The following is an in-depth discussion about the research methodology for The Land Reform Deception. The following information is intended to supplement the brief methods discussion included in the appendix of the book in order to provide additional details and rationale in the research decision-making.
There are three unique datasets underpinning The Land Reform Deception. I personally conducted 111 in-depth interviews with stakeholders ranging from government ministers, to military intelligence agents, farmers and farm workers. I identified and mapped 21,491 instances of conflict over an eight year period and used these data to create a unique longitudinal study of nationwide violence and intimidation. I also undertook a global survey of 34% (n=1,442) of Zimbabwe’s commercial farmers who had since migrated to 24 countries worldwide.
The following Mixed Methods Flowchart illustrates the three different methodologies, the data behind them, and how the datasets informed the overall study of the land seizure era.
Mixed Methods Flowchart
Study 1: Quantitative Violence Mapping
Prior to this research little was known about where violence proliferated and what types of conflict predominated during the land seizure era. Assessing the mechanisms of violence and how violence was used politically required a systematic and longitudinal assessment of type, frequency, and location of acts of violence.
Data from the Zimbabwe Human Rights NGO Forum was especially valuable because it is the most substantial source for records of violence since 1998. Indeed, prior to the land seizure era, few data collectors realised how severe, punitive and pervasive government strategies would become. At the time of research, the Forum was made up of sixteen member organisations who are, “concerned with the level and nature of organized violence and torture in the country perpetuated mainly, though not exclusively, by state agents and their ancillaries.” Their data forms the only systematic source of monthly violence information and is of sufficient quality to be cited by prominent NGOs, such as Human Rights Watch and the International Crisis Group.
Forum data included monthly reports on political violence in the land seizure era that provided the type, location, date and victim(s) for each act reported. The group’s special reports focused on issues such as election violence and torture. These monthly and special reports were selected as the basis for the violence mapping data set for this project because they offered the most detailed and systematic contemporaneous accounts available.
The decision to rely on secondary data for the mapping exercise was made for two reasons. First, my research study began in 2006, which meant it was unfeasible to gather comprehensive nationwide monthly accounts of violence several years after land seizures had begun, especially with vulnerable respondents such as farm workers. Second, with modification, comprehensive Forum data could provide insights into the proliferation of violence in the land seizure era. While the mapping data primarily functioned to visually identify where incidents occurred, it also played an important role in informing and complementing the other data sets by identifying potential patterns of violence, victim and non-victim types and regional variation that were further investigated using the other, more in-depth research methods. This research design exemplifying Morgan’s recommended approach of complementary assistance when combining methods.
Quality and Reliability
Data compiled by the Forum were gathered in an environment hostile to both victims and researchers. This inevitably resulted in some imperfections in the quality, completeness, and reliability of the data. Reports were gathered nationwide through a network of member organisations that were “very loosely bound but effective in terms of receiving information countrywide.” Some member organisations had brick-and-mortar locations that victims could travel to, while others relied on sending out data collectors to rural and urban locations in order to gather victim testimonies.
It must be emphasised that the Forum data by no means captured all incidents of conflict in the land seizure era; these records do not exist. The Forum’s violence data is the best available, given the exceptionally demanding and risky data collection environment. It provides a snapshot insight into the overall conflict environment present during the land seizure era.
The Forum reports made every effort to be systematic, detailed and consistent. They included dates, precise numbers of victims and perpetrators, and offering legally-framed descriptions of each event. Reports using media information indicate this use. Forum reports included offences committed by the MDC. Based on Tuchman’s guidelines for assessing and corroborating secondary sources, published Forum data represents an impressive and laudable best-effort. However, the data that fell short of the standards required for direct utility in this specific research project.
The limited coverage of data collection centres cast doubt on representativeness, a problem exacerbated by reliance on self-reporting. Self-report data is notoriously prone to underreporting, especially in environments where victims have limited resources to travel from distant areas to reporting centres. In a Forum report on election violence in Mberengwa, underreporting was estimated to be 30:1 between cases recorded in a community survey and those reported to data collection centres by victims seeking help. Moreover, self-report data is also prone to over-reporting where individuals have an incentive to overstate events. For instance, members of the political opposition might overstate the severity of violence in order to portray the government in a negative light.
The inclusion of news media sources, which provided biased views, in Forum reports likely introduced bias into their data. Selection and partisan bias were particularly problematic in a country where the dominant newspaper, The Herald, was a government-owned mouthpiece, making it likely to suppress information about violence, while independent newspapers more likely to have publicised these events, had been subjected to years of intense government criticism, investigations, and violence. Therefore, I excluded data captured from news media sources by the Forum from my datasets.
There was little indication of pre-defined, social scientific methodology in data collection or any centralised, consistent oversight of the analysis. This is unsurprising and understandable, given that the Forum’s data collection evolved organically as the land seizure era unfolded. However, for the purposes of inclusion in this specific research project, I found variation in the quantity and quality of data over time. Reporting became more substantial over time, probably because of changes in collection or analysis. Additionally, there were inconsistencies in the description of violent events; for example, “torture” indicated discrepant forms of violence across cases. Lastly, monthly Forum tabulations of violence and intimidation by type were sometimes inconsistent.
In referring to contemporary intra-state conflict, Kalyvas states that “violence tends to be disproportionably located in the countryside. Yet, these wars tend to be viewed through a heavily urban lens by both scholars and practitioners.” This “urban bias” was not a strong feature of Forum data. While a Forum representative noted that, “It is true to some extent that Harare would have received the most cases because of the proximity of the reporting centres,” their data showed a significant number of incidents occurring in rural areas. This rural reporting occurred nationwide for the period of the study.
Selected Application of Forum data in this Research Project
Given the limitations in the Forum’s data what utility did it offer this research project? The US Department of State noted the Forum’s “lead in coordinating reports on human rights violations and abuses,” and their reports repeatedly cite Forum data. The Zimbabwe Officer at the State Department and the Zimbabwe Desk at the UK Foreign and Commonwealth Office stated that the Forum presented the most reliable data available. Both officials pointed out that the data, although incomplete, effectively represented the profile of violence. Moreover, the inclusion of Forum data by prominent NGOs such as Human Rights Watch and International Crisis Group all suggest that the data is of high quality.
