Changes in stakeholder dynamics and salience during a mining disaster

The objective of the study was to illustrate the consequences of management oversight, as an element of poor corporate governance, of timeous stakeholder identification and engagement during a South African mining crisis. A secondary objective was to apply Quantitative Narrative Analysis (QNA), a methodology thus far mainly used in sociological research, to the understanding of this governance problem.


Introduction
In August 2012, a crisis in the South African mining industry evoked international attention.It involved an illegal labour strike at the Lonmin platinum mine, the third largest platinum producer in the country, based in the bushveld town of Marikana in the North-West province of South Africa.The police shot striking employees and protestors, killing 34 people.An additional ten people died in violent protests in the build-up to the main shooting event.
Earlier in 2012, leading up to the strike in August, labour and community unrest in the platinum mining industry escalated into a struggle over union recognition between the established and recognised National Union of Mineworkers (NUM) and the new-comer union, the Association of Mineworkers and Construction Union (AMCU) (Creamer Media, 2012).Rock Drill Operators (RDOs) perform a critical function in mining operations, thus making this group of employees an easy target for instigators (Tiro, 2012) and for changing union allegiance (Jones, 2012).
The triggering factor in the Marikana disaster was the approach by AMCU to the management of Lonmin with a request to discuss the salaries of the RDOs.Management refused to entertain these discussions as the recognised union, NUM, was not involved in the request and because management believed that the current wage agreement reached with NUM was still effective for another year (Mokqata, 2012).
The objective of the case study was to illustrate the consequences of poor governance in terms of management oversight of early stakeholder identification and engagement during a mining crisis.The study provided an opportunity to explore the application of stakeholder theory within the South African mining context, in particular, with a view to understanding how a crisis affected the company, its main stakeholders and the interactions between all parties.The use of Quantitative Narrative Analysis (QNA) (Franzosi, 1989;2010) afforded the application of a methodology primarily used in sociological studies, to the understanding of an organisational problem.
This article proceeds with a discussion of the literature relating to stakeholders and stakeholder theory within the context of corporate governance, the management of stakeholders to address sound corporate governance, and the theory underlying QNA as the chosen methodology to understand the variety and strengths of stakeholder relationships that were active in the Marikana mining disaster.Recommendations are later furnished to leaders regarding the importance of early stakeholder identification and management as elements of sound corporate governance.

