haunschild and Miner, 1997; Greve, 1998). Although past research has tended to emphasize learning from success (e.g., learning by doing in which cumulative experience reduces costs or improves performance; imitating the practices of other organizations that appear successful), we focus on chains' learning from their own and other chains' naming strategy failures. Chains can potentially learn about a range of practices from their failures, for example, viable locations, component management and strategy, and staffing policies. Our interest, however, is in whether one of the lessons they learn is about the differential effectiveness of the common and local naming strategies and, in particular, the operational and competitive advantages of emphasizing coordination of a standard set of routines or capabilities in multiple locations to achieve the scale economies and reputation benefits associated with common naming of components relative to the capabilities for local adaptation associated with the local naming of components.

Because organizational and strategy failures are salient and well-publicized events rich in information that matters for competitive strategy (Ingram and Baum, 1997b), decision makers attend to them naturally (Ocasio, 1997), comparing both their own and others' failures and successes to gain insight into what causes failure and to learn what not to do. While organizational performance measures often come without clear definitions of what outcomes are acceptable (Haunschild and Miner, 1997; Greve, 1998), significant failures--here, the failure of one or more chain components--represent unambiguous and salient indicators of poor performance. We analyzed the effects of component failures in an empirical study of the adoption of a common naming strategy by the 32 chain nursing homes that operated in Ontario from January 1971 to December 1996. At the start of 1971, seven nursing home chains operated in Ontario, three of which used common naming, with one naming all its components in common. By the end of 1996, 12 of the 18 nursing home chains operating in the province used common naming, with six naming all their components in common. This dynamic, combined with prior research showing the benefit of common naming for Ontario nursing home chains (Baum, 1999), make it a useful empirical setting for the current study.

LEARNING ABOUT STRATEGY FROM NAMING FAILURES

An organization's name is infused with meaning and reputation and identifies the organization to internal and external audiences (Glynn and Abzug, 1998). Because a name represents an identity, reputation, culture, and other characteristics that require resources and time to develop (Fombrun and Shanley, 1990), naming a multiunit chain's components is not an inconsequential, cosmetic event. Rather, a chain's naming strategies may greatly affect internal and external audiences' responses to and support for the chain and its components (Dutton and Dukerich, 1991). Naming strategy is critically important for multiunit chains because naming components in a way that identifies them with each other and with the chain increases the chain's credible commitment to customers (Ingram, 1996). In turn, this commitment can generate reputation benefits for the chain, particularly in health and human service sectors, where customers face uncertainty about the quality of services and products (Baum, 1999). Combined with chains' emphasis on standardization, common naming helps reduce uncertainty about a given component's quality (Ingram, 1996). Standardization raises consumers' confidence in the chain's ability to produce a given quality of service repeatedly across its components and their perceptions of its accountability, because interdependence among commonly named components puts pressure on each component to maintain and enhance its chain's standards--poor quality service in any component can damage the entire chain's reputation (Ingram, 1996). This interdependence increases customers' beliefs in the quality and trustworthiness of the chain, reducing their search and monitoring costs.

For some chains, the benefit of common naming derives from the chain's commitment to providing good service. For example, because travelers are unlikely to return to the same hotel repeatedly and are unable to gauge its service quality without prior experience, hotels have no incentive to provide good service in order to attract future business. Hotel chains that name their components similarly solve this problem by creating the opportunity for repeat business at different locations. Travelers who have stayed at one of a hotel chain's components will choose another of its components over a hotel with which they have no experience, provided, of course, that service quality at the chain's other component was satisfactory (Ingram, 1996).

In health and human service fields, the value of common naming stems from its role in fostering trust and signaling service quality to consumers. Trustworthiness is particularly important because consumers cannot evaluate the quality of complex services, either because they defy assessment and monitoring or because the beneficiaries of the services are unreliable witnesses to service quality (e.g., a nursing home resident with Alzheimer's disease). Under such conditions, consumers look for other signs of quality and trustworthiness, such as credible commitments to a standard, for example, in the form of certification (DiMaggio and Powell, 1983; Baum and Oliver, 1991) or, in the case of chains, common naming. Thus, common naming not only fosters commitment and loyalty when there is a possibility of repeated interactions but also beliefs about quality and trustworthiness in situations in which consumers have great difficulty ascertaining what good choices are (Ingram, 1996).

Common naming has been shown to reduce dramatically the failure of U.S. hotel chains and their components (Ingram, 1996; Ingrain and Baum, 1997a, 1997b) and Ontario nursing home chain components (Baum, 1999). Ingram and Baum (1997a) found that a hotel named in common with 50 percent of the other hotels belonging to the chain had a failure rate 85 percent lower than that of a hotel that belonged to a chain that did not name any of its components in common. Similarly, Baum (1999) reported that a component nursing home that shared a common name with 50 percent of the other homes in its chain was 80 percent less likely to fail than a comparable component that did not share a common name with any other member of its chain.

Despite the apparent benefits of common naming, some multiunit chains adopt a local naming strategy, naming components to identify them with local markets rather than with the chain (Ingrain, 1996). Adaptation to local market conditions, even on a dimension such as the name, can allow components to avoid the perception of chains as large, cold, impersonal bureaucracies, an image inconsistent with quality service. This has led some multiunit chains (e.g., hotels, nursing homes, funeral homes, day care centers) to hide the fact that their components are part of a chain, instead, naming their components in ways that identify them with the community in which the component is located and making appropriate local market adjustments to each component. The choice between these two naming strategies corresponds to a more general strategic tension: whether to adopt strategies that allow for flexibility and adaptation to local conditions or to adopt strategies that enhance consistency across locations. Learning theory suggests that chains make this choice by drawing lessons from their own and other chains' experience with naming strategies.

Learning theorists characterize organizations as experiential learning systems that adapt incrementally to past experience (Levitt and March, 1988; Cyert and March, 1992). At the organizational level, boundedly rational decision makers attend to their own past experience, particularly failures, to reduce uncertainty and the cost of search. At the interorganizational level, facing insufficient information to learn from their own experience, the same boundedly rational decision makers attend to the actions and outcomes of other organizations to reduce their uncertainty and determine their course of action (Miner and Haunschild, 1995; Haunschild and Miner, 1997).

Learning from the Failure of One's Own Strategy

Experience is one of the fundamental mechanisms facilitating organizational learning (Huber, 1991; Cyert and March, 1992). Organizations adjust their behavior based on their goals and past performance. Such learning by doing is widely held to be a source of organizational knowledge, capabilities, and improved organizational performance (Argote, Beckman, and Epple, 1990). Through such learning by doing, multiunit chains' decision makers may gain insight into the performance consequences of their naming strategies and adjust them accordingly.

Learning theory suggests that change is triggered by performance below aspirations: satisfactory or superior outcomes tend to result in the reinforcement of lessons drawn from earlier experience, while unsatisfactory outcomes call existing practices and strategies into question (Levitt and March, 1988; Cyert and March, 1992; Lant and Mezias, 1992). Failure upsets the status quo, draws attention to potential problems, and stimulates a search for possible solutions. Failures are thus vital engines for change, initiating exploration of new practices, strategies, and courses of action, rather than reinforcing continued use and refinement of current ones. A range of studies provides evidence that poor performance leads to organizational change (e.g., Lant, Milliken, and Batra, 1992), while satisfactory and superior performance does not (e.g., Singh, 1986; March and Shapira, 1992). Greve (1998), for example, showed that U.S. radio station chains were more likely to change their market strategies when their performance was below their aspirations.

Many measures of organizational performance do not have clear definitions of what outcome level is acceptable (Greve, 1998). For chains, however, an unequivocal and salient measure of poor chain performance is the failure of one or more of its components. Component failures differ from performance indicators that are internally constructed because the rules used to assess performance cannot be easily changed to provide feedback that can be interpreted complacently (Greve, 1998). Given this measure, the forgoing arguments suggest that the failure of a chain's component is a salient indicator of poor performance that can motivate the chain to reevaluate its strategy. Consequently, a chain's own commonly named component failures should lead the chain's decision makers to reexamine the common naming strategy, stimulating exploratory search and leading the chain to decrease its use of common naming in favor of local naming among its surviving components. Failure of a chain's locally named component would have the opposite impact. Component failures thus facilitate recognition and interpretation of otherwise ambiguous outcomes (Sitkin, 1992), providing a chain's management with information on what not to do, as well as the potential value of the alternative naming strategy that did not fail. In this sense, a chain's components provide the chain with opportunities for ongoing experiments to assess the outcomes associated with different naming strategies (Audia, Sorenson, and Hage, 2001). These ideas suggest the following hypotheses:

Hypothesis 1a (H1a): Failures of a chain's own commonly named components will decrease its use of common naming.

Hypothesis 1b (H1b): Failures of a chain's own locally named components will increase its use of common naming.

Learning from the Failure of Others' Strategy

Organizations may also learn by observing the experiences of others in their field. Organizational learning theorists have long contended that, like individuals, organizations can learn vicariously, imitating or avoiding specific strategies or practices based on their perceived impact (Levitt and March, 1988; Cyert and March, 1992). Learning from performance experience may thus often occur in organizational populations as well as in individual organizations (Miner and Haunschild, 1995).

By observing other organizations, decision makers can potentially learn a range of strategies, practices, and technologies produced by the ongoing explorations of others in their industry (Levitt and March, 1988; Levinthal and March, 1993). Faced with insufficient information to learn from their own experience, decision makers can use this learning mode to reduce uncertainty by turning to other organizations' actions and experiences for clues about how to interpret their own situation and act. Decision makers attend to the outcomes of particular strategies and practices, imitating those that appear beneficial for other organizations and avoiding those that appear harmful, a mode of learning that Haunschild and Miner (1997) have termed outcome-based imitation. Conell and Cohn (1995), for example, showed that the success of strikes increased the chances of French coal miners striking, while Haunschild and Miner (1997) showed that very low acquisition premiums increased the chances of an acquirer selecting particular investment bankers associated with that favorable outcome. More indirect evidence comes from studies suggesting that nuclear power plants (Zimmerman, 1982), shipyards (Argote, Beckman, and Epple, 1990), and hotels (Baum and Ingram, 1998) starting later imitate the successful practices of earlier entrants.