I therefore determined that, with modification, the Forum data could still be invaluable to the research, in spite of its deficiencies. Although some of the relatively unsystematic treatment of data by the Forum could not be retroactively addressed, I was able to introduce a consistent new framework for operationalisation, tabulation and analysis. This way more ‘raw’ Forum data could then be used for the mapping research. I accomplished this by first separating the Forum data from the Forum analysis.
First, I discarded the Forum’s monthly violence tabulations because they were not always accurate and I applied new coding guidelines to the data. I initially deduced ten codes (‘violence types’) from my assessment of Forum reports: Abduction/Disappearance, Assault, Displacement, Intimidation/Verbal Threat, Murder, Attempted Murder, Property Damage/Theft, Rape, Torture, and Unlawful Arrest. Second, I discarded suspect news media-sourced reports to avoid potentially biased advocacy data. Third, I undertook a pilot assessment of the 2002 data and refined my process, such as by narrowing the definitions of the codes. Finally, I then ‘coded’ year 2000 to 2008 and recorded the results.
Coding is a process where written data, such as victim testimony about violence, is converted into quantitative data, such as for use in the violence mapping for this research. To code the Forum data I first read each incident in the Forum reports, which were written in prose, to determine the location, month, and frequency of each violence/intimidation type. For example, the following is an excerpt from a March 2001 Forum report:
A.M. was kidnapped [coded: 1 – Abduction/Unlawful Arrest] and assaulted [coded: 1 – Assault] by a member of the CIO on 1 March 2001, while bathing in the Rozva River. He was threatened with death [coded: 1 – Verbal Threat] for having supported the MDC in the run-up to the Bikita West by-election.
Once I had coded all relevant Forum reports, I repeated the process for a second time. I compared the results using Excel and in any instances where there were discrepancies between the two passes I re-visited the relevant violence accounts and made a final determination. This process of coding the reports twice increased the reliability of the process.
I coded all instances of political violence while all violence that was not clearly political was excluded. I relied upon a definition for political violence used by the UNHCR, “organized violent activity for political goals,” to determine whether accounts were political or not. As Tilly points out, “a good deal of violent behaviour occurs under the cover of law. Government agents and allies regularly employ violence as they pursue their own ends.” This was certainly true in Zimbabwe during the land seizure era. For this reason, all violence that was clearly reported was included, regardless of its legality under Zimbabwean law or the involvement of state agents authorised to use force.
I coded incidents undertaken by ZANU-PF only if the acts were committed by active state agents or those tacitly operating with state approval. I did not include intra-party violence. I also did not code violence undertaken by the MDC – and Forum reports indicated instances, highlighting the group’s efforts to be impartial. The reason why I did not include MDC violence is because the mapping portion of The Land Reform Delusion seeks to understand how political violence influenced the seizure of commercial farms and the suppression of the MDC. The research examined how and why violence was used to achieve political power, not to assess the relative levels of government and opposition violence. As such, it would be inappropriate to include MDC incidents.
It should be noted that evidence from Forum reports shows that the overall volume of violence by MDC members against ZANU-PF members, workers or farmers was relatively small, accounting for just two percent of incidents in one report. This finding that only a low volume of violence committed by the MDC was supported by the Solidarity Peace Trust report, which found MDC members committing only one percent of violence. Violence against farmers and workers was overwhelmingly undertaken by ZANU-PF agents.
In any instances where the political affiliation of the individuals undertaking violence or other key details of the incident were unclear, I did not include the account.
After I populated a spreadsheet with the resultant coded data for the nine-year period of the study, I combined the code for “Abduction or Disappearance” with the code for “Unlawful Arrest.” The distinction between these codes became pointless because Forum accounts demonstrated that victims were often handed back and forth between state agents and farm invaders who were holding them against their will. Additionally, the “Displacement” code was redacted due to substantial underreporting; Forum data insufficiently correlated with other data on displacement, such as from the UNDP.
The final eight codes and operational definitions are set out in the table below.
Quantitative Violence Mapping Codes
|Abduction/Unlawful Arrest||Forcible and unlawful restriction or elimination of an individual’s freedom, including brief occurrences.|
|Assault||An act of intended and unlawful violence against another person.|
|Murder||The unlawful, intentional killing of another person.|
|Murder (Attempted)||Malicious and unlawful intention to kill.|
|Rape||Unlawful sexual intercourse (vaginal, anal or oral penetration).|
|Torture||“Any act by which severe pain or suffering, whether physical or mental, is intentionally inflicted on a person for such purposes as obtaining from him or a third person information or a confession, punishing him for an act he or a third person has committed or is suspected of having committed, or intimidating or coercing him or a third person, or for any reason based on discrimination of any kind, when such pain or suffering is inflicted by or at the instigation of or with the consent or acquiescence of a public official or other person acting in an official capacity.”|
|Intimidation/Verbal Threat||An act in which no physical violence has occurred to the person, but which unlawfully and specifically implies or directly states the threat of future physical violence.|
|Property Damage/Theft||“Property damage” is the damage or destruction of property by someone other than the owner and not by natural phenomena. “Theft” is the illegal seizure of property without the owner’s freely given consent.|
Geographic mapping of longitudinal violence data facilitated the identification of patterns such as incident clusters, proliferation or absence in unexpected areas, and distribution around key cities.
Maps were based upon 21,491 individual acts of violence coded from 90 Forum reports and seven special reports covering the 99 months of the land seizure era. Each incident was depicted by a colour circle on a geographical map corresponding to the month in which the incident occurred. Given that data were unavailable for January 2000 and from July to December 2000, the mapping resulted in 92 geographical maps for the 99 month period.
While maps provided unique insight into the land seizure era, they have limitations. First, maps are only as accurate as the source data. The aforementioned inadequacies in the data made it difficult to accurately depict some incident locations. For example, some accounts only indicated approximate locations in provinces. In such cases I placed the incident symbol in the approximate location. If the location was completely unknown, the data point was not included.
Additionally, limitations of scale prevented distinction between incidents occurring close together, especially in population centres such as Harare and Bulawayo. This resulted in symbol clusters extending beyond city confines, which may be misleading regarding the precise locations of events. Nonetheless, every effort was made to cluster incidents as close to the location where they occurred as was possible given the limitations in the source data and the ability to depict incidents on relatively small maps.
This research only made claims about mapping data to the provincial level, reflecting that it was only at the provincial level that the data was most accurate. The incidents depicted on the maps were indicative of the general locations of where they occurred. As such, the utility of the maps – to provide basic geographic trends of the location as well as the proliferation and dissipation of conflict – including the identification of conflict “hotspots” – remains sounds.
