AAAJ 15,3

312

Received July 2000 Revised January 2002 Accepted January 2002

An examination of the

corporate social and

environmental disclosures of

BHP from 1983-1997

A test of legitimacy theory

Craig Deegan, Michaela Rankin and John Tobin

School of Accounting and Law, RMIT University, Melbourne, Australia Keywords Disclosure, Environment, Management, Business policy, Perception, Annual reports

Abstract This study examines the social and environmental disclosures of BHP Ltd one of the largest Australian companies) from 1983 to 1997 to ascertain the extent and type of annual report social and environmental disclosures over the period, and whether such disclosures can be explained by the concepts of a social contract and legitimacy theory. This research is also motivated by the opportunity to compare and contrast results with those of Guthrie and Parker, in whose study the social and environmental disclosures made by BHP Ltd were also the focus of analysis. In testing the relationship between community concern for particular social and environmental issues as measured by the extent of media attention), and BHP's annual report disclosures on the same issues, significant positive correlations were obtained for the general themes of environment and human resources as well as for various sub-issues within these, and other, themes. Additional testing also supported the view that management release positive social and environmental information in response to unfavourable media attention. Such results lend support to legitimation motives for a company's social and environmental disclosures. A trend in providing greater social and environmental information in the annual report of BHP in recent years, and its variable pattern, was also evidenced.

1.Introduction

This study examines whether an organisation, specifically BHP Ltd one of Australia's largest diversified public companies), discloses social and environmental information in response to particular social expectations, expectations which typically change across time[1]. The research updates the work undertaken by Guthrie and Parker 1989) which studied the annual reports of BHP Ltd for the 100 years to 1985 and which has become a highly cited and respected paper in the area of social and environmental reporting. The current study examines the extent and type of corporate social and environmental reporting by BHP Ltd over a 15-year period from 1983-1997. As with Guthrie and Parker 1989), this study correlates social disclosure to a measure of public concern. However, this study uses a different measure of public concern than that used by Guthrie and Parker 1989). Specifically, this study uses the extent of media attention devoted to the activities of BHP Ltd

The authors would like to acknowledge the helpful comments made by Rob Gray, Lee Parker, James Guthrie, Markus Milne and the anonymous referees.

Accounting, Auditing & Accountability Journal, Vol. 15 No. 3, 2002, pp. 312-343. 5 MCB UP Limited, 0951-3574 DOI 10.1108/09513570210435861

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

over the period of the study. It does this on the basis of a belief that media attention reflects or perhaps shapes) community concerns see Smith, 1987; Zucker, 1978; Ader, 1993). In undertaking this study, we seek to establish if there have been changes in BHP Ltd's social and environmental disclosures over the period of the study and to investigate whether, consistent with legitimacy theory, specific social and environmental disclosures can be associated with specific societal concerns as reflected by the media attention). 313

In contrast to the method used in this paper, Guthrie and Parker 1989, p. 347) used a ``data bank of all major events and issues relating to BHP''. This ``data bank'' came from the contents of 11 publications addressing BHP's, and its industry's history. One of the publications was compiled by BHP, whilst the other publications were produced by independent researchers. Guthrie and Parker sought to match peaks and troughs in corporate annual report disclosure with social events and issues identified by the various publications. As they state Guthrie and Parker, 1989, p. 347):

For each major category of disclosure environment, energy, human resources and community involvement), the timing of observed peaks of disclosure was compared with any apparently related BHP activities or socio-economic environmental conditions occurring immediately before or during peak periods. A majority of peak disclosures associated with relevant events is considered evidence of a legitimising explanation for BHP corporate social reporting.

In the Guthrie and Parker 1989) study the authors were unable to confirm legitimacy theory. However, as we would argue, and as they concede, this may have been due, at least in part, to deficiencies in the way they constructed their measure for community concern. As Guthrie and Parker 1989, p. 348) acknowledge, their measure of community concern may exclude some important events or activities in BHP's history. They also acknowledge the possibility that their testing procedures may have failed to detect disclosure reactions if those disclosure reactions lagged behind the various social and environmental events.

Whilst Guthrie and Parker failed to provide results to support legitimacy theory, many other papers have tended to support the theory for example, Dowling and Pfeffer, 1975; Patten, 1992; Gray et al., 1995a; Deegan and Rankin, 1996; Deegan and Gordon, 1996; O'Donovan, 1999; Brown and Deegan, 1998). We are left to wonder whether there is something different about BHP, or whether Guthrie and Parker's measure for community concern was mis- specified. As we show in this paper, using our method of defining community concern, there does appear to be a strong association between BHP's disclosure policies and community concern a result consistent with legitimacy theory, but inconsistent with Guthrie and Parker 1989).

The balance of the paper proceeds as follows: the next section investigates the role of the media in either shaping or reflecting community concerns. Reference will be made to media agenda setting theory. Section 3 reviews literature which suggests that corporate social and environmental disclosures are reactive to community concerns. The development of hypotheses are

A test of legitimacy theory

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

outlined in section 4, followed by details of the research method in section 5. Section 6 discusses the results of hypotheses testing, with some concluding comments presented in section 7.

2.Theroleofthemediainshapingcommunityconcerns

As Brown and Deegan 1998) explain, media agenda setting theory posits a 314 relationship between the relative emphasis given by the media to various topics referred to as the ``media agenda''), and the degree of salience these topics have for the general public as reflected by the ``public agenda'')[2]. In terms of causality, increased media attention is believed to lead to increased community concern for a particular issue. The media are not seen as mirroring public priorities; rather, they are seen as shaping them, and in turn, shaping the public agenda. In further exploring the notion of the ``public agenda'', McCombs et al.

1995, p. 282) state:

Walter Lippmann 1965) defined the public agenda as that array of issues concerning which the well-being of numerous individuals is dependent upon mutual action, cooperation, or, at least, tacit consent. He also noted that this array of issues is largely beyond direct experience:

For the real environment is altogether too big, too complex, and too fleeting for direct acquaintance. We are not equipped to deal with so much subtlety, so much variety, so many permutations and combinations. And although we have to act in that environment, we have to reconstruct it on a simpler model before we can manage it. To traverse the world, men must have maps of the world p. 16).