Organizations may not learn from all outcomes, however, attending selectively to very visible or salient ones (March, Sproull, and Tamuz, 1991), and outcome-based imitation may only be sustained by highly salient outcomes (Haunschild and Miner, 1997). Thus, while the successes of a strategy or practice can be used to gauge its value, its failures may prove even more salient and informative (Ingram and Baum, 1997b). In the same way that the failure of a chain's own components can trigger a change in strategy, the failure of other chains' components may lead a chain to adopt an alternative strategy to avoid experiencing the same fate itself (Miner et al., 1999). A focus on learning from others' failures may also reduce blind imitation of successful organizations' strategies without attention to context (Sitkin, 1992). Thus, a chain's observation of failures of other chains' commonly named components may serve to discredit the common naming strategy, leading the chain to decrease its use of the strategy. Failure of other chains' locally named components would have the opposite effect. These ideas suggest the following hypotheses:

Hypothesis 2a (H2a): Failures of other chains' commonly named components will decrease a focal chain's use of common naming.

Hypothesis 2b (H2b): Failures of other chains' locally named components will increase a focal chain's use of common naming.

Learning from the Failure of One's Own and Others' Strategy

To this point, we have considered a chain's own and others' strategy failure experience separately and independently, but chains' decision makers may consider and react to these two strategy performance indicators simultaneously and interactively. If they do, consistency in the two indicators should produce a reinforcing effect, resulting in a more pronounced, interactive effect on naming strategy. Consequently, when combined with the failure of other chains' commonly named components, a chain's own commonly named component failures may create doubt about the strategy. Failures of other chains' locally named components, combined with failure of a chain's own locally named components, would result in the opposite. Therefore, we hypothesize the following interactions between the failure of a chain's own and others' component failures:

Hypothesis 3a (H3a): Combined, a failure of a chain's own and other chains' commonly named components will decrease its use of common naming more than either alone.

Hypothesis 3b (H3b): Combined, failures of a chain's own and other chains' locally named components will increase its use of common naming more than either alone.

Organizational Momentum and Learning from Failure

The preceding hypotheses assume that the effects of failures of one's own and others' components on a chain's naming strategy do not depend on the chain's current naming strategy, but research suggests that a chain's decision makers can become bound up in their experience with a strategy, influencing how they respond to a failure of that strategy (Amburgey and Miner, 1992). Although organizational decision makers cannot initially be certain of the outcomes of their actions, with repetition they gain experience and confidence, and uncertainties lessen as their understanding and capabilities develop. Given initial success with a strategy (i.e., few immediate negative consequences), decision makers are likely to repeat it, because they know increasingly well how to and because it is less risky and more rewarding to repeat a strategy than to try out potentially superior alternatives with which they have limited experience (Levitt and March, 1988). The consequences of changing are thus usually less well known than the consequences of not changing, and as a result, organizations do not necessarily change when performance feedback suggests they should. Learning by doing thus improves an organization's competencies in its current activities but draws attention and resources away from innovation and change (March, 1991). Over time, the more an organization engages in a particular strategy, the more that strategy becomes institutionalized, embedded in routines and political structures, making it less likely to be questioned or challenged and even more difficult to change (Pfeifer and Salancik, 1978; Hannan and Freeman, 1984).

Decision makers' attributions of success to their own abilities, and to the strategies and practices (correct or not) they previously adopted, further limit the likelihood that they will initiate experimentation. This is especially true if there are few negative results, a situation that increases the likelihood that any false or superstitious beliefs that decision makers hold will be reinforced and that their learning will be self-serving, or biased toward confirmation (Miller, 1990, 1999; Levinthal and March, 1993). Thus, once experience with a strategy has accumulated, advocates of the strategy are less likely to interpret failures as an indication that the strategy is incorrect than as an indication that the strategy has not been pursued vigorously enough (Levitt and March, 1988).

This bias toward strategic persistence creates organizational momentum, the tendency to maintain the direction and emphasis of prior choices and actions in current behavior (Miller and Freisen, 1980). Empirical studies provide robust evidence of organizations' tendency to stick to strategies and courses of action they employed in the past (Kelly and Amburgey, 1991; Amburgey and Miner, 1992). As a result, organizations experienced with a given strategy tend to persist in that strategy, despite evidence that they should not in which case the organization suffers 'from the so-called competency trap (Levitt and March, 1988). Multiunit chains' emphasis on the diffusion of standard practices across components makes them especially prone to organizational momentum and potential competency traps by creating cost structures and incentives that favor replication of practices and strategies adopted early on (Baum, Li, and Usher, 2000; Greve and Baum, 2001).

The tendency toward organizational momentum in a chain's strategic action suggests that its decision makers will interpret the failure of components through the lens of their experience. Given initial success with local naming, a chain's decision makers are likely to repeat it, as their confidence in the strategy grows and it becomes increasingly risky and costly to experiment with common naming, with which they have limited experience. As a chain's use of local naming expands, the chain's decision makers will react less to failures of its own locally named components, as their experience and commitments lead them to interpret the failures not as a challenge to their chosen strategy but as an indication that they need to pursue it with greater resolve. In contrast, the chain's decision makers will react increasingly strongly to failures of their own commonly named components, which they take not only as evidence of the inefficacy of the common naming but also as proof (rightly or wrongly) of the correctness of their chosen strategy. Therefore, we hypothesize:

Hypothesis 4a (H4a): The greater a chain's use of local naming, the more the chain will decrease its use of common naming in response to failures of its own commonly named components.

Hypothesis 4b (H4b): The greater a chain's use of local naming, the less the chain will increase its use of common naming in response to failures of its locally named components.

The momentum of a naming strategy may similarly influence how a chain responds to the failures of other chains' components. Chains experienced with local naming should be less likely to increase their use of common naming in response to other chains' locally named component failures, as their decision makers interpret the failures not as an indication that their chosen strategy is incorrect but as an indication that the other chains did not pursue the strategy vigorously enough. In contrast, they should be more likely to decrease their use of common naming in response to the failure of other chains' commonly named components, which provides further impetus for doing so by discrediting common naming while (implicitly) corroborating their past choices. Therefore, we hypothesize:

Hypothesis 5a (H5a): The greater a chain's use of local naming, the more the chain will decrease its use of common naming in response to failures of other chains' commonly named components.

Hypothesis 5b (H5b): The greater a chain's use of local naming, the less the chain will increase its use of common naming in response to failures of other chains' locally named components.

Support for H4a-5b would contribute importantly to an explanation for why, despite its apparent superiority, not all chains adopt common naming: a chain's experience with local naming biases it against the adoption of common naming, even in the face of evidence that local naming is an inferior strategy. This is an example of a competency trap in which a chain's use of the local naming strategy prevents its decision makers from accumulating sufficient experience with the superior common naming strategy to adopt it without incurring a loss (possibly short term) in performance (Levitt and March, 1988; Fang and Levinthal, 2002).

METHODS

To test our hypotheses, we collected data on all 557 nursing homes that operated in Ontario at any time between January 1971 and December 1996. The data were compiled from two archival sources: the Ontario Ministry of Health (MOH) licensing records and the Ontario Hospitals' Association (OHA) Directory. The data cover the period in which chains first began to appear in Ontario and after April 1972, when the Ontario Ministry of Health (MOH) extended public insurance for nursing home residents under the Ontario Health Insurance Plan and took regulatory control of licensing nursing homes and setting fees paid for their services.

In 1971, there were 479 nursing homes operating in Ontario. Among them were 32 components of seven chains, representing 6 percent of all nursing homes operating at the time. Thus, the life histories for these seven chains founded before the study period were left-censored. During the observation period, chains' share of nursing home beds in Ontario increased from less than 5 percent to more than 50 percent. The chains grew primarily through acquisition, acquiring 170 homes (158 independent, 12 chain components) while founding only 17. Costs of licensure and the advent of public extended-care insurance appear to have triggered this shift from independent to chain ownership (Tarman, 1990; Baum, 1999). The passage of Medicare and Medicaid is thought to have had a similar impact on the U.S. nursing home industry (Light, 1986).

During the study period, 32 different chains operated in the province. In 1971, three of the seven nursing home chains operating in Ontario used common naming, with one naming all its components in common. By the end of 1996, 12 of the 18 nursing home chains operating in the province had adopted common naming, with six naming all their components in common. Forty-three components failed during the study period; 15 had common names, 28 had local names. A nursing home was defined as founded in the year it was first licensed and as failing in the year it was de-licensed by the MOH. Licensing records also indicate the year of founding for homes established prior to 1971.

Dependent Variable and Analysis

Our theoretical arguments and hypotheses focus on the likelihood of a chain increasing or decreasing its use of the common naming strategy. Therefore, we defined our dependent variable, change in common naming, as the yearly net change in the number of a chain's components that shared a name that identified them with each other and the chain. We defined the variable at the chain rather than component level of analysis because our interest was in chains' naming strategies. Net change was computed based on existing components' name changes and the names given to newly founded or acquired components during a year, thus capturing the chain's overall increase or decrease in common naming. (1) A positive value indicates a net increase in the number of components sharing a common name, a negative value, a net decrease.