The sometimes inexact source data and technical limitations of being able to densely place symbols in a confined location means that the maps are only accurate to the provincial level. Despite these limitations, the maps provide unique insights for this research because I only made claims based exclusively on them to the provincial level. The maps were indicative of where violence proliferated and dissipated and provided a useful tool to enable me to locate interview respondents in regions and provinces where violence was most concentrated.
Study 2: Qualitative Interviews
I conducted 110 in-depth, semi-structured qualitative interviews in order to investigate experiences on commercial farms during the land seizure era. A semi-structured interview design, where some questions are prepared in advance while others arise out of topics and issues raised in the interview, is ideal for this kind of research because it allows a set of themes to be explored and probed as the interview evolves. In-depth semi-structured qualitative interviewing of the kind used in this study has been a well-established means of data collection in the social sciences for decades. There are hundreds, if not thousands, of methodological sources supporting the value of qualitative interviewing in leading journals across the social sciences. The Land Reform Delusion has benefitted from the strength of this methodological background, enabling the most robust approaches to be used.
I conducted all interviews in person except for one telephone interview with a respondent living in Australia. Interviews were conducted with three sample groups: with 24 farm workers; second, with 55 commercial farmers, farmers’ wives, and those with privileged knowledge of the farming community; and third, with 31 members of government or those with privileged knowledge of government, including perpetrators of violence. The worker and farmer samples predominantly consisted of victims, whereas the third sample provided the countervailing government perspective.
Qualitative Interview Sample One: Farm Workers
To investigate experiences on commercial farms, I interviewed 24 senior farm workers, again focusing on Mashonaland, where evidence from Alexander, Reeler and Lebas indicated that conflict predominated; I did not select interviews in these areas because of any convenient proximity to Harare. Senior farm workers presented a major victim group on farms due to their close association with the farmer. Because of their senior, generally respected, positions overseeing other labourers, they would likely be privy to events occurring in the compound, making them knowledge sources and thus important respondents for this research. Senior workers across a range of professions were located, but all worker respondents were black, since whites rarely worked in non-managerial positions.
Worker Interview Demographics (n = 24)
No. of Respondents
% of Sample
Identifying, locating, and securing interviews with former workers was challenging: few farms were still operating during the fieldwork period, and the environment of severe violence against these workers made approaching unknown members of the public in search of workers risky for all concerned and unlikely to yield willing respondents. Snowball sampling with farmer gatekeepers and privileged “insiders,” such as prominent members of the farming community, who identified possible worker respondent candidates, addressed this challenge. This method is one of the few means of sampling “when a population is widely distributed or elusive,” highly stigmatised, or vulnerable.
Some worker respondents were recommended by farmers, while others were recommended by a security consultant who worked with commercial farmers, a religious worker who operated in the community, a war veteran or other farm workers. As such, there are numerous instances where farm worker respondents were not connected to farmer recommendations at all. Moreover, it should be noted that ‘farmers’ were not a monolithic group anyway (as I detail elsewhere in the book) who would somehow all recommend similar farm worker respondents; indeed, Selby provides nuanced discussion dispelling the myth that farmers comprised a kind of undifferentiated entity. Farmers’ political views, ages, relative wealth and types of crops farmed varied considerably, thus yielding variation in the farmer sample.
In the instances where farmers select worker respondents a selection bias was introduced. Specifically, farmers would recommend worker respondents who were English-speaking, respondents whom they perceived as trustworthy and sufficiently receptive to minimise security risks, and who were at senior employment levels to satisfy the research design. Farmers selected individuals who had experiences with violence, and respondents with no such experiences. Additionally, worker respondents shared some characteristics associated with their job requirements: managers were likely among the most conscientious, competent individuals on the farm, possessed some leadership ability, and had the farmer’s trust, which was not necessarily given to other workers. These individuals were also more likely to identify with the farmers’ perspectives.
“Interviewer effect” was assessed for worker respondents. I approached worker respondents stating my relationship to the farming community but stressed my current residency in the UK. Reactions were mixed. The fact that I was not living in Zimbabwe, was from a UK-based university, and would be taking the data out of the country seemed to reassure respondents. Likewise, my white ethnicity paradoxically contributed positively. There was widespread fear of undercover CIO agents infiltrating commuter buses, bars, and social areas where they can blend in and survey. Respondents considered it very unlikely that, as a white person, I was a CIO agent. I was perceived as a fellow victim because of my ethnicity and farming background who, like the worker respondents, needed to take safety precautions. Similarly, I was apparently foreign and my surname would likely only be recognised by workers local to my family farm’s former location, which appeared to provide security for some worker respondents. However, there is a significant risk that worker respondents will have presented a perspective more favourable to the white farmers due to my status and ethnicity.
Qualitative Interview Sample Two: Commercial Farmers
As of 2000, there were approximately 4,300 farmers; by October 2008 nearly 4,000 of them had been evicted. During this time twelve farmers were murdered, some tortured, and many experienced violence and intimidation.
Commercial farms were the focal point for invaders in the land seizure era. They served as a “nexus” for various larger processes and political aims. For this reason, a higher research focus on commercial farmers was warranted. Fifty-five interviews (50% of all interviews) with farmers (in some cases with their spouses, and in other cases only with spouses) were completed. While participants came from communities nationwide, Mashonaland was overrepresented. This is because violence maps and researchers such as Alexander, LeBas, and Reeler, indicated that violence predominated in Mashonaland. To ascertain possible provincial variation and to gain views from those in ostensibly less violent areas, respondents were also sought from the Eastern Districts, Midlands, and Matabeleland.
Farmers were sought from a variety of cropping and livestock backgrounds, and from the broadest range of political perspectives possible. The sample included 52 farmers that experienced a range of intimidation, theft, vandalism, or arson. Three respondents directly experienced the murder of a family member.
While Zimbabwean commercial agriculture was highly masculinised, male farmers were not the only targets of farm invaders. Interviews demonstrated that women played substantial roles in agricultural operations. Some were homemakers, while others contributed through administrative tasks and specialised oversight such as livestock and horticulture management. Given these key roles and personal connections to male commercial farmers, women were a likely factor in farmers’ and farm invaders’ decision-making and thus probable targets. I also sought interviews with the wives of commercial farmers and from the only Zimbabwean female commercial farmer that I could locate. Criteria for selecting female farmers were the same as for male.
This book aims to understand why the commercial farms were seized by the government beginning in 2000, when the effect of such violent seizure would almost certainly have severe negative socio-economic consequences for the country. Almost all commercial farmers in 2000 were white, which means that the views sought for the research ended up being mostly of white male farmers. However, I endeavored to gain a range of perspectives. The ethnicity and gender breakdown of the commercial farmer sample was driven by the ethnicity and gender of the overall commercial farmer population in Zimbabwe as of 2000.
Definitive nationwide statistics on the ethnic and gender composition of commercial farmers do not exist, but insight can be gained from data provided by the Commercial Farmers Union (CFU), the largest and dominant commercial farming advocacy group. According to CFU data, black commercial farmers made up 9.9% of the CFU database in 1999, and black commercial farmer respondents made up 2% of the interview sample., The sample in this research is admittedly less than the CFU data, but locating black commercial farmers during the fieldwork proved to be exceptionally challenging in the politically charged context.
The perspectives of white female commercial farmers were also sought. According to Neil Wright, a CFU representative, it is not possible to provide a definitive figure on the number of female commercial farmers in Zimbabwe as of 2000. This is because of multiple and often overlapping identities held by women in the commercial farming sector. Some women were commercial farmers while others played discrete roles on the farm such as being in charge of the dairy portion of an otherwise diversified commercial farm to which they sometimes had title. In some instances women were directors or partners, or had inherited farms from a deceased husband.
A variety of women involved in the commercial farming sector were interviewed in order to gain a broad range of perspectives from this stakeholder group. Females involved in commercial farming constituted 27% of the commercial farming interview sample. Stakeholders included a female commercial farmer, women who performed specific farming roles such as being in charge of accounting and cattle, and widows of commercial farmers.
As has been noted, this research focused on large-scale, commercial farmers. The reason for this decision is that commercial farmers – not small-scale farmers – were a primary target of the government during the farm takeovers. This has been strongly demonstrated in existing literature published prior to the commencement of the research for this book. Commercial farmers were targeted by the state because they were predominantly white owned (i.e., because whites historically opposed the political platform of the ruling ZANU-PF party). These farms were also specifically sought by the government because they were relatively wealthy commercial operations (i.e., they contained moveable assets sought by senior government officials). By contrast, small-scale and relatively asset-poor farms were not owned by opponents of ZANU-PF (they were mostly owned by black ZANU-PF supporters) and did not contain the moveable assets sought by government officials. For this reason, the focus in the research on commercial farmers is the correct population for an investigation into the causes and mechanisms of the land seizure era.
The table below provides an overview of commercial farmer respondents and those with insights into the farming community, such as the wives of farmers.
Farmer Interview Demographics (n=55)
No. or Respondents
% of Sample
|Occupation||Male Commercial Farmer||38||69|
|Female Commercial Farmer||1||2|
|Female Spouse of Farmer||13||24|
|Spouse of Security Consultant||1||2|
Qualitative Interview Sample Three: Farm Invasion Group
Multifarious individuals were involved in the design and execution of the land seizure era. Opportunists, and decision makers from all branches of government were vying for assets and power. ‘Farm invasion group’ therefore covers a professionally diverse collection of those directly involved in the seizures of commercial and those familiar with their methods. Individuals were sought from senior legislative positions, state security (military, police, CIO), war veterans (genuine and youth imposters seeking veteran advantages), government workers, “new” farmers, and members of ZANU-PF.
All but two respondents in this group were black (n=29), which reflected the high numbers of black people in senior government positions (the subsequent table provides demographic details). Interviews were also conducted with individuals possessing regional or national knowledge of events. For the same reasons as farmers and workers, respondent distribution sought views from respondents nationwide but focused predominantly on Mashonaland. Given that many farm invaders worked in government with their offices and homes in or around Harare, this decision was especially appropriate.
The farm invader respondent selection processes relied on theoretical sampling in order to, “maximize opportunities to compare events, incidents, or happenings to determine how a category varies in terms of its properties and dimensions.” Moreover, given that this sample included individuals undertaking secretive, “hidden” and potentially illegal actions, and that the full parameters of the population are therefore unknown, more standardised forms of sampling (e.g., random sampling) were unfeasible. In this circumstance, it is accepted practice for researchers to explicitly acknowledge the positions from which they are seeking out participants and the concomitant limitations.
Farm Invasion Group Sample Demographics (n=31)
Race, Gender and Ethnicity
No. of Respondents
% of Sample
Ethnicity of Black Respondents
I conducted the challenging task of identifying, approaching and gathering data from farm invader respondents in three different ways. First, I selected well known public figures, such as politicians, or those with prominent positions, such as senior members of the police, who could likely provide government-level explanations and insights. As public figures, most of these individuals had well-known offices or places of business where they were approached directly. Second, I selected war veterans because they could provide the perspectives of land invaders. I identified and approached war veteran respondents through snowball sampling. In some cases farmer and worker respondents suggested potential farm invader respondents, with some farmers identifying invaders who had seized their farms. Third, I specifically sought respondents who were current or former members of the state security services who could provide “hidden” perspectives of government decision-making, such as stated versus actual objectives and covert operations against political opposition during the land seizure era. These individuals were typically those working or formerly employed in military intelligence and the CIO.
I approached these individuals using snowball sampling from farmers, workers and farm invader respondents. Two out of the 31 respondents in this sample were recommended by farmers. The rest of the respondents were located from recommendations made by war veterans, politicians, and members of the state security apparatus amongst others. In other cases I approached individuals from the government-connected sample without any recommendations, for example, by locating publicly known figures at their official offices and places of business.
The individuals selected for this sample were chosen in order to present a broad cross-section of views on events during the land seizure era. Many farm invaders were self-proclaimed ZANU-PF supporters but were from varied professional backgrounds. However, I also gained perspectives from MDC officials. I also sought interview respondents who would shed light on the research questions from different angles, enable comparisons, and expose different aspects of the research problem.
Interviewer effects were strongest in the government-perspective sample of respondents. First, my identity influenced my perceived credibility in both positive and negative ways. As a white person with a British accent and ties to the commercial farming community, I possessed many attributes that could make some respondents in this sample more suspicious of my intentions with the interview data. These characteristics and my age, which was over thirty and therefore higher than some might expect for a student, also raised suspicions that I may be working for the news media. At least one potential respondent who eventually declined the interview after informing the CIO of my identify and location. In another instance, a respondent who worked in military intelligence believed that I was a member of the Mossad, Israel’s intelligence service.
Second, my white ethnicity likely biased respondents towards expressing less critical views about whites in Zimbabwe. Lee defines the “militancy scale,” by which black respondents were less likely to express hostile attitudes about white people when the interviewer was white.,
Assessing the Veracity of Interview Accounts
I designed the interview sample in order to allow for the triangulation of workers, farmers and those with insight into government decision-making. The inclusion of data from members of the government, such as ZANU-PF officials and members of the state security apparatus who in many instances opposed workers’ interests and would normally oppose farmers’ interests, exemplifies triangulation.
Given that not all workers were recommended by farmers, and no government-connected respondents were recommended by farmers, there is no reason to assume a priori that interview accounts from workers, farmers and respondents giving insights into the government would align. Indeed, worker respondents were often highly critical of farmers – which I document in book – suggesting they were communicating their genuine beliefs independent of farmer interests. In addition, government-connected respondents were often fiercely critical of farmers, further suggesting a clear separation from farmer interests.
If claims from famers, workers and government-connected respondents align – which they largely do – then it is far more likely that this has happened because all three interview samples were discussing actual events that they all saw play out from different perspectives, rather than this being attributable to methodological shortcomings. That is, the congruence in interview accounts between disparate samples is evidence of the strength of the research and associated data.
In addition, the findings from interviews were triangulated against the mapping data (92 maps depicting 21,491 incidents of violence over a nine year period) and the large-scale survey of 34% of the commercial farming community. Once again, claims made by interview respondents tended to closely align with the survey data – and vice versa – as well as dovetail with the mapping data set, once again suggesting that accounts were aligned with actual events.
Interviews undertaken with state security officials, members of the military, and Mugabe’s legal team were completed outside Zimbabwe. These respondents provided some of the most unique insights into often secret conversations and hidden decision-making in the land seizure era. I strongly believe that one reason why this research was able to elicit such detailed, insightful and often incriminating evidence was that by undertaking interviews outside Zimbabwe these respondents were far less concerned about repercussions from Zimbabwe security officials from their testimony. In addition, as with all respondents I provided robust assurances about protecting their identities, and my claims were likely more plausible because I too was not in Zimbabwe during the interviews. I believe I was therefore able to gain more genuine views that were not constrained by respondent concerns over repercussions from Zimbabwe’s state security services. This belief that the research captured a broad range of genuine views is made more plausible because while some individuals indicated support for the opposition MDC party and for gradual land reform, others stated strong support for ZANU-PF, Robert Mugabe and for large-scale farm seizures.
In addition, I strongly believe that my university affiliation contribution to more open interviews. Several respondents asked if I was affiliated with any news media outlet, and insisted that they did not trust, and would not conduct interviews, with journalists whom, they believed, operated with overt biases. Given that I could demonstrate a university affiliation and provided a detailed overview of how respondents’ contributions would form the basis of what, at the time was my doctorate being undertaken at the University of Oxford, accounts tended to be more detailed and forthcoming.
Finally, also provided evidence of steps to ensure anonymity. Prior to interviews I asked respondents not to use their names, farm names or any other identifying information. I also provided evidence to interview respondents about how I encrypted each interview using the same level of encryption relied upon by financial institutions to protect customer data during transaction. Thus, by respondents not providing identifiable information and all recorded data being immediately encrypted, interviews undertaken in Zimbabwe were robustly protected. Similarly, I sued the same level of encryption and anonymity for the survey of farmers. Survey and interview respondents both related to me that they recognised my effort to rely upon technology to protect their data and that it contributed positively in their decision to participate in this research.
Informed Consent and Protection of Subjects
I recorded 97% of interviews; one individual asked not to be recorded and two others could not be recorded due to security considerations. Each interview was openly recorded with the respondent’s permission. Individuals wishing to remain anonymous were asked to omit identifying information while being recorded. After each interview, the digital data were encrypted on a laptop for security reasons.
Anonymous interview respondents cited in this project were always referenced by their anonymised pseudonym. For individuals providing data such as personal letters or photographs who were not interviewed, names were used only with permission.
The 110 in-depth qualitative interviews resulted in extensive output of recorded data. The 107 of the 110 interviews that were recorded yielded 138 hours of recorded interviews. When transcribed, these interviews totalled 2,317 pages of single-space, 12 point font typed transcripts. Interviews averaged 1 hour and 29 minutes in length, with the shortest interview lasting 21 minutes and the longest interview lasting 3 hours and 13 minutes.
Study 3: Quantitative Survey of Commercial Farmers
After completing the pilot fieldwork interviews, I determined that the qualitative data would be further supported by quantitative survey. Farmer and worker testimonials provided rich, personal narratives with personal and contextual detail, yet they routinely could not answer questions about possible relationships between phenomena on their own farms. Experiences among properties clustered in a similar area varied considerably, and often in a counter-intuitive manner. For instance, it was unclear what factors determined which farms were targeted. In the Mazowe, Mvurwi, and Bindura triangle, most farms experienced violence, but some were attacked with exceptional persistence and aggression. Testimonials related this but could not explain why individuals who were known within their local community for their altruism and service projects had been severely attacked, and yet a notoriously racist farmer located a few miles away experienced little trouble. It was these kinds of questions that the survey sought to better understand.
A survey of farmers was designed and undertaken in order to systematically gather demographic, farm and violence data. Maxim notes that mail surveys are “often most successful with small, clearly defined samples (such as professionals), where the issues of jargon are salient.” Given my familiarity with the region and community, I felt confident in my awareness of these critical jargon issues, making survey use unproblematic. Contact details were sought for all farmers residing on their land at any point from January 2000 until December 2008, regardless of cropping type, provincial location, kinds and frequency of violence experienced, and current residence (including emigrants), targeting a total population of 4,300.
Establishing a Survey Method
The survey, conducted by post and internet, targeted all possible farmer respondents. Early fieldwork illustrated the difficulties associated with contacting a highly transitory group, in which even strong social ties are occasionally broken. With widespread diaspora, an email/internet approach was the only means of reaching some respondents. In order to avoid possible selection bias with an online survey that may favour younger, technical, and possibly more affluent people, printed questionnaires were sent by post to those without internet. Both methods, as compared to telephone or in-person surveys, allowed respondents “the time and opportunity to consider the issues involved and to collect pertinent data,” an important consideration given the depth of survey questions.
Compiling a Contact Database
Compiling contact details for respondents was a formidable task. As of September 2008, approximately 300 of the estimated 4,300 farmers remained on their properties. Most had experienced multiple moves since leaving their farms, and many had emigrated. The personal and political nature of the research also increased the risk to participants and thus made it less likely that contact details would be given by gatekeepers and friends, and also less likely that respondents would participate if contacted. Even for those remaining on their farms, their vulnerable position made it likely that they would not respond favourably to requests for their contact details.
To overcome these challenges, I enlisted the support of an individual connected to the farming community through a farmer support organisation who provided me with contact details for some farmers. Additionally, I contacted the Commercial Farmers Union (CFU) in order to gain access to their farmer database; the CFU is Zimbabwe’s largest commercial farmers union and the de-facto body representing the interests of most commercial farmers. Due to confidentiality policies the CFU stated that they could not provide me with their contact list, but the organisation agreed to send their members the survey email with a link to the questionnaire website, thus allowing respondents to participate online.
Sampling and Administration
The farmer support organisation provided 1,017 email and 421 postal addresses, which formed the initial database. A preliminary email to these individuals announcing the upcoming survey and asking for additional respondent names resulted in 247 undeliverable emails, 24.3% of the online database, reducing the online sample to 770 possible respondents; the number of undeliverable emails suggested the high level of transition in the farming community at the time. A similar measure of determining whether postal addresses in the database were accurate was not possible, but was presumed to be higher due to portability of email addresses. An appeal to the initial contact database and CFU for more respondents resulted in 151 new email and 100 new postal contacts. The final sample was 921 email and 521 postal contacts (the CFU sample is unknown) for a total of at least 1,442 respondents sampled.
Both the online and postal versions of the 77-question survey were identical except for minor, non-substantive variations necessary to accommodate online dissemination. Survey questions tended to be long and open-ended, a strategy that complicated response processing but tended to illicit more frank responses regarding threatening topics. Especially when investigating “hidden” topics that were largely un-researched, such as possible counter-intuitive farmer responses to state violence, this method allowed respondents to answer in their own words and to reveal unknown and unpredicted aspects of violence.
I downloaded the online survey responses and manually coded the printed and online responses. I deleted duplicated or incomplete responses. After a three week incubation period I repeated the process in a separate spreadsheet. I then compared the accuracy of the two passes and where there was a discrepancy I located the original ‘raw’ data and made a final determination until both spreadsheets were identical. Inter-rater reliability did not apply to this method because I was the only person undertaking data entry. A single coder, deeply familiar with the topic, interpreting codes on two different passes with an appropriate incubation period was considered sufficient and acceptable.
By keeping the online and printed samples separated in the data gathering stage, I could compare relevant statistics to determine possible bias or other variation in the samples. However, merging these samples facilitated data analysis. Before merging, I assessed similarity among the samples by comparing the key demographics of age, country of residence, and farm location. Upon examination, merging was deemed appropriate since the samples were relatively small and no troubling demographic differences existed.
Sample Envelopes from Survey Respondents
A key consideration when undertaking this doctoral research was the difficulty in obtaining research data. Approaching respondents about personal data and sensitive experiences is fraught with difficulty. These problems were compounded in the LSE where state surveillance posed threats to both the researcher and respondents participating in research on state-sponsored political violence and asset expropriation. Moreover, given Zimbabwe’s extremely serious economic deterioration during the mid to late 2000’s, where eventually the country experienced the second highest inflation in recorded worldwide history, sending respondents an unsolicited survey where the respondents had to pay the return postage costs, further mitigated against broad participation (despite extensive efforts to do so, the researcher was ultimately unable to send payment internationally, especially when the cost of stamps in Zimbabwe changed constantly).
The cost and inconvenience of returning an unsolicited questionnaire is shown in the following two images. In the first example the respondent’s envelope required 32 stamps. In the second example the respondent paid ZWD$270,000 (January 2008 value). These examples reinforce the extensive difficulty of undertaking research in conditions present in Zimbabwe during the LSE. Without the generosity of respondents evidenced in these examples, this doctoral research would not have been possible.
For more details on the research methodology and findings from this doctoral research, see other data on this website and click here for information on the forthcoming peer-reviewed book.
Source of Images: Charles Laurie © 2015
As discussed, there was significant risk for respondents in participating, and as well as significant difficulties in locating the population of current or former commercial farmers. Due to these constraints, probability sampling was not possible. Instead, a non-probability strategy examined selection bias in the respondents, determined a response rate for the survey, and enabled an undertaking of sample quality assessment.
The contact list provided by the farmer support organisation was national and included all farms and crop types. There was no political, racial, or gender affiliation associated with the sample, although it is noted that almost all farmers were white males. Since most farmers suffered some kind of economic loss, and the organisation specialised in ascertaining farmer financial losses, this primary database offered a sample with no obvious biases politically or demographically. Those subsequently referred through the CFU or from snowball sampling of the compensation database were entirely combined with the database to form a single list, and within this list, all individuals were approached, ensuring no additional selection bias.
It was not possible to determine an exact response rate because the number of questionnaire recipients in sample B was unknown. I estimated a response rate of 30.5% by averaging the response rates samples “A” and “C” only.
 Zimbabwe Human Rights NGO Forum Representative, Email, 20 June 2009; Email, 8 May 2009.
 Zimbabwe Human Rights NGO Forum, “Homepage,” http://www.hrforumzim.org/, http://www.hrforumzim.org/.
 Human Rights Watch, “Fast Track Land Reform in Zimbabwe,” (UNHCR, 2002).; Crisis Group, “Zimbabwe: Time for International Action,” (Harare/Brussels: Crisis Group, 2001).
 D. L. Morgan, “Practical Strategies for Combining Qualitative and Quantitative Methods: Applications to Health Research,” Qualitative Health Research 8, no. 3 (1998); Integrating Qualitative and Quantitative Methods: A Pragmatic Approach (Sage, In Press).
 Zimbabwe Human Rights NGO Forum Representative.
 Gaye Tuchman, “Historical Social Science: Methodologies, Methods, and Meanings,” in Strategies of Qualitative Inquiry, ed. Norman K. Denzin and Yvonna S. Lincoln (Thousand Oaks: Sage Publications, 1998), 250-52.
 Zimbabwe Human Rights NGO Forum, “Report on Political Violence in Bulawayo, Harare, Manicaland, Mashonaland West, Masvingo, Matabeleland North, Matabeleland South and Midlands,” (Harare: Zimbabwe Human Rights NGO Forum, 2000), 3.
 Adrienne LeBas, “Polarization as Craft: Party Formation and State Violence in Zimbabwe,” Comparative Politics 38, no. 4 (2006): 428; Stathis N. Kalyvas, The Logic of Violence in Civil War, ed. Margaret Levi, Cambridge Studies in Comparative Politics (New York: Cambridge University Press, 2006), 51.
 BBC, “Mugabe Seeks Media Monopoly,” http://news.bbc.co.uk/2/hi/africa/1721399.stm.
 Kalyvas, The Logic of Violence in Civil War, 39.
 Zimbabwe Human Rights NGO Forum Representative.
 US Department of State, “2004 Country Reports on Human Rights Practices,” (US Department of State, Bureau of Democracy, Human Rights, and Labor, 2005).
 Human Rights Watch, “Fast Track Land Reform in Zimbabwe.”; Crisis Group, “Zimbabwe: Time for International Action.”
 2,849 media sourced reports were coded but not included in tabulations.
 Zimbabwe Human Rights NGO Forum, “Political Violence Report, December 2001,” (Harare: Zimbabwe Human Rights NGO Forum, 2001), 6.
 Sarah Kenyon Lischer, “Refugee Involvement in Political Violence: Quantitative Evidence from 1987-1998,” in New Issues in Refugee Research (Cambridge, Massachusetts: Center for International Studies, Massachusetts Institute of Technology, 2000).
 Charles Tilly, Politics of Collective Violence (Cambridge: Cambridge University Press, 2003), 19.
 E.g., Zimbabwe Human Rights NGO Forum, “Who Is Responsible? A Preliminary Analysis of Pre-Election Violence in Zimbabwe,” (Harare: Zimbabwe Human Rights NGO Forum, 2000), 3, 6.
 Solidarity Peace Trust, “Desperately Seeking Sanity: What Prospects for a New Beginning in Zimbabwe?,” (Port Shepstone (South Africa): Solidarity Peace Trust, 2008), 25.
 UNDP, “Comprehensive Economic Recovery in Zimbabwe,” (United Nations Development Programme, 2008), 157-58; Amnesty International, “Time for Accountability,” (London: Amnesty International, 2008), 18; Zimbabwe Human Rights NGO Forum and JAG, “Adding Insult to Injury,” (Harare: Zimbabwe Human Rights NGO Forum and the Justice for Agriculture Trust in Zimbabwe, 2007), 26-27.
 UN, “Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment,” United Nations, http://untreaty.un.org/english/treatyevent2001/pdf/07e.pdf.
 I interviewed 41 farmers and 20 farm workers during the first fieldwork trip, which took place from July to September 2006. I interviewed 11 farmers, 4 farm workers, and 19 farm invaders on the second fieldwork trip, from December 2006 to April 2007. I undertook two additional interviews of farmer respondents and 12 of farm invaders in Europe after the two major fieldwork trips.
 Jocelyn Alexander, Unsettled Land: State-Making and the Politics of Land in Zimbabwe 1893-2003 (Oxford: James Currey, 2006), 186, 88.; A. P. Reeler, “Role of Militia Groups in Maintaining Zanu Pf’s Political Power,” http://www.kubatana.net/docs/hr/reeler_militia_mar_030331.pdf.; LeBas, “Polarization as Craft: Party Formation and State Violence in Zimbabwe,” 428.
 Zimbabwe Human Rights NGO Forum and JAG, “Adding Insult to Injury.”; Zimbabwe Human Rights NGO Forum, “Political Violence Report, October 2001,” (Harare: Zimbabwe Human Rights NGO Forum, 2001), 8; “Who Was Responsible? Alleged Perpetrators and Their Crimes During the 2000 Parliamentary Election Period,” (Harare: Zimbabwe Human Rights NGO Forum, 2001), 12; “Political Violence Report, 16-31 March 2002,” (Harare: Zimbabwe Human Rights NGO Forum, 2002), 25; “Political Violence Report, November 2002,” (Harare: Zimbabwe Human Rights NGO Forum, 2002), 10; “Political Violence Report, May 2005,” (Harare: Zimbabwe Human Rights NGO Forum, 2005), 11.
 John W. Creswell, Qualitative Inquiry and Research Design (Thousand Oaks: Sage, 1998), 117-18.
 Tim May, Social Research: Issues, Methods and Process, Third ed. (Maidenhead: Open University Press, 2001), 95-96.; Raymond M. Lee, Doing Research on Sensitive Topics (London: Sage, 1993), 67.
 Angus Selby, “Commercial Farmers and the State: Interest Group Politics and Land Reform in Zimbabwe” (University of Oxford, 2006).
 Zimbabweans commonly speak English across all education levels, therefore English was not a proxy for education.
 Alexander, Unsettled Land: State-Making and the Politics of Land in Zimbabwe 1893-2003, 186, 88.; LeBas, “Polarization as Craft: Party Formation and State Violence in Zimbabwe,” 428.; Reeler, “Role of Militia Groups in Maintaining Zanu Pf’s Political Power.”
 This is a conservative estimate that does not account for black farmers from missions, churches, or from those farms managed by government departments, municipalities, or members of the military.
 Commercial Farmers Union, “Cfu Licence Database Information,” (Harare1999).
 Neil Wright, 12 September 2014.
 Anselm Strauss and Juliet Corbin, Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory, 2nd ed. (Thousand Oaks: Sage Publications, 1998), 202.
 Lee, Doing Research on Sensitive Topics, 100.
 Although respondents never specifically addressed this connection, it was tangentially referenced.
 Timothy Ngavi, Interview with author, 25 January 2009; Agnes Sengwe, Interview with author, 23 March 2009; Samuel Limpopo, Interview with author, 17 October 2008; Edison Bubye, Interview, 24 July 2008; Hastings Lundi, Interview with author, 9 April 2009; Christian Mucheke, Interview with author, 17 May 2009; Edward Chidembo, Interview, 26 October 2008.
 The survey cover letter asked for anyone who was a “commercial farmer at any point from January 2000 until the present day [March 2008].” Prospective respondents asked whether farm managers and leaseholders could participate. Since violence primarily targeted removal of “farm owners,” meaning individuals who held legal title to the farm and whose presence prevented the seizure of land and assets, the researcher excluded the former and include the latter from the definition of “commercial farmer” and extended the final date through to December 2008. It must be noted, however, that many white managers suffered extensive violence during the land seizure era.
 Paul S. Maxim, Quantitative Research Methods in the Social Sciences (New York: Oxford University Press, 1999), 283.
 Telephone surveys were impractical given Zimbabwe’s expensive and unreliable system, the possibility of surveillance.
 Maxim, Quantitative Research Methods in the Social Sciences, 282-83.
 Of the 521 postal surveys sent, only one undeliverable letter was returned. It was postmarked in Chipinge six months after posting. Given that most farmers had been evicted by 2008, there were probably many other inactive post boxes. The lack of returned mail indicated the breakdown of the Zimbabwe postal system, and the associated difficulty of evaluating the percentage of targets reached.
 Lee, Doing Research on Sensitive Topics, 76-77.
Alexander, Jocelyn. Unsettled Land: State-Making and the Politics of Land in Zimbabwe 1893-2003. Oxford: James Currey, 2006.
Amnesty International. “Time for Accountability.” London: Amnesty International, 2008.
BBC. “Mugabe Seeks Media Monopoly.” http://news.bbc.co.uk/2/hi/africa/1721399.stm.
Bubye, Edison. Interview, 24 July 2008.
Chidembo, Edward. Interview, 26 October 2008.
Commercial Farmers Union. “Cfu Licence Database Information.” Harare, 1999.
Creswell, John W. Qualitative Inquiry and Research Design. Thousand Oaks: Sage, 1998.
Crisis Group. “Zimbabwe: Time for International Action.” Harare/Brussels: Crisis Group, 2001.
Human Rights Watch. “Fast Track Land Reform in Zimbabwe.” UNHCR, 2002.
Kalyvas, Stathis N. The Logic of Violence in Civil War. Cambridge Studies in Comparative Politics. edited by Margaret Levi New York: Cambridge University Press, 2006.
LeBas, Adrienne. “Polarization as Craft: Party Formation and State Violence in Zimbabwe.” Comparative Politics 38, no. 4 (2006): 419-38.
Lee, Raymond M. Doing Research on Sensitive Topics. London: Sage, 1993.
Limpopo, Samuel. Interview with author, 17 October 2008.
Lischer, Sarah Kenyon. “Refugee Involvement in Political Violence: Quantitative Evidence from 1987-1998.” In New Issues in Refugee Research. Cambridge, Massachusetts: Center for International Studies, Massachusetts Institute of Technology, 2000.
Lundi, Hastings. Interview with author, 9 April 2009.
Maxim, Paul S. Quantitative Research Methods in the Social Sciences. New York: Oxford University Press, 1999.
May, Tim. Social Research: Issues, Methods and Process. Third ed. Maidenhead: Open University Press, 2001.
Morgan, D. L. Integrating Qualitative and Quantitative Methods: A Pragmatic Approach. Sage, In Press.
———. “Practical Strategies for Combining Qualitative and Quantitative Methods: Applications to Health Research.” Qualitative Health Research 8, no. 3 (1998): 362-76.
Mucheke, Christian. Interview with author, 17 May 2009.
Ngavi, Timothy. Interview with author, 25 January 2009.
Reeler, A. P. “Role of Militia Groups in Maintaining Zanu Pf’s Political Power.” http://www.kubatana.net/docs/hr/reeler_militia_mar_030331.pdf.
Selby, Angus. “Commercial Farmers and the State: Interest Group Politics and Land Reform in Zimbabwe.” University of Oxford, 2006.
Sengwe, Agnes. Interview with author, 23 March 2009.
Solidarity Peace Trust. “Desperately Seeking Sanity: What Prospects for a New Beginning in Zimbabwe?”. Port Shepstone (South Africa): Solidarity Peace Trust, 2008.
Strauss, Anselm, and Juliet Corbin. Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory. 2nd ed. Thousand Oaks: Sage Publications, 1998.
Tilly, Charles. Politics of Collective Violence. Cambridge: Cambridge University Press, 2003.
Tuchman, Gaye. “Historical Social Science: Methodologies, Methods, and Meanings.” In Strategies of Qualitative Inquiry, edited by Norman K. Denzin and Yvonna S. Lincoln. Thousand Oaks: Sage Publications, 1998.
- “Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment.” United Nations, http://untreaty.un.org/english/treatyevent2001/pdf/07e.pdf.
UNDP. “Comprehensive Economic Recovery in Zimbabwe.” United Nations Development Programme, 2008.
US Department of State. “2004 Country Reports on Human Rights Practices.” US Department of State, Bureau of Democracy, Human Rights, and Labor, 2005.
Wright, Neil. 12 September 2014.
———. “Political Violence Report, 16-31 March 2002.” Harare: Zimbabwe Human Rights NGO Forum, 2002.
———. “Political Violence Report, December 2001.” Harare: Zimbabwe Human Rights NGO Forum, 2001.
———. “Political Violence Report, May 2005.” Harare: Zimbabwe Human Rights NGO Forum, 2005.
———. “Political Violence Report, November 2002.” Harare: Zimbabwe Human Rights NGO Forum, 2002.
———. “Political Violence Report, October 2001.” Harare: Zimbabwe Human Rights NGO Forum, 2001.
———. “Report on Political Violence in Bulawayo, Harare, Manicaland, Mashonaland West, Masvingo, Matabeleland North, Matabeleland South and Midlands.” Harare: Zimbabwe Human Rights NGO Forum, 2000.
———. “Who Is Responsible? A Preliminary Analysis of Pre-Election Violence in Zimbabwe.” Harare: Zimbabwe Human Rights NGO Forum, 2000.
———. “Who Was Responsible? Alleged Perpetrators and Their Crimes During the 2000 Parliamentary Election Period.” Harare: Zimbabwe Human Rights NGO Forum, 2001.
Zimbabwe Human Rights NGO Forum and JAG. “Adding Insult to Injury.” Harare: Zimbabwe Human Rights NGO Forum and the Justice for Agriculture Trust in Zimbabwe, 2007.
Zimbabwe Human Rights NGO Forum Representative. Email, 20 June 2009.
———. Email, 8 May 2009.