The Justification of Organizational Performance in Annual Report Narratives

In this paper, the nature of organizational discourse is theoretically underpinned by the concept of self-serving attributions, a type of causal reasoning that allows the writer to take credit for good news and avoid blame for bad news. We incorporated signaling theory to the extant theoretical framework for self-serving attributions in order to develop hypotheses for the expected levels of attributional bias in the justification of organizational performance. A sample of 49 companies was selected, both from a bad year and a good year regarding the capital market context. Each company’s Letter to Shareholders was content analyzed in order to test our propositions concerning the presence and intensity of self-serving attributions in that section of annual reports. The results partially corroborate the proposed theoretical hypotheses, but the sample size is an issue in terms of robustness. Nevertheless, the results indicate that companies attempt to create a positive corporate image to external stakeholders even when negative performance occurs in a clearly favorable external context. Moreover, we observed that companies with positive performance in a good external context blame negative effects on the environment in a proportion equivalent to that observed for companies with positive performance in a bad year.


INTRODUCTION
Self-serving attributional patterns are usually manifested in the justification of organizational performance by a tendency to associate positive events to internal causes and negative events to external ones.The pattern is identified as opportunistic, biased or hedonic, because it allows one to take credit for successes and to avoid blame for failures (Staw, Mckechnie, & Puffer, 1983).It has also been associated with impression management, a term that embraces a diverse set of strategic behaviors aimed at controlling others' perception of oneself (Gardner & Martinko, 1988;Schelenker & Weigold, 1992).
By and large, research on the use of self-serving attributions is based on psychological theories that postulate either motivational or informational explanations for this organizational behavior.The motivational theory is associated with retrospective rationality and ego-defensive behavior, observed in situations of unfavorable outcomes (Bettman & Weitz, 1983;Staw, 1980).An informational explanation has been derived either from bounded rationality premises or from attributional principles of discounting and augmentation (Aerts, 2001;Bettman & Weitz, 1983;Tsang, 2002).The motivational explanation is commonly associated with attempts to manage the corporate image (Staw, 1980;Salancik & Meindl, 1984).The informational explanation, in turn, is based either on biased internal information processing capabilities (Miller & Ross, 1975) or on other reasoning processes related to the interpretation of environmental events (Huff & Schwenk, 1990;Kelley, 1971).
Results from empirical studies have been controversial.Bettman and Weitz's (1983) results pointed in mixed directions, and Staw et al. (1983), Clatworthy and Jones (2003) and Aerts (2005) presented evidence that supports the motivational explanation.Tsang (2002) found evidence to support the informational hypothesis in a sample of Singaporean companies' reports, the same tendency observed by Clapham and Schwenk (1991) in annual reports from US companies.However, Tsang's (2002) rationale is based on cross-cultural variations, while Clapham and Schwenk (1991) suggest an interpretation based on sense making processes.This work is the first to propose and test hypotheses derived from both motivational and informational theories, rather than attempting to eliminate one of them.We combine an analysis of contextual events with propositions from both motivational and informational theories in order to predict the presence and intensity of self-serving attributions in the justification of organizational performance.Contextual events, in the form of mixed combinations of good (bad) year and high (low) performers, are assumed to exert influence on the relative weights of motivational and informational drivers in the process of justification of organizational performance.We also incorporate propositions of signaling theory (Spence, 1973) to the rationale of self-serving attributions.
This study also contributes to the literature on self-serving attributions by providing results produced in a different institutional setting, based on Aerts ' (2005) suggestions that the explanation patterns displayed by companies from different countries are subject to cultural influences.Brazil's institutional setting is usually depicted as a weak institutional environment (Anderson, 1999), and the country is held to have a poor legal regime, enforcement and transparency (Durnev & Kim, 2005).Additionally, Lopes (2006) suggests that Brazilian firms rely on private deals to obtain funding, which reduces the informativeness of accounting reports, and Lopes, Tukamoto and Galdi (2007) conclude that the high level of discretion associated with a poor institutional environment and low level of monitoring creates the conditions for earnings management to emerge.Thus, Brazil's institutional setting leads us to suspect that managers can exercise higher levels of discretion in the process of corporate image management.
Our sample consists of 49 companies from the Economatica database.We analyzed the Letter to Shareholders from the annual reports for each company in the years of 2002, classified as a bad year, and 2003, a good year.Each Letter to Shareholders was content analyzed in order to identify and code attributions presented in sentences that discussed performance issues.
The results partially support the theoretical hypotheses developed here.Nevertheless, we observed that the narrative sections of annual reports in Brazil are marked by the presence of significant levels of self-serving attributions.The results indicate that companies attempt to create a positive corporate image to external stakeholders even when negative performance occurs in a clearly favorable external context.Additionally, we observed that companies with positive performance in a good external context blame negative effects on the environment in a proportion equivalent to that observed for companies with positive performance in a bad year.Staw (1980) and Staw et al. (1983) were probably the first authors to introduce the management of public impressions rationale to analyze the textual portion of corporate annual reports.Using the impression management theory that had been previously developed by psychology researchers, Staw (1980) argued that both individuals and organizations strive for rational and goal-oriented behavior.Nevertheless, actions generally fall short of these ideals, which motivate individuals to rationalize or justify their course of action.The farther the results are from the ideal, the greater the forces that drive the justification process (Staw, 1980).This process involves both self-justification as well as an external form of justification termed impression management (Staw, 1980).

DEVELOPMENT OF HYPOTHESES
Moving from the individual to the organizational level, Staw et al. (1983) tested for the presence of self-serving attributions in organizations' reporting of performance information.One of their research goals was to determine whether self-serving attributions are best explained by either an internal form of justification, expressed by the use of defensiveness attributions, or by an external form, which involves the use of enhancing attributions (Staw et al., 1983).Moreover, defensive attributions are observed as a pattern of crediting positive events to internal sources and negative events to external factors (Staw et al. 1983).Staw et al. (1983) successfully demonstrated the existence of self-serving attributions in the Letter to Shareholders, but they did not find organizational performance to determine causal attributions, as they expected.Bettman and Weitz's (1983) study was centered on the analysis of reasoning patterns in the justification of corporate performance in order to shed light on the nature of self-serving attributions.They developed motivational hypotheses for the use of self-serving attributions based on egodefensive rationalizations.Also, they used Kelley's (1971) attributional principles of discounting and augmentation, which involve the search for plausible arguments to explain the occurrence of performance related events, to develop the informational hypotheses for the use of self-serving attributions.
The informational rationale of discounting and augmentation proposes that, when an unfavorable outcome occurs in a good year there are fewer plausible external causes to assign to the outcome (Bettman & Weitz, 1983).On the other hand, if an unfavorable outcome occurs in a bad year, the role of external causes would be more relevant (Kelley, 1971).Bettman and Weitz (1983) observed the typical self-serving pattern of attributions in the Letter to Shareholders, but neither a purely informational nor a purely motivational explanation was supported by these attributions.The results for unfavorable outcomes supported the informational explanation, while the results for favorable outcomes appeared to be more consistent with the motivational explanation (Bettman & Weitz, 1983).Tsang (2002) analyzed the Letters to Shareholders of 94 firms listed on the Singapore Stock Exchange from 1985, classified as a bad year and 114 companies in 1994, a good year, in an attempt to replicate Bettman and Weitz's (1983) study.According to Tsang, although Bettman and Weitz's data did not allow them to clearly demonstrate that only one hypothesis prevailed, the informational explanation was strongly supported by Tang's data.
In addition, Tsang (2002) provided solid evidence for the informational explanation, and made a significant contribution to the motivational-informational debate by bringing cross-cultural differences to the analysis.He suggested that a critical cultural difference exists between East Asian managers and Western managers in explaining self-serving attributions.Tsang interpreted that Asian managers are prone to adopt a more holistic perspective than their Western counterparts in the decision making process, which results in more objective and higher quality decisions.Also, "Singapore managements, compared with their Western counterparts, tended to have a more stable amount of causal reasoning across different outcome scenarios" (Tsang, 2002, p. 62).
Despite his discussion about cross-cultural differences, Tsang (2002) did not translate it into hypotheses subject to empirical testing.The author also presented motivational and informational motives as rival explanations but he did not report how he tried to eliminate one of them.Moreover, he did not discuss whether the explanations overlap or interact.Salancik and Meindl (1984) presented a longitudinal study that examined the reasons given by CEOs to explain their firms' performance in the Letter to Shareholders over an 18-year period, comparing firms with stable and unstable performance.They observed that, contrary to psychological theories, managers of firms with unstable performance claim responsibility for both positive and negative outcomes more than the managements of firms with stable performance do.Managers of firms with unstable performance also seemed reluctant to attribute poor performance to uncontrollable environmental events.They argued that this provides evidence that, as the lack of real control over organizational outcomes increases, managers of these firms strategically manipulate causal attributions to manage impressions of their control (Salancik & Meindl, 1984).
A tendency of managers to credit themselves for positive outcomes and blame negative effects on the environment was also observed in Salancik and Meindl's (1984) results.The authors argued that the low correlation between attributional tendencies with past performance suggests that these management tendencies are more likely to result from presentational biases.They also suggested that "the evidence points to the possibility that attributional styles result from intentional strategic attempts to create a sense of management's effectiveness and control over the welfare of the corporation" (Salancik & Mendl, 1984, p. 252).Clapham and Schwenk (1991) explored whether the use of self-serving attributions represents attempts to manage corporate image.They investigated annual reports from heavily regulated companies with the premise that "attempts at impression management through the use of self-serving attributions would be more readily detected by the regulatory agency and less likely to be effective" (Clapham & Schwenk, 1991, p. 221).Thus, one should expect a weaker and more subtle pattern of self-serving attributions in regulated industries, which could be interpreted as evidence that the use of these attributions is due to impression management aims (Clapham & Schwenk, 1991).Conversely, if the pattern of attributions is derived from informational issues, the regulatory context should not necessarily affect it and one could expect the same levels of self-serving attributions in annual reports (Clapham & Schwenk, 1991).
Their results (1991, p. 226) showed the same basic pattern of attributions, since they observed that "executives tended to take credit for good outcomes and lay blame on the environment for poor outcomes".They also brought up Huff and Schwenk's (1985) propositions to suggest that the attributional pattern often found in annual reports results from a type of cognitive bias which affects how managers recall events that occurred before positive and negative outcomes.Aerts (1994), more than a decade after Bettman and Weitz (1983) and Staw et al. (1983), was the first author to add an accounting dimension in the research about self-serving attributes.The author proposed the compelling argument that accounting logic is the source of technical-calculatory relationships that can be rhetorically transformed into attributions of causality (Aerts, 1994).For instance, he pointed out that managers explain financial actions and results using the internal logic of the financial accounting model.
In the context of self-serving attributions, Aerts (1994) introduced the accounting bias as "a tendency to explain negative performance more in technical accounting terms" (p.341), with positive performance being expressed more in strict cause-effect terms.His overall results reinforced findings in other studies.He also argued that the accounting explanation interacts with the self-serving attributional pattern, in which the former "obscures the perception of the tendency to use (external) excuses and justifications" (Aerts, 1994, p. 349).
In a subsequent study, Aerts (2001) used a research design suitable for evaluating the relative strength of consistency and inertial forces on the attributional behavior in annual reports.He proposed that an interaction between listing status and performance history constrains the variability of the attributional content over time.This implies that the attributional content, as well as other explanation patterns, would be very similar year after year (Aerts, 2001).Aerts (2001) drew on the propositions of Gibbins, Richardson, and Waterhouse (1990) to explain and predict corporate financial disclosures.Gibbins et al. (1990) developed a theory supported by two initial dimensions of the disclosure process: "an uncritical acceptance of rules and norms and a propensity to seek firm-specific advantage in how disclosures are made and interpreted" (p.122).These two dimensions are affected by "market factors as well as firm-specific factors" (Gibbins et al., 1990, p. 122).The authors also argued that the disclosure attributes are managed not only with relation to what is presented or absent from the narratives, but also with respect to their timing and interpretation.
Furthermore, a firm's disclosure history, what Gibbins et al. (1990) termed the disclosure position's ritualistic dimension, also affects the prevalent disclosure position, in an effect similar to the concept of path dependence.Based on this, Aerts (2001) argued that disclosure procedures can be embedded in organizational routines, marked by an uncritical adherence to prescribed disclosure norms that cause these procedures to tend toward persistence.According to Aerts (2001), organizational inertia derives from spontaneous habits, the existence of structures and routines, professional standards, education and training, precedents, rituals and traditions.Aerts (2001) elaborated further on the financial performance forces that trigger corporate verbal behavior.He asserts that high economic performance and low risk are assumed to signal the intrinsic quality of management.In addition, accounting data constitute the primary source of information for the interpretation of a company's economic performance.Thus, the presentation of accounting numbers tangibly associated with an unstable performance should be accompanied by some kind of narrative justification, since management's reputation is at stake.The results of Aerts' (2001) longitudinal study indicated that the attributional content presented in accounting narratives showed a high degree of stability over time, arguing against a purely calculative view of attributional behavior.Also, evidence of an inertial effect of company listing status and performance history was also deemed to affect the assertiveness aspects of attributional behavior and the differential use of accounting language in the explanation of financial accounting outcomes (Aerts, 2001).Clatworthy and Jones (2003) motivated their study on the importance and usefulness of accounting narratives.The authors continued the current of research established by Aerts (1994Aerts ( , 2001)), which incorporates an accounting rather than a managerial perspective.Their results buttressed the idea that accounting narratives are an important data source to study the management of corporate image.
In consonance with Aerts' (1994) propositions, Clatworthy and Jones (2003) found that companies in general avoid explicit causal attributions.The prevalent attributional strategy observed was to avoid specific blame for bad news.They also argued that improving performers are more assertive in the language they use in their annual reports.Aerts (2005) treated the capital market context as a critical variable to discern situations of strong and weak motivational influences on the use of self-serving attributions.He considered that previous studies overstated the relative importance of the informational explanation in their attempts to understand the presence of biased patterns of attributions.This mistake is derived from the fact that most previous research did not consider "the specifics of the social and organizational environment in which attributional behaviour occurs" (Aerts, 2005 p. 494).For instance, increased accountability demands, as invoked by the capital market context, and retrospective scrutiny are significant contextual forces in shaping motivational attributional behavior in listed companies (Aerts, 2005).
Results from Aerts' (2005, p. 495) research suggested that "listed companies offer more attributional explanations on a wider range of accounting outcomes, although these are not more extensive or profound."It was also observed that "listed companies exhibited a higher degree of defensiveness in explaining negative accounting outcomes" (Aerts, 2005, p. 514).Interestingly, the moderate degree of attributional defensiveness, in comparison with previous research that used US data, pointed to potentially significant cultural influences on the explanation patterns displayed by companies from different countries (Aerts, 2005).
All the previous studies discussed so far documented the asymmetry in the attribution of causality in the justification of organizational performance.Our aim here is to use the theoretical rationale presented in this section in order to predict the expected levels of attributional bias in the justification of organizational performance for a sample of Brazilian listed firms.We also bring signaling theory to suggest that companies whose performance is better than that of the market as a whole will seek ways to signal the superiority of that performance (Smith & Taffler, 1992) if the signaling costs are lower than the expected benefits (Spence, 1973).
We first consider the case of negative performers in a bad year.Higher levels of ego-defensive behavior are expected, since the company's performance is negative.Even though the egodefensiveness is not as high as in the combination of negative performance in a good year context, we suggest that the situation of bad performance in a bad year will elicit the highest levels of self-serving discourse, since the external conditions can be used to augment the role of external causes.
Let us now consider the combination of negative performance in a good year.Although egodefensiveness is expected to be preeminent, the principle of discount suggests that allegations of external causes to justify performance will be discounted by rational readers.A similar situation should be observed in the combination of positive performance in a good year context.From the signaling perspective, companies in this situation will aim to signal their superior performance, but a rational reader will discount excessive allegations to internal causes if they are used to justify performance.
The last case is the combination of positive performance in a negative year.We expect that companies in this situation will display the lowest level of self-serving attributions in their discourse, since there is neither a need to defend one's ego nor to signal superior performance.
From the preceding discussion we present the following hypotheses: H1: The presence of self-serving attributions in the CEO's Message of companies with negative performance in a bad year is the highest in comparison with all other combinations of performance and context.

H2:
The presence of self-serving attributions in the CEO's Message of companies with positive performance in a bad year is the lowest in comparison with all other combinations of performance and context.
In the next section we present the procedures used to gather data and test the hypotheses proposed.

DATA, VARIABLES AND METHODS
The annual reports of companies listed on the São Paulo Stock Exchange [Bovespa] were sampled for the years of 2002 and 2003.Content analysis was performed in the Letter to Shareholders section of each annual report.Content analysis has frequently been used in accounting research (Abrahamson & Amir, 1996;Bryan, 1997;Jones & Shoemaker, 1994;Smith & Taffler, 2000).
The Letters to Shareholders were first parsed in sentences for subsequent coding and classification according to previously defined rules.The aim of the analysis was to identify and code occurrences of causal attributions in a company's explanations of performance.
Even though the narrative sections of corporate annual reports as a data source can be regarded as comparable between companies, their form of presentation varies a great deal.Since the Letter to Shareholders is the most frequently read portion of the annual reports, and is more standardized (Clatworthy & Jones, 2006), we chose it as the object of our analysis.

Sampling Procedures
In order to provide variation in the context variable within the sample, a good year and a bad year were selected using the per capita gross domestic product and the Ibovespa (Bovespa index) as proxies.
Within a scope to the last ten years, we initially chose 1998 as the bad year.However, annual reports were not readily available for that specific year in a satisfactory number, so instead we chose 2002 as the bad year.
In 2002 the per capita GDP decreased 0.32% and the Ibovespa showed a negative variation of 17 percentage points.We chose 2003, with a 4.19% increase in the per capita GDP and a positive variation of 97.33 points in the Ibovespa, as the good year.
We selected 49 companies to compose the sample based on the criteria that: (1) the company presented the Letter to Shareholders in the annual report and (2) information about net earnings was available for 2001, 2002 and 2003 in the Economatica database.
Net income was chosen as a proxy for performance due to its focus on shareholders, who are remunerated according to it.It is also the proxy for performance that is easiest for an investor to understand.We believe that managers exercise some discretion in net earnings presentation, but since we are precisely concerned with the management of corporate image as disclosed by managers, the use of net income is justified by the fact that managers will have to provide interpretation for the numbers they present.

Coding of Attributions
In each Letter to Shareholders we focused specifically on causal attributions present in performance justifications, individually considered.A causal attribution is defined here as a sentence, phrase or paragraph in which an argument is built to connect performance results, or effects, with their respective presumed causes.The unit of analysis was the specific instance of causal reasoning in the Letter to Shareholders, following Bettman and Weitz (1983) and Tsang (2002).
All instances of causal attribution were retained.A total of 1957 phrases were analyzed, and 234 causal attributions relative to each company's performance were identified.In these, causes and effects were identified and coded.The effects were coded as good news or bad news.The causes were classified according to the locus of causality as internal or external.
An effect was considered good news when reporting an increase in revenue, sales, profits, investments, productivity or company growth.On the other hand, reporting a decrease or reduction in revenues, sales, profits, productivity, the presence of any kind of loss, closure of plants, etc, was considered as bad news.
Regarding the locus of causality, a cause was considered internal when referring to factors internal to the organization, such as company strategies, management decisions, know-how and human resources, among others.Otherwise, it was considered external, arising from factors external to the organization, such as inflation, market prices, government policies, climate, and so on.
The coding was performed in a three-phase process: I.Ten Letters to Shareholders, from annual reports for 2006 were coded with the purpose of establishing the general procedure for identifying and coding causal attributions; II.Afterwards, two undergraduate students in accounting and one of the authors independently coded the 2002 and 2003 Letters to Shareholders; III.The results of the coding were compared and differences resolved by discussions between the authors and coders.
A total of 234 attributions were coded, with an intercoder reliability of .75 in the coding process.Most of the disagreements were related to what should be codified as a cause and effect argument.This measure of reliability is a little less consistent than in previous studies, but it is in accordance with studies about readability of corporate annual reports, which have displayed reading ease scores on the borderline between difficult and very difficult to read (Rutherford, 2003).
Following Bettman and Weitz (1983) and Tsang (2002), we considered attributions individually when analyzing the use of self-serving attributional patterns.The next section presents the results and discussion.

Attributional Patterns
To delve into the patterns of argumentation observed in the sample of listed Brazilian companies, we prepared contingency tables for each variable and its respective associations.Each table was submitted to Chi-square tests to determine whether the association between variables was due to chance.
Contingency tables are used for categorical data analysis and they consist of tables of frequencies classified according to two or more sets of categorical variables.Table 1 presents the aggregated results for the coding of effects and their respective locus of causality.The null hypothesis predicts no association between the coding of effects and the locus of causality.Under this hypothesis, expected frequencies in each cell are calculated by multiplying together the two marginal totals and dividing the product by the grand total (Morettin & Bussab, 2003).For instance, the expected number in the cell at the first row and first column of Table 1 is given by multiplying 171 by 180 and dividing this number by 234, which results in 131.5.
In order to obtain the test statistic one needs to compute (Morettin & Bussab, 2003): The self-serving pattern of attributions is apparent in the shaded gray cells, since 90% of the good news is associated with internal causes.Also, 83.3% of the bad news is associated with external causes.This pattern of association is significant at 1%, as indicated by the Chi-square statistics (pvalue < 0.01).Following the same logic, we disaggregated the data by year in order to investigate the occurrence of self-serving attribution in each year.Tables 2 and 3 present the association between the locus of causality and the polarity of news observed for the years of 2002 and 2003.Again, the association between internal causes and good news and between external causes and bad news is significant in both years, as seen in Tables 2 and 3 (p-value < 0.01).
One can observe in Table 2 that even in a good year companies tend to blame external causes for their performance-related bad news.Another noteworthy observation is that the proportion of good news relative to the overall coded effects is not different from a good year to a bad year (0.85 and 0.67, p-value = 0.23).This result is consistent with previous studies, which found that no matter how bad the performance is, the tone of the Letter to Shareholders is predominantly positive (Clatworthy & Jones, 2003;Hildebrandt & Snyder, 1981;Staw et al., 1983).
Next we proceed to the test of our hypotheses.Table 4 reports the frequencies of attributions observed in the Letter to Shareholders for companies in the sample separated by their respective combination of performance and external context.Self-serving attributions are shaded grey.The frequencies of the shaded grey cells in each column of Table 4 were totalized and transformed in a proportion of self-serving attributions relative to the total amount of attributions of their respective column.
The results of the tests for differences in proportions obtained in Table 4 are presented now in Table 5.For companies with a negative performance in a bad year (NB), the proportion of self-serving attributions presented in the Letter to Shareholders (1.00) is significantly higher than that of companies with a negative performance in a good year (NG) (p-value = 0.04).Moreover, the proportion of self-serving attributions of NB companies is higher than companies both with a positive performance in a good year (PG, p-value = 0.01) and with a positive performance in a bad year (PB, pvalue = 0.01).These results support hypothesis H1, which posits that the presence of self-serving attributions in the Letter to Shareholders of companies with a negative performance in a bad year (NB) is the highest in comparison with all other combinations of performance and context.
It is conceivable that negative performers exhibited more instances of self-serving attributions in a bad year due to an effort to provide more explanation for their disappointing operational results.This is not the case, since we found that negative performers in a bad year (NB) prepared shorter Letters to Shareholders (significant at 0.01 level) and also displayed less occurrences of causal reasoning related to their performance (significant at 0.1 level), in comparison with positive performers in a bad year context (PB).Also, negative performers in a bad year (NB) presented shorter Letters to Shareholders (significant at 0.05 level) and less instances of causal explanation (significant at 0.01 level) in comparison with negative performers in a good year (NG).
In H2 we propose that the presence of self-serving attributions in the Letter to Shareholders of companies with a positive performance in a bad year is the lowest in comparison with all other combinations of performance and context.H2 is not supported by the results, since in Table 5 one can see that the proportion of self-serving attributions displayed by companies with a positive performance in bad year (PB) is not different from the proportions presented both by companies with negative performance in good year (NG, p-value = 0.47) and by companies with positive performance in a good year (PG, p-value = 0.61).
We did not test for differences in proportions of self-serving attributions between companies with a negative performance in a good year (NG) and companies with a positive performance in a good year (PG) because this comparison relates to none of our hypotheses.
The results must be interpreted with caution because the sub-sample sizes are not large enough to ensure robustness.However, since nonparametric tests for difference in proportions are not currently available in the literature, the results indicate that H1 is corroborated, although H2 is not.
In order to understand why hypothesis H2 was not corroborated, we performed tests with disaggregated proportions within the sample.The tests revealed that companies with negative performance in a good year (NG) take credit for the good news in a proportion equivalent to that presented by companies with positive performance in bad year (PB), since the p-value for the null hypothesis is 0.26.This result implies that NG companies might try to sway the reader's attention to the good news instead of their bad performance.
Additionally, companies with a positive performance in a good year (PG) blame negative effects on the environment in a proportion that is equivalent to that displayed by companies with positive performance in a bad year (PB, p-value = 0.445).

CONCLUSION
This work explored the nature of organizational discourse in different combinations of contexts and company performances.The nature of organizational discourse was theoretically underpinned by the concept of self-serving attributions, which is a specific instance of causal reasoning that allows the writer to take credit for good news and avoid blame for bad news.Two samples were selected, one in a bad year and the other in a good year, regarding the capital market context.Based on the theoretical propositions on self-serving attributions in annual reports, we added contributions from signaling theory in order to develop hypotheses for the expected levels of self-serving attributions in the justification for organizational performance.
Testing the theoretical hypotheses advanced in this study proved difficult, since the size of the subsamples used in the comparisons were not as large as needed to ensure robustness.Nevertheless, the results corroborated the hypothesis that the presence of self-serving attributions in the Letter to Shareholders of companies with a negative performance in a bad year is the highest in comparison with all other combinations of performance and context.Yet, contrary to our second hypothesis, companies with a positive performance in a bad year did not display the lowest level of self-serving attributions in the Letter to Shareholders in comparison with all other combinations of performance and context.
The results also indicate that companies attempt to create a positive corporate image to external stakeholders even when negative performance occurs in a clearly favorable external context.Additionally, we observed that companies with positive performance in a good external context blame negative effects on the environment in a proportion equivalent to that observed for companies with positive performance in bad year.Both observations are incompatible with the premise of rational readers and they suggest opportunities for additional investigation.
Our conclusions are subject to a number of limitations.As mentioned, sample size is the first.We also suggest that other studies might be developed considering the presence of self-serving attributions conditional to corporate governance indexes.This might be an interesting way to incorporate the institutional setting in the discussion.
Contemporaneity is also a motivating issue with respect to Brazilian companies' annual reports.Due to the theoretical hypotheses advanced here, this study used data from 2002 and 2003.To the best of our knowledge, no study has been conducted about self-serving attributions in the context of the Brazilian capital market, which is gaining increasing economic relevance.
Moreover, other verbal strategies plausibly associated with impression management may be intertwined in a structure that includes legitimacy needs, impression management tactics, accounting explanations, the use of attributions, and elements of metadiscourse.It is a challenge to develop research designs that allow uncovering the complex relation between these elements.