It is the news media, noted Lippmann, that provides these maps of the world. Through their selection and display of the daily news, journalists provide major cues about what are the important topics of the day. Over time, many of the issues receiving major emphasis in the news become the major issues on the public agenda. Although this agenda-setting role of the news media is a secondary and unintentional by-product of the necessity to select a few issues for attention, it is one of the most significant effects of mass communication.

McCombs et al. 1995) stress that public awareness is the first step in the formulation of public opinion and that the media clearly shapes this awareness. There have been numerous studies of media agenda-setting effects, many of which have adopted media agenda setting theory. Research indicates that the media influences the public's perceived salience for issues Smith, 1987; Brosius and Kepplinger, 1990; Ader, 1993), and that the media agenda typically precedes public concern for particular issues McCombs and Shaw, 1972; Funkhouser, 1973; Trumbo, 1995; Neuman, 1990). Research also shows that public concerns and the media agenda are not necessarily reflective of ``real world'' conditions Funkhouser, 1973; Ader, 1993). For example, in a review of newspaper articles, a ``real-world'' pollution indicator[3], and opinion polls from 1970-1990, Ader 1993) found that the amount of media attention devoted to pollution influenced the degree of public salience for the issue, but the ``real- world'' pollution indicator was negatively correlated with the amount of media coverage. According to Ader 1993, p. 310), ``the public needs the media to tell them how important an issue the environment is. Individuals do not learn this from real world cues.''

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

A review of the literature suggests that a number of variables mediate the relationship between media activity and public salience of an issue. These variables include: the obtrusiveness of issues; how the issue is framed as positive or negative); and associated time lags. Zucker 1978) defined the concepts of obtrusiveness people's direct experience of an issue) and unobtrusiveness people may not have direct experience of an issue). Studying six issues, he concluded that the less direct experience the public has with a given 315 issue area, the more it will have to depend on the news media for information

about that area. News media coverage preceded the rise of importance of an issue in public opinion polls for the unobtrusive issues, while for the obtrusive issues media coverage and importance to the public seemed to increase together. In a number of studies the environment has been deemed to be an unobtrusive issue, an issue about which the media appears very capable of influencing public concern Blood, 1981; Eyal et al., 1981; Zucker, 1978). Other issues, such as the ongoing activities of politicians, are also deemed to be unobtrusive. According to Lippmann 1965, p. 18), the world people have to deal with politically is out of reach, out of sight and out of mind. It has to be explored, reported and imagined. He argued that ``the pictures in our heads F F F of things we have not experienced personally are shaped by the mass media''[4].

The intensity of the media coverage has also been found to affect the likelihood that particular media coverage will impact the public agenda, although it is not clear what extent of coverage is required before an agenda- setting effect is created Brosius and Kepplinger, 1990). The way in which the media covers the issue can also affect the likelihood of whether it impacts public attitudes. Dearing and Rogers 1996, p. 64) found that an issue presented in a negative light is more likely to be regarded by the community as an important concern. That is, negative media attention is more likely to have an effect on the public's salience for a particular issue relative to positive, or favourable, attention.

According to McCombs and Shaw 1994, p. 380) any acceptance of the ``agenda setting hypothesis'' requires that a matching public agenda lags behind the media coverage of the issue. Such a lag has been found in many studies including some of those referred to above). In relation to the issue of time lags, Stone and McCombs 1981) took the results of two public opinion surveys and compared them with a content analysis of the US national news magazines Time and Newsweek. As well as supporting the cumulative effect of mass communication on public perception, they demonstrated ``a time lag in the movement of issues' salience from the media agenda to the public agenda'' Stone and McCombs, 1981, p. 53) of two to six months. In a similar vein, a content analysis of environmental news stories appearing in the three largest circulation newspapers in the Lansing Michigan area over a 239-day period, and measures of the public agenda from a number of surveys, led Salwen 1988, p. 100) to suggest that the public begin to adopt the media agenda from the five to seven week mark of a particular issue's coverage. This is also consistent with a review of the literature undertaken by McCombs et al. 1995). Other

A test of legitimacy theory

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

studies indicate that an issue's salience for the public can commence from within four weeks of the media coverage see for example, Winter and Eyal 1981) who studied the relationship between public concern about civil rights between 1954-1976 and related media coverage). Of course, it would be reasonable to expect that time lags will vary depending upon the issues in focus, but the evidence does suggest that lags do exist thereby suggesting that

316 media coverage shapes community perceptions. There are various forms of news-media, including newspapers, television

and radio. Research supports the view that newspapers tend to have a greater ability to set the public agenda McCombs, 1981). A survey undertaken by Bogart 1984) to establish the relative impact of daily newspapers and television on public perceptions documented that:

F half the public is exposed daily to both newspapers and television news; F newspapers are part of the life of nearly nine out of ten Americans; F newspapers touch two out of three on a typical day; and F nearly four out of five readers report looking at any given page.

According to Bogart 1984, p. 719), ``for a majority of the audience, the two media complement each other, and newspapers' ability to cover the news of the area in detail and in depth remains a major advantage''. In a study of voting behaviour, Stempel and Hargrove 1996, p. 557) indicated that ``it is newspaper reading, not TV news use, which relates most to voting''. Such a result is also confirmed by McCombs and Shaw 1994, p. 382) who explain that by framing a story within a larger context, the print media is better able to point out its significance to a reader as opposed to television news which is deemed to be more in the nature of a ``headline service'').

Related to some of the above studies, convergence in public opinion has also been attributed to media coverage. For example, Shaw and Martin 1992) studied the opinions of people with different demographic characteristics, namely men and women, young and old, educated and uneducated, and Caucasian and non-Caucasian. They found convergence on issues associated with education, pollution, housing and poverty as the exposure to related news items increased.