For the analysis, we pooled the yearly chain data and estimated a single model on the pooled cross sections using time series regression. Each chain is represented in the sample for the years in which it operated in Ontario. The sample contained 443 yearly observations on 32 different chains. Pooling repeated observations on each chain is likely to violate the assumption of observation independence, leading to autocorrelation in the model's residuals and incorrect variance estimates, which renders OLS estimates biased and inefficient. To obtain unbiased and efficient estimates, we used random-effects GLS models, which correct for autocorrelation of disturbances (Greene, 2000).

Independent Variables

All independent variables, defined below, were updated annually and lagged one period to avoid simultaneity problems. Lagging the independent (and control) variables reduced the effective sample size to 411 observations.

Own strategy performance experience. To test H1a-1b, we constructed two variables to capture a chain's recent strategy performance experience. Rather than simply counting each chain's number of component failures, we weighted component failures by their size to capture differences in their magnitude and salience. (2) Own commonly named failures and own locally named failures were therefore defined, respectively, as the focal chain's number of commonly and locally named component failures, weighted by the number of beds in each component. Thus, the value of the variables is equal to the total number of beds in the failed components. Support for H1a requires a negative coefficient for own commonly named failures and H1b a positive coefficient for own locally named failures.

Others' strategy performance experience. To test H2a-2b, we computed a second, analogous pair of variables to capture other chains' strategy performance experience. The two variables, others' commonly named failures and others' locally named failures, were defined, respectively, as the number of commonly and locally named component failures experienced by chains other than the focal chain, weighted by the number of beds in each failed component. The value of the variables is thus equivalent to the total number of beds in other chains' failed components. For these variables, the size weighting is consistent with research on trait-based imitation, which suggests that decision makers are more influenced by actions of and events affecting large organizations because of their greater salience and significance (e.g., Haunschild and Miner, 1997; Baum, Li, and Usher, 2000). (3) Support for H2a requires a negative coefficient for others' commonly named failures, while a positive coefficient for others' locally named failures supports H2b.

Own and others' strategy performance experience. To test H3a-3b, we created two own x others' interaction terms, one for commonly named component failures and one for locally named component failures. H3a predicted a negative coefficient for the commonly named component failure interaction; H3b predicted a positive coefficient for the locally named component failure interaction.

Own chain's strategy use. To test H4a-4b and H5a-5b, we measured a chain's naming strategy experience, with own local name concentration defined as the proportion of a chain's components that did not share a common name in the prior year. Own local name concentration ranges from zero, indicating all components are commonly named, to one, indicating all components are locally named. To test our momentum hypotheses, we computed four interactions: (1) Own local name concentration x Own commonly named failures, (2) Own local name concentration x Own locally named failures, (3) Own local name concentration x Others' commonly named failures, and (4) Own local name concentration x Others' locally named failures. Negative coefficients for interactions 1 and 3 support H4a and H5a. Negative coefficients for interactions 2 and 4 support H4b and H5b. We included own local name concentration in the models to control for the main effect of organizational momentum and obtain unbiased coefficients for the interactions (Cohen and Cohen, 1983).

Control Variables

We controlled for a variety of features of nursing home chains and their context, as many other factors than those we theorized may influence a chain's naming strategy. All the control variables were updated annually and lagged one year for the analysis.

Nursing home chain characteristics. Because age and size of a nursing home chain may affect its likelihood of strategic change (Hannah and Freeman, 1984), we controlled for the effects of each chain's age and size on change in naming strategy. We defined chain age as the number of months the chain had been in operation. We defined chain size as the number of components a chain operated in a given year. (4) A chain's growth may also provide an opportunity and impetus to change its naming strategy, making it important to distinguish empirically the effects of chain growth on the adoption of common naming from the effects of component failures. Because the nursing home chains in our sample grew primarily through acquisitions, we controlled for the number of acquisitions a chain made each year to further rule out the possibility that the component-failure variables would spuriously capture the chain growth process.

We also controlled for the effects of a chain's spatial strategy on its naming strategy, since chains with more widely dispersed components may tend to adopt a different naming strategy than those with more proximate components. We measured the proximity of a chain's components using spatial compactness, defined as the average Euclidean distance (based on the latitude and longitude of a chain's components' postal addresses) between any pair of the chain's components.

As multiunit organizations, nursing home chains likely consider the implications of the adoption of the common naming strategy for their multimarket relationships with other nursing home chains--in the same way that common naming identifies the chain's components to consumers, it also identifies them to competitors. Therefore, we controlled for the effect of a chain's multimarket contact on its naming strategy. We defined a chain's multimarket contact as the total number of multimarket competitors (i.e., a competitor the chain meets in more than one market) that the chain meets in each of the local markets in which it operates in a given year. We defined local markets using provincial census district boundaries to approximate local markets for nursing home services. Our approach is analogous to the use of county boundaries to represent local markets for healthcare services in the U.S. (e.g., Banaszak-Holl, Zinn, and Mor, 1996), which has been found to be a reasonable approximation for nursing home markets (Gertler, 1989).

Although common and local naming were the two dominant naming strategies, three chains in our sample adopted a hybrid naming strategy, identifying their components with both their chains and specific component locations (e.g., Extendicare-Ottawa). Only one chain, Extendicare, used this hybrid strategy fully for the entire observation period; the others adopted it only for a subset of their components and for a relatively short time. All three appear to have adopted hybrid naming primarily to distinguish neighboring components and not to gain benefits of identification with a local community. They did not appear to make any substantive local adaptations to their chain-wide standards. Given the small number of hybrid-named components in our sample and their lack of local adaptation, we considered these components to be commonly named. We nevertheless included a time-varying dummy variable, hybrid naming, coded one for all chains with one or more hybrid-named components in a given year, and zero otherwise, to account for any differences in naming strategies of chains that did or did not use hybrid names.

Finally, we included a dummy variable, left censored, to examine the possibility that the seven left-censored chains operating prior to 1971 exhibited systematically different naming behavior, perhaps because of their early pioneering roles in the industry (Baum, 1999).

Environmental characteristics. We controlled for several factors influencing competition and the demand for nursing home services, which may influence a chain's naming strategy. These variables included chain density, defined as the number of chains in a given year, component density, defined as the number of components operated by all chains in a given year, and independent nursing home density, defined as the number of independent (i.e., non-chain) nursing homes operating in a given year, to control for possible effects of competition among chains and with independent nursing homes. Since the effects of competition are relative to demand, we also included the provincial census population aged 65 and over in each year.

Other chains' use of common naming may also influence a chain's naming strategy. Since other organizations' choices and strategies suggest that they have positive opinions of those strategies, observers' evaluations of options are influenced by their choices and actions. Thus, more frequent use provides information that others find a particular practice or strategy valuable, which, in turn, may increase the likelihood of its use (Haunschild and Miner, 1997). We therefore controlled for other chains' common name concentration as the proportion of all chains' components, excluding the focal chain, that shared a common name. Research also suggests that organizations imitate the successful practices of established organizations at the time of their entry (Zimmerman, 1982; Argote, Beckman, and Epple, 1990; Baum and Ingram, 1998), which several theorists refer to as congenital learning (e.g., Huber, 1991). Therefore, we also controlled for other chains' common name concentration at founding, defined as other chains' common name concentration in the year prior to a chain's entry, to account for the possibility that chains entering when other chains were making use of common naming would be more likely to adopt the strategy.

We also controlled for the extended care per diem, in 1986 dollars, which sets the basic fee paid for nursing home services covered under the provincial extended care plan. Nursing homes are required to provide extended care for at least 75 percent of their beds and are prevented by law from offsetting the cost of providing extended care services by charging their residents above the provincial government's fee schedule. As a result, the per diem is an important determinant of revenues and profitability. Lastly, to ensure that our main findings were not simply a result of unobserved factors related to the passage of time, we included a time trend control, calendar time.

Descriptive statistics for the independent and control variables are given in table 1. Although correlations among several of the control variables are quite high, correlations between the control and theoretical variables and among the theoretical variables themselves are small in magnitude (the largest is .31 or 9.6 percent shared variance). Such levels of multicollinearity among explanatory variables can result in larger standard errors for the highly correlated variables but will not bias parameter estimates. So, although this does not pose a serious estimation problem, it can make it difficult to draw inferences about the effects of adding specific variables to the models. Therefore, when estimating results, we followed a strategy of estimating hierarchically nested models to check that multicollinearity was not causing less precise parameter estimates (Kennedy, 1992).

RESULTS

Table 2 presents estimates for random-effects generalized least squares models of nursing home chains' change in common naming strategy and gives log-likelihood-ratio test statistics, which compare the fit of nested models. Model 1 presents a baseline model, and models 2-7 test our hypotheses. The baseline model includes lagged chain characteristics as well as the environmental control variables.

Model 2, which adds variables to test for effects of a chain's own strategy performance experience (H1a and H1b), is a significant improvement over model 1. As predicted in H1a, there is a significant negative coefficient for own commonly named failures. Thus, the larger the number of commonly named component failures a chain experienced in the prior year, the more it decreased its use of common naming in the current year. H1b is also supported by the significant positive coefficient for own locally named failures, which indicates that chains experiencing more locally named component failures in the prior year increased their use of common naming to a greater extent in the current year.

Model 3 adds other chains' strategy performance experience variables to test H2a and H2b; however, neither was supported. The coefficient for others' locally named failures is not significant, and the coefficient for others' commonly named failures is positive and significant (p < .10), opposite to our prediction. Model 4 adds Own x Other strategy performance experience interactions for both commonly and locally named failures to test H3a and H3b. Again, however, neither is supported. The coefficient for Own x Others' commonly named failures is positive and significant, opposite to our prediction in H3a. And Own x Others' locally named failures is not significant, failing to support H3b.