Stakeholder identification and engagement
Corporate governance is founded on ethics (Young & Thyil, 2008) or "the sets of relatively shared values and norms that are expressed and negotiated" (Fleming & McNamee, 2005: 137).Governance invokes the values and norms of behaviour that direct the relationship between companies and the societies within which they operate (Francis & Armstrong, 2003).Corporate governance constitutes the thread that binds together the management of the organisation, its board and its broad stakeholder groupings through the definition of mutual objectives and the strategies by which such objectives will be achieved (Organisation for Economic Co-operation and Development, 2005) in ways that are "ethically defensible" (Fleming & McNamee, 2005: 137).Accordingly, Rossouw (2009: 6) proposes the concept of "the governance of ethics" where various organisational behaviours, such as sound stakeholder management, ensure that the organisation is ethically governed.
The on-going assessment of stakeholders and their attributes is an important management practice as the dynamics of stakeholder engagement change over time according to the shifts in issues (Clarkson, 1995).Stakeholder engagement demands that leaders establish processes and structures to ensure mutual commitment to solving issues between the organisation and the environments within which it operates (Manetti & Toccafondi, 2012).It is critical that management appreciates the concept of stakeholder salience or the degree of prominence to accord competing stakeholder claims and desires (Mitchell, Agle, & Wood, 1997).It is argued that, had the management of Lonmin appreciated the importance of early stakeholder identification and meaningful stakeholder engagement, major tragedies that occurred in this mining disaster may have been averted.
A stakeholder is an individual or a group of individuals who impact the objectives of an organisation or who are affected by the achievement of such objectives (Philips, Freeman & Wicks, 2003;Tullberg, 2013).The concept of stakeholder is, however, a contested one with definitions emphasising central themes or elements that are pertinent to the specific context within which the construct applies (Miles, 2012).Freeman et al. (2010) suggest that a refined and universally accepted definition of a stakeholder would be practically and conceptually useful, but acknowledge that no single definition would necessarily fit every situation to which it needed to be applied.
The most popular classifications of stakeholders are based on the extent of stakeholder influence on organisational survival (Miles, 2015) as well as on the existence or absence of contractual agreements between stakeholders and the organisation (Savage, Nix, Whitehead & Blair, 1991).Primary stakeholders are those upon whom the organisation is dependent for its ongoing survival (Clarkson 1995), who have a direct stake (Carroll & Buchholtz, 2012) and who exert an economic impact on the organisation (Savage et al. 1991).
Primary stakeholders include shareholders and investors, employees and managers, customers, local communities, governments, suppliers and other business partners (Carroll & Buchholtz, 2012).Secondary stakeholders hold an indirect stake in the organisation yet can still be influential in that they can influence or affect, or be influenced or affected by, the organisation (Thijssens, Bollen & Hassink, 2015).However, they are not essential to the survival of the organisation (Clarkson, 1995).Secondary stakeholders may include governments and regulators, civic institutions, social pressure groups, media and academic commentators, trade bodies and competitors (Carroll & Buchholtz, 2012).Phillips (2003) differentiates stakeholders according to the nature of the legitimacy of the relationship between them and the organisation.Normative stakeholders are those to whom the organisation "has a moral obligation, an obligation of stakeholder fairness" (Phillips 2003: 124) that extends beyond all the other social actors.Derivatively legitimate stakeholders are those who have the potential to impact normative stakeholders, and therefore warrant management's attention, for example, members of the media (Benn, Abratt & O'Leary, 2016).Dangerous and dormant stakeholders, described in the typology of Mitchell et al. (1997), can be compared to these stakeholders since they have the power to influence the organisation and its normative stakeholders but the organisation does not have any moral obligation to them.Carroll and Buchholtz (2012: 63) further deconstruct the concept of stakeholder by suggesting that it is important to understand the notion of 'stake', or an "interest or a share in an undertaking".Employees and shareholders hold direct influential stakes in the organisation and can, therefore, be considered primary stakeholders.Similarly, the media does not hold a direct stake in the organisation but can exert an influence on the organisation, thereby becoming a secondary stakeholder.
Various methodologies and typologies of stakeholder attributes have been proposed to describe the forms of stakeholder classification and the importance that management should accord such stakeholders over time and with due regard to the issue confronting the organisation (Fassin, 2012).Alpaslan, Green and Mitroff (2009) add that understanding stakeholder theory and adopting a stakeholder model is useful as a corporate governance tool that can be applied and used in crisis situations.
The present study adopted the typology of stakeholder categories proposed by Fassin (2009;2012) which includes stakeowners (stakeholders), stakewatchers (pressure groups), stakekeepers (regulators) and stakeseekers (activists).The typology is based on eight attributes: legitimacy, dependence, influence, power dominance, loyalty, responsibility, fairness and reciprocity (Fassin, 2012).These attributes vary depending on how the stakeholder is classified.The stakeowner has a legitimate stake in the organisation and can, therefore, be equated to a primary stakeholder.The interests of stakeowners are protected by the stakewatchers such as unions who protect employee rights or consumer protection associations.Stakekeepers have no stake in the organisation but can influence and control the organisation through regulations and restrictions and include groups such as auditors and government departments.Stakeseekers are activists who identify new stakes to be claimed, and who attempt to involve themselves in organisational decisionmaking.Freeman's (1984) original hub-and-spoke stakeholder model has been enhanced to reflect different stakeholder attributes, for example, by the use of visual aids to graphically represent stakeholder interaction (Fassin, 2008;2010).In this regard, stakeholders and their relationships can be depicted through models that utlise size and shape variations, connecting lines, directional arrows, colour, and intensity of shading.The dynamism of stakeholder salience can be illustrated through a sequence of successive stakeholder network diagrammes that show, over time, how actions taken by stakeholders affect the organisation or other stakeholders, or how they indirectly affect the organisation through other stakeholders (Fassin, 2010).Such modelling and graphical representation provide a deeper understanding of stakeholders.
Quantitative Narrative Analysis (QNA) Franzosi (1989) developed QNA as a story grammar framework methodology to gather textual data to understand social issues.QNA, as a methodology, is growing in its application in the social sciences, with its real power residing in its relational properties (Franzosi, De Fazio & Vicari, 2012).
QNA involves interrogating the body of a text and applying the narrative principles of invariant linguistic structural properties, namely Subject-Verb-Object (S-V-O), in order to convert words to numbers (Franzosi 2010).This conversion of words to numbers occurs by dissecting the narrative clause into grammar objects.Franzosi (1989;2010) refers to the S-V-O relationship as a semantic triplet, where words are given meaning, incorporating the elements of who, what, where, when, why and how.A series of semantic triplets, collectively, amount to an event or "a set of actions performed at a particular place and time by some actor(s) against or in favour of some other actor(s)" (Franzosi, 1989: 276), and a collection of events contributes to a main dispute.The narrative expresses the meaning of events by decomposing an event into its parts and gaining an understanding of the significance of the parts in relation to the event (Elliott, 2005).In QNA, the coding categories, therefore, relate to each other through the specific S-V-O linguistic structure (Franzosi, 2010).
Based on the data detailing the presence of relationships between social actors, network models can be developed, visually depicting the direction of the relationship (i.e.subject to object) via an arrowhead (Franzosi, 2010), with the width of the arrowhead representing the frequency of the action (Franzosi et al., 2012).Network models can also reflect relations for a given time and place, provided the information has been linked to the recorded semantic triplets (Franzosi, 2010).

