Crisis and the content of managerial communications: a study of the focus of attention of top managers in surviving and failing firms.
Do managers of failing firms pay more attention to the internal or external environment? With respect to the external environment, do these managers pay more attention to the input or output side of the environment? Do managers of firms headed toward bankruptcy attend to different aspects of the environment than managers of successful firms in the same industry?
This exploratory study attempts to shed light on these questions by focusing on the allocation of attention in important managerial communications. Effective communications from senior executives are critical during the period preceding bankruptcy, because communications are important tools in managing the impressions of key stakeholders (Bettman and Weitz, 1983; Staw, McKechnie, and Puffer, 1983; Salancik and Meindl, 1984). In this study, we examine letters to shareholders from senior managers of firms that ended up filing for bankruptcy and compare the topics of their statements with those of the executives of matched firms that thrived in the same declining environment.
To date, there has been very little empirical work on the focus of attention in communications by organizational leaders who are in severe trouble. There have been a few highly significant studies of the communications issued by top managers when performance declines or becomes unstable (e.g., Bettman and Weitz, 1983; Staw, McKechnie, and Puffer, 1983; Salancik and Meindl, 1984). These studies, however, did not look at cases in which firms were approaching bankruptcy, and they were intentionally limited to statements that attributed blame or used causal reasoning. Further, these studies did not address the amount of attention paid to the internal and external environments in communications issued by successful and failing firms undergoing external crises, such as declining demand, that are associated with failure. Communications theory offers few clues for predicting the content of managerial communications during external crises. Several authors have asserted that differences in language reflect differing cognitions and perceived realities and that language mirrors those mental processes (e.g., Sapir, 1956; Whorf, 1956; Berger and Luckmann, 1967; Silverman, 1971; Chomsky, 1972; Kress and Hodge, 1979; Pfeffer, 1981; Pondy and Mitroff, 1979:25; Choe, 1987). It is not currently known, however, how communications reflect the crisis conditions associated with failure.
Decline and crisis theories do not provide much additional clarification of these issues. In fact, as we discuss in more detail below, at first blush, these theories lead to conflicting predictions about the attention of senior managers during the crises associated with decline and failure. This empirical study, although admittedly exploratory, is designed to consider and begin to reconcile the conflicting predictions surrounding this issue.
CRISIS AND ATTENTION TO THE EXTERNAL AND INTERNAL ENVIRONMENTS
We define a crisis as any event or condition that threatens the survival of the organization (Starbuck, Greve, and Hedberg, 1978). One such crisis is declining or stagnant demand, which frequently precedes organizational bankruptcy (Hambrick and D'Aveni, 1988). Declining demand and slowing demand growth rates are symptomatic of larger problems such as a niche contraction or a downturn in the general economy that are often threats to organizational survival (Hannan and Freeman, 1977; Harrigan, 1980a, 1980b; Zammuto and Cameron, 1985).
Several theories predict how senior managers might respond to an external crisis such as decline in demand, but these theories appear to be in conflict. The threat-rigidity response (Staw, Sandelands, and Dutton, 1981), crisis-denial (Starbuck, Greve, and Hedberg, 1978), and environmental-scanning (Daft, Sormunen, and Parks, 1988) models predict different patterns of attention in response to crisis. Under the threat-rigidity response and crisis-denial theories, a crisis is expected to divert a manager's attention away from the locus of the crisis because it creates "noise" that may keep the manager from considering relevant information about elements in the organization's environment that are the source of the crisis (Kiesler and Sproull, 1982). Environmental scanning (Hambrick, 1981, 1982) and stress theories (Janis and Mann, 1977; Lazarus and Folkman, 1984) predict that managers will pay more attention to the external crisis because of the importance, immediacy, and uncertainty of the issue (Dutton, 1986).
Despite the apparent conflicts among these predictions, there is no reason to doubt the quality of these earlier studies. It may be that there is a way to reconcile these theories by considering another factor that has not been considered previously. Perhaps managers of successful and failing firms facing a severe external crisis pursue fundamentally different response patterns. It is this assumption that is tested in our present study.
We suggest that managers of failing firms react to a crisis in the output environment by engaging in maladaptive behaviors that are predicted by the decline, crisis-denial, and threat-rigidity response models. Managers of successful firms, in contrast, engage in different behaviors after the crisis strikes. These different behaviors enable the survivors to adapt appropriately, and they are predicted by the environmental scanning and stress theories. As we argue below, at the heart of these differences in behavior is the focus of attention of top management in the face of an externally induced crisis. If these predicted differences are not observed after the crisis, then our attempt to reconcile the various theories has failed. If the predicted patterns are observed, then this exploratory study will have opened an interesting avenue for further, more systematic research.
Allocating Attention between the internal and External Environment
Two of the prominent theories that contribute to our understanding of maladaptation and adaptation are the crisis-denial and environmental-scanning theories. They also suggest differences in behavior with respect to the allocation of attention between the internal and external environments.
Crisis denial and maladaptation. The crisis-denial model suggests that managers will not change their focus of attention in response to an externally induced crisis. The model suggests a causal relationship that leads managers of failing firms to ignore external crises. Thus, after the crisis strikes, managers act as if the external crisis does not exist, and they do not pay more attention to external factors than internal factors (Argenti, 1976; Holsti, 1978; Starbuck, Greve, and Hedberg, 1978; Whetten, 1980). Years of earlier success, for example, lead many managers to ignore external crises because they perceive them to be temporary or inconsequential. Starbuck and his colleagues (1978) cited several cases in which managers failed to react to external crises. These managers were insensitive to changing external conditions because they were focusing on internal methods that were successful in the past. They perceived no need to change, because they had overestimated the strength of their strategy or had dismissed the seriousness of changes in the marketplace (Argenti, 1976; Holsti, 1978).
Environmental scanning and strategic adaptation. The normative strategy literature suggests that effective adaptation requires scanning of the external environment so that managers may mediate between the external environment and the firm to bring about fit between strategy, structure, and external constraints (Mintzberg, 1978; Andrews, 1980; Tushman and Romanelli, 1985). Managers of low-performing firms do not scan the external environment as vigilantly as those of high performers do in response to a crisis stimulus (Aguilar, 1967; Schoderbek, Kefalas, and Schoderbek, 1975; Hambrick, 1981, 1982). They are, therefore, unable to lead their firms to adapt as well as those firms whose managers scan (and hence understand) the external environment. This lack of scanning can be a crucial error because it inhibits a firm's ability to adapt to some of the most frequent and serious external threats facing failing firms, like declining demand (Harrigan, 1980a, 1980b; Zammuto and Cameron, 1985).
Taken together, these two theories allow us to suggest that reactions differ for successful and failing firms. When faced with a sudden, external threat, maladaptive managers do not react to the external change. Adaptive managers are able to take stock of the external situation and determine an appropriate evasive action.
Proposition 1a: After an external crisis hits, managers of firms that survive become engaged in adaptive scanning behavior. Therefore, after the crisis strikes they will pay more attention to the external environment than to the internal environment. The managers of firms that eventually go bankrupt are trapped in maladaptive crisis denial after an external crisis hits. Therefore, after the crisis they will not pay more attention to the external environment than to the internal environment.
Proposition 1b: After an external crisis hits, managers of successful firms will pay more attention to the external environment than will the managers of failing firms.