To test H4a and H4b, model 5 adds interactions between the chain's own use of local naming (own local name concentration) and the variables for own strategy performance experience. H4a, which predicted that organizational momentum would lead the chains with local naming experience to react strongly to failures of their own commonly named components by abandoning common naming, is not supported. H4b is supported, however, by the significant negative coefficient for Own name concentration x Own locally named failures. This interaction counteracts the positive main effect of own locally named failures, indicating that a chain experienced in local naming increased its use of common naming less in response to failures of its own locally named components than a chain experienced in common naming. Thus, chains' experience with local naming biased them against adopting common naming, even in the face of evidence indicating the inferiority of local naming, contributing to an explanation for why not all chains adopted common naming.

Model 6 includes interactions between a chain's own use of common naming (own local name concentration) and variables for others' component failures, to test H5a and H5b, which predicted that a chain's use of local naming would decrease its use of common naming more in response to other commonly named component failures (H5a) but increase its use of common naming less in response to failures of other chains' locally named components (H5b). Only the first of these predictions is supported. The significant negative coefficient for Own local name concentration x Others' commonly named failures indicates that chains experienced with local naming decreased their use of the common naming strategy more in response to others' failure with it. Again, chains' experience with local naming biased them against adopting common naming.

In model 7, we removed the two insignificant interaction terms from model 6 to check that the two significant interactions (H4b and H5a) provided a significant improvement over model 4. As the likelihood ratio test statistic given in table 2 shows, they do. Two additional results are notable in model 7. First, adding the Own local name concentration x Others' commonly named failures interaction increased the significance level of the positive main effect of others' commonly named failures, opposite to H2a. Together with the interaction, this main effect suggests a very strong momentum bias: the greater a chain's use of common naming (and so the smaller the weight of the interaction), the more the chain increased its use of common naming when other chains failed at the strategy. Chains' experience with common naming thus appears not only to have led their decision makers to interpret other chains' common naming failures as an indication that the others had not pursued the strategy vigorously enough, but also as an opportunity to take advantage of competitors' failure to execute the strategy effectively. This suggests that competitive dynamics may interact with organizational learning processes in the naming strategy adoption process (Greve, 1995).

Second, the coefficient for the Own x Others' locally named failures interaction, which became significant and negative (p < .10) after adding the Own local name concentration x Own locally named failures interaction in model 5, is now significant at p < .05. Thus, the estimates in model 7 run counter to both H3a and H3b, which predicted that own and others' naming strategy performance experience would mutually reinforce a chain's tendency to adopt the common naming strategy. These interactions indicate that, although responding as anticipated in H1 and H1 b to their own naming strategy failures, a chain's reaction to its own component failures weakened when other chains had similar experiences. A chain increased its use of common naming more after a failure of one of its own commonly named components if other chains also experienced such failures. Conversely, a chain increased its use of common naming less after experiencing a failure of one of its own locally named components when other chains also experienced such failures. Thus, the failure of a chain's own commonly and locally named components appears to become more tolerable when other chains are also experiencing component failure of the same type.

Several effects of the control variables are worth mentioning. The main effect of own local name concentration is negative and significant, reinforcing the role of strategic momentum in the strategy adoption process. The positive coefficient of chain size indicates that large chains tend to use the common naming strategy to link their components to each other and to the chains to build up the chain's reputation for its components. Somewhat surprisingly, then, the coefficient for number of acquisitions is negative. Inspection of our data suggests that this finding reflects a lag in the renaming of newly acquired components. This may occur because, wanting to avoid negative chain-wide repercussions of operating a poorly performing component, chains pursuing the common naming strategy wait until an acquired component's facilities and operations have been brought in line with the chain's standards before changing its name to identify it with the chain. (5)

To illustrate how the main and interaction effects influence a chain's naming strategy and observe their magnitudes, we summarized the findings graphically in figure 1. Using estimates from model 7 in table 2, we considered a chain experiencing combinations of commonly and locally named component failures ranging from 0/1 (i.e., own commonly named failures = 0 and own locally named failures = 1) to 1/0 (i.e., own commonly named failures = 1 and own locally named failures = 0). Line H1a-1b shows the estimated change in a chain's number of commonly named components as a function of its own commonly and locally named component failures. As the line's downward-right slope indicates, the greater the value of a chain's own commonly named failures in the prior year, the larger the decrease in its use of common naming in the current year, and the greater the value of a chain's own locally named failures, the larger the increase in its use of common naming. Across the range of commonly and locally named failures, the estimated change in a chain's number of commonly named components ranges from an increase of 3 to a decrease of 2.5, which is large relative to 6.1, the mean (s. d. = 5.1) number of components for the sample chains.

Using line H1a-1b as a baseline, line H4b adds the effect of the Own name concentration x Own locally named failures interaction, with own name concentration set to .5 (i.e., half a chain's component share a common name). The flatter slope of line H4b shows how a chain's experience with local naming blunts the impact of a chain's own locally named failures on its use of common naming. At its maximum, the interaction cuts the estimated change in the number of commonly named components 50 percent, from 3 to 1.5.

Using line 4b as a baseline, lines H3a and H3b add the unanticipated moderating effects of Own x Others' commonly named failures and Own x Others' locally named failures, respectively, with others' commonly/locally named failures both set equal to two. Line H3a increases relative to line H4b as a chain's own commonly named failures increase, illustrating that in the presence of other chains' commonly named component failures, a chain's own commonly named component failures lead to smaller reductions in its use of common naming. In contrast, line H3b decreases relative to line H4b as a chain's own locally named failures increase, showing how in the presence of other chains' locally named component failures, a chain's own locally named component failures result in smaller increases in its use of common naming. Again, the interaction effects are substantial. At its maximum, H3a lowers the estimated decrease in number of commonly named components from 2.5 to just over 1.0; H3b lowers the estimated increase from 1.5 to less than 0.5 at its maximum.

Finally, again using line H4b as a baseline, line H5a shows the effect of Own name concentration x Others' commonly named failures, with others' commonly named failures set to two. In contrast to the other interaction effects, this interaction shifts the line downward (by roughly 0.5 components) but does not alter its slope. Thus, the interaction serves to increase (or decrease) the estimated decrease (or increase) in common naming.

DISCUSSION AND CONCLUSION

Although past research has demonstrated that the common naming strategy greatly enhances the survival chances of chains and their components, no study had yet explored what leads chains to adopt (or not) common naming. To better understand chains' naming strategies, we theorized and modeled their adoption of the common naming strategy as a product of organizational and interorganizational learning from failure. Our theoretical focus was on how a chain's own and other chains' commonly and locally named component failures, and their interactions with each other and the chain's naming strategy experience, influenced the naming strategy adoption process. Two fundamental results in particular emerge from our empirical analysis. One is that the nursing home chains in our sample learned from their own and others' failures, and the second is that they learned less from such failures when they had a historical investment in the failing strategy. The analysis also yields several counterintuitive findings and nonfindings that are interesting and instructive. As an early investigation into organizational learning from failure, however, there is much to take from this study.

Although past research has shown how a bank holding company's divestment of related (or unrelated) acquisitions lowers its future related (or unrelated) acquisition rate (Ginsberg and Baum, 1994), how learning from the failure of other organizations enhances survival chances for organizations observing them (Ingram and Baum, 1997b; Kim, 2000), and how an airline's accidents can contribute to reducing its future accident rate (Haunschild and Ni, 2002), little is known about the role of learning from failure in the strategy adoption process. We can now say much more about the role of failure-induced learning in strategy adoption and, in particular, why Ontario nursing home chains adopted (or did not) the common naming strategy. Our findings reveal a multifaceted learning process influenced by contingencies that can enhance or detract from a chain's ability to learn from failure. These contingencies include the interaction of a chain's own and other chains' component failures, which appears to have led decision makers to adjust their aspirations in ways we did not anticipate, and organizational momentum, which biased decision makers' interpretation of failures in favor of their own prior choices. Thus, while prior research emphasizes knowledge transfer and rapid diffusion as sources of learning advantage for multiunit chains, our findings also reveal complications arising from organizational momentum and competency traps that may detract from a chain's ability to learn, complications to which chains may be especially prone because of their emphasis on the replication and diffusion of practices and strategies.

Our results are broadly consistent with several well-established learning theory ideas (e.g., experiential and vicarious learning and organizational momentum) but also offer insight into learning research by providing some of the first systematic evidence that highly salient negative outcomes (i.e., component failures) can produce learning at organizational and interorganizational levels. We found clear influences of experiential learning on Ontario nursing home chains' naming strategies. As predicted, chains that experienced failure with the common naming strategy during a given year lowered their use of common naming in the next year (H1a). In contrast, chains that experienced failure with the local naming strategy during a given year increased their use of common naming in the following year (H1b).

Interorganizational learning processes had a more complex influence on the naming strategy adoption process. We found no support for H2a and H2b, which predicted parallel effects to H1a and H1b of other chains' component failures on a chain's naming strategy. Instead, our estimates are consistent with the possibility that managers of chains experienced with common naming--and thus likely to comprehend the value of the strategy--view others' difficulties as a chance to gain a competitive advantage. While we conceptualized strategy adoption as a function of organizational learning processes, the strategy adoption process may thus also be driven by competitive dynamics (Greve, 1995). Future research might therefore fruitfully examine more closely the interplay between organizational learning and competitive dynamics in the strategy adoption process.

Other chains' failures also affected a chain's naming strategy through their interactions with the chain's own component failures. A chain was less likely to decrease its use of common naming in response to its own commonly named component failures if other chains also experienced such failures and, conversely, was less likely to increase its use of common naming in response to its own locally named component failures if other chains also experienced such failures. A chain's own failures appeared to be less of a failure when other chains experienced similar failures. Although counter to our predictions (H3a and H3b), these findings are consistent with the idea that, through a process of social comparison, organizations' decision makers adjust their performance aspiration levels downward in response to the poor performance of other organizations (e.g., Festinger, 1954; Herriott, Levinthal, and March, 1985; Greve, 1998). Chains' decision makers thus seem to interpret and react to their own performance failures based in part on a social comparison process (e.g., Festinger, 1954; Porac et al., 1995) in which other chains' performance failures serve as a basis for determining appropriate aspiration levels for themselves. Such aspiration adjustments may help keep chains from decreasing their use of the common naming strategy too quickly in response to idiosyncratic failures of commonly named components. But they can also contribute to chains persisting with the local naming strategy despite its inferiority. Although our interpretation is speculative, linking the failure-induced learning perspective to aspiration-level adjustment models seems a promising direction for future research.