Data bases
Secondary textual data, stored in electronic databases, were sourced from English language newspaper articles that appeared in South Africa (Newsmonitor) and the United Kingdom (NewsBank) for the period 1 to 24 August 2012.The filtering keywords of 'Lonmin' and/or 'Marikana' were the criteria used to identify relevant articles.UK newspapers were included as Lonmin is listed on the London Stock Exchange as well as on the Johannesburg Stock Exchange.The data comprised 418 articles.QNA is well suited to analysing newspaper articles (Wada 2005).

Coding structure
A coding structure was designed to guide the coding of the categories of the content of the articles and to simplify the data analysis.The coding structure was based, firstly, on the hierarchical structure of a story grammar and then on a deep structural schema for news reports (Van Dijk, 1988).Figure 1 illustrates an adapted version of Van Dijk's structural schema, with some additions for the purpose of the present study.
In essence, news reports provide a summary of the story and/or further information relating to the story, such as sketching the situation and/or providing commentary.Sketching the situation may consist of the provision of background information, typically covering the current context or circumstances within which the story takes place, and may even include historical information related to the story.The main episode in the story describes the events (in terms of who, what, when, why and how) as well as the consequences of the events or even the main episode.For the purposes of the study, the consequences were adapted to accommodate further classifications according to actions (what stakeholders did), criticisms (what criticism was evoked), reactions (what stakeholders said in response), and market indicators and performance (such as changes in share price, commodity price or exchange rate).In Figure 1, the comments grouped under 'the situation' include opinions, criticisms, reactions and speculation regarding any of the events that took place as part of the main episode.The principle of coding the data as semantic triplets in the S-V-O format applied to events, consequences and comments.
A basic Microsoft Excel spreadsheet design was used to capture the coded items located in the articles.