From the discussion above, and a review of other literature, it does appear that the media attention directed towards particular issues particularly coverage included within newspapers) can shape and change community concern for many issues. Obviously, for corporate managers to react to media publicity and in this study we are interested in corporate disclosures) they must perceive that the media publicity will impact community concerns and that the media is simply not a ``passive transmitter of a reality that has an existence of its own'' Severin and Tankard, 1992). Managers are obviously not expected to be cognisant of the research undertaken within the context of media agenda setting theory. Nevertheless, consistent with the view that corporate managers do perceive that newspapers can impact public attitudes

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

which is consistent with the findings of researchers working within the theoretical perspective of media agenda setting theory) it is common for corporations to subscribe to information services which provide daily information about how or whether the corporation has received media coverage in nominated national newspapers. Further, research by O'Donovan 1999), which we will consider in more depth later in the paper which is embedded within the theoretical perspective provided by legitimacy theory), also shows 317 that corporate managers consider previous newspaper coverage when

determining the disclosures they will make in their subsequent corporate annual reports. This evidence is consistent with a view that not only is management aware of media coverage pertaining to their organisation, but that they feel a necessity to respond to it from a disclosure perspective. The next section will briefly consider the possible relationship between community concern and corporate disclosures[5]. Legitimacy theory will be utilised to explain how corporate disclosures might be used by management to change community perceptions about the disclosing corporations. Hence, whilst media agenda setting theory argues that the mass media can shape community perceptions about certain issues, legitimacy theory provides arguments consistent with the view that corporations can also impact community perceptions through their disclosure practices.

3.Corporatesocialdisclosuresasareactiontocommunity expectations Studies which have examined social and environmental disclosure within annual reports indicate that it has been increasing across time, both in number of companies making disclosures and in the amount of information being reported Ernst & Ernst, 1978; Harte and Owen, 1991; Gray et al., 1995a; Deegan and Gordon, 1996). Reporting has been generally qualitative in nature and favourable to the company concerned, even to the point of increasing positive disclosures around the time of negative events Deegan and Rankin, 1996; Deegan et al., 2000).

In research specifically related to the current study, Hogner 1982, p. 249) accorded the growth, decline, and evolution of US Steel's social reporting over an 80-year period to a ``concentration on the reporting of activities that society is perceived as valuing most at the time''. While concluding that the matrix of forces affecting corporate reporting practice arose from a legitimation motive, he did not empirically test the proposition.

Guthrie and Parker 1989) undertook a study of the social and environmental disclosure practices of BHP Ltd between the years 1885-1985 and compared their observations to Hogner's 1982) examination of US Steel annual reports for a similar period. They observe a history of growth, decline and change in social disclosure over the period studied, rather than a period of growth and development. Human resource disclosures were found to be BHP's primary form of social disclosure, although it was quite inconsistent. The authors also observed a total absence of environmental disclosures until around 1950, with a

A test of legitimacy theory

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

recurrence in the early 1970s and 1980s, although they remained at a relatively low level.

Guthrie and Parker sought to determine if disclosures were consistent with legitimacy theory and in doing so they compared the disclosure practices of BHP with major events and issues which affected BHP throughout its history as documented within the various publications they reviewed. The authors

318 concluded that the peak in environmental disclosures in the 1970s was associated with a time when mining, steel and oil industries became targets for criticism by conservationists. However, legitimacy theory in relation to environmental activities was not supported in earlier periods when disclosures were rare and not a reaction to public pressure or other external events. The authors also concluded their evidence failed to support a legitimacy perspective for other categories of social disclosure. They acknowledged, however, that they may not have accurately captured the ``events'' in BHP's history which they were attempting to match with the company's reporting, and that the

study may have also suffered from unidentified time lags. In another study which sought to test legitimacy theory, Brown and Deegan

1998) adopted media coverage as a proxy for community concern. They found support for legitimacy theory in relation to their review of corporate environmental disclosures. In some industries the environmental disclosure strategies of management appeared to be tied to the extent of media attention devoted to environmental issues. Further, changes in media attention, not the level of media coverage per se, appeared to explain variations in corporate environmental disclosure strategies[6].

In a further study which considered the role of the media and its impact on corporate disclosures, O'Donovan 1999) conducted interviews with senior management of three major Australian corporations, including BHP, and found that the managers consider that they use the annual report to respond to perceived public concerns, with reports in news media affecting what information they disclosed. The corporate managers' responses were linked to a perception that media attention devoted to particular issues impacts the community's concern about such issues. Management perceived that media reports of a continuing nature, particularly negative or unfavourable reports, were most likely to result in a response in the annual report. O'Donovan's 1999, p. 82) data analysis suggested that ``corporate management believe, to some extent, that the annual report is an effective way for informing and educating the public of the corporation's view about certain environmental issues''.

As indicated above, in explaining the practice of corporate social reporting, many authors have adopted legitimacy theory for example, Patten, 1992; Guthrie and Parker, 1989; Deegan and Gordon, 1996; Deegan and Rankin, 1996; Neu et al., 1998; Buhr, 1998; O'Donovan, 1999). Legitimacy theory is a theory that, as applied in the social and environmental reporting literature, is rather simplistic but nevertheless appears to be the theoretical basis most frequently used in attempts to explain corporate social and environmental disclosure policies[7]. Legitimacy theory relies upon the notion of a social contract and on

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

the maintained assumption that managers will adopt strategies, inclusive of disclosure strategies, that show society that the organisation is attempting to comply with society's expectations as incorporated within the social contract).

A test of legitimacy theory

Pursuant to legitimacy theory, managers' choices of legitimising strategies are based on the perceptions of the particular managers involved, and different managers will be likely to have different ideas about what society expects that is, what the terms of the social contract are), and whether the organisation is 319 perceived by community members as complying with these expectations. Nevertheless, when significant events such as a major environmental disaster occur, or when there is sustained mass media interest, then it is reasonable to assume that most managers would perceive that the organisation's ongoing legitimacy is threatened. Conceivably, there will also be different views about appropriate strategies to adopt when legitimacy is threatened and of course, some managers might perceive a problem when in fact none might exist, and vice versa). Working out the effects of all the various judgements and perceptions with any precision is a difficult if not impossible task. Consequently, researchers have used simplifying assumptions: for example, that evens such as the Exxon Valdez disaster Patten, 1992) or proven environmental prosecutions Deegan and Rankin, 1996) are assumed to create legitimacy problems for an industry, and that the managers will use disclosure strategies to reinstate damaged legitimacy as inferred by increasing disclosures around the time of the events).

In this paper we also make assumptions, grounded in media agenda setting theory, that media attention directed at corporations can impact society's views of such corporations. Also, we make the explicit assumption, based on previous research such as O'Donovan 1999), that managers also perceive that the media can impact community perceptions, and that the annual report is perceived by managers to be one means to shift community perceptions back in favour of the organisation. Hence, we predict that if the mass media is focusing upon particular attributes of the organisation's operations then this is potentially legitimacy threatening and therefore likely to provoke a reaction by management specifically, a disclosure reaction.

As noted above, legitimacy theory posits that organisations seek to ensure that they act, or at least appear to act, within the boundaries and norms of the societies in which they operate. The ``social contract'' also frequently referred to as the ``community licence to operate'') is an implicit agreement between an organisation and society Shocker and Sethi, 1974)[8]. Failure to act in accordance with the social contract is construed as being detrimental to the ongoing operations of the entity. Community expectations are not considered static, but rather, change across time thereby requiring organisations to be responsive to the environment in which they operate. An organisation could, accepting this view, lose its legitimacy even if it has not changed its activities from activities which were previously deemed acceptable legitimate). Because community expectations will change across time it is argued that the organisation must also make disclosures to show that it is also changing or

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

perhaps to justify why it is not changing). Changing activities without communicating such changes is considered to be insufficient. In relation to the dynamics associated with changing community expectations, Lindblom 1994, p. 3) states:

Legitimacy is dynamic in that the relevant publics continuously evaluate corporate output,

methods, and goals against an ever-evolving expectation. The legitimacy gap will fluctuate 320 without any changes in action on the part of the corporation. Indeed, as expectations of the relevant publics change the corporation must make changes or the legitimacy gap will grow

as the level of conflict increases and the levels of positive and passive support decreases.

Again, organisational legitimacy a perceived state) is deemed to occur when there is congruence between what the community expects of an organisation, and whether they believe the organisation is basically complying with those expectations[9].

Using the report Tomorrow's Company as a reference, Solomon and Lewis 2001, p. 7) identify eight ``forces'' which impact community perceptions about whether an organisation is complying with its social contract. Whilst not emanating from the mass media literature, one such ``force'' is considered to be the media[10]. Given a perspective that business organisations exist as a result of compliance with their social contract, the argument is that they must establish congruence between ``the social values associated with or implied by their activities and the norms of acceptable behavior in the larger social system of which they are a part'' Dowling and Pfeffer, 1975, p. 122). Legitimacy, which is deemed to exist when the entity's value system appears to be congruent with the value system of the larger social system of which the entity is a part Lindblom, 1994) is, in a sense, treated like a resource provided by parties outside the organisation, much like financial capital or labour Dowling and Pfeffer, 1975). Pursuant to legitimacy theory, management are deemed to use disclosure media, such as the annual report, to allay community concerns, or more particularly, what they perceive to be the community concerns Lindblom, 1994). Indeed, Hurst 1970) argues that one of the functions of accounting and related reports is to legitimise the existence of the organisation. Further, the annual report has been deemed to be an important document for an organisation seeking to shape its own ``social imagery'' Gray et al., 1995b).

According to legitimacy theory, the disclosures might be made to show that the organisation is conforming with community expectations, or alternatively, they might be made to alter societal expectations. Of course, studies which argue that disclosures can change, or perhaps are perceived to change, community perceptions many such studies being grounded in legitimacy theory) are based upon an assumption that corporate disclosures, such as those in annual reports, do actually impact community concerns. This assumption is consistent with the views held by senior managers interviewed by O'Donovan 1999) who, when interviewed, responded that annual report disclosures are used as a strategy to change perceptions about the organisation[11].

In summary, to this point we have argued that the media agenda impacts society's level of concerns for particular issues from media agenda setting

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

theory) and we have provided evidence that managers believe that the mass media has the ability to shape community expectations. We have also argued that the extent and type of corporate social disclosure, in the annual report, is likely to be directly related to management's perceptions about the concerns of the community from legitimacy theory). The next section of the paper develops the hypotheses of this study.

4.Hypothesesdevelopment

The importance that the public ascribe to an issue is influenced by the amount of media attention it receives for example, see McCombs and Shaw, 1972; Funkhouser, 1973; Ader, 1993). Public salience for an issue increases with the number of media articles between ``takeoff'' and ``tapering'' thresholds Neuman, 1990). A certain ``critical'' number of articles are required to move an issue to one of public concern, and the pattern of evolving public awareness varies for different types of issues Neuman, 1990, p. 159). The response function varies according to the issue covered, but there is consistent evidence of a relationship between the volume of media coverage and the level of public concern.

To this point, a varying pattern of social disclosures across both companies, and time, has been displayed for example, see Pang, 1982; Guthrie, 1982; Guthrie and Parker, 1990; Deegan and Gordon, 1996). If management perceive that in the opinion of the ``relevant publics'' the organisation is not meeting its ``social contract'' with society, it is likely that the organisation will take steps to demonstrate its legitimacy and relevance to society and so avoid potential constraints and sanctions. Research has shown that management consider that the media can influence community concern, and that management will use the annual report to counter unfavourable media coverage O'Donovan, 1999).

The discussion above has focused on establishing that disclosures, inclusive of social and environmental disclosures, are likely to be made to legitimise corporate behaviour in response to public concern, and that public concern is impacted by the media agenda, as expressed by the print media. The underlying proposition is that changes in society concerns, reflected by changes in the themes of print media articles, will be mirrored by changes in the social and environmental themes disclosed, and to the extent of the disclosure made. Applying this specifically to BHP leads to the development of the following hypothesis.

H1. Higher lower) levels of the print media coverage given to specific attributes of BHP's social and environmental performance will be associated with higher lower) levels of specific social and environmental disclosures made by BHP in its annual reports.

The above hypothesis does not address whether the media coverage and annual report disclosures are favourable or unfavourable in nature. As noted previously, the framing of an issue in the media as positive or negative influences its salience for the public Schoenbach and Semetko, 1992; Dearing and Rogers, 1996). The managers in O'Donovan's study indicated that they

A test of legitimacy theory

321

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

were most likely to respond to media coverage that depicted their companies in an unfavourable light. Other studies indicated an increase in positive or self- laudatory disclosures around the time of events that depicted the organisation in an unfavourable light Patten, 1992; Deegan and Rankin, 1996).

In circumstances where media attention is of a negative or unfavourable nature, organisations have a greater incentive to provide more positive 322 disclosure Brown and Deegan, 1998) to affirm or re-establish their legitimacy. This is consistent with the disclosure strategies described by Lindblom 1994). One way to convince society that the organisation is meeting its social contract

is to furnish positive perspectives of its activities. Accordingly:

H2. Higher lower) levels of unfavourable print media coverage given to specific attributes of BHP's social and environmental performance will be associated with higher lower) levels of specific positive social and environmental disclosures made by BHP in its annual reports.

5.Researchmethod

Content analysis is used in this study to measure both media attention and corporate social disclosure. Krippendorff 1980) defines content analysis as ``a research technique for making replicable and valid inferences from data to their context''. The success of the process depends on the reliability and validity of the procedures employed. While there are a number of measures of reliability, Krippendorff's 1980) alpha[12] will be used to assess the replicability of the results. The coding rules are developed in detail to yield standard classifications over each of the 15 annual time-periods of this study.

Ingram and Frazier 1980) state that categories used in content analysis should result from a systematic application of a set of rules to identify exhaustive and mutually exclusive categories. Accordingly, in order to draw valid and reliable inferences from the measurement process, the method suggested by Weber 1985) to create and test a coding scheme has been followed.

First, the recording units are defined. The reliability of different measures of social disclosure has received much attention in the literature. Ingram and Frazier 1980) suggest the sentence as the unit of analysis, as it is easily identified and is less subject to intercoder variation than other measures, such as words and pages. The advantages of sentences are in not needing to standardise words, in obtaining more reliable intra- and inter-rater coding, and in allowing more detailed analysis of specific issues and themes. ``Sentences are to be preferred if one is seeking to infer meaning'' Gray et al., 1995b, p. 84). Whilst we use sentences in this study to measure the amount of annual report disclosure it should be noted that many other studies use measures such as words, or proportion of pages. These different measures have been found to be highly correlated Hackston and Milne, 1996), hence the results should not be greatly influenced by the choice of sentences instead of words, or proportion of pages.

Second, the categories to be classified are defined to allow an item to be allocated to a particular category. They are intended to be mutually exclusive

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

to avoid confounding of the subsequent statistical analysis. This study primarily employs the content classifications of Hackston and Milne 1996), which are based on earlier schemes developed by Ernst & Ernst 1978), Guthrie 1982), and Gray et al. 1995b). Accordingly, the dimensions of the content analysis for both media articles and BHP's annual reports broadly embrace the classifications of environment, energy, human resources, community involvement and other.

Third, test coding is undertaken on a sample of the texts by one team member. This reveals some ambiguities in the proposed coding rules, and suggests some minor, but helpful, revisions of the classifications. Fourth, the accuracy or reliability of the coding is assessed by a test-retest, by the same team member, of a sample of the data at different times, and by supplying the coding rules and data samples to a second reviewer for analysis. Fifth, the coding rules are revised to increase the expected reliability of the coding of all texts, and sixth, all relevant text is coded.

Finally, the second reviewer again codes samples of the data to assess achieved reliability and accuracy, using Krippendorff's 1980) alpha. The results of our testing indicate that our coding procedures appear reliable using the guidance provided by Krippendorff 1980)).

Societal pressure/community concern is measured by the number of relevant articles in the print media. Previous discussion has suggested that coverage of issues in the print media parallels exposure in the overall communication media. Summaries of all articles in the CD-ROM index, the Australian Business Index database ABIX), mentioning BHP for the period 1983 to 1997, are obtained[13]. The index provides an up-to-date and accurate guide to published information from a wide cross-section of business, finance and trade resources by indexing approximately 85 newspapers and journals Brown and Deegan, 1998). The number of those newspapers and journals in the ABIX database has steadily increased during the 15-year period. To enable a consistent comparison across the years 1983-1997, only print media services present in the database for those years, and which are available for review in the authors' home university library, are selected. The papers used had the largest circulations of major newspapers in Australia and included The Age, The Australian, The Australian Financial Review, Westralian, The Bulletin, Business Review Weekly, Courier Mail and The Sydney Morning Herald. It is proposed that this selection includes an adequate number of national and regional papers in Australia to reflect adequately media attention to, and public concern with, the issues of this study.

Following selection of the articles mentioning BHP in the Corporate Name/ Subject area of the ABIX search menu, the abstract of each article is examined for information concerning any of the themes outlined in the previous section. Relevant articles are then examined from microfilm records of the publications, and where relevant, are coded according to the categories or themes provided in the Appendix.

A test of legitimacy theory

323

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

324

This classification is sufficient to test H1, but additionally, to test the H2, and following the definitions of Hogner 1982) and Brown and Deegan 1998), each print media article is categorised as:

F unfavourable: where the content indicates that the operations/strategies/ performance of BHP are detrimental to, or not in harmony with, the social environment;

F favourable: where the content indicates that the operations/strategies/ performance of BHP are beneficial to, or in harmony with, the social environment;

F other: where the content does not indicate that operations/strategies/ performance of BHP are beneficial or detrimental to the social environment.