Interorganizational learning processes also depended on a chain's own use of the common naming strategy, which we anticipated based on the idea of organizational momentum. Chains experienced with local naming increased their use of common naming less in response to their own locally named component failures (H4b) and decreased their use of common naming more in response to other chains' commonly named component failures (H5a). Chains' experience with local naming thus biased them toward persisting with their chosen strategy even as evidence of its shortcomings mounted, helping explain why not all chains adopted the apparently superior common naming strategy. At the same time, however, momentum also led chains choosing common naming initially to accelerate their exploitation of that superior strategy. And, like aspiration-level adjustment, momentum also kept chains using common naming from abandoning the strategy too quickly (and likely detrimentally) in response to idiosyncratic failures of commonly named components.

More generally, momentum prevented chains from engaging in potentially costly strategic changes whose consequences were not well understood. While such prevention was likely of benefit to chains using common naming, whether it was also beneficial for chains using local naming depends on the magnitude of costs and risks associated with changing from local to common naming. Changing their naming strategy is not inconsequential for chains. A chain's naming strategy shapes its external identity and reputation among consumers. Altering it may affect consumers' responses to and support for the chain and its components. A chain using local naming must invest in a new identity and commitment to quality to generate external reputation and internal cost advantages for the chain. In the language of organizational ecology, such change may produce a renewed liability of newness, robbing a chain's history of its survival value (Hannan and Freeman, 1984). Even if changing to common naming ultimately enhances a chain's performance, attempting the change may lower the chain's short-run performance, exposing it to an increased risk of failure. A chain may bring about its own failure (or the failure of individual components) as a direct result of its attempts to improve its performance and survive (March, 1981). Thus, while common naming confers a significant survival advantage on chains in this (Baum, 1999) and several other empirical settings, whether the costs and risks of changing naming strategies are greater or smaller than the ultimate benefits of doing so remains an open empirical question.

While our study reveals some new dimensions of learning from failure in general, and effects on strategy adoption in particular, much work remains to be done. Our findings suggest a number of directions for future research. Interorganizational learning theorists contend that three modes of learning--frequency-, trait-, and outcome-based--represent key mechanisms facilitating vicarious learning (Miner and Haunschild, 1995; Haunschild and Miner, 1997). Our focus here on outcome-based learning, the least studied of the three modes, suggests the need for further research examining how these modes are interrelated. As Haunschild and Miner (1997) pointed out, frequency- and trait-based imitation are social in character, while outcome-based learning is more technical. A practical concern raised by these modes of imitation is whether such social mechanisms lead organizations to adopt (or abandon) strategies and practices effectively (Greve, 1995). Although imitating others' behavior enables an organization to take advantage of their information, and so is a reasonable response to decision making under ambiguity, it is more problematic when other organizations also imitate. Then, each imitation will transmit either the information of the imitated organization or of the organization that it, in turn, imitated. Imitation of adoption or abandonment may thus cause adoption or abandonment with very little information indicating that the practice is valuable or obsolete, leading to possibly faddish behavior (Abrahamson, 1991). Attention to outcomes may help avoid such faddish adoption (Sitkin, 1992). Future studies examining the role of outcome-based learning in counteracting faddish adoption driven by frequency- and trait-based imitation may thus prove particularly valuable.

Turning to outcome-based learning itself, while our findings help correct the success bias in empirical research on organizational learning, organizations learn from strategic successes as well. Consequently, future research attending to the combined influence of a firm's success and failure with a particular strategy on its use of that strategy would be particularly valuable. We wonder, for example, whether organizations learn more effectively from success or failure. What balance of success and failure is optimal for organizational learning? We also wonder how success and failure interact. For example, heroism tends to be attributed to leaders who ultimately succeed in their efforts after a series of failures, and their visions are embraced despite the initial failures (Medcof and Evans, 1986). Does experiencing or observing such a pattern of strategic success and failure lead organizational decision makers to become similarly overcommitted to their chosen strategies?

Other avenues suggested by our analysis concern refinements to our measures of others' failure experience. Several studies highlight that experiential and vicarious organizational learning processes can be moderated by space and time (e.g., Baum, Li, and Usher, 2000; Greve, 2000). As decision makers look for role models, they tend to monitor the behavior of a reference group of comparable organizations in similar situations, and their opinions and actions evolve toward those in their reference group (Fiegenbaum and Thomas, 1995; Lant and Baum, 1995; Porac et al., 1995). By focusing their attention on comparable organizations, decision makers increase the potential relevance to their own situation of the experiences and actions being observed. As a result, for another organization's experience to influence a decision maker, the organization and its context must be seen as sufficiently similar to the decision maker's that information that it has acted in a particular way is viewed as having diagnostic value for the imitator. Despite the salience of component failures, it may be that a chain's decision makers may attend to and interpret these events differentially as a function of how similar the chain experiencing the failure is in size (i.e., number of components), strategy (i.e., common or local naming) and market location (i.e., urban or rural markets) to the decision maker's chain. Comparability effects may be reinforced by multiple market contacts, which can increase chains' knowledge about each other.

Relatedly, while our results provide evidence of a chain's learning from its own recent failures, research suggests that history also matters for organizational learning (Amburgey and Miner, 1992). While learning theory suggests that the impetus for change is performance failure, mitigating the tendency toward momentum, as our results show, organizations resist change, preferring to reemploy strategies they have used before, even in the face of their failure. Repeated failures may provide an accumulation of evidence disconfirming decision makers' chosen strategy, however, permitting them to distinguish idiosyncratic failures from meaningful ones, making it more difficult to sustain false or superstitious beliefs, and promoting serious exploration of alternatives (Nystrom and Starbuck, 1984; Ginsberg and Baum, 1994). A chain's cumulative history of prior component failures may therefore prove vital to dislodging decision-making biases associated with organizational momentum and therefore be a fruitful opportunity for future research.

An analogous issue arises at the interorganizational level. While our results provide evidence of organizations learning from others' recent failures, research suggests that organizations also learn congenitally, adopting successful practices prevalent at the time of their founding, and so firms founded later tend to outperform their predecessors (e.g., Zimmerman, 1982; Argote, Beckman, and Epple, 1990; Baum and Ingram, 1998). Although we did not examine congenital learning from failure, estimates for our control variable, other chains' common name concentration at founding, were not significant. This is surprising, because the cost and disruption involved in changing a chain's naming strategy after it is already in place suggests that it may become increasingly difficult for chains to alter their naming strategies after their founding. Perhaps chains' ability to acquire and divest components provides opportunities for redefinition and adaptability that are unavailable to single-unit organizations. This possibility is consistent with findings suggesting that congenital learning dominates incremental learning at the component but not the chain level (Ingram and Baum, 1997b; Baum and Ingram, 1998). Future empirical work might thus productively examine more closely the mechanisms through which chains alter their naming strategies and adapt more generally, as well as organizational factors that affect the relative balance of vicarious learning at the time of founding and thereafter.

Finally, the strategy adoption process may be driven by competition as well as learning, and these processes may interact. For multiunit chains, the multimarket character of competition is particularly relevant (Greve and Baum, 2001). The benefits of multimarket contact predicted by mutual forbearance theory (e.g., Edwards, 1955) suggest that organizations will enter markets in which such contacts can be established (Scott, 1989, 1991; Korn and Baum, 1999). In the same way that common naming helps consumers identify a chain's components, it helps competitors identify each other's components. By facilitating identification of components' corporate affiliations, common naming may play an important role in facilitating the emergence of mutual forbearance among multiunit chains. Chains using common naming may create more market contacts with each other than with chains using local naming and benefit more from a given level of multimarket contact than chains using local naming. In turn, chains with more market contacts may make greater use of common naming to identify themselves to competitors. Corroboration of such conjectures in future research would expand the scope of common naming benefits beyond consumer reputation benefits.