Allocating Attention between Aspects of the External Environment
When paying attention to the external environment, managers allocate their attention between the input and output aspects of that environment. Four other theories that contribute to our understanding of maladaptation and adaptation suggest differences in behavior with respect to the allocation of attention between the input and output environments. The literatures on critical success factors and stress suggest that the managers of adaptive firms will pay more attention to the output environment than the input environment. The literatures on threat-rigidity response and decline suggest that the managers of maladaptive firms will attend to the input environment more than to the output environment.
Critical success factors and adaptation. The strategy literature argues that, under normal and crisis conditions, managers of successful firms pay more attention to the output environment than to the input environment because they attend to "critical success factors" (Hambrick and Lei, 1985). This literature asserts that the most important critical success factors are output related. The two most frequently identified critical factors are customer needs (e.g., Peters and Waterman, 1982) and demand growth rate, which is often called the product life cycle (e.g., Hofer, 1975). In an empirical study of the critical contingency factors affecting the relationship between strategy and performance, Hambrick and Lei (1985) found that five of the six most significant contingency factors were related to the output environment, including buyer power, demand, customer characteristics, and the type of sales transaction.
Stress and adaptation. The managers of successful firms are also expected to be more output oriented when they face a decline in demand because they pay more attention to the specific problem areas facing them, a process labeled "problem sensing" by Kiesler and Sproull (1982). The crisis is a stimulus for organizational information-search routines that attend to the locus of the crisis. Dutton (1986) found that a crisis issue caused firms to expend greater resources on resolving the issue than did noncrisis issues. At the individual level, the stress literature suggests that crisis causes stress that, in turn, results in individuals focusing increased attention on task-relevant cues and the source of the crisis (Janis and Mann, 1977; Lazarus and Folkman, 1984). This literature implies that, since problem sensing and resolution are necessary for effective adaptation, the managers of successful firms will attend to the output environment, more than to the input environment, if this is the part of the environment from which the crisis stems.
Taken together, the critical-success-factor and stress literatures suggest the following proposition: Proposition 2a: Under normal circumstances, and after a demand-decline crisis hits, the managers of firms that survive are engaged in problem-sensing behaviors that attend to critical success factors, and they therefore pay more attention to the output environment than to the input environment.
Threat rigidity and maladaptation. The threat-rigidity response model, as outlined by Staw, Sandelands, and Dutton (1981), argues that crisis-induced stress has a different effect because it changes information-processing patterns at the individual, small-group, and organizational levels. They argued that a crisis does generate increased search behavior, but they also noted that this often results in information overload. The resultant confusion disguises the real cause of the crisis and forces the premature abandonment of search activity aimed at the locus of the crisis. Consequently, information processing is restricted and decision makers turn their attention to simplistic efficiency concerns, such as identifying lower-cost inputs and finding ways to use internal resources differently. As a consequence of this high level of attention to the input and internal environments, managers under crisis attend to the output environment less than to the input and internal environments (Sutton, 1990).
Staw, Sandelands, and Dutton (1981) noted that some firms that experience threat-rigidity response are successful despite, or perhaps because of, the threat-rigidity response of their managers. It is not always dysfunctional to focus on inputs and internal operations, because efficiency-oriented reactions are sometimes enough to save a firm from failure. However, in the face of maturing or declining demand, emphasizing improved use of inputs and internal resources is usually insufficient to solve the more severe environmental problems that the firm confronts (Hambrick and Schechter, 1982). This suggests that only the failures are associated with threat-rigidity response when demand declines. Moreover, Staw, Sandelands, and Dutton (1981) noted that threat rigidity is also stimulated by declining performance.
Decline and maladaptation. The organizational decline literature also suggests that managers of failing firms will pay more attention to the input and internal environments than to the output environment. If the declining demand causes a firm's performance to decline, managers must also cope with the adverse consequences of declining performance as well as with the declining demand (Whetten, 1980). Performance decline creates several problems that must be attended to, including high employee turnover (Schein, 1978; Mobley et al., 1979), departing key managers (Gamson and Scotch, 1964; D'Aveni, 1990), and dissatisfied stockholders, creditors, and suppliers (Altman, 1983). In order to maintain the cooperation of worried internal participants and input providers needed for domain consensus (Barnard, 1938; March and Simon, 1958; Evan, 1966; Thompson, 1967), managers of failing firms must pay attention to them. Thus declining performance creates its own crises, which force managers of failing firms to pay more attention to the short-run problems associated with declining performance. In contrast, an output crisis like declining demand has long-term effects that must be responded to with long-term solutions, such as changing strategies or environments (Harrigan, 1980a, 1980b; Zammuto and Cameron, 1985). But the short-run problems make managers of failing firms unable to plan for the long term (Hall and Mansfield, 1971; Anderson, 1976; Smart and Vertinsky, 1977; Holsti, 1978; Whetten, 1980). Thus managers of failing firms are expected to pay more attention to the short-run problems in the input and internal environments than to the long-run problems in the output environment.
Taken together, the threat-rigidity response and decline models suggest the following propositions:
Proposition 2b: The managers of firms that eventually go bankrupt become trapped in maladaptive threat-rigidity response after an external crisis hits, or they engage in decision processes to solve short-run problems due to declining performance. Therefore, despite the fact that the crisis is external, the managers of failing firms will gradually stop paying more attention to the output environment than to the input environment.
Proposition 2c: After an external crisis hits, the managers of successful firms will pay more attention to the output environment and less attention to the input and internal environments than will the managers of failing firms.
We have attempted to reconcile the predictions from this confusing array of theories by postulating distinctly different attentional reactions by managers of successful and failing firms. An empirical study of the attentional patterns of senior managers in response to an external crisis could provide some indication of which of these predictions are borne out for failing and successful firms and whether the apparently conflicting predictions are reconcilable in the way we have postulated above. The empirical study described below, despite its obvious limitations, provides some insights into these issues.
Content analysis of written communications is useful for reconstructing perceptions and beliefs of their authors (Holsti, 1968, 1969). Pfeffer (1981:26) noted that analyses of language provide "evidence on the origination and diffusion of various concerns and organizational responses to these concerns." Differences in language reflect differing cognitions and perceived realities, since language mirrors mental processes (e.g., Sapir, 1956,, Berger and Luckmann, 1967; Chomsky, 1972; Kress and Hodge, 1979).
To determine the attentional patterns of senior managers, we have treated letters to shareholders in annual reports as manifestations of the perceptual focus of attention of these leaders and have analyzed their contents. Content analysis of this type has been used successfully in many past studies (e.g., Lentz and Tschirgi, 1963; Bowman, 1976, 1978; Bettman and Weitz, 1983; Staw, McKechnie, and Puffer, 1983; Salancik and Meindl, 1984). Letters to shareholders are particularly good indicators of the major topics that organizational managers attend to (Watzlawick, Beavin, and Jackson, 1967; Bradley and Baird, 1977), and they reflect the perceptions of organizational stewards because they are the product of the inputs of many individuals. In addition, they are closely reviewed by top managers (Goodman, 1980; Petzinger, 1982). Letters to shareholders reveal how much attention is paid to various aspects of the environment, relative to others. Here, attention is defined as "allocating information-processing capacity (receiving, cognitive processing, disseminating) to environmental stimuli over time" (Sproull, 1984: 10, based on the work of Simon, 1971, 1973). As Sproull (1984: 10) also wrote in defining attention:
When a manager is attending to a particular stimulus he is processing information relevant to it.... This [Sproull's] definition is broader than the psychological concept of attention. It includes the overt, visible steps of receiving and disseminating information that are captured in the common-place phrase, "paying attention to," as well as the mental steps of noticing and encoding.... Although the internal, mental aspects of attention remain hidden, the concept can prove useful in systematic investigations of managerial information processing. [Emphasis added.]
We used a matched-pair design and measured the topics attended to by senior managers as indicated in their letters to shareholders. This was done for each failing firm and a matched survivor over the last five years of the failing firm's life. This design allowed us to compare the managers of failing firms with matched survivors at various points in time preceding the bankruptcy. It also allowed us to monitor how the managers of successful and bankrupt firms changed their attentional patterns in relation to an external crisis and changes in performance.
Sample. The sample was composed of 57 large firms who filed a Chapter X or XI bankruptcy petition between 1972 and 1982 and 57 matched (on size and product/market environment) survivors. We limited the study to large bankruptcies so as to maximize the availability of letters to shareholders and to eliminate smaller bankruptcies, which are likely to happen so quickly (Argenti, 1976) that the letters to shareholders would not capture the phenomena under study. Our population was the 1,000 largest bankruptcies from the period, drawn from numerous sources, including records from the major stock exchanges, Dun and Bradstreet's private files, the Securities and Exchange Commission (SEC), prior research (Altman, 1983), and newspaper articles indexed in Funk and Scott's Index of Corporate Change. To be chosen for the sample, the bankrupts must have been listed in the Dun and Bradstreet (D&B) Directory of Corporate Management for five years prior to the bankruptcy, and the bankruptcy must not have been "intentional." The D&B source was necessary for other parts of this study not reported here. Of the 63 firms that met our initial criteria, one "intentional" bankruptcy was eliminated. Intentional bankruptcies were defined as those used as legal strategies to cope with (1) onerous labor unions, (2) substantial antitrust or product liability suits, or (3) hostile takeover attempts.
Each bankrupt firm was then matched with a surviving firm that was closely positioned in size and scope of activities and in the same environment five years prior to the bankruptcy (T - 5). Thus, we matched the firms with respect to industry (product/market mix) and position (sales size) in the industry. Second, we sought to ensure that the matches were prospering (i.e., not in the process of failing) by selecting only matches that (a) exhibited profitability in the top half of their industries in the year of the bankruptcy or (b) survived for at least three years beyond the bankruptcy date. The matching process was performed by three independent judges, as described in Appendix A.
The 57 matched pairs came from the following sectors: 33 manufacturing, 16 retailing, and 8 transportation. The mean sales for the 114 firms at T-5 was $404 million and their mean age was 56 years, indicating that these were well-established firms.
The success of the matching process was verified by a logistic regression analysis performed on the bankrupts and their matches. This analysis showed that the bankrupts and survivors were operating in the same environments and in relatively the same position within those environments in all periods, T-5 to T-1. There were no significant differences with respect to the real demand growth rate and demand variability in the primary and secondary lines of business of the bankrupts and their matches or with respect to their size (as measured by total assets, sales, or number of employees).
Comments on the sampling procedure. Matched-pair sampling procedures are sometimes criticized because they may involve sampling on the dependent variable (Berk, 1983). This is less problematic in this study because our dependent variable is the focus of attention of managers and, hence, we are not attempting to predict bankruptcy. In addition, the most obvious advantage of the matching procedure is that it provides controls for confounding factors, such as industry and position within the industry. Moreover, in many cases, using matched-pair designs may be the only practical way to study a rare phenomenon (Zmijewski, 1984), especially when the costs and availability of data prohibit large random samples, as they might in this study. See Hambrick and D'Aveni (1988) for a summary of the limitations and advantages of this sample.
Measuring attention. Attention was measured as the attention paid to a specific element of the environment in a letter to shareholders. We counted the number of sentences in each letter referring to seven elements of the environment: customers, suppliers, creditors, owners, employees, top managers, and general economic conditions that affect demand. General economic conditions affecting demand included references to business cycles, interest rates, inflation, and unemployment.
We coded a total of 490 of the 570 shareholder letters potentially available for the 114 firms in the sample over the five-year period leading up to the bankruptcy and over the same five-year period for each matched survivor. A small number of missing letters (21) from the survivors resulted from discrepancies in the data sources that made the letters unavailable. The larger number of missing letters from the failures (59) may have resulted because failing firms did not write letters or they delayed their publication, sometimes with promises that they would release them in the near future. Once in bankruptcy, these firms were exempt from SEC filings and, hence, may never have released a letter for the period studied. We took great pains to maximize the number of letters coded, including going to the SEC's Washington archives, searching court records, purchasing letters from commercial suppliers of letters, and tracking down surviving firms and remnants of the bankrupts for copies of old letters.
While content analysis does not identify the reason for mentioning any particular aspect of the environment, it indicates that managers are paying attention to it according to the definition provided by Sproull (1984: 10). Since it is impossible to know who or how many people were involved in drafting the letters, we decided to take the conservative approach taken by Salancik and Meindl (1984), among others, who attributed the letters to top managers as a group. Chief executive officers must sign the letters and are held fiduciarily responsible for their accuracy, so, at a minimum, they can be assumed to have final say over their contents. in addition, since letters are widely distributed to security analysts, stockholders, creditors, and other constituents of the firm, there are legal and practical reasons that force top managers to review their general tone, perceived accuracy, and emphasis.
Content analysis procedure. The letters were evenly divided between two teams, composed of three business Ph.D. or M.B.A. students each. In total, the teams coded 22,593 sentences, using uniform instructions to decide whether each sentence referred to a particular element of the environment. These instructions provided definitions of the seven elements and a list of common synonyms. After the sentences had been coded independently, disputed sentences were negotiated among the members of each team. A description of the methods used to increase coding reliability is contained in Appendix B.
To test the reliability of the results, the two teams coded an overlapping subsample of 18 letters. As indicated in Table 1 [omitted], the two teams reached substantially the same conclusions for all the variables. (1) Table 1 [omitted] also reports independent and post-negotiation disagreement rates, which are discussed in Appendix B.
Measures of output, input, and internal attention. Based on a stakeholder approach to organizations (e.g., Freeman, 1984) and open-systems theory (Katz and Kahn, 1966), we combined the seven elements of the environment into three summative measures, as follows: (1) internal attention--the number of sentences referring to owners, employees, and top managers; (2) output attention--the number of sentences referring to customers or general economic factors affecting demand; and (3) input attention--the number of sentences referring to creditors and suppliers. The internal category included the participants who share in the control of the organization. Stockholders (owners) were not included in the input category because, even though they provide capital, they differ from input providers in that they also participate in the legal control of the firms. Rivals and unions were not included in any of these measures, even though we measured references to rivals and unions. The number of sentences referring to rivals was not included because it tended to make the input and output measures too highly correlated, perhaps because rivals can be perceived as affecting both the input and output environments. Similarly, the number of sentences referring to unions was not included in any of the measures because it tended to make the input and internal measures too highly correlated, perhaps because a union can be perceived as either a supplier or as an internal organizational participant.