Each media article is then given a score of 1, and allocated to the specific issue within each general theme, with an attendant favourable, unfavourable, or "other'' classification[14].

As noted in a previous section, measurement of corporate social disclosure has largely focused on information provided in companies' annual reports. This medium is considered the preferred information source for a number of stakeholder groups Tilt, 1994; Deegan and Rankin, 1997). Use of the annual report enables comparisons to previous studies Hogner, 1982; Guthrie and Parker, 1989; Brown and Deegan, 1998), and allows examination of whether managements' perceptions O'Donovan, 1999) of using information in the annual report to respond to public concerns is supported.

The annual reports of BHP are examined for the years 1983-1997. The content themes and issues, and reasons for the selection of sentences as the recording unit, have already been presented. Positive and negative disclosures are defined similar to Hogner 1982) and Brown and Deegan 1998) as:

F positive: referring to information about corporate social activities which have a positive or beneficial impact on society;

F negative: referring to information about corporate social activities which have a negative or deleterious impact on society;

F neutral: referring to information about corporate social activities whose impact on society cannot be determined as either positive or negative.

The positive/negative measurement is central to the testing of H2.

6. Results and interpretation

6.1 Print media coverage and levels of corporate social disclosure

In examining this association, media articles were collated by year for each of the 49 issues shown in the Appendix, and compared to the corresponding annual report sentences on the same issues in the same years[15]. The data are now described for all of the categories of the study.

Summary aggregated totals over the 15-year period from 1983-1997 are displayed in Table I for each theme. Human resources, environment and

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

Total annual Total media Positive annual Theme report sentemces articles report sentences

Environment 347 87 276 Energy 11 8 11 Human resources 403 460 382 Community 95 31 79

involvement Other 36 2 36

Total 892 588 783

Unfavourable media articles

48 0 400 16

0 464

A test of legitimacy theory

325

Table I.

Total social and environmental disclosure by general themes 1983-1997

Notes: Spearman's rank-order correlation between total annual report sentences and total media articles = 0.9000 p = 0.019); Spearman's rank-order correlation between positive annual report sentences and unfavourable media articles = 0.9747 p = 0.002); Spearman's rank-order correlation coefficient one-tailed) expresses the correlation of the number of annual report sentences on the five general themes, with the number of media articles on the same five themes, for 15 years

community involvement rank in that order in both the print media and the annual reports, with the categories of ``energy'' and ``other'' receiving minimal attention in both media. At this aggregated level the results are consistent with the hypotheses in this paper. The two issues attracting the most media attention human resources and environment which together account for 93 percent of the media coverage) account for 84 percent of the total corporate social and environmental disclosures.

The results in Table I are generally consistent with Guthrie and Parker 1989). They found limited disclosures in relation to community involvement. They also found that human resource disclosures accounted for the highest proportion of total disclosures across the period of their study 1885-1985).

In considering the trends in the disclosure of different categories of social and environmental information, we found that environment related disclosures made by BHP were fairly minimal one or two sentences) until 1989. This is consistent with Guthrie and Parker who found fairly minimal disclosures across the period of their study with the period of their study ending in 1985). In 1989, environmental disclosures increased to 11 sentences and then tended to increase in each year to 1997, wherein total environmental disclosures amounted to 140 sentences in only two years during this period was there any downturn in environmental disclosures, but even then, disclosure did not return to 1989 levels). The disclosures in relation to energy were fairly minimal throughout the period of observation, which again was consistent with Guthrie and Parker. Disclosures pertaining to human resources, community involvement and ``other'' tended to increase and decrease from period to period with no discernible trends of continued growth, or otherwise) being apparent. Again, this is generally consistent with the results provided by Guthrie and Parker.

Table I demonstrates the association between public concern as measured by the proxy of media attention) and annual report social and environmental disclosures, by five themes over the 15 years from 1983-1997. This correlation of total media attention and corporate disclosures produces a correlation coefficient

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

326

of 0.900, p < 0.05, while positive annual report disclosures and unfavourable media articles provides a correlation coefficient of 0.9747, p < 0.01.

A view of the overall pattern of attention, in both the media and the annual report, to the sum of the social and environmental issues of this study, is presented in Table II.

For the annual report, there is a general trend of increasing total social and environmental disclosure over time, with comparatively few such disclosures in the 1980s, followed by pronounced upward shifts in 1989 and 1996. Whilst Table II aggregates various social and environmental disclosures and media coverage of different issues it does show that total social and environmental disclosures, which have remained predominantly voluntary within Australia, are generally increasing across time. In noting the general increase in the extent of media coverage with various peaks and troughs across time), 1989, 1995 and 1997 represented major peaks. In 1989 and 1995 the media concerns are primarily related to employee health and safety issues. In 1995, particular attention was directed to the Moura mine disaster in which a number of BHP employees were killed. The higher level of media attention in 1997 arises primarily from attention directed at the Newcastle plant closure. The Newcastle plant was a major employer of people from the Newcastle area and had been in operation for many years. Overall, the associations between total annual report sentences and total

Table II.

Aggregated social and environmental disclosure and media attention by year

Notes: Spearman's rank-order correlation between total annual report sentences and total media articles = 0.5201 p = 0.023); Spearman's rank-order correlation between positive annual report sentences and unfavourable media articles = 0.4508 p = 0.046); Spearman's rank-order correlation coefficient one-tailed) expresses the correlation of the number of annual report sentences by year, with the number of media articles in the same years

Total annual Total media Year report sentences articles

1983 16 20

1984 36 15

1985 41 11

1986 270

1987 215

1988 201

1989 51 47

1990 437

1991 491

1992 743

1993 54 19

1994 39 29

1995 69 139

1996 148 57

1997 204 234

Total 892 588 Minimum 16 0 Maximum 204 234 Mean 59 39

Positive annual report sentences

11 36 40 25 18 20 51 39 49 74 52 37 61

142 128

783 11 142 52

Unfavourable media articles

18 15 8 0 5 1 47 6 1 2 6 26 123 35 171

464 0 171 31

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

media articles; and positive annual report sentences and unfavourable media articles, is strong, with correlation coefficients of 0.520, p < 0.05, and 0.450, p < 0.05, respectively. While this year-by-year correlation is interesting, it may be that specific issues in the media generate a greater corporate annual report response than other issues covered by the media. This issue will be investigated shortly.

6.2 Results of testing H1 The first hypothesis posits a relationship between the total print media coverage given to attributes of BHP's social and environmental performance and the total levels of corporate social disclosure in their annual reports. Table III provides the results of the testing of the hypothesis for each of the five major themes.

The above table presents results of tests for the correlation across each of the 15 years for each general theme. H1 is supported in the themes of environment correlation coefficient 0.644), and human resources correlation coefficient 0.578), both with p < 0.05. These are the two issues with the greatest level of media attention and annual report disclosures.

A measure of correlation was also obtained for 15 of the 49 specific issues under examination see the Appendix for details of each of the issues identified under each of the five major themes). These 15 issues recorded 694 of the total annual report sentences 78 percent of all sentences), and 570 media articles 97 percent of the total). With hindsight, the application of 49 categories to this one company was excessive, and as a result, a number of issues could not correlate. Perusal of the 34 issues which attracted no, or very minimal, media attention indicated that, with only three exceptions[16], they received no, or minimal annual report disclosure. This is what would be expected in terms of the hypothesis with limited concern as proxied by media attention) there is an expectation that there is limited incentive to provide disclosures. Table IV presents the relationship and level of significance for each of the specific issues for which a correlation statistic was calculated.

Environment. There was a modest rise in media attention and annual report disclosures in the years 1989-1995, and a pronounced rise in both occurring from 1996. A correlation of 0.6441, significance 0.005, is found for the environment category.

A test of legitimacy theory

327

Total Environment Energy Human resources Community involvement Others

Spearman rank-order correlation coefficient

+0.9000 +0.6441 +0.3236 +0.5786 +0.0481 0.3270

Significance one-tailed test)

0.019 0.005 0.120 0.012 0.432 0.117

Table III.

Correlation by theme between annual report sentences and media articles H1

Note: This table correlates the number of annual report sentences over the 15 years of the study, with the number of media articles in the same years, for each of the five themes

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

328

Individual issues

Environment

Pollution issue no. 1) Standards and regulations issue no. 2) Prevent/repair damage issue no. 4) Awards issue no. 9) Review/impact studies issue no. 14)

Energy

Conservation awards issue no. 24)

Human resources: Employee health and safety issue no. 27) Employment of minorities issue no. 28) Employee training issue no. 29) Employee remuneration issue no. 31) Employee morale issue no. 34) Industrial relations issue no. 35) Other issue no. 36)

Community involvement

Donations issue no. 37) Aid or compensation issue no. 47)

Spearman rank-order correlation coefficient

+0.3850 +0.7787 +0.6235 +0.1722 +0.3660

+1.0000

+0.5786 0.2709 +0.3484 +0.0628 0.3347 +0.2752 +0.0576

0.3359 +0.9898

Significance one-tailed test)

0.078 0.000 0.006 0.270 0.090

0.000

0.012 0.164 0.102 0.412 0.111 0.160 0.419

0.096 0.001

Table IV.

Correlations between annual report sentences and number of media articles for individual issues H1

Note: This table correlates the number of annual report sentences over the 15 years of the study, with the number of media articles in the same years, for each individual issue that recorded a correlation in testing

Within the environment theme, Table IV shows that four specific sub-issues correlate with values of p < 0.10, these relating to pollution, standards and regulations, prevention or repair of damage, and review or impact studies. In considering the respective levels of media attention it is interesting to note that issues associated with the prevention or repair of damage, pollution, and review or impact studies attracted the greatest level of media attention from among the 16 environment sub-issue disclosure categories. The fourth issue with a significant correlation, standards and regulations, ranked as number six in terms of media attention. These four sub-issues within the general theme of environment were also the four issues that attracted the highest level of annual report disclosure issues associated with repair or damage to the environment attracted the highest level of media attention as well as representing the environmental issue with the greatest amount of annual report disclosure).

These results, for the environment theme in total, and for four of its issues, support H1 in that these categories record positive correlations ranging from 0.366 to 0.778, with values of p < 0.10. Those environmental issues not directly supporting the hypothesis are those with minimal media coverage and minimal annual report disclosures. Interestingly, whilst the company dedicated some space in the annual report to environmental awards, the awards attracted little media attention, perhaps reflecting a view that the media tends to focus on

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

negative attributes of corporate performance or perhaps, that the media considered the awards to be trivial).

Energy. The theme of energy attracts minimal attention, with media articles appearing only from 1995 and an annual report response in 1996 alone. The Spearman's Rank-order correlation coefficient of 0.3236 significance 0.12), results entirely from the information on energy conservation awards in 1996. While the general lack of attention to the overall energy theme is interesting, 329 the general absence of data over the years for the theme does not allow a

conclusion which supports the first hypothesis. Human resources. The data for human resources resulted in a Spearman's

rank-order correlation coefficient between total media articles and total annual report sentences of 0.5786 at a significance level of 0.012. The sub-issue associated with employee health and safety also produced an association significant at the 10 percent level. The analysis of the total result offers evidence to support H1, with further confirmation supplied by health and safety issues. Employee health and safety attracted the greatest level of media attention within the human resources theme as well as representing the human resources issue with the greatest amount of annual report disclosure.

Community involvement. Media articles mentioning BHP's involvement with the community are infrequent, with the majority of the 1995 and 1996 articles arising from compensation and support considerations for the people of the Fly River who were severely affected by the mining operations of Ok Tedi, an organisation in which BHP has an ownership interest. The sporadic pattern of media attention and annual report disclosure led to an absence of an overall association correlation coefficient 0.048, p = 0.432). However, disclosure of aid and compensation issues to communities affected by corporate activities correlated strongly coefficient 0.989, p < 0.001) in the years when the Ok Tedi operations were receiving adverse media exposure. A conclusion supporting H1 cannot be drawn from the results obtained in this theme, although there is a significant association for the disclosures relating to the aid and compensation issues the issues attracting the greatest level of media attention within the theme of community involvement.

Other. Under the ``other'' theme, two media articles on the subject of corporate governance, and an average of 3.5 sentences for ten of the 15 annual reports on BHP's general social responsibility were recorded. Testing did not reveal an association between media articles and annual report disclosures for this theme.

As a concluding comment in relation to H1 the results show that the specific issues attracting the greatest level of media attention were also generally the issues which had the greatest amount of annual report disclosure.

6.3 Results of testing H2 H2 predicts that higher levels of unfavourable media attention given to BHP's social and environmental implications will result in higher levels of positive social and environmental disclosure in the annual reports. Of the total 892

A test of legitimacy theory

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

330

annual report sentences collated for the study, 783 were positive in nature, as compared to 464 of 588 media articles portraying an unfavourable impact on society or the environment or both)[17]. Table V presents the results of the aggregate testing of the two variables across the five general themes, where the overall relationship resulted in a Spearman's rank-order correlation coefficient of 0.4508, significance 0.046.

The results for the total, and the environment and human resources themes, are significant p < 0.05). Additionally, nine of the individual issues produced an association, of which five were significant p < 0.10), as shown by Table VI. These nine issues recorded 401 51 percent) of the positive annual report sentences and 457 98 percent) of the unfavourable media articles.

Table V.

Correlation by theme between positive annual report sentences and unfavourable media articles H2

All themes Environment Energy Human resources Community involvement Others

Spearman rank-order correlation coefficient

+0.9898 +0.6640 +0.4604 0.4901

Significance one-tailed test)

0.046 0.003 0.042 0.032

Note: This table correlates the number of annual report sentences over the 15 years of the study, with the number of media articles in the same year, for each theme; Sufficient data were not available to produce a correlation for the themes of energy zero unfavourable media articles) and others zero unfavourable media articles)

Table VI.

Individual issues

Environment

Pollution issue no. 1) Standards and regulations issue no. 2) Prevent/repair damage issue no. 4) Review/impact studies issue no. 14)

Energy

Human resources: Employee health and safety issue no. 27) Employment of minorities issue no. 28) Industrial relations issue no. 35) Other issue no. 36)

Community involvement

Aid or compensation issue no. 47)

Spearman rank-order correlation coefficient

+0.3684 +0.8126 +0.5179

0.3167

+0.4188 0.1590 +0.3367 +0.0577

0.6274

Significance one-tailed test)

0.088 0.000 0.024 0.125

0.060 0.286 0.110 0.419

0.006

Correlations between positive annual report sentences and number of unfavourable media articles for individual issues H2

Note: This table correlates the number of annual report sentences over the 15 years of the study, with the number of unfavourable media articles in the same years, for each individual issue

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

Environment. At the general level of environment a Spearman's rank-order correlation coefficient 0.6640 is produced, significance p = 0.003. Three specific sub-issues are also strongly correlated p < 0.10), namely:

1) pollution Spearman's rank-order coefficient of 0.3684, significance 0.088);

2) standards and regulations Spearman's rank-order coefficient of 0.8126, significance 0.000);

3) prevention or repair of damage Spearman's rank-order coefficient of 0.5179, significance 0.024).

The above three issues attracted the greatest amount of negative media attention and were also the issues within the theme of environment that recorded the greatest amount of positive annual report disclosure. The theme's results, and those of three major categories, offer support for H2.

Energy. Only 11 annual report sentences were provided in total across all years all positive). Consistent with the hypothesis and the minimal corporate disclosures) there were no negative media articles on the theme of energy for BHP, and consequently no association was derived. Hence, for this theme, there was no statistical support for H2.

Human resources. Prominent over the period were safety issues in the oil industry 1989), the Moura disaster 1995), and the proposed Newcastle plant closure 1997). Significant results were obtained for the overall total of the human resources theme correlation 0.4604, significance 0.042) and health and safety issues correlation 0.4188, significance 0.060). Health and safety issues attracted the greatest amount of negative media attention within the human resources theme as well as providing the greatest amount of positive annual report disclosures within the theme. The aggregate result for the human resources theme and the result for the sub-issue of health and safety support H2.

Community involvement. Similar to the overall total for this theme, unfavourable media articles and positive annual report responses are very infrequent, only occurring in a limited number of years. For eight of the 11 sub-issues within the community involvement theme there was no unfavourable media attention recorded in any of the years, with these same eight issues each attracting a maximum of one sentence of positive annual report disclosure across all the years of the study. The resulting association between the unfavourable media attention and the positive annual report disclosures is negative and hence not in accordance with H2. Despite the total theme result, the issue associated with aid or compensation to local communities affected by operations e.g. Ok Tedi), produced a Spearman's rank-order correlation coefficient of 0.6274, significance 0.006. Issues associated with aid or compensation attracted the greatest amount of negative media attention within the community involvement theme. Hence while the results pertaining to the sub-issue of aid or compensation issues provides results consistent with H2, the results of the theme in total do not support H2.

A test of legitimacy theory

331

Downloaded by Federation University Australia At 22:30 17 July 2018 (PT)

AAAJ 15,3

332

6.4 Time lags

As prefaced earlier, one objective of this study is to examine the possible existence of time lag effects in social and environmental disclosure. Both Brown and Deegan 1998) and O'Donovan 1999) suggest that there could be time lags from media attention to eventual annual report disclosure, and this is explained on the basis that managers react to media attention. This is also consistent with media agenda setting theory which proposes that media attention precedes shifts in the ``public agenda''. Accordingly, testing was conducted again, relating media attention in time t, to annual report disclosure in time t + 1. The statistical results a