Evidence is mounting on the learning capabilities of multiunit organizations. Our study provides some of the first systematic empirical evidence of outcome-based learning, and in particular, failure-induced learning, by chains. Our theoretical account and empirical results identify key organizational and interorganizational learning processes through which chains' experiences interact with their own and each other's strategy failures to shape--for better or worse--their strategic choices. Our findings reveal some of the consequential complications that can arise as chains endeavor to learn and adapt, reinforcing the view that organizational learning is often an errorful process subject to biases and distortions imposed by boundedly rational decision makers. Beyond the chain organizational form and strategy adoption and abandonment, we see many ways to build on the failure-induced perspective on organizational learning we have outlined here.
Table 1

Means, Standard Deviations, and Bivariate Correlations for Control
and Theoretical Variables

Variable                                  Mean     S.D.     1      2

1. Own local name concentration             .51     .42
2. Age/1000 (in months)                     .11     .08    .37
3. Size (number of components)             6.06    5.09   -.06    .32
4. Multimarket contact/10                  6.9     3.94   -.01    .56
5. Spatial compactness                    79.2     3.95    .09   -.13
6. Independent nursing home density      261      72.95   -.01   -.54
7. Chain component density               111.5    34.82    .00    .57
8. Nursing home chain density             16.59    3.57    .03    .43
9. Other chains' common name                .51     .06    .38   -.44
   concentration
10. Others' common name concentration       .56     .04   -.01    .15
    at founding
11. Extended-care per diem                27.31    4.2    -.02    .38
12. Census population >65 (mil.)           1.05     .31   -.02    .55
13. Calendar time                         14.58    6.59    .00    .58
14. Hybrid naming                           .09     .29   -.20    .15
15. Number of acquisitions                  .23     .81   0.07   -.03
16. Left censored                           .37     .48   -.40    .32
17. Own commonly named failures/100         .04     .27   -.14    .05
18. Own locally named failures/100          .03     .15    .09   -.07
19, Others' commonly named failures/100     .65    1.02    .01   -.02
20. Others' locally named failures/100      .52     .75   -.03   -.02

Variable                                   11       12      13     14

12. Census population >65 (mil.)           .51
13. Calendar time                          .64     .90
14. Hybrid naming                         -.02    -.03   -.03
15. Number of acquisitions                -.05    -.07   -.03    .17
16. Left censored                         -.26    -.24   -.32    .25
17. Own commonly named failures/100        .01    -.02   -.03   -.05
18. Own locally named failures/100         .04    -.05   -.02   -.06
19. Others' commonly named failures/100    .13     .02    .00   -.04
20. Others' locally named failures/100     .20    -.10    .02   -.01

Variable                                     3      4      5      6

1. Own local name concentration
2. Age/1000 (in months)
3. Size (number of components)
4. Multimarket contact/10                  .19
5. Spatial compactness                    -.06   -.02
6. Independent nursing home density       -.14   -.83    .05
7. Chain component density                 .18    .93   -.04   -.95
8. Nursing home chain density              .07    .63   -.05   -.90
9. Other chains' common name              -.27   -.55    .06    .54
   concentration
10. Others' common name concentration      .07   -.42   -.07    .32
    at founding
11. Extended-care per diem                 .11    .63   -.04   -.68
12. Census population >65 (mil.)           .17    .91    .02   -.73
13. Calendar time                          .18    .96   -.03   -.93
14. Hybrid naming                          .66   -.02   -.10    .02
15. Number of acquisitions                 .14   -.07    .01   -.02
16. Left censored                          .15   -.27   -.13    .35
17. Own commonly named failures/100       -.01   -.02    .04    .05
18. Own locally named failures/100        -.03   -.03    .02   -.01
19. Others' commonly named failures/100   -.04    .02    .02    .02
20. Others' locally named failures/100     .03   -.02   -.05   -.12

Variable                                   15     16     17     18

12. Census population >65 (mil.)
13. Calendar time
14. Hybrid naming
15. Number of acquisitions
16. Left censored                         -.07
17. Own commonly named failures/lO0       -.03    .06
18. Own locally named failures/lO0         .05   -.04   -.02
19. Others' commonly named failures/lO0    .02   -.01    .20     .02
20. Others' locally named failures/lO0     .02   -.04    .02     .31

Variable                                     7      8      9     10

1. Own local name concentration
2. Age/1000 (in months)
3. Size (number of components)
4. Multimarket contact/10
5. Spatial compactness
6. Independent nursing home density
7. Chain component density
8. Nursing home chain density              .80
9. Other chains' common name              -.61   -.45
   concentration
10. Others' common name concentration     -.39   -.21    .23
    at founding
11. Extended-care per diem                 .66    .61   -.37   -.23
12. Census population >65 (mil.)           .82    .52   -.42   -.38
13. Calendar time                          .98    .76   -.56   -.40
14. Hybrid naming                         -.03   -.03   -.29    .20
15. Number of acquisitions                -.01    .06    .00    .03
16. Left censored                         -.33   -.34   -.02    .32
17. Own commonly named failures/100       -.04   -.01   -.09    .01
18. Own locally named failures/100         .01    .00    .02    .01
19, Others' commonly named failures/100    .01    .20   -.14    .02
2. Others' locally named failures/100      .14    .10   -.20   -.05

Variable                                   19

12. Census population >65 (mil.)
13. Calendar time
14. Hybrid naming
15. Number of acquisitions
16. Left censored
17. Own commonly named failures/100
18. Own locally named failures/100
19. Others' commonly named failures/100
20. Others' locally named failures/100     .12

Table 2

Random Effects Models of Chains' Change in Common Naming Strategy *

Variable                       Model 1        Model 2        Model 3

Theoretical variables (t-1)
H1a (-): Own commonly
         named failures/100                  -0.590 (C)     -0.634 (C)
                                             (-0.228)        (0.230)
H1b (+): Own locally named
         failures/100                          0.817 (B)      0.834 (B)
                                               (0.398)       (0.406)
H2a (-): Others' commonly
         named failures/100                                   0.131 (A)
                                                             (0.084)
H2b (+): Others' locally
         named failures/100                                   0.029
                                                             (0.126)
H3a (-): Own x Others'
         commonly named
         failures/10000

H3b (+): Own x Others'
         locally named
         failures/10000

H4a (-): Own local name
         concentration x
         Own commonly
         named failures/100
H4b (-): Own local name
         concentration x
         Own locally named
         failures/100
H5a (-): Own local name
         concentration x
         Others' commonly
         named failures/100
H5b (-): Own local name
         concentration x
         Others' locally
         named failures/100
Control variables (t-1)
Own local name concen-        -.0593 (C)      -0.624 (C)     -0.656 (C)
  tration                      (0.190)        (0.189)        (0.191)
Age/1000 (in months)           -0.004 (C)     -0.004 (C)     -0.004 (C)
                               (0.002)        (0.002)        (0.002)
Size (number of compo-          0.049 (C)     -0.051 (C)      0.051 (C)
  nents)                       (0.017)        (0.017)        (0.017)
Multimarket contact/10          0.000          0.001         -0.001
                               (0.008)        (0.008)        (0.008)
Spatial compactness            -0.012         -0.010         -0.010
                               (0.015)        (0.015)        (0.015)
Independent nursing home
  density                      -0.002          0.000         -0.004
                               (0.004)        (0.004)        (0.005)
Chain component density         0.031 (B)      0.032 (B)      0.026
                               (0.018)        (0.018)        (0.024)
Nursing home chain density      0.062          0.077 (A)      0.015
                               (0.049)        (0.048)        (0.065)
Other chains' common name
  concentration                 0.803          0.351          0.923
                               (1.737)        (1.736)        (1.805)
Others' common name con-
  centration at founding        1.775          1.797          1.669
                               (2.267)        (2.251)        (2.247)
Extended-care per diem          0.010          0.013          0.007
                               (0.020)        (0.020)        (0.021)
Census population >65           3.020 (C)      3.100 (C)      2.640 (B)
  (millions)                   (0.813)        (0.813)        (0.921)
Calendar time                  -0.315 (B)     -0.325 (B)     -0.266 (B)
                               (0.133)        (0.134)        (0.162)
Hybrid naming                  -0.373         -0.458 (A)     -0.419 (A)
                               (0.310)        (0.310)        (0.311)
Number of acquisitions         -0.287 (C)     -0.288 (C)     -0.291 (C)
                               (0.076)        (0.075)        (0.075)
Left censored                   0.182          0.174          0.172
                               (0.183)        (0.182)        (0.181)
Constant                       -2.827         -3.424         -1.130
(3.484)                        (3.484)        (3.481)        (3.887)

Log likelihood               -671.42        -658.08        -656.87
Likelihood ratio test                         26.69           2.41
(D.f.)                                       (2) (C)         (2)ns
versus nested model                          vs. M1         vs. M2

Variable                          Model 4         Model 5

Theoretical variables (t-1)
H1a (-): Own commonly
         named failures/100         -2.412 (C)      -2.661 (C)
                                    (0.880)         (0.976)
H1b (+): Own locally named
         failures/100                1.784 (B)       2.658 (B)
                                    (0.893)         (1.498)
H2a (-): Others' commonly
         named failures/100          0.119 (A)       0.132 (A)
                                    (0.083)         (0.084)
H2b (+): Others' locally
         named failures/100          0.026           0.030
                                    (0.125)         (0.125)
H3a (-): Own x Others'
         commonly named
         failures/10000              0.942 (C)       1.026 (C)
                                    (0.354)         (0.382)
H3b (+): Own x Others'
         locally named
         failures/10000             -0.354          -0.506 (A)
                                    (0.319)         (0.339)
H4a (-): Own local name
         concentration x                             0.941
         Own commonly
         named failures/100                         (1.505)
H4b (-): Own local name
         concentration x                            -2.590 (B)
         Own locally named
         failures/100                               (1.527)
H5a (-): Own local name
         concentration x
         Others' commonly
         named failures/100
H5b (-): Own local name
         concentration x
         Others' locally
         named failures/100
Control variables (t-1)
Own local name concen-              -0.676 (C)      -0.629 (C)
  tration                           (0.191)         (0.194)
Age/1000 (in months)                -0.005 (C)      -0.004 (C)
                                    (0.002)         (0.002)
Size (number of compo-               0.053 (C)       0.052 (C)
  nents)                            (0.017)         (0.017)
Multimarket contact/10              -0.001          -0.002
                                    (0.008)         (0.008)
Spatial compactness                 -0.009          -0.009
                                    (0.015)         (0.015)
Independent nursing home            -0.004          -0.003
  density                           (0.005)         (0.005)
Chain component density              0.028           0.028
                                    (0.024)         (0.024)
Nursing home chain density           0.019           0.018
                                    (0.065)         (0.065)
Other chains' common name
  concentration                      0.766           0.554
                                    (1.806)         (1.810)
Others' common name con-
  centration at founding             2.365           2.340
                                    (2.261)         (2.255)
Extended-care per diem               0.010           0.011
                                    (0.021)         (0.021)
Census population >65                2.840 (B)       2.860 (B)
  (millions)                        (0.923)         (0.923)
Calendar time                       -0.281 (B)      -0.276 (B)
                                    (0.161)         (0.161)
Hybrid naming                       -0.483 (A)      -0.476 (A)
                                    (0.312)         (0.311)
Number of acquisitions              -0.285 (C)      -0.283 (C)
                                    (0.075)         (0.075)
Left censored                        0.239           0.225
                                    (0.183)         (0.183)
Constant                            -1.896          -2.057
(3.484)                             (3.884)         (3.890)