In general, the three measures of attention were independent constructs and letters that frequently mentioned one of the three aspects of the environment did not necessarily mention the others frequently. In an analysis not reported here, we obtained correlation coefficients for the bankrupt and matched firms for the three attention measures in absolute terms (i.e., the actual number of sentences referring to the output, internal, and external environments). These correlations indicated that the three measures were orthogonal in most years and, in years in which they were correlated, the correlations were not high enough to indicate that the three measures should be treated as if they were one construct. (2)
We measured attention in four ways: (1) absolute attention was measured simply by the number of sentences referring to a facet of the environment, say output; (2) proportionate attention was used to control for letter length by counting sentences devoted to a facet of the environment, say output, as a proportion of the total number of sentences in the letter; (3) focus of attention measured how much management focused the letter on a particular facet of the environment, say output, measured by sentences devoted to output as a proportion of only those sentences devoted to the environment (i.e., output + input + internal); and (4) emphasis, which was measured in two ways, captured the relative allocation of attention among alternative aspects of the environment, to see where more focus was placed by the managers. The two emphasis measures were (1) external/internal emphasis and (2) output/input emphasis. External/internal emphasis was measured to tap whether management focused more attention on the external environment than on the internal environment. We first determined output, input, and internal attention for each letter and then calculated the relative emphasis on external over internal as follows: (output focus plus input focus) minus internal focus. A high score on external/internal emphasis means management emphasized the external environment over the internal environment. Output/input emphasis was used to assess whether management focused its attention on the output aspects rather than the input aspects of the external environment. We calculated the relative emphasis on output over input as follows: output focus minus input focus. A high score on output/input emphasis means management emphasized the output environment over the input environment.
The absolute and proportionate measures of attention distinguish between "amount" and degree," as Staw, McKechnie, and Puffer (1983) did in their study of self-serving attributions. The focus measures are consistent with Sproull's (1984) definition of attention, which treats the total amount of attention as finite and distributed among the three aspects of the environment. The emphasis measures are based on measures of focus, but the qualitative conclusions that can be drawn from the results presented below are the same irrespective of whether emphasis is based on measures of focus, absolute, or proportionate attention.
The external/internal emphasis variable allowed us to use t-tests to see if the failing and surviving firms emphasized the external environment over the internal environment and then to use matched-pair t-tests to see whether they did so more than their matched survivors. In a similar vein, the output/ input emphasis variable allowed us to use t-tests to see if the failing and surviving firms emphasized the output portion of the external environment over the input portion of the external environment and then to use matched-pair t-tests to see whether they did so more than their matches did.
Descriptive Statistics of the Sample
Statistical analysis of the sample demonstrated that the sample satisfied the two key criteria related to the theories cited above: (1) output crisis and (2) performance characteristics.
Output crisis. The first criterion was that both bankrupts and their matches should experience an external (in this case, output) crisis (i.e., declining or maturing demand). For failing firms real demand growth rates declined significantly (p < .001) from T-5 (mean = 6.33, s.d. = 10.62) to T-1 (mean = - .91, s.d. = 9.87). Similarly, for the successful firms, real demand growth rates declined significantly (p < .01) f rom T - 5 (mean = 5.73, s.d. = 10.4) to T - 1 (mean = -.76, s.d. = 9.01). The environmental downturns for both bankrupts and survivors were most abrupt going f rom year T - 3 to T - 2, with the T - 2 demand growth rate of zero significantly (p < .05) lower than the year T - 3 rate of approximately 7 percent.
Performance characteristics. The second criterion was that the failing firms should exhibit performance downturns that the successful firms should avoid. The return on assets (ROA), net income over total assets, of failing firms declined significantly (p < .0001) from T - 5 (mean = .01, s.d. = .07) to T- 1 (mean = -.14, s.d. = .19). Most of the performance downturn occurred during T - 2 and T - 1. In contrast, the survivors' performance did not change significantly from T - 5 to T- 1. A second performance downturn experienced by the bankrupts was increased ROA instability, namely a jump in the intertemporal coefficient of variation in ROA plus a constant, i.e., [variance in: ROA + 1000]/[mean of: ROA + 1000]. The constant was added to compensate for the negative and zero returns of the failing firms. Sensitivity analysis indicated that the size of the constant did not influence the conclusions that follow. From T - 5 to T - 3 (i.e., before the crisis), there was no difference between the survivors' and failing firms' average performance instability. However, in the period of crisis from T - 3 to T - 1, the failing firms showed significantly more performance instability than the surviving firms (p < .001). Furthermore, the failing firms experienced significantly (p < .01) more performance instability during T - 3 to T- 1 than they did from T - 5 to T - 3, while the survivors showed no change in performance instability from the early period to the later period.
Top-Management Attention Measures
Table 2 [omitted] reports the means and standard deviations for the absolute attention, proportionate attention, and focus of attention measures for bankrupts and survivors for each of the five years preceding the bankruptcy of the failing firms. The table also includes the results of t-tests comparing the bankrupt and survivor means in each time period. Nonparametric tests yielded similar qualitative conclusions as those shown in Table 2.
Table 2 [omitted] shows that the bankrupts and survivors used letters of similar length, except in T - 2, when the bankrupts used somewhat longer letters. Thus, in general, differences in letter length do not appear to be significant. Table 2 also reports input, output, and internal attention in absolute, proportionate, and focus terms. in general, before T-3, the bankrupt and survivor averages are not significantly different. However, after the crisis has struck, by T - 1 (1) the survivors pay more attention to the output environment than the bankrupt firms do, and (2) the bankrupt firms pay more attention to the input and internal environment than the survivors do. No matter which method of measuring attention is chosen, these results were similar. Using attention to the seven elements of the environment, measured as a proportion of the number of sentences referring to any of these seven elements, we also did a more detailed investigation into which of the seven aspects of the environment contributed the most to the input and internal orientations of the managers of failing firms and the output orientation of the managers of surviving firms in years T - 2 and T - 1. Paired-comparison t-tests indicated three results: (1) Less attention both to customers and to general economic conditions contributed most to the lower output orientation of letters of failing firms (compared to surviving firms) in both T- 1 and T - 2; (2) Higher attention both to creditors and to suppliers contributed most to the higher input orientation of failing firms (compared to surviving firms) in both T - 2 and T - 1, attention that creditors and suppliers are likely to demand when their customer is heading for bankruptcy; and (3) Higher attention to top managers contributed most to the higher internal orientation in the letters of failing firms (compared to surviving firms) in T- 1. The only inconsistent finding was that, in T-2, the managers of failing firms paid less attention to stockholders than did the managers of surviving firms. Perhaps, this finding is not surprising, given that bankruptcies can be viewed as resulting from agency problems D'Aveni, 1989).
Overall, the t-test results in Table 2 support proposition 2c, which asserted that, compared with managers of failing firms, the managers of surviving firms were more output oriented and less input and internally oriented. Furthermore, with only a small number of exceptions, none of these differences are observed prior to T - 2, when the output crisis had either first struck or had intensified.
The above results do not address the emphasis of management, as the external/internal and output/input emphasis variables do. We therefore turn to the emphasis measures.
External/internal emphasis in the letters. The results for external/internal emphasis and output/input emphasis measures are shown in Table 3 [omitted]. Proposition 1a stated that after an external crisis hits, managers of surviving firms will emphasize the external environment over the input environment, while the managers of failing firms will not. The results, as shown in Table 3 [omitted], are consistent with this proposition.
Until the crisis occurred, both bankrupts and survivors maintained a "balance" between external and internal emphasis (i.e., the external/internal emphasis measures do not differ significantly from zero.) However, in T - 1 after the crisis has struck, survivors are found to have significantly (p < .05) transferred their emphasis to the external environment over the internal. Moreover, consistent with proposition 1b, the survivors are found to have emphasized the external environment marginally (p < .10) more than the bankrupts have.
Output/input emphasis in the letters. Propositions 2a and 2b stated that, under normal circumstances and after an output crisis hits, surviving firms will emphasize the output environment over the input environment, while bankrupts will gradually stop emphasizing the output environment over the input
environment. The results for T - 5 to T - 1 in Table 3 are consistent with these propositions. Year after year, and even after the output crisis has struck, survivors are found to have consistently and significantly (p < .001) placed more emphasis on output than input.
In addition, in four of the five years, survivors emphasized the outputs over inputs significantly more than the bankrupts. In contrast, the bankrupts started off with greater emphasis on the output environment than the input environment, but they gradually became distracted by the input environment and shifted their emphasis to it. Table 4 [omitted] summarizes the results of the analysis of emphasis.
The gradual changes in emphasis found in Table 3 prompted further investigation, so we explored the changes in focus that led to the shifts in emphasis.
Changes in focus. The shifts in output/input emphasis resulted from distinctly different shifts in output focus by survivors and input focus by bankrupts. From T - 3 to T - 1 the managers of surviving firms significantly increased their focus on outputs (mean increase = .061, s.e. = .030, p < .05). The largest contributor was the increase in focus on the general economy (mean increase = .066, 5.c. = .024, p K .01), which is an appropriate concern for anyone facing a demand decline. This increased focus on outputs occurred at the expense of focus on the internal environment. From T - 5 to T - 1, the survivors significantly decreased their internal focus (mean decrease = -.071, s.e. = .035, p < .05)
From T - 5 to T - 1, however, managers of failing firms significantly increased their input focus (mean increase = .064, s.e. = .030, p < .05). Not surprisingly, the largest contributor to this change was an increase in the focus on creditors (mean increase = .082, s.e. = .029, p < .01). Furthermore, from T - 5 to T - 1, the managers of failing firms significantly decreased their focus on customers (mean decrease = -.061, s.e. = .030, p < .05). Thus, although both survivors and bankrupts emphasized the output environment at first, the bankrupts apparently became increasingly embroiled with creditors and reduced their focus on customers. Meanwhile, survivors decreased their focus on internal factors and increased their focus on the output side--and the economy, which is the source of declining demand.
These intertemporal comparisons yielded findings that are consistent with the cross-sectional, cross-firm results. The comparisons to historical levels of attention indicate that the survivors become more focused on the output environment, while the failures become more focused on the input environment. Although the reasons for the above shifts in focus cannot be discerned from this study, it indicates that successful and failing firms may respond differently to a crisis of declining demand growth. Somehow managers whose firms succeed are able to maintain an external and output orientation, while those whose firms fail end up locked into internal, input, and creditor-driven concerns.
DISCUSSION AND CONCLUSIONS
Given the potential for bias in the estimates that emanate from nonrandom samples, the results of this study are limited and should be construed as indicative of a need for more research that will allow us to reconcile the various theories in more detail. We are able to assert that we found only an association between output crisis and output focus in the letters of successful firms. Further research on much larger and random samples is needed to establish causal direction.
The underlying reasons for the differences between surviving and failing firms are also not at all clear. They may involve the effects of problems normally associated with performance decline, but which we could not identify, perhaps even coupled with the pressures for impression management that occur during a crisis or declining performance.
In addition, since a sample of successful turnarounds was not examined, no detailed observations can be made about avoiding failure. Future research should, for example, explore whether an input focus or an internal focus contributes to, or interferes with, a troubled firm's adaptive behavior. Perhaps an output focus helps managers to identify the need for significant strategic changes more quickly because perceptions influence strategic planning and goal-setting processes (Barnes, 1984; Schwenk, 1984; Bourgeois, 1985).
The results of this study do suggest, however, that previous turnaround studies (e.g., Schendel and Patton, 1976; Schendel, Patton, and Riggs, 1976; Hambrick and Schecter, 1982) might have ignored the importance of crisis-induced perceptual shifts among top managers that could influence turnaround success. As Weitzel and Jonsson (1989) suggested, future research should be directed toward understanding how inadequate scanning behavior, short-term planning, and output-crisis denial inhibit turnaround efforts.
Moreover, the results are consistent with the suggestion that, for reasons yet to be determined, firms that fail in the face of an externally induced crisis behave very differently from firms that survive. The annual report can be seen as a series of snapshots that capture the amount of top-management attention devoted to various facets of the firm's environment year by year. Most of the differences between survivors and bankrupts are minimal until after the demand crisis has struck. In T - 2 and T - 1, after the output crisis has struck, the snapshots show distinctly different patterns of attention by survivors and bankrupts.
While the findings do not resolve the conflicting explanations proposed by the various theories, they do suggest that the various theories may converge on a single phenomenon. It may be that some of the theories operate simultaneously or that they describe different aspects of a larger picture. For example, in reconciling the predictions of the various theories, it is now possible to understand how a maladaptive firm might deny a crisis and simultaneously experience threat-rigidity response. Crisis denial may cause the firm to maintain a balanced emphasis on the external and internal environments, but threat-rigidity response may induce a shift in the emphasis on the output and input environments. Both are simultaneously possible because maladaptive firms reallocate their attention from the output to the input portion of the external environment.
If the bankrupts had emphasized the external environment over the internal environment after the crisis hit, then there would be no reason to pursue further the argument that bankrupts could fall prey to crisis denial, while survivors are able to increase their scanning vigilance and adapt. Furthermore, if the input emphasis of the bankrupts did not exceed that of the survivors, there would be no reason to pursue further the argument that bankrupts could be the victims of threat-rigidity response or distracted by the short-term demands of their creditors and suppliers, while survivors are somehow able to avoid this and focus on the real, long-term source of the problem-declining demand. The fact that these patterns did occur indicates that the postulated reconciliation among several apparently conflicting theories is worth further investigation.
The results are also important for studies of impression management. The emphasis data indicate, for example, that surviving firms emphasized the external environment over the internal environment in T - 1, while the failing firms did not do so. These results appear contrary to the findings of Bowman (1976, 1978), Bettman and Weitz (1983), Staw, McKechnie, and Puffer (1983), and Salancik and Meindl (1984), who found that troubled firms attributed their poor performance to the external environment more than successful firms did. However, these previous studies looked at excuse making and self-serving attributions, not attention, as defined here. When we stripped away the causal attribution language from references to various aspects of the environment, we found results that were not predictable from the findings of these earlier studies of impression management.
At present, no specific impression-management theory (e.g., Schlenker, 1980) predicts what topics will be attended to during crises. Managers may deny or ignore crises, for example, by omitting references to them for the purposes of avoiding raising problematic issues in the minds of the readers or avoiding the alarming appearance that the situation is not as favorable as in the past. However, our results seem to indicate that this did not occur. Both the bankrupts and survivors seem to address the most pressing crisis that they face by paying attention to it. This conforms to the legal norms set by the SEC and the expectations of investors, who demand that managers reveal their knowledge of the most important factors affecting the value of stocks and bonds on public exchanges. This also suggests that managers may have little discretion with respect to the symbolic manipulation and distortion of the focus of attention of their public communications, even though they may attempt to put the best face on the problems that they emphasize in their communications.
Even the lack of focus on the output environment by the managers of failing firms may reflect a distorted view of the environment rather than a deliberate attempt to deceive stakeholders about the conditions of the market. Furthermore, it would be difficult to suggest that changing the content of letters to shareholders would have changed the fate of the firm. It seems highly unlikely that the letters could manage the impressions of stakeholders so well that they would prevent bankruptcy. Thus, the results do not provide any unequivocal evidence that managers can or do intentionally distort the focus of their letters so as to manipulate and reassure stakeholders. More research is needed to determine whether the "attention" phenomena observed here are explained by impression management.
Finally, this study might be used to guide further work on reconciling the various theories related to the attention of managers in trouble. One possibility would be an in-depth study on the effects of crises as reflected in the internal communications of managers in firms in decline. Studies of less public internal communications in field settings may be promising. Assuming that a more private set of communications--such as intraorganizational memos--were available from bankrupt firms, one could investigate the shifting focus of top managers' more private thoughts and behaviors. Moreover, careful selection of field settings would allow a test of the impact of organizational structure, information-processing capabilities, and intraorganizational power distributions on the focus of attention of top managers. Careful selection of different field settings would also allow a test of different types of environmental adversity that might pose qualitatively different threats and trigger different crisis reactions. Our sample was based primarily on declining industries, but new field settings might include highly competitive but growing industries, stable but stagnant environments, and predictably cyclical industries entering a downturn. Assessing the generalizability of our findings will require random samples of new types of communications in different crisis settings. Such future efforts will add considerably to our understanding of the relationships between information-processing capabilities, focus of top-managerial attention, crisis reaction, adaptation, and organizational survival and should contribute to an integration of the various decline, crisis, and adaptation theories surveyed in this study.
1. Spearman correlations were slightly lower, but they did not yield different conclusions.
2. Results are available from the first author.
REFERENCES Aguilar, Francis J. 1967 Scanning the Business Environment. New York: Mac-Millan. Altman, Edward I. 1983 Corporate Financial Distress. New York: Wiley. Anderson, Carl R. 1976 "Coping behaviors as intervening mechanisms in the inverted U, stress-performance relationship." Journal of Applied Psychology, 61: 30-34. Andrews, Kenneth R. 1980 The Concept of Corporate Strategy, rev. ed. Homewood, IL: Richard D. Irwin. Argenti, John 1976 Corporate Collapse. New York Halstead. Barnard, Chester I. 1938 The Functions of the Executive. Cambridge, MA: Harvard University Press. Barnes, James, H., Jr. 1984 "Cognitive biases and their impact on strategic planning." Strategic Management Journal, 5: 129-137. Berger, Peter L., and Thomas Luckmann 1967 The Social Construction of Reality. Garden City, NY: Anchor Books. Berk, Richard A. 1983 "An introduction to sample selection bias in sociological data." American Sociological Review, 48: 386-398. Bettman, James R., and Barton A. Weitz 1983 "Attributions in the board room: Causal reasoning in corporate annual reports." Administrative Science Quarterly, 28: 165-183. Bourgeois, L. J., Ill 1985 "Strategic goals, perceived uncertainty, and economic performance in volatile environments." Academy of Management Journal, 28: 548-573. Bowman, Edward H. 1976 "Strategy and the weather." Sloan Management Review, 17(2): 49-62. 1978 "Strategy, annual reports and alchemy." California Management Review, 20(3): 64-71. Bradley, Patricia Hayes, and J. E. Baird, Jr. 1977 "Management and communication style: A correlation analysis." Central States Speech Journal, 28: 194-203. Choe, Man Kee 1987 "Organizational performance and communication contents in annual reports." Unpublished dissertation, Graduate School of Business, New York University. Chomsky, Noam 1972 Language and Mind. New York: Harcourt Brace Jovanovich. Daft, Richard L., Juhani Sormunen, and Don Parks 1988 "Chief executive scanning, environmental characteristics, and company performance: An empirical study." Strategic Management Journal, 9: 123-139. D'Aveni, Richard A. 1989 "Dependability and organizational bankruptcy: An application of agency and prospect theory." Management Science, 35:1120-1138. 1990 "Top managerial prestige and organizational bankruptcy." Organization Science, 1(2): 121 -142. Dutton, Jane E. 1986 "The processing of crisis and non-crisis strategic issues." Journal of Management Studies, 23: 501 -517. Evan, William M. 1966 "The organization-set: Toward a theory of interorganizational relations." In James D. Thompson (ed.), Approaches to Organizational Design: 173-191. Pittsburgh: University of Pittsburgh Press. Freeman, R. Edward 1984 Strategic Management: A Stakeholder Approach. Boston: Pitman. Gamson, William A., and Norman R. Scotch 1964 "Scapegoating in baseball." American Journal of Sociology, 70: 69-76. Goodman, Ronald 1980 "Annual reports serving a dual marketing function-Report as survey." Public Relations Quarterly, 36(Summer): 21-24. Hall, Douglas T., and Roger Mansfield 1971 Organizational and individual response to external stress." Administrative Science Quarterly, 16: 533-547. Hambrick, Donald C. 1981 "Environment, strategy, and power within top management teams." Administrative Science Quarterly, 26: 253-275. 1982 "Environmental scanning and organizational strategy." Strategic Management Journal, 3: 159-174. Hambrick, Donald C., and Richard A. D'Aveni 1988 "Large corporate failures as downward spirals." Administrative Science Quarterly, 33: 1-23. Hambrick, Donald C., and David Lei 1985 "Toward an empirical prioritization of contingency variables for business strategy." Academy of Management Journal, 28: 763-788, Hambrick, Donald C., and Steven Schecter 1982 "Turnaround strategies for mature industrial product business units." Academy of Management Journal, 26: 230-248. Hannan, Michael T., and John H. Freeman 1977 "The population ecology of organizations." American Journal of Sociology, 82: 929-964. Harrigan, Kathryn Rudie 1980a Strategies for Declining Businesses. Lexington, MA: Lexington Books. 1980b "Strategy formulation in declining industries." Academy of Management Review, 5: 599-604. Hofer, Charles W. 1975 "Toward a contingency theory of business strategy." Academy of Management Journal, 18: 784-810. Holsti, Ole R. 1968 "Content analysis." In Gardner Lindzey and Elliot Aronson (eds.), The Handbook of Social Psychology, 2d ed., 2:596 - 692. Reading, MA: Addison - Wesley. 1969 Content Analysis for the Social Sciences and Humanities. Reading, MA: Addison - Wesley. 1978 "Limitations of cognitive abilities in the face of crisis." Journal of Business Administration, 9(Spring): 39-55. Janis, irving L., and Leon Mann 1977 Decision Making: A Psychological Analysis of Conflict, Choice, and Commitment. New York: Free Press. Katz, Daniel, and Robert L. Kahn 1966 The Social Psychology of Organizations. New York: Wiley. Kiesier, Sara, and Lee Sproull 1982 "Managerial response to changing environments: Perspectives on problem sensing from social cognition." Administrative Science Quarterly, 27: 548-570. Kress, Gunther, and Robert Hodge 1979 Language as Ideology. London: Routiedge and Kegan Paul. Lazarus, Richard S., and Susan Folkman 1984 Stress, Appraisal, and Coping. New York: Springer. Lentz, Arthur, and Harvey Tschirgi 1963 "The ethical content of annual reports." Journal of Business, 36: 387-393. March, James G., and Herbert A. Simon 1958 Organizations. New York: Wiley. Mintzberg, Henry 1978 "Patterns in strategy formation." Management Science, 24:934-948. Mobley, William H., Roger W. Griffeth, Herbert H. Hand, and B. M. Meglino 1979 "Review and conceptual analysis of the employee turnover process." Psychological Bulletin, 86: 493-522. Peters, Thomas J., and Robert H. Waterman, Jr. 1982 in Search of Excellence: Lessons from America's Best - Run Companies. New York: Harper and Row. Petzinger, Thomas, Jr. 1982 "Double talk grips business reports as firms try to sugarcoat bad news." Wall Street Journal, March 31: 31, column 4. Pfeffer, Jeffrey 1981 "Management as symbolic action: The creation and maintenance of organizational paradigms." In L. L. Cummings and Barry M. Staw (eds.), Research in Organizational Behavior, 3: 1 -52. Greenwich, CT: JAI Press. Pondy, Louis R., and Ian I. Mitroff 1979 "Beyond open systems models of organizations." in L. L. Cummings and Barry M. Staw (eds.), Research in Organizational Behavior, 1: 3-39. Greenwich, CT: JAI Press. Salancik, Gerald R., and James R. Meindl 1984 "Corporate attributions as strategic illusions of management control." Administrative Science Quarterly, 29: 238-254. Sapir, Edward 1956 Culture, Language and Personality: Selected Essays. David G. Mandelbaum, ed. Berkeley, CA: University of California Press. Schein, Edgar H. 1978 Career Dynamics: Matching Individual and Organizational Needs. Reading, MA: Addison-Wesley. Schendel, Dan E., G. R. Patton, an James Riggs 1976 "Corporate turnaround strategies: A study of profit decline and recovery." Journal of Economics and Business, 28(3): 236-241. Schendel, Dan E., G. R. Patton, an James Riggs 1976 "Corporate turnaround strategies: A study of profit decline and recovery." Journal of General Management, 3(3): 3-11. Schlenker, Barry R. 1980 Impression Management: The Self-Concept, Social Identity, and Interpersonal Relationships. Monterey, CA: Wadsworth. Schoderbek, Peter P., Asterios G. Kefalas, and Charles G. Schoderbek 1975 Management Systems: Conceptual Considerations. Dallas: Business Publications. Schwenk, Charles R. 1984 "Cognitive simplification processes in strategic decision-making." Strategic Management Journal, 5: 111-128. Silverman, David 1971 The Theory of Organizations. New York: Basic Books. Simon, Herbert 1971 "Designing organizations for an information-rich world." In Martin Greenberger (ed.), Computers, Communications and the Public Interest: 37-52. Baltimore: John Hopkins Press. 1973 "Applying information technology to organization design." Public Administration Review, 33: 268-278. Smart, Carolyne, and Ilan Vertinsky 1977 "Designs for crisis decision units." Administrative Science Quarterly, 22: 640-657. Sproull, Lee S. 1984 "The nature of managerial attention." In Lee S. Sproull and Patrick D. Larkey (eds.), Advances in Information Processing in Organizations, 1: 9-27. Greenwich, CT: JAI Press. Starbuck, William H., Arent Greve, and Bo L. T. Hedberg 1978 "Responding to crisis." Journal of Business Administration, 9(2): 112-137. Staw, Barry M., Pamela I. McKechnie, and Sheila M. Puffer 1983 "The justification of organizational performance." Administrative Science Quarterly, 28: 582-600. Staw, Barry M., Lance E. Sandelands, and Jane E. Dutton 1981 "Threat-rigidity effects in organizational behavior: A multi-level analysis." Administrative Science Quarterly, 26: 501 -524. Sutton, Robert I. 1990 "Organizational decline processes: A social psychological perspective." In Barry M. Staw and L. L. Cummings (eds.), Research in Organizational Behavior, 12: 205-253. Greenwich, CT: JAI Press. Thompson, James D. 1967 Organizations in Action. New York: McGraw-Hill. Tushman, Michael L., and Elaine Romanelli 1985 "Organizational evolution: A metamorphosis model of convergence and reorientation." In L. L. Cummings and B. M. Staw (eds.), Research in Organizational Behavior, 7: 171-222. Greenwich, CT: JAI Press. Watziawick, Paul, Janet H. Beavin, and Don D. Jackson 1967 Pragmatics of Human Communication: A Study of interactional Patterns, Pathologies, and Paradoxes. New York: W. W. Norton. Weitzel, William, and Ellen Jonsson 1989 "Decline in organizations: A literature integration and extension." Administrative Science Quarterly, 34: 91-109. Whetten, David A. 1980 "Sources, responses, and effects of organizational decline." in John R. Kimberly, Robert H. Miles and Associates (eds.), The Organizational Life Cycle: Issues in the Creation, Transformation and Decline of Organizations: 342-374. San Francisco: Jossey-Bass. Whorf, Benjamin L. 1956 Language, Thought, and Reality: Selected Writings of Benjamin Lee Whorf. John B. Carroll, ed. Cambridge, MA: MIT Press. Zammuto, Raymond, and Kim S. Cameron 1985 "Environmental decline and organizational response." In L. L. Cummings and Barry M. Staw (eds.), Research in Organizational Behavior, 7: 223-262. Greenwich, CT: JAI Press. Zmijewski, Mark E. 1984 "Methodological issues related to estimation of financial distress prediction models." Journal of Accounting Research, 22 (Supplement): 59-86.
APPENDIX A: The Matching Process
The matching process was done in three stages by three independent judges: a senior strategy professor and the two authors. First, a research assistant identified a comprehensive set of potential matches for each bankrupt. The set consisted of all firms of roughly the same size as the bankrupt in T - 5 (five years prior to the bankruptcy), as listed in Ward's Business Directory under the firm's top two (by sales volume) 3-digit Standard Industry Classification (SIC) codes and Standard and Poor's published lists of firms by 4-digit SIC codes. As with the bankruptcies, a potential match was excluded if it was not listed in the Dun and Bradstreet Directory of Corporate Management The research assistant was otherwise lenient, such that each list usually consisted of between 5 and 20 potential matches. Data on each firm's size, products, and markets (from Moody's Industrial and OTC Manuals) for year T- 5, as well as its profitability in year T were included in a comprehensive file.
Second, the three judges worked independently to select their first and second choice for matches from each file. The judges considered size and product/market mix (both in T - 5) as co-equal primary matching criteria. Then they turned to profitability in year T and ruled out poor performers in that year. The profitability condition was imposed to eliminate survivors in year T that might have gone bankrupt after the time frame of the study.
The independent matching process resulted in agreement in first choices of at least two of the three judges for 53 of the cases. Of these 53 cases, all three judges selected the same first choice in 28 cases and, in 18 more cases, the dissenting judge had the majority's choice as his second choice. The final step was to resolve the nine disputed cases. The judges reached a negotiated consensus for four of the cases but concluded that the other five had no satisfactory matches.
A fuller description of the sampling and matching procedures, plus a list of the bankrupt firms and their matches, is provided in Hambrick and D'Aveni (1988).
APPENDIX B: The Content-Coding Process
Several steps were used to control coding reliability. First, coders were instructed to read each sentence separately and to decide if it made reference to a stakeholder, then to reread the sentence to see if it made reference to the next stakeholder. Coding was thus done as a set of simple, sequential, dichotomous choices to reduce coding complexity: reference to a specific stakeholder versus no reference to the stakeholder.
Second, the dichotomous choice process was carefully controlled. Coding instructions were written to reduce the coding task to a clerical activity. The instructions provided a list of synonyms for each stakeholder that were to be coded as an explicit reference to that stakeholder. Frequently occurring borderline cases were identified by multiple coders who pretested the rules on 50 letters from firms not in the sample. Explicit coding rules were devised to resolve these problems. (These coding instructions are available from the first author.)
The general coding rule was to code only explicit references to a stakeholder. Three exceptions to this general rule were allowed: (1) References back: Every sentence with an implicit reference that resulted from referring back to another sentence mentioning the stakeholder was coded as a separate reference to the stakeholder. For example, customers are mentioned in both of these sentences: "Customers are important to us. They are our life blood." (2) Inclusion under a title: Every sentence that was contained in a section of a letter that was headed with a boldface or underlined title referring to a stakeholder was coded as a reference to that stakeholder unless the sentence was clearly irrelevant to the title. (3) Actions by external stakeholders: Actions taken by an external stakeholder were coded as references to the stakeholder when the implicit reference was unambiguous. For example, this sentence was coded as an implicit reference to suppliers: "Faster methods of delivering raw materials are being employed everyday."
Two types of frequently occurring, implicit references to stakeholders were specifically not coded because of their ambiguity: (1) Actions of the focal firm implying an internal stakeholder: Actions by the focal firm imply someone was the actor, However, we did not code implicit references to internal stakeholders unless an internal social entity was explicitly mentioned in connection with the action. Consider the following: "We undertook an advertising campaign." in this case, it was impossible to tell if the implicit actors were marketing department employees or an outside advertising agency. Given the ambiguity, it was decided not to code such cases at all. (2) Actions by the focal firm implying an external stakeholder: Actions by the focal firm sometimes imply that someone outside the firm was affected by the action. However, we did not code these implicit references unless an external social entity or group was explicitly mentioned. For example, "Our new buying practices have affected our inventory shortage problems." was not coded as a reference to suppliers because it was hard to say if suppliers were the object of the sentence. The inventory may have been finished goods, affecting delivery to customers. Creditors using inventories as collateral may be affected. Given the ambiguity, it was decided not to code such cases at all.
As can be seen, the rules for handling the borderline cases made coding a complex task, especially since it was difficult to provide simple rules for coding all the subtle nuances of words and the multiple ways to express the same thought. Even though the complex coding rules were necessary to limit unpredictable coding judgements, they also created reliability problems. Therefore, a third method to increase reliability was used: simplification of the rules. During the pretest phase, rules creating a lot of disagreement were simplified to improve agreements between coders. Thus, in some cases, less valid (but simpler) rules were adopted to boost reliability. This is a common tradeoff that is made in all coding and scale construction.
Fourth, the problems of coding complexity also were reduced by careful administration of the coding process. Coder recruitment, training, retention, and evaluation were carefully planned. Coders were trained for over a month and kept on the project for two to three semesters. Coders were provided with descriptions of the firms' businesses, products, and internal subunits from Moody's Industrial and OTC Manuals, so that they could familiarize themselves with typical customers and suppliers of the focal firm before starting the coding process. All letters from a given company and its match were assigned to the same coders. Companies in similar sectors or lines of business were assigned to the same coders. Finally, Holsti (1969) recommended firing coders who do not demonstrate adequate skills to comply with the coding instructions. This was done early in the coding phase. We hired three teams, with the plan of firing the least reliable team and told the teams this when they were hired. The fired team's coded data were recoded by the more reliable coders of the remaining two teams.
Fifth, each letter was coded by three independent coders. Sentences were tallied as a reference to the stakeholder only if at least two of the three coders indicated that the sentence referred to the stakeholder. As the coding process continued, the agreement rates between the independent coders were monitored, and big errors and recurrent problems were identified so that they would not be repeated during the independent coding process.
Sixth, a noncoder (fourth person) identified all sentences for which one coder was odd-man-out-for which only one of the three coders indicated that the sentence referred to a stakeholder. Each of these sentences was discussed by the three coders, and a negotiated result was then arrived at. The purpose of the negotiation process was to see if the odd-man-out coder had identified something overlooked by the two other coders and, if not, to clean up flagrant errors made by the odd-man-out coder. They were instructed to use the coding instructions to see what insight it brought to the dispute. The negotiations were also used to discuss borderline, implied references that were not listed in the instructions. However, redefinition of the categories was prohibited during the negotiation process. Each team designated one person who was responsible to keep detailed notes on the negotiated decisions so that the team would be consistent with respect to similar problems encountered when coding the matched firm and other firms in similar lines of business. A fifth person (author of the coding instructions and one of the authors of this paper) frequently observed the negotiation processes and reviewed the post-negotiation results to insure adequate conformity to the coding instructions. Problems were pointed out to the coders so that errors would not reoccur in future coding.
Seventh, all the letters of both the bankrupt and its match were assigned to the same team for coding. This allowed us to control (at least in part) coding biases and errors by using analyses that compared the matched pairs to each other. That is, some biases and errors were controlled because any team-specific systematic coding bias or coding error would be reflected in both the bankrupt and its match.
Eighth, we attempted to eliminate any cultural biases of the coders that might influence the interpretation of the letters. We did this by hiring teams that were culturally, ethnically, racially, and sexually mixed.
Finally, we attempted to insure test-retest reliability by asking the teams to recode a small number of letters after some time had passed. We terminated this effort, because the results for the first few of these letters were almost identical to the previous results and because the total cost of the coding was so high that we were running out of funds.
Overall, this extremely careful, costly, and time-consuming process resulted in very high interrater reliability. The two teams did not perform very differently on their respective halves of the sample. Each of the two teams showed post-negotiation disagreement rates across all variables of less than 1 percent of the 22,593 sentences that were coded. The intercoder disagreement rate is the percentage of all sentences coded in which one coder indicated that the sentence referred to the stakeholder, but the other two disagreed. Table 1 [omitted], above, reported the disagreement rates by variable. It indicates that there was considerable agreement among the coders. For any given sentence, the random probability of one coder affirmatively indicating a reference that the other two did not recognize is .375. The coders did considerably better than random guessing at both the independent and post-negotiation coding stages. That is, in a series of Bernoulli trials, the probability of only one hit in three tries, with each try having a random success rate of .5, is 3/8. Thus over a large number of sentences, the expected disagreement rate from random guessing is 37.5 percent, compared with our less than 1 percent disagreement rate.
The negotiations did not substantially change the results of the independent coding process. Table 1 [omitted] showed that the pre-and post-negotiation results are highly correlated (>.95, p < .0001). Yet the negotiations did clean up the data by reducing the percentage of coded references for which one coder was odd-man-out. The negotiations reduced the odd-man-out rate to 1.6 percent for employees, 1.1 for top managers, 1.8 for customers, 4.6 for suppliers, 1.5 for creditors, .6 for owners, and 3.2 for general economic conditions. The odd-man-out rate is the percentage of all relevant sentences in which one coder indicated that the sentence referred to the stakeholder, but the other two did not. A sentence was considered relevant if at least one coder judged the sentence as referring to the stakeholder.
|Printer friendly Cite/link Email Feedback|
|Author:||D'Aveni, Richard A.; MacMillan, Ian C.|
|Publication:||Administrative Science Quarterly|
|Date:||Dec 1, 1990|
|Previous Article:||Technological discontinuities and dominant designs: a cyclical model of technological change.|
|Next Article:||Comparing organizational sampling frames.|