Data gathering
Once the content was coded and recorded, based on the S-V-O relationship (Franzosi, 2010), i.e who did what to whom and when, an historical storyline of the sequence of events was created.The sequence of events supported the understanding of how the events leading up to and after the Marikana disaster started and unfolded.It was assumed that, in order to conduct a stakeholder analysis, there should be clarity regarding the 'when' of events.The events were further classified according to either the creation of new events or the consequences identified, enriching the understanding of the impact of how the salience of stakeholders can change over time.The content relating to the main episode was classified against an adapted deep structural schema for newspaper reports (Van Dijk, 1988) in order to capture both the events and the consequences of the events.Dates were recorded to understand the timing of these events and their consequences.
The next step involved identifying the main participating actors involved in the main episodes using Fassin's (2012) stakeholder model and noting the frequency of the occurrences/appearances of the different actors.
The relationships between the different actors and the events were mapped to assist in the stakeholder analysis, for example, double arrows highlight the precipitating or trigger events.The width of the arrows in relation to the other connecting lines, illustrates the relative strength (data count) of these reported events.In addition, variation in shape size and positioning was graphically represented for the different stakeholder models.Each relationship was considered within the context in which the interactions took place, requiring a qualitative interpretation of the sequencing of events.Finally, based on the outcomes of the above, sequential graphic stakeholder models were prepared to facilitate the stakeholder analysis of the Marikana disaster.
Using QNA methodology, the data can be evaluated independently of the subjects involved and unobtrusively collected (Bertrand & Hughes, 2005), thus avoiding any reactive responses from the subjects who could adapt their behaviour in response to being observed (Cooper & Schindler, 2011).QNA methodology also allows for the relationships between the subjects and actors to be recorded and counted for each appearance in different newspaper articles.The focus on the relationship between the actors thus presents the opportunity to conduct a stakeholder analysis where stakeholders are regarded as the actors or subjects in the newspaper content.It is at this juncture that QNA and stakeholder theory meet.
In the coding of the data, a number of assumptions were made.Where the initial articles consulted contained incorrect information, such information was corrected when new information came to light.This implied that once the data were analysed, the corrected semantic triplets were linked to the original earlier articles.A single semantic triplet was not captured more than once per article, even if it was repeated in the newspaper articles and, therefore, assumptions of the stakeholder involved had to be made in some cases.The events that evoked opinions about what should have happened in the past or what should take place in future to rectify the situation potentially influenced other stakeholders, the image and reputation of Lonmin or other stakeholders involved.Accordingly, they were deliberately included.

Phase 1: Sequencing of events
Using the guidelines provided by Franzosi (2010) for a sequence analysis, an historical timeline was constructed of the sequence of events surrounding the main episode.Sequence analysis assists in identifying the roles (impacts) that the respective actors (stakeholders) enact as well as the order in which they unfold.Events were classified as either precipitating events (those that occurred potentially as a consequence of previous events, but that served to trigger further events) or 'other' events.The sequence of events, as well as the precipitating events, were captured by data counts based on frequency tables.

Phase 2: Stakeholder identification
The data were first organised in a manner aligned to Fassin's (2012) typology by conducting an initial grouping of all the actors (as subjects and objects) recorded in the dataset.Franzosi (1989) advises that the coding of data from texts should be aggregated at the lowest possible level and data were coded for all actors mentioned in the text, provided the issue being discussed related, in some way, to the main event, that being the Marikana disaster.Once all actors were recorded as either actors as subjects, or actors as objects, the actors were re-aggregated into logical groups to simplify the application of the mass of data to Fassin's (2012) stakeholder framework.This was followed by a sub-grouping within the constituent grouping.Thereafter, each grouping was classified according to Fassin's (2012) typology of stakeowners, stakewatchers, stakekeepers or stakeseekers.The data counts of the data lines per group and subgroup assisted in identifying the frequency of mention of the main actors in the newspaper articles.

Phase 3: Mapping relationships between stakeholders and events
After the identification of the main stakeholders, further data counts and frequencies were analysed to identify the main relationships between stakeholders.This information was captured using the categories allocated (type and sub-type) according to the deep structural schema adopted for newspaper articles, and then by time period.The data count informed the strength of the documented relationships per semantic triplet and the assumption was made that the relative frequency of a relationship would determine the strength of that relationship.
Finally, this analysis was translated into graphic illustrations of the sequence of events as they unfolded to determine the effects of the incident on Lonmin and the main stakeholders involved.Such illustrations highlighted the connection between actors through directional arrows and the strength of the relationships through the relative width of the directional arrows derived from the volume of recorded content (frequencies of data lines).

Results
The findings are reported according to the phases of the methodology.

Phase 1: Sequencing of Events
The data were clustered into four logical time-periods to simplify the analysis for the sequencing of the events and are presented in Table 1.
 The period leading up to the illegal strike: Period 1 to 10 August 2012.The precipitating event was the dispute about the RDO salaries. The period when the illegal strike began, starting with a protest march on 10 August: Period 11 to 15 August 2012.
Various events took place, and during this period, ten lives were lost due to violence. The day on which the main shooting occurred: Period 16 August 2012.On this day, a number of incidences of violence were identified, and the approach taken was to treat the single day as a time period. The events and consequences following the strike-related violence: Period 17 to 24 August 2012.Phase 2: Stakeholder identification Twenty-one stakeowners were identified, for example, the local community, Lonmin employees, and the families of the victims who had died.A further 28 stakeholders were classified as stakewatchers for example, investors and business groups.Twenty stakeholders were classified as stakekeepers, including members of the media, national government entities and market analysts.Finally, three stakeseekers were identified -the criminal element in the form of the 'unknown persons' who perpetrated the violence, a Sangoma (a traditional healer in some South African cultures) and Julius Malema (the rising leader of a new political party).These classifications assisted in aggregating the data to promote further analysis and interpretation.
Stakeholders were isolated through a simple data count performed on all 2225 data lines.For the actors as subjects, the top five high-level groups identified were management, the media, governmental agencies, employees and unions.Governmental agencies, in particular, included the police, involving 180 data lines.For the actors as objects, employees, management, unions, government agencies and shareholders, were the most frequently mentioned groups.These collective top ten groups were further analysed, providing insight into the specific actors involved, resulting in the identification of the main actors: management, Lonmin employees (in various capacities), the two main unions (AMCU and NUM), the police, national government and competitors.Additional actors warranting consideration included shareholders, investors, the media, Malema, the families of victims and citizens of the country.A 'non-stakeholder' group featured in the top ten list and included the unknown persons (who perpetrated the violence) and the Sangoma.

Phase 3: Mapping relationships between stakeholders and events
To record the relative frequency of the mention of a relationship as a means of determining the strength of the relationship, each actor was cross-referenced against all other actors.These cross relationships between actors as subjects and actors as objects, and by stakeholder group according to Fassin's (2012) typology, were then tabulated (for both the frequency of total data lines and for unique recorded semantic triplets).In total, 390 semantic triplets were identified across the 24 days.Some semantic triplets were reported on more than one occasion, collectively comprising the 2225 data lines.In order to understand the relationships between actors, each semantic triplet was qualitatively reviewed in the sequence in which the event took place, and consideration was given to the count of the data lines for each recorded triplet relative to other events taking place in that time period.Therefore, the data were separated according to the semantic triplets that were recorded and/or reported within the four logical time-periods previously identified.
The classifications of the main identified stakeholders for the stakeholder analysis were juxtaposed against Fassin's (2012) typology, as indicated in Table 2.  2) was applied in the sequential stakeholder models in the form of graphical representations to illustrate the differentiation between stakeowners, stakewatchers, stakekeepers and stakeseekers.A fifth classification was deliberately added, namely that of nonstakeholder, in order to illustrate how actors who are initially classified as non-stakeholders can, over time, fall into one of the other stakeholder classifications.The direction of the arrow reflected the direction of the event (i.e.actors as subjects to objects, as recorded in the semantic triplets).The nature of the relationship was indicated by the strength of the connecting line; the wider the line, the greater the relationship strength based on the relative frequency of the data counts.
In each stakeholder model illustrated in Figures 3 to 8, a brief description is provided to label the connecting numbered lines.These numbers do not correspond to the semantic triplets, but rather represent how the order in which events that took place must be read.The numbering is sequential from one Figure to the next.On 11 August, a second protest march took place (8), and members of AMCU shot at NUM members (9).On 12 August, unknown persons, allegedly striking employees, marched to the NUM offices on the Lonmin premises and torched a security vehicle, killing two security guards and damaging seven company vehicles (10).The body of a Lonmin employee was found near the gathering of the protesting employees.In response, Lonmin called on the police for protection (11).The next morning, three members of the Solidarity Union (the oldest union in the country and one with no political affiliation) were assaulted on their way to work (12), resulting in this Union demanding protection for its members (13).As a consequence of the violence, the Lonmin premises and surrounding areas were heavily guarded by the police ( 14).However, striking employees continued to gather illegally (15).On 13th August, the police attempted to control the 3000 protestors and, in a clash, the police allegedly shot and killed three protestors ( 16).In the same clash, two policemen were killed (17).The following day, the body of a NUM shop steward was found close to the gathering protestors (18).As a consequence of the violence, production was halted and the Lonmin share price dropped (19).Lonmin management refused to negotiate with the protestors (20).

Period 16 August
The sequence of events described, have demonstrated how various actors were involved in the main episode, either as actors (as subjects), contributing to the unfolding of events, or as recipients, experiencing the consequences of these actions (actors as objects).The growing role of AMCU was evident in how the story grammar unfolds, supporting the view that the power exerted over Lonmin by AMCU was increasing and, as such, so was the salience of AMCU. Figure 6 illustrates the stakeholder model for the events that took place on the day of the main shooting.On 16th August, Lonmin issued a statement to shareholders and to the public, announcing the CEO's serious illness and related hospitalisation and warned that the strike action may impact production targets, further affecting the share price (21).Lonmin also officially issued an ultimatum to the striking employees, threatening dismissal if they did not return to work (22).The strike action continued, with protestors waving traditional weapons and placards (23).A traditional healer, or Sangoma, administered ritualistic medicine to the protestors (24).The respective union presidents addressed the protestors, appealing to them to disperse and to return to work (25).The union presidents received different responses from the crowd, reinforcing the rivalry between the two unions (26).Later that day, the police took action and, in their attempts to disperse the group, resorted to firing live ammunition at the protestors when it appeared that the protestors were attacking the police ( 27).The consequences of the attack were 34 deaths, 78 injuries, and 259 protestors arrested for various crimes.International media camera crews filmed the clash (28).

Period 17 to 24 August
Figure 7 illustrates the effects and consequences of the Marikana incident and the related violence directed at Lonmin and some of the main stakeholders.Lonmin is located at the centre and the data illustrate how the management of Lonmin reacted to the crisis and the effect of these actions.Lonmin issued a statement expressing condolences to the victims' families (29).The police issued a statement claiming self-defence (30).The victims' families reacted to Lonmin with anger and blame (31) and also directed anger and blame at the police (32).The media blamed the police for the deaths (33).Lonmin announced the extension of the ultimatum issued to the striking employees to return to work (34).The extension evoked criticism and accusations of insensitivity from striking employees (35), AMCU leadership (36), the media (37), and even the South African President, Jacob Zuma (38).The President also appointed a Commission of Inquiry to investigate what had taken place (39).NUM and AMCU blamed each other and Lonmin for the violence (40).Due to the extended illegal strike, production was halted and Lonmin's share price continued to drop (41).
Figure 8 illustrates the stakeholder model for other stakeholders involved in the disaster and identifies the effect on these stakeholders.Some of the events took place simultaneously with those listed above during the same period.Julius Malema (the young leader of the new political party) visited the protesting employees to show support for their position (42) and blamed the National President, Jacob Zuma, for the violence (43).Two days later, Malema and his colleagues appeared in court and offered financial assistance to the arrested protestors to assist with their legal representation (44).On the same day, employees from neighbouring platinum mines presented wage demands, proceeding to strike illegally two days later (45).Almost a week after the police shooting, President Zuma visited the protesting employees and addressed them to hear their demands and to extend his condolences to the victims' families (46), evoking criticism for taking so long to address them (47).On the 23rd August, across the country, memorial services took place for the victims of the violence; one service was hosted by Lonmin (48).Malema allegedly paid for the main memorial service that was held (49), with a ministerial delegation attending, representing the national government (50).Once again, Malema addressed the mourning crowd and openly criticised the ruling party and the President, resulting in the crowd becoming unruly and the ministers leaving before the service was over (51).Representatives from neighbouring competitor, Impala Platinum, attended the memorial service (52).The media and other entities criticised Malema and accused him of being opportunistic by using the Marikana disaster as a means to promote his political agenda (53).

Discussion
The overall objective of the study was to illustrate the consequences of management oversight of early stakeholder identification and engagement during a mining disaster.A secondary objective was to apply QNA methodology (Franzosi 1989;2010) to assist in the understanding of this organisational problem.
From the findings, it can be seen that the stakeholder salience and the dynamics between stakeholders changed substantially over the period of 24 days.For example, the shape size of AMCU, relative to NUM, changed from the period 1-10 August to the period 17-24 August (compare Figure 3 and Figure 7), representing the relative salience of each union over a short time period.This change in relative shape sizes also signalled the change in dominance of the two unions where the importance of AMCU (previously the nonrecognised union), overtook that of NUM.However, the management of Lonmin appeared not to recognise the increasing salience and importance of AMCU as a stakeholder.
Similarly, the salience of the police increased from period 11-15 August to period 16 August (compare Figure 4 to Figure 6).Initially, the police were considered a statekeeper only.However, as Lonmin increasingly came to depend on the police for the safety of its employees and its property, their salience to Lonmin increased.However, the management of Lonmin did not factor in the negativity towards the police held by the members of AMCU, the unknown persons, the families of victims and the media.
The stakeowner group changed from being applied exclusively to RDOs to include protestors, on the basis that they shared a desired stake in Lonmin, namely the wage demand (see the change from Figure 3 to Figure 4).Unknown persons were classified as criminals in the newspaper articles and, therefore, initially, were categorised as non-stakeholders (see Figure 3).However, according to Fassin (2012), these unknown persons should be considered to be stakeseekers (activists), even though their desired stake in the company was unknown at the time.With increasing reports of the involvement of the unknown persons in perpetrating violence and their affiliation to the protesting group, from period 11-15 August (Figure 4) onwards, these unknown persons, while still represented as a separate group of actors, were portrayed as floating within a shape symbolising the stakeseekers.Over a period of a few days, they changed from being nonstakeholders to being stakeseekers.This group also overlapped with the RDO group since reports alleged they were from the same group.The Sangoma was not considered to have a stake in Lonmin and, therefore, was classified, initially, as a non-stakeholder.However, his presence had a direct influence on the protestors, and, as such, he had an indirect influence on the appeals by the company management for employees to return to work.Accordingly, he was also represented, later, as a stakeseeker (Figure 6).The move of the Sangoma from non-stakeholder to an influential stakeseeker was not considered by the management of Lonmin.
Figure 7 illustrates the consequences of the violence on Lonmin, the police, the families of the victims, the unions and shareholders.While, initially, Malema was not classified as a stakeholder, it emerged that he was developing a growing, intended stake in the organisation, although it was an indirect one.Figure 8 illustrates a networked picture with Malema, not Lonmin, at the centre of the various interactions.
Stakeholder dynamism is illustrated in Figure 9 in which the typology of Mitchell et al. (1997) is used to compare the relative positions of the two main unions, NUM and AMCU, involved in the Marikana disaster.NUM was the recognised, legitimate union due to attaining the legislatively required 50% membership, and had the power to act on the formalised stake it enjoyed in Lonmin.Therefore, it was a dominant stakeholder, demanding and expecting management's attention.Earlier in 2012, AMCU evidenced low membership at Lonmin, to the extent that Lonmin did not consider AMCU to be a stakeholder and refused to engage with this union or to entertain its requests to discuss the RDO demands (see Figure 3).At that time, AMCU was therefore, at best, a demanding stakeholder.However, as AMCU chose to represent the RDOs, its membership and its power dominance over Lonmin increased, changing its status into a dangerous stakeholder, with demands from the protesting employees for greater levels of engagement from management (Figure 7).Therefore, AMCU's salience changed from that of a minor stakeholder to one of a definitive stakeholder (Figure 9), illustrating how the salience of an actor can change in a short time period according to the changing environment in which an incident occurs.Such movement, arguably, is an important development that could inform leadership negotiating decisions and actions in crisis negotiations.
The study highlighted the importance for organisational leaders to develop a comprehensive stakeholder typology against which to plot and track the changes in stakeholder attributes over time.Such identification and measurement may assist management in addressing sound and ethical governance with stakeholders, especially during times of crisis (Laplume, Sonpar & Litz, 2008).The consequences of the absence of such action can result in the lack of recognition of and engagement with stakeholders who can exercise a material influence on the organisation and, as such, is a governance oversight.An understanding of the sequence of events and how respective stakeholders are involved in either influencing events or are affected by the events, can assist leaders in initiating preventive discussions and interventions to avert evolving crises during a dispute such as the one discussed in this paper.In this way, relationships with stakeholders can be effectively managed (Elijido-Ten Kloot, & Clarkson, 2010).QNA methodology holds promise for application in the field of management, especially during periods of crisis and, it is suggested, can be used by organisational leaders to track the changes in their stakeholder compositions to guide decisions relating to stakeholder engagement.
A limitation relating to the use of QNA methodology in the present study was the accuracy with which the secondary data, namely the content of the newspaper articles, is reported (Mouton, 2011), resulting in key messages and relevant information possibly being excluded.Similarly, a second limitation was that the data were limited to that contained in English newspaper articles, published in two countries.Accordingly, additional data that could have informed the study may have been omitted.
It is suggested that the methodology employed in the present study, be replicated in other industries in order for the value to be established of developing stakeholder typologies juxtaposed against the stakeholder landscape (Bundy, Shropshire & Buchholtz, 2012).In addition, future studies, integrating both quantitative and qualitative approaches when conducting stakeholder analyses, may enrich the understanding of stakeholder dynamics and salience during disputes.Qualitative input could be obtained from interviews with management and participating stakeholders to supplement the weakness of relying solely on newspaper content.

Conclusion
Stakeholder theory, at its core, embodies the interactions between actors in various contexts.It also includes the collection of concepts, theories and ideas related to how organisations approach the value created by stakeholders, and how they are managed in a transparent and ethical manner that balances the business needs with legitimate stakeholder rights (Freeman et al., 2010).
The effective management of stakeholders rests with the focused efforts and competence of management teams, requiring that stakeholder management becomes institutionalised as a form of good governance and that leaders identify stakeholders and develop stakeholder engagement strategies (Guerci & Shani, 2013).In this regard, the principles of accountability, transparency, responsibility and independence should guide such strategies (Meyer, 2011).The present study underscores these sentiments.

Figure 2 :
Figure 2: Legend used in the Lonmin stakeholder model Period 1 to 10 August In July 2012 (estimated date), Lonmin increased the allowances for RDOs outside the existing wage agreement (1).Violence erupted at a neighbouring platinum mine (Aquarius) (2), rumoured to have been instigated by AMCU or NUM -both unions denied any involvement (3).The AMCU president claimed he approached Lonmin management to open wage discussions (4) and the management of Lonmin reportedly refused to engage with AMCU (5).As a result, the RDOs staged an illegal protest

Figure 3 :
Figure 3: Stakeholder model for Lonmin (Period 1 to 10 August) Period 11 to 15 AugustFor this period, the stakeholder model is separated into two parts to illustrate the dynamism of stakeholder salience, in particular as it related to the police.Figure4illustrates the first part.

Figure 8 :
Figure 8: Stakeholder model for other stakeholders (Period 17 to 24 August)