Log likelihood                    -651.86         -650.19
Likelihood ratio test               10.03            3.33
(D.f.)                             (2) (B)         (2) ns
versus nested model                vs. M3          vs. M4

Variable                          Model 6         Model 7

Theoretical variables (t-1)
H1a (-): Own commonly
         named failures/100         -2.916 (C)      -2.438 (C)
                                    (0.978)         (0.874)
H1b (+): Own locally named
         failures/100                2.916           2.966 (B)
                                    (1.524)         (1.492)
H2a (-): Others' commonly
         named failures/100          0.223 (C)       0.259 (C)
                                    (0.110)         (0.110)
H2b (+): Others' locally
         named failures/100          0.043           0.021
                                    (0.154)         (0.125)
H3a (-): Own x Others'
         commonly named
         failures/10000              0.924 (C)       0.910 (C)
                                    (0.381)         (0.352)
H3b (+): Own x Others'
         locally named
         failures/10000             -0.513 (A)      -0.573 (B)
                                    (0.340)         (0.337)
H4a (-): Own local name
         concentration x             0.694            --
         Own commonly
         named failures/100         (1.512)
H4b (-): Own local name
         concentration x            -2.527 (B)      -2.974 (B)
         Own locally named
         failures/100               (1.592)         (1.519)
H5a (-): Own local name
         concentration x            -0.259 (B)      -0.274 (B)
         Others' commonly
         named failures/100         (0.140)         (0.138)
H5b (-): Own local name
         concentration x            -0.016            --
         Others' locally
         named failures/100         (0.199)
Control variables (t-1)
Own local name concen-              -0.435 (B)      -0.451 (B)
  tration                           (0.231)         (0.212)
Age/1000 (in months)                -0.004 (C)      -0.004 (C)
                                    (0.002)         (0.002)
Size (number of compo-               0.053 (C)       0.052 (C)
  nents)                            (0.017)         (0.017)
Multimarket contact/10              -0.021          -0.002
                                    (0.008)         (0.008)
Spatial compactness                 -0.009          -0.009
                                    (0.015)         (0.015)
Independent nursing home            -0.003          -0.003
  density                           (0.005)         (0.005)
Chain component density              0.025           0.028
                                    (0.024)         (0.024)
Nursing home chain density           0.021           0.021
                                    (0.065)         (0.064)
Other chains' common name
  concentration                      0.181           0.529
                                    (1.816)         (1.797)
Others' common name con-
  centration at founding             2.301           2.262
                                    (2.255)         (2.245)
Extended-care per diem               0.012           0.011
                                    (0.021)         (0.021)
Census population >65                2.700 (B)       2.870 (B)
  (millions)                        (0.923)         (0.917)
Calendar time                       -0.250 (A)      -0.280 (B)
                                    (0.161)         (0.160)
Hybrid naming                       -0.505 (B)      -0.466 (A)
                                    (0.311)         (0.310)
Number of acquisitions              -0.292 (C)      -0.296 (C)
                                    (0.075)         (0.075)
Left censored                        0.232           0.226
                                    (0.183)         (0.182)
Constant                            -1.946          -2.079
(3.484)                             (3.892)         (3.859)

Log likelihood                    -648.39         -648.71
Likelihood ratio test
(D.f.)                               3.61 (2)ns      6.28(2) (B)
versus nested model                 vs. M5          vs. M4

(A) p < .10; (B) p < .05; (C) p < .01

* Standard errors are in parentheses.


(1) It seems unlikely that a chain's decision makers would close components as a means of altering its naming strategy. Rather, as an anonymous reviewer suggested to us, component failures more likely reflect environmental selection processes, the very processes from which we suggest chains' decision makers can learn the value of common naming. Consequently, to distinguish the effects of learning from those of selection, we computed, change in common naming without considering the names of components that failed. We checked the robustness of findings based on this measure by using an alternative specification that did include information on the names of components that failed and thus treated failure as a source of change in naming strategy. The estimates for this alternate specification did not differ substantively from those reported below, with the exception that coefficients for a chain's own failure experience were larger, suggesting that this specification may indeed conflate adaptation and selection processes. Two additional specifications, one excluding information on the names of newly founded components, the other excluding both names of newly founded and failed components, also yielded analogous findings. These additional results are available from the authors.

(2) We are grateful to an anonymous reviewer for suggesting this weighting scheme.

(3) Coefficients based on raw counts of the number of own and others' failed components are similar to those for the weighted counts reported below but are less efficient.

(4) Using a chain's total number of beds to define its size does not alter the findings.

(5) In supplementary analysis, we found that, lagged two years, the coefficient for number of acquisitions became positive but was insignificant. Consistent with our interpretation, this suggests that after two years, chains using common naming had changed acquired components' names, while chains using local naming had not, effectively canceling each other out.

REFERENCES

Abrahamson, E.

1991 "Managerial fads and fashions: The diffusion and rejection of innovations." Academy of Management Review, 16: 586-612.

Amburgey, T. L., and A. S. Miner

1992 "Strategic momentum: The effects of repetitive, positional, and contextual momentum on merger activity." Strategic Management Journal, 13: 335-348.

Argote, L.

1999 Organizational Learning: Creating, Retaining and Transferring Knowledge. Norwell, MA: Kluwer Academic.

Argote, L., S. L. Beckman, and D. Epple

1990 "The persistence and transfer of learning in industrial settings." Management Science, 36: 140-154.

Audia, P. G., O. Sorenson, and J. Hage

2001 "Tradeoffs in the organization of production: Multiunit firms. geographic dispersion and organizational learning." In J. A. C. Baum and H. R. Greve (eds.), Advances in Strategic Management, 18: 75-108. Stamford, CT: JAI Press.

Banaszak-Holl, J., J. S. Zinn, and V. Mor

1996 "The impact of market and organizational characteristics on nursing care facility service innovation: A resource dependency perspective." Health Services Research, 31 97-117.

Baum, J. A. C.

1999 "A rise of chain nursing homes in Ontario, 1971-96." Social Forces, 78: 543-584.

Baum, J. A. C., and P. Ingram

1998 "Survival-enhancing learning in the Manhattan hotel industry, 1898-1980." Management Science, 44: 996-1016.

Baum, J. A. C., S. X. Li, and J. M. Usher

2000 "Making the next move: How experiential and vicarious learning shape the locations of chains' acquisitions." Administrative Science Quarterly, 45: 766-801.

Baum, J. A. C., and C. Oliver

1991 "Institutional linkages and organizational mortality." Administrative Science Quarterly, 36: 187-218.

Bradach, J. L.

1997 "Using the plural form in the management of restaurant chains." Administrative Science Quarterly, 42: 276-303.

Cohen, J., and P. Cohen

1983 Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, 2d ed. Hillsdate, NJ: Lawrence Erlbaum.

Conell, C., and S. Cohn

1995 "Learning from other people's actions: Environmental variation and diffusion in French coal mining strikes, 1890-1935." American Journal of Sociology, 101: 366-403.

Cyert, R. M., and J. G. March

1992 A Behavioral Theory of the Firm, 2d ed. Cambridge: Blackwell.

Darr, E. D., L. Argote, and D. Epple

1995 "The acquisition, transfer, and depreciation of knowledge in service organizations: Productivity in franchises." Management Science, 42: 1750-1762.

DiMaggio, P. J., and W. W. Powell

1983 "The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields." American Sociological Review, 48: 148-160.

Dutton, J. E., and J. M. Dukerich

1991 "Keeping an eye on the mirror: Image and identity in organizational adaptation." Academy of Management Journal, 34: 517-554.

Edwards, C. D.

1955 "Conglomerate bigness as a source of power." In NBER Conference Report, Business Concentration and Economic Policy: 331-339. Princeton, NJ: Princeton University Press.

Fang, C., and D. A. Levinthal

2002 "Building bridges to a distant peak: Limits to optimization for boundedly rational actors." Paper presented at the Academy of Management Meetings, Denver, CO, August.

Festinger, L.

1954 "A theory of social comparison processes." Human Relations, 7: 117-140.

Fiegenbaum, A., and H. Thomas

1995 "Strategic groups as reference groups: Theory, modeling and empirical examination of industry and competitive strategy." Strategic Management Journal, 16: 461-476.

Fombrun, C., and M. Shanley

1990 "What's in a name? Reputation-building and corporate strategy." Academy of Management Journal, 33: 233-258.

Gertler, P. J.

1989 "Subsidies, quality and regulation in nursing homes." Journal of Public Economics, 39: 33-53.

Ginsberg, A., and J. A. C. Baum

1994 "Evolutionary processes and patterns of core business change." In J. A. C. Baum and J. V. Singh (eds.), Evolutionary Dynamics of Organizations: 127-151. New York: Oxford University Press.

Glynn, M. A., and R. Abzug

1998 "Isomorphism and competitive differentiation in the organizational name game." In J. A. C. Baum and H. Rao (eds.), Advances in Strategic Management, 15: 105-128. Stamford, CT: JAI Press.

Greene, W. H.

2000 Econometric Analysis, 4th ed. Englewood Cliffs, NJ: Prentice-Hall.

Greve, H. R.

1995 "Jumping ship: The diffusion of strategy abandonment." Administrative Science Quarterly, 40: 444-473.

1996 "Patterns of competition: The diffusion of a market position in radio broadcasting." Administrative Science Quarterly, 41: 29-60.

1998 "Performance, aspirations, and risky organizational change." Administrative Science Quarterly, 43: 58-68.

2000 "Market niche entry decisions: Competition, learning and strategy in Tokyo banking, 1894-1936." Academy of Management Journal, 43: 816-836.

Greve, H. R., and J. A. C. Baum

2001 "Introduction: A multiunit, multimarket world." In J. A. C. Baum and H. R. Greve (eds.), Advances in Strategic Management, 18: 1-28. Stamford, CT: JAI Press.

Hannan, M. T., and J. Freeman

1984 "Structural inertia and organizational change." American Sociological Review, 49: 149-164.

Haunschild, P. R., and A. S. Miner

1997 "Modes of interorganizational imitation: The effects of outcome salience and uncertainty." Administrative Science Quarterly, 42: 472-500.

Haunschild, P. R., and B. Ni

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

Haveman, H. A.

1993 "Organizational size and change: Diversification in the savings and loan industry after deregulation." Administrative Science Quarterly, 38: 20-50.

Herriott, S., D. A. Levinthal, and J. G. March

1985 "Learning from experience in organizations." American Economic Review, 75: 298-302.

Huber, G. P.

1991 "Organizational learning: The contributing processes and the literatures." Organization Science, 2: 88-115.

Ingram, P.

1996 "Organizational form as a solution to the problem of credible commitment: The evolution of naming strategies among U.S. hotel chains, 1896-1980." Strategic Management Journal, 17 (Summer Special Issue): 85-98.

Ingram, P., and J. A. C. Baum

1997a "Chain affiliation and the failure of Manhattan hotels, 1898-1980." Administrative Science Quarterly, 42: 68-102.

1997b "Opportunity and constraint: Organizations' learning from the operating and competitive experience of industries." Strategic Management Journal (Special Issue), 18: 75-98.

2001 "Interorganizational learning and the dynamics of chain relationships." In J. A. C. Baum and H. R. Greve (eds.), Advances in Strategic Management, 18: 109-139. Stamford, CT: JAI Press.

Kelly, D., and T. L. Amburgey

1991 "Organizational inertia and momentum: A dynamic model of strategic change." Academy of Management Journal, 34: 591-612.

Kennedy, P.

1992 A Guide to Econometrics, 3rd ed. Cambridge, MA: MIT Press.

Kim, J.-Y.

2000 "Crash test without dummies: A longitudinal study of interorganizational learning from failure experience in the U.S. commercial banking industry, 1984-1998." Unpublished Ph.D. dissertation, University of Wisconsin-Madison.

Korn, H. J., and J. A. C. Baum

1999 "Chance, imitative and strategic antecedents to multimarket contact." Academy of Management Journal, 42: 171-193.

Lant, T. K., and J. A. C. Baum

1995 "Cognitive sources of socially constructed competitive groups." In W. R. Scott and S. Christensen (eds.), The Institutional Construction of Organizations: 31-47. Thousand Oaks, CA: Sage.

Lant, T. K., and S. J. Mezias

1992 "An organizational learning model of convergence and reorientation." Organization Science, 3: 47-71.

Lant, T. K., F. J. Milliken, and B. Batra

1992 "The role of managerial learning and interpretation in strategic persistence and reorientation: An empirical exploration." Strategic Management Journal, 13: 585-608.

Levinthal, D. A., and J. G. March

1993 "The myopia of learning." Strategic Management Journal, 14: 95-112.

Levitt, B., and J. G. March

1988 "Organizational learning." Annual Review of Sociology, 14: 319-340.

Light, D. W.

1986 "Corporate medicine for profit." Scientific American, 225: 38-45.

March, J. G.

1981 "Footnotes to organizational change." Administrative Science Quarterly, 26: 563-577.

1991 "Exploration and exploitation in organizational learning." Organization Science, 2: 71-87.

March, J. G., and Z. Shapira

1992 "Variable risk preferences and the focus of attention." Psychological Review, 99: 172-183.

March, J. G., L. Sproull, and M. Tamuz

1991 "Learning from samples of one or fewer." Organization Science, 2: 1-13.

Medcof, J. W., and M. G. Evans

1986 "Heroic or competent, a second look." Organizational Behavior and Human Decision Processes, 38: 295-304.

Miller, D.

1990 The Icarus Paradox: How Exceptional Companies Bring about Their Own Downfall. New York: Harper Collins.

1999 "Selection processes inside organizations: The self-reinforcing consequences of success." In J. A. C. Baum and B. McKelvey (eds.), Variations in Organization Science: In Honor of Donald T. Campbell: 93-109. Thousand Oaks, CA: Sage.

Miller, D., and P. H. Freisen

1980 "Momentum and revolution in organizational adaptation." Academy of Management Journal, 4: 591-614.

Miner, A. S., and P. R. Haunschild

1995 "Population level learning." In L. L. Cummings and B. M. Staw (eds.), Research in Organizational Behavior, 17: 115-166. Greenwich, CT: JAI Press.

Miner, A. S., J.-Y. Kim, I. W. Holzinger, and P. R. Haunschild

1999 "Fruits of failure: Organizational failure and population-level learning." In A. S. Miner and P. Anderson (eds.), Advance in Strategic Management, 16: 187-220. Greenwich, CT: JAI Press.

Mitchell, W., J. A. C. Baum, J. Banaszak-Holl, W. B. Berta, and D. Bowman

2002 "Opportunity and constraint: Chain-to-component transfer learning in multiunit chains of U.S. nursing homes, 1991-1997." In C. W. Choo and N. Bontis (eds.), Strategic Management of Intellectual Capital and Organizational Knowledge: A Collection of Readings: 555-573. New York: Oxford University Press.

Nystrom, P., and W. H. Starbuck

1984 "To avoid organizational crisis, unlearn." Organizational Dynamics, 12 (Spring): 53-65.

Ocasio, W.

1997 "The attention-based view of the firm." Strategic Management Journal, 18 (Special Issue): 187-206.

Pfeffer, J., and G. R. Salancik

1978 The External Control of Organizations. New York: Harper & Row.

Porac, J. F., H. Thomas, F. Wilson, D. Paton, and A. Kanfer

1995 "Rivalry and the industry model of Scottish knitwear producers." Administrative Science Quarterly, 40: 203-227.

Scott, J. T.

1989 "Purposive diversification as a motive for merger." International Journal of Industrial Organization, 7: 35-47.

1991 "Multimarket contact among diversified oligopolists." International Journal of Industrial Organization, 9: 225-238.

Singh, J. V.

1986 "Performance, slack, and risk taking in organizational decision making." Academy of Management Journal, 29: 562-585.

Sitkin, S.

1992 Learning through Failure: The Strategy of Small Losses. Greenwich, CT: JAI Press.

Tarman, V. I.

1990 Privatization and Health Care: The Case of Ontario Nursing Homes. Toronto: Garamond Press.

Zimmerman, M. B.

1982 "Learning effects and the commercialization of new energy technologies: The case of nuclear power." Bell Journal of Economics, 13: 297-310.

This research is supported by a grant from the Social Sciences and Humanities Council of Canada. We are grateful to Reed Nelson, Don Palmer, and three anonymous ASQ reviewers for their careful and cogent advice on how to improve our work, and to Linda Johanson for editorial assistance. We thank Whitney Berta, Jackie Hick, and Leisa Sargent for assistance with data collection and coding.

Joel A. C. Baum [coauthor, "It's All in the Name' Failure-Induced Learning by Multiunit Chains"] is Canadian National Chair in Strategy and Organization at the Rotman School of Management (with a cross-appointment to the Department of Sociology), University of Toronto, 105 St. George St., Toronto, ON M5S 3E6, Canada (e-mail: baum@rotman. utoronto.ca). His current research focus is firms' network strategies and the dynamics of interfirm networks and cliques. Recent publications include "Alliance-Based Competitive Dynamics," with B. S. Silverman (Academy of Management Journal, 45: 791-806), "Where Do Small Worlds Come From?," with T. J. Rowley and A. V. Shipilov (Industrial and Corporate Change, vol. 12, forthcoming, 2003), and "Competing in Groups," with T. J. Rowley, A. V. Shipilov, H. R. Greve, and H. Rao (Managerial and Decision Economics, vol. 24, forthcoming, 2003). He has also recently completed editing Geography and Strategy, with O. Sorenson (Advances in Strategic Management, vol. 20, forthcoming, 2003). He is founding co-editor, with R. Greenwood and R D. Jennings, of the new journal Strategic Organization, which published its inaugural issue in February 2003, and currently serves as chair of the Organization and Management Theory Division in the Academy of Management. He received his Ph.D. in organizational behavior from the University of Toronto.

You-Ta Chuang [coauthor, "It's All in the Name' Failure-Induced Learning by Multiunit Chains"] is an assistant professor in management at the School of Administrative Studies, Atkinson College, York University, 4700 Keele Street, Toronto, ON M3J 1 P3, Canada (e-mail: ychuang@yorku.ca). He is completing his Ph.D. in strategic management at the Rotman School of Management, University of Toronto. Research interests include the effects of organizational and interorganizational learning on population evolution, multimarket competition, and corporate governance. Recent publications include "Racing for Market Share: Hypercompetition and the Performance of Multiunit-multimarket Firms," with S. X. Li (Advances in Strategic Management, 18: 329-356). His thesis concerns how performance feedback and interorganizational learning influence organizational diversity.
COPYRIGHT 2003 Published for Cornell University, Johnson Graduate School.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2003 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Chuang, You-Ta; Baum, Joel A.C.
Publication:Administrative Science Quarterly
Date:Mar 1, 2003
Words:15015
Previous Article:Getting by with the advice of their friends: CEOs' advice networks and firms' strategic responses to poor performance.
Next Article:In the bud? Disk array producers as a (possibly) emergent organizational form.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters