6. Writing It Up

In 2017, scientists in 600 cities undertook the first ever March for Science, believing that scientific ‘values are currently at risk’, and that ‘(w)hen science is threatened, so is the society that scientists uphold and protect’ (Principles and Goals, March for Science 2017). Actually, it is not clear that there is a general decline of trust in science (see, for instance, Pew Center 2017). Nonetheless, social scientists, like others, have felt under increased pressure to explain and justify why they deserve public funding and how their methods hold up to scrutiny. In political science, one response to this has been a change in the Ethics Guide of the American Political Science Association in 2012, reflecting the so-called Data Access and Research Transparency (DA-RT) principles. Subsequently, a number of leading political science journals have adopted a Journal Editors’ Transparency Statement (JETS) (https://​www.​dartstatement.​org), which constitutes an operationalization of the DA-RT principles form the perspective of journal editors. DA-RT states that ‘researchers should provide access to … data or explain why they cannot’, and JETS operationalizes this by committing journal editors to ‘(r)equire authors to ensure that cited data are available at the time of publication through a trusted digital repository’. While journal editors would be at liberty to grant exemptions, the new standard intended by JETS is full publication of raw data.

From late 2014, these statements became subject to increasing controversy, and a petition signed by many leading political scientists requested a delay in the implementation of DA-RT and more specifically JETS. The ethical and epistemological implications of these statements for various types of qualitative research, they argued, had not been sufficiently thought through. A lively debate on the implications of DA-RT has since ensued especially among primarily US-based political scientists (see https://​dialogueondart.​org and https://​www.​qualtd.​net).

Whereas US debates on transparency have been motivated by concerns about scientific legitimacy and reliability probes, recent European initiatives promoting ‘open science’ have been more government-driven, arguing that data-sharing accelerates innovation, and could give Europe a competitive edge. In practical terms, they have focused on developing institutional digital storing and archiving capabilities rather than on changing editorial practices of journals (Directorate-General, Research & Innovation 2016). In a recent position paper, Germany and the Netherlands proposed the fast-track development of a ‘European Open Science Cloud’ (EOSC), which is to be ‘a trusted, open environment for European researchers for the handling of all phases of the data life cycle and generated results’. The principle underlying the cloud is ‘to make research data findable, accessible, interoperable and re-usable (FAIR)’ (Joint Position Paper 2017, 1). What these European policy plans have in common with the US initiatives is the sense of urgency and universal applicability with which its proponents contend that all data should become open to all, as soon as possible.

Some scholars have already published their reflections, particularly in response to DA-RT, on tensions between transparency obligations and protection of respondents in specific authoritarian contexts (Driscoll 2015; Shih 2015; Lynch 2016). However, these comments tend to focus more on why DA-RT and JETS are problematic than on what should be considered best practice. And unlike DA-RT and JETS, the European ‘open science’ initiatives have yet to generate extensive debate within the political science profession. Hence, we find more extensive reflection, and making recommendations, on standards relating to writing up and publishing research-based authoritarian fieldwork desirable.

While, as we have reflected in previous chapters, the sources we collect in the field are much broader than interviews, we will focus the discussion on interview practices, because this is the area where the tension between transparency and the ‘do no harm’ imperative is most evident. When we conduct interviews, we always begin with a little opening speech explaining who we are and what kind of research we are doing and explaining that we will transcribe the interview, the transcript stays with us, but we may quote from it in academic publications. We always deal with the matter of informed consent orally. None of us has ever used informed consent forms. To our knowledge, they are not customary in authoritarianism research. They can cause distrust among respondents (‘why do I need to sign something?’), as well as bringing about a potentially risky paper trail during fieldwork. At this point, our practices diverge, depending on the type of respondent. We discern roughly three categories of respondents. The first type, ‘ordinary people’, we typically inform that we intend to anonymize the transcript of the interview and not use their real name if we should cite them. With the second type, ‘expert informants’, we typically have an exchange about whether and if so how they would like to be anonymized. The third type are ‘spokespersons’, whom we typically ask for their permission to be cited by name.

The default option of anonymity we find most appropriate when we interview certain categories of ‘ordinary people’. Thus, our Kazakhstan researcher has used this when interviewing young Kazakhs who had studied or were studying abroad. Our Malaysia researcher used it when interviewing people about their decision-making as to whether to join a demonstration. We find the default option of anonymity appropriate in these cases for three reasons. First, as ‘ordinary citizens’, these people have typically not chosen to be professionally engaged with politics, they are usually not accustomed to being interviewed and cited, and they thus deserve a high level of protection of their privacy. Secondly, while some respondents might simply refuse to speak to us if we were to cite them by name, others might well agree to be interviewed, but we believe that the validity of their answers to our questions might suffer if they knew they could be quoted by name. This is an issue that is not unique to authoritarianism research, it would apply to interviews with vulnerable groups (i.e. victims of sexual abuse, undocumented migrants, or drug users) in democratic societies as well. In an authoritarian context, all ordinary citizens are in the ‘vulnerable’ category when we ask them questions that relate to their views of their government or dissident behavior. This is not to suggest that they would be in immediate fear of arrest or worse. In both cases we mentioned here, Kazakhstani students and Malaysian potential demonstrators, respondent concerns related more to their professional environment, their relation to their university, or even their family, all of which might disapprove of dissident views or behavior. Finally, we think anonymity is not problematic in this particular category because the respondents do not have access to unique information as individuals. These kinds of interviews are more akin to surveys in that sense. What makes them different from surveys is that we do not claim that the people we speak to are representative for a broader group, and we do not attempt to quantify their opinions or experiences. Instead of reliability, it is validity we are after, trying to reconstruct and reflect their thought processes in relation to at least somewhat politically sensitive issues. In these instances, our sampling method may be questioned, but not the anonymity as such.

Most of the interviews we conduct in authoritarian settings fall into a different category, one that leads us to give respondents a choice when it comes to anonymity. We find this appropriate when interviewing lower-level government or party officials, corporate executives, journalists, local academics, opposition politicians, and activists. The reason we want to interview these people is usually that they can give us some insight into how the authoritarian system works in practice: within the bureaucracy or the party, in its dealings with other politically relevant actors, or in its dealings with critics. Revealing such information can make them vulnerable, although this does not always need to be the case. By anonymizing them (in such a way that they are genuinely unrecognizable, see below) we exclude any such risk. But it requires us to relax the ideal of complete transparency in how we come by our findings. If we cannot tell readers who we spoke to, they cannot trace whether we quoted and interpreted these sources accurately. We will reflect on this trade-off in more detail below.

Our experiences with the degree to which this second, broad category of respondents take us up on our offer of anonymity is very varied, depending on the repressiveness of the regime, the type of respondents, and the nature of our research question. Our Malaysia researcher found that the activists he interviewed were all but one entirely comfortable with being named. Their status as dissidents and members of a social movement in opposition to the government was well known, and while the answers to questions asked by our researcher were more specific than what they have publicly said online, they were not more incendiary. Our China researcher by contrast finds that both public officials and corporate employees almost always prefer anonymity. She still asks them, but she already knows what the answer will be. Most of us have experienced getting mixed responses in this category. When we have a mixed response, we have a difficult choice. It is clear that we cannot name respondents who have asked to be anonymized, but should we always name the ones who have told us they are comfortable with doing so? From the perspective of transparency, this would be the best option, but we find that there may be three reasons to act otherwise.

The first is homogeneity. In her research on municipal governance, for instance, our Morocco researcher felt that it did not make sense to anonymize a civil servant for one municipality, but not someone in the same position for another municipality. Likewise, our China and Iran researchers, who both interviewed people working for Internet companies who deal with government agencies, did not believe there was value added in providing some of their names but not others. Others in our group however believe that stylistic homogeneity should not be a priority, and every respondent who can be named represents a transparency gain and should therefore be named.

A second consideration is whether we sometimes feel that what we are being told is quite sensitive, and whether perhaps the respondent requires more protection than he or she is asking from us. This was a choice consistently made by Driscoll (2015), who interviewed Tajik or Afghan former militia members in an authoritarian but also volatile environment. He writes that ‘(i)n a few cases, the subject insisted that I record his full name. For my own safety, and that of my respondents, I never complied with these requests’. If we make such a judgment, we are in effect second-guessing our respondent’s own judgment, as well as foregoing transparency. Nonetheless, some of us have occasionally made such a judgment. In the municipal research quoted above, our Morocco researcher came across a respondent who gave her highly sensitive information, really explaining how the administrative system exerted power over elected officials. He did not think he required anonymity, but she believed this was perhaps a little naïve, and he was saying things that would undoubtedly make his boss unhappy, and his boss’s boss, all the way up to the top. So she decided that it was better to be overprotective than sorry and anonymized his statements. Our Mexico researcher has anonymized all journalist-respondents in a journal article, whether they had asked him to or not, but has yet to decide what to do in his dissertation. On the one hand, as discussed in Chap. 5, the context is very repressive, with murder a regular outcome for critical journalists. On the other hand, the journalists in question are already openly critical of the government and do not reveal much to our researcher that they have not said or written before, so the additional risk flowing from his published work may be quite limited. In fact, some activists actually seek—preferably international—visibility, not only to advertise their cause but also because they believe it gives them some level of protection. When a respondent articulates such a strategy, it would not make sense to overrule him or her for their protection.

A final reason to anonymize a respondent without being asked to do so can be the long delays in the publishing process, and a change in the situation on the ground during this process. Information that was not sensitive when provided may become sensitive by the time it is published. One of us, for instance, had the consent of (some) Egyptian activists to be mentioned by their first name in early 2013, but decided because of the subsequent military crackdown, that in subsequent publications they needed to be given aliases for their protection.

The third category of interviews is where we speak to high-profile politicians or civil society figures on their official stance, as our China, Kazakhstan, and Morocco researchers have all done on occasion. We may still put anonymity on the table as an option, especially if we do not know in advance exactly how the interview is going to go, but we do not usually expect them to take it up. These are public figures who are used to media exposure (albeit in the constrained circumstances of their authoritarian context), and in some cases, they may have already spoken of the same topics on public occasions. They will know exactly what they want to say to us, and how to say it. Moreover, their quotes are only meaningful in the context of who they are. What they say to us is interesting not because they give us insight into their thought process, or into the inner workings of the bureaucracy or the political process, but because they are the head of the Islamist Party or the president of the biggest women’s rights association. These kinds of interviewees will often give their consent to be quoted by name in our published work as a matter of course, although they may also give us some off-the-record information at the same time. Even if they only give us their official views, this can be of interest, because they give us insight into authoritarian legitimation strategies. But they typically only give us one face of authoritarian political processes: the public face. For some of our research, that is all we need, but for many other research questions, it is not enough.

Not mentioning a name in our published work is only a small part of our anonymization practices. When we do not use real names, we use different kinds of descriptors instead. When respondents have very similar profiles, sometimes we just number them. Our Mexico researcher, for instance, just refers to journalist 1, journalist 2, and so on. Sometimes, especially when it comes to ‘ordinary people’ interviews such as those done by our Kazakhstan and our Malaysia researcher, we give our respondents aliases, fake first names. This improves readability and makes it possible for a reader to track particular respondents in a publication. When we have interviewed people in the second category, we try to convey some additional information, so that the reader can understand why the respondent in question would have a uniquely relevant perspective on the matter at hand, while still not making them traceable. We might refer to them as ‘senior manager at Baidu, Beijing’, or ‘journalist, target of phishing attempts’. Sometimes, we need to omit more than a name to keep someone’s identity a secret (see Shih 2015, 22 for some very concrete examples on how to be protective while still conveying to a reader why a respondent could be considered as a well-informed source). A Malaysian civil servant who had attended an anti-government demonstration, for instance, insisted that it was not enough to delete his name; any detailed information concerning his workplace could make him traceable, so the researcher used the vague reference ‘works for the government’.

But anonymization is not just about what eventually gets published. When a respondent asks to remain anonymous, we always keep their real names separate from our transcript. The real names may be found in our notes, in an old diary, or in a document kept separate from the transcripts. Contact details are also kept separate from transcripts (see also Shih 2015, 22). We are well aware that none of these practices is completely secure. While we are in the field, we are in possession both of a set of contact details with identifiers (real names or not) and a set of transcripts. If someone were to steal or forcibly seize all our notes, transcripts, and recordings (which is in some contexts much more likely than high-tech electronic surveillance, as graphically depicted in Driscoll 2015, 6), and study them attentively, they would probably be able to trace the respondents. In some cases, the transcripts themselves are actually revealing, as, for instance, in the case of our Mexico researcher, whose journalist-respondents sometimes refer to their own published work. At other times, it would be a matter of putting together the transcripts and our other information. None of us have been in a position where our material was taken from us during fieldwork (although a bag containing fieldwork material was once stolen during transit, see Chap.2), but it can happen. Our protective practices would make it time-consuming and difficult to trace respondents, but we cannot guarantee that it would be impossible.

As Bellin (2016) has pointed out, such risks put us under an ethical obligation to be transparent in quite a different way from that intended by DA-RT or ‘open science’: transparency to our research subjects (see also Loyle 2016), resulting in ‘a negotiation of the level of risk and disclosure that the respondents are comfortable with’ (https://​www.​qualtd.​net, IV.1). While we endorse the spirit of this, we do not think that her proposal that in the writing-up phase ‘the respondents work with the researcher to specify what identifying information can be written about and what should be removed or altered’ is always practicable. Precisely when working with respondents who may be at risk, we cannot assume that our—usually digital—communications with them about such matters would be safe. We may sometimes have to take decisions about altered risk conditions for them, as in the case of Egyptian activists referred to above.

We also come across information in our research that we cannot use at all, not even on condition of anonymity. Sometimes, such information is really of no interest to us, so we just ignore it. Our Malaysia researcher, for instance, found that his activist respondents would sometimes disparage each other off the record, but their internal relations had little to do with his research question. It becomes more difficult when they do give us information that is important, and is either new, or corroborates other evidence. Our India researcher has literally had the experience of a respondent who had given permission for a recorded interview changing his mind and asking for the recording to be erased. When such a request is made, whether we think it is reasonable or not, no visible trace of the information should remain in our published work. But what cannot be asked of us is that we erase the information from our minds. If it is indeed important, it will inform our analysis, and we may look for other sources for the same information. In this case, other interviews confirmed the story the respondent had told. Our Kazakhstan researcher has likewise had relevant information from an opposition source who emphatically asked that what he said be kept confidential, since the information ‘would put him in a situation of risk’. She tried to find written sources to corroborate the factual information he had given and kept the opinions he had expressed in the back of her mind during her analysis. In both of these cases, the off-the-record information was of some use to the researchers, but in both cases, it resulted in the written analysis looking less solid than it actually was, because it rested on an additional source that could not be mentioned at all.

The primary reason not to insist on divulging sources has already been mentioned and is widely acknowledged as taking precedence over the merits of research transparency: respondents in precarious circumstances require our protection (Ahram and Goode 2016; Bellin 2016; Driscoll 2015; Lynch 2016; Shih 2015; Stroschein 2016). The value of anonymity is not unique to authoritarianism research, or even to research on vulnerable groups in the social sciences. It also applies to medical research, or public opinion research. In this sense, transparency is never boundless in academia. As we have argued above, we think anonymity is relatively unproblematic when it comes to research on random members of a subgroup of the general population. Anonymity becomes more controversial when we rely on respondents who have specific, privileged knowledge of the workings of the authoritarian system and who are not interchangeable with others. Using what these kinds of respondents tell us under condition of anonymity poses a dilemma between transparency and anonymity. Betraying their confidence goes against the do-no-harm principle and is ethically unconscionable. So the only other alternative would be not to publish anything that would have to rely on anonymous sources, which raises its own ethical challenges, since it furthers the interests of authoritarian powerholders in opaqueness and potentially ignores voices that can and want to tell us about abusive practices. It is possible in principle to do authoritarianism research entirely based on named sources, for instance, by focusing on historic cases (Art 2016). But we believe that in our field of research—as well as many others—too much would get lost. Each of us, in many different contexts, has at times relied on anonymity. In our experience, those of our ‘authoritarianism’ colleagues who rely on field research as a primary source have almost all done so too. We believe that it is fair to say that the field could not exist without it.

None of us think reliance on anonymous sources is unproblematic from a scientific point of view. Whether or not we believe in actual replicability of our kind of research (we are divided on this), we all think that transparency is required for other academics to judge our work, and to build on it. None of us have published work that rests entirely on anonymous sources. Indeed we agree with the advice given by Shih (2015) and Bellin (comment on https://​www.​qualtd.​net, IV.1, 2016) that authoritarianism research should not fetishize the interview as the only or best source of information but triangulate information from interview material with public documents or online sources. Some of us even think work that relies entirely on anonymous sources should not be published because it is not even partially verifiable. Others think that under very specific circumstances, when the author can argue why there was no safe alternative way of gaining relevant insights, such publications can be permissible. But instead of arguing about precisely how much a publication should be allowed to rely on anonymous sources, we find it more helpful to shift the way we think about transparency from a primary focus on the identity of our sources to a focus on increasing transparency about our methods of working.

Publications that rely on qualitative research are wildly variable in the attention they give to research methods, depending on their disciplinary or subdisciplinary traditions and epistemological orientations. In some journals, the standard is for qualitative research to be written up in ways that approximate as closely as possible the manner in which quantitative research is conducted and described, which is not always appropriate and helpful. In other subdisciplines and associated journals, there is simply no tradition of requiring a methodology section giving attention to how the empirical material was gathered and analyzed. We think that there are many ways of doing good authoritarianism research, but regardless whether it aims to substantiate causal claims or whether it is more exploratory or interpretive in nature, it always benefits from transparency about how we do things. We would argue for more transparency than is currently customary in our field of research. The spotlight should be not on the identity of the sources but on the practices of the researcher. We already typically share when an interview took place (where it took place is occasionally sensitive, see Shih 2015, 22), so that at a minimum we could, when challenged, prove that we were ‘in the field’ on that day, rather than behind our desk inventing respondents, but that is just fraud-proofing. We can do more: share how we came by respondents, and what biases there might be in that process, give insight into the kinds of questions we asked, into the informed consent-related conversations we had with respondents, whether we recorded, how we treated our material, and so on. There should also always be a justification, which can be brief if it is relatively obvious, of why certain sources need to be kept anonymous. As Shih proposes, scholars of authoritarianism could also be more explicit about the other ways in which research has been tailored to meet constraints imposed by the regime, making it clear, for instance, that in undertaking a survey, we might ‘ask proxy questions that are highly correlated with the sensitive questions’ (Shih 2015, 20–21). Our choices regarding methods, ethics, and integrity could all be treated in one section, or if being transparent eats too much into our word count, they can be elaborated in an online appendix.

When it comes to sharing of sources, one might think that while the identity of our respondents often needs to be secret, the material itself could be shared with all, just as the raw data underlying medical research or population statistics can be made public. Why do we not just put anonymized transcripts online? While one of us has indeed done so in the past, we think that too often, doing so would still put our respondents at risk. Precisely because they are not random respondents, but people with specific expertise or privileged information, a good secret service can come to understand who you have been talking to, either from the transcripts alone, or by combining it with their other information about you, or your respondents (see also Tripp’s comment on https://​www.​qualtd.​net, IV.1, 2016). We also believe it unlikely that respondents who want to remain anonymous would readily give their consent to having the entire transcript of the interview made publicly available. And if they did, the information they would give us might be a lot less valuable: having a conversation with us, after carefully having built a relation of trust (see Chap. 4) is not the same as making a broadcast—even an anonymized broadcast—to the world. Transcripts cannot therefore be available to everybody.

A final reason for not making interview transcripts publicly available is that we believe anonymized transcripts would in fact be of limited value to other scholars. Transcripts are faithful transcriptions of a conversation, but they cannot be readily interpreted without the requisite contextual knowledge, the relation to off-the-record comments possibly made during the interview itself, the connection with other conversations that could not be transcribed, and so on. They are not equivalent to quantitative data, and our process of drawing conclusions from them cannot be replicated in the same way quantitative procedures can be replicated on the basis of the data and code by anyone with the requisite methodological skills.

Nonetheless, we think that the current practice of saying ‘just trust us’, and keeping transcripts entirely to ourselves, is not good for our collective reputation as academics. We believe a culture of qualified sharing of anonymized transcripts should be fostered in our field, and perhaps also in relation to qualitative research with vulnerable respondents more widely. We will describe two concrete ways in which we imagine that this can work, which should be read as complementary to each other.

The first is sharing between colleagues, usually but not necessarily within the same department. Within our project, we share all anonymized transcripts with each other. We do not share real names with each other: the only benefit of sharing real names we can think of would be to further reduce the likelihood of fraud, but as we discuss below, we do not think it plausible that researchers can and will invent reams of pages of false transcripts. What sharing means to us in practice is that transcripts are all stored together on an offline laptop in our office. It is a system to guarantee that a small number of people have seen the interviews and can confirm their existence. In case of a doctoral candidate struggling to turn his material into an argument, moreover, supervisors can actually review the material and offer better advice. This is an obvious and attractive solution for research groups such as ours. Such groups are increasingly prevalent in Europe due to the current nature of funding, which favors personal grants to mid-career or senior scholars, intended for building a group around a project. It may also work for region-based research centers, that is, centers for Middle Eastern studies and China or Russia studies, where there is an institutional awareness of the specificities of our work. We would be more hesitant to recommend it as a solution in all circumstances: in general political science departments, there may not be the same understanding of the sensitivity of the material, or conversely, the practice might lapse because nobody polices it. A drawback of sharing within a group or center is that, in case there are concerns over authenticity, close colleagues may have a personal or institutional stake in covering for each other. But this kind of sharing is still to be preferred over not sharing at all, which we believe to be the current standard, and it can be combined with other sharing practices as described below.

A second sharing practice could emerge in the context of the publication process of a journal article or book manuscript. One form this could take is peer review: either transcripts could be shared with reviewers as a matter of course, or there could be a designated ‘source reviewer’. We think most authoritarianism researchers would be reluctant to accept such a system: given that most peer review is double-blind, it would require authors to hand over transcripts without having any idea to whom, other than that these people are presumably also academics. Researchers might well feel that submitting transcripts in this way would breach their obligation to their respondents. Moreover, it would place a heavy burden of responsibility on journal editors or book publishers, who would then be responsible not only for the academic quality of the reviewer but also for her integrity with regard to neither using the transcripts for their own purposes nor sharing them with third parties.

In our conception, only confidential materials explicitly referred to in publications should be subject to sharing. We concur with Lynch (2016, 38) and Tripp (https://​www.​qualtd.​net, IV.1, 2016) that it makes little sense to share our multilingual fieldnote scribbles, which will not be intelligible to anyone. Nor should we be under an obligation to make them intelligible, any more than we should be obliged to reconstruct inspirational thoughts we may have in the shower before writing them up. We acknowledge that editors will not be in a position to fully interpret transcripts. As we explained above, seeing transcripts does not imply that you can ‘replicate’ the analysis. Finally, there are practical challenges, for which we do not yet have adequate solutions to offer, concerning how to securely transmit transcripts to an editor. But our position is that we should move toward a culture where it would be considered natural and legitimate for editors to ask to see anonymized transcripts that we refer to in publications, and we would share them on request. It cannot be the case that while quantitative researchers are increasingly being asked to make raw data publicly available, we will not share any of our material with anyone and just insist on being trusted.

As we explained in our introduction to this chapter, recent European policy initiatives aim to create a ‘network of networks’ of digital data repositories. It is still quite unclear at what point a researcher would be expected to place data in a digital repository, and to what extent access would indeed be open to all. The notion in a recent policy paper that all data should be available to all, in all phases of the research cycle (Joint Policy Paper 2017) reflects a poor understanding of how scientists work, and governmental overreach in terms of transforming their ways of working. Nonetheless, social science researchers in Europe may soon come under institutional pressure to comply with mandatory data storage in digital repositories. Open access repositories are subject to exactly the same objections as the DA-RT and JETS initiatives, specifically but not exclusively from the perspective of the authoritarian field, so we need not rehearse our arguments here.

But we want to go a step further and state our objections even to digital repositories with restricted and/or embargoed access. Again, our primary objection concerns risk to respondents. As Marc Lynch points out, ‘the difficulty of guaranteeing confidentiality for materials deposited in a trusted repository are not hypothetical to those of us who conduct research in the Middle East and North Africa’ (Lynch 2016, 37), and the same is true for other authoritarian contexts. We discern three aspects to this risk of breaching confidentiality, and hence risking harm: political contingency, legal risk, and digital risk. The first aspect relates to the apparently very reasonable suggestion that we could place our material in digital repositories under embargo, to be made public after, for instance, five or ten years, subject to our consent. The problem with this is that we cannot predict the future, and hence we cannot assume that publishing transcripts would gradually become less sensitive over time. It can also become more dangerous. Lynch gives the example of Egypt, already referred to above: what were ‘bold but safe’ statements made by activists in 2012 or the first half of 2013 quickly became very dangerous from the latter half of 2013. Stroschein gives the similar example of Turkey, which has become much more repressive after the coup attempt of 2016 (https://​www.​qualtd.​net, IV.1). A disembargoed interview with a Turkish respondent from, say 2011, could well land her in trouble in 2017. And as we discussed in Chap. 3, our China researcher has seen a more subtle but discernible shift in the ‘red lines’ of permissibility in China over the past years, that could have implications for disembargoed transcripts.

Second, there is the possibility of transcripts becoming subject to legal subpoena, a particular concern with US scholars (Driscoll 2015, 6; Lynch 2016). We have not given attention in this book to the risk of legal subpoena, because we have no personal experience with it, and it still seems to be a rare occurrence. But what we can say is that when we store materials in a digital repository, we give up control: the (difficult) choice of weighing responsibility toward respondents against legal obligations and possible criminal liability would no longer be ours to make. Driscoll (2015, 6) records actually having burnt some of his field materials in order to guard against the risk of subpoena. None of us have gone this far, but we are aware that ethical review boards sometimes insist on the destruction of data to protect respondents. A blanket destruction requirement would be just as extreme as a blanket transparency requirement, but the fact that social scientists can be subject to both contradictory prescriptions at the same time illustrates the unhelpfulness of blunt, one-size-fits-all solutions to research dilemmas.

Finally, even if deposited transcripts were to remain in restricted access, it would be naïve, in the age of hacking, to believe that academic repositories can be made fully secure. We would like to believe that most secret services most of the time have other priorities than getting access to our transcripts, but we can never be certain. In Chap. 4, we quoted the forthright answer one of us got from a Moroccan activist when she asked him a sensitive question: ‘if you can assure me that you can protect me I will give you my answer … but since you cannot, I will not’. Here, we paraphrase him to state our position on storing sensitive interview transcripts in digital repositories: if the institution can assure us that it can protect our respondents, we will give it our transcripts, but since it cannot, we will not.

As academics, we all want our work to be paid attention, by our peers but perhaps also beyond academia. We may even dream of being famous as academics. But for a researcher on authoritarianism, academic fame is a double-edged sword. If more than a handful of colleagues are taking notice of our work, the regime may be doing so too. As we described in earlier chapters of this book, we all do research in contexts where there is some degree of space for, and understanding of, social science research. But this space is constrained, and we do carefully consider what we publish, where we publish, and how we disseminate our work.

Our Kazakhstan researcher suspects that the regime would not be happy about some of her work, especially that which focuses on the workings of the party in power. She did consider this when writing, but she believes that most political leaders will not read it, and even if they should read it, they would still assume that, as an academic paper, it would be mostly ignored or considered harmless because it does not communicate directly with a large public. She faced a dilemma when an assistant to the prime minister specifically asked to be sent a copy of her work on the political leadership’s legitimation strategies (Del Sordi 2016). Since the assistant had been very helpful and had agreed to be interviewed herself, our researcher could not refuse, but she did have some concern that her access to the country could be jeopardized by this move. The prime minister in question has a reputation for academic curiosity, however, and she has not in fact had difficulties with her most recent visa. She has even seen colleagues taking a more public critical position without consequences for their access, but as the authorities are always weighing the reputational consequences of denying access against those of being criticized, one cannot rely on being able to combine public criticism with continued access.

Likewise, our China researcher believes that her description of the Chinese political system as ‘fragmented’ might not please the government, but since it does not aim to undermine the Chinese Communist Party, she does not believe she would really be denied access to her home country. Again, the relative obscurity of academic work also makes a difference: journalists from the west are much more regularly denied access than scholars. This difference seems to be confirmed by the recent experience of a colleague in China, who was recently ‘invited for a cup of tea’ by security agents because of (English-language) news coverage of an academic publication of hers.

Our Iran researcher by contrast has, in the very specific context of the repressive aftermath of the 2009 election protests, initiated an activist-oriented edited volume (Michaelsen 2011) that he thought could compromise future access. He found the situation in Iran so dramatic at that time that he wanted to take a position. He decided that the story told by this book, edited together with 11 journalists who left the country and wrote about their experiences during and after the protests, was more important than going back to Iran. The book was published in English and Farsi, and he gave interviews about it to Farsi language online media in the diaspora considered inimical by the regime. When, six years later, he prepared for another trip to Iran, he did briefly wonder whether this publication might compromise his access to and security in the field, but he still thinks that there are times when academic researchers should take a clear and principled position.

Another way of disseminating our work, perhaps the most effective way in numerical terms, is by acting as commentators in western media. The increasing emphasis on societal engagement, moreover, may propel scholars to think that all publicity is good publicity. We do sometimes give radio interviews, or allow ourselves to be quoted in newspapers, but we are very careful about the exact wording. In case of print media, we always insist on seeing and being allowed to correct quotes before publication. A more negative phrasing than we are comfortable with, sometimes desired by journalists, not only interferes with the nuance of what we want to say, it can also have consequences for our access to the country and to sources in the country.

A final consideration, when it comes to weighing publicity against future access, is the extent to which our careers and our lives are bound up with one country and its political system. Our India and Mexico researcher and our Malaysia researcher have not been much concerned about future access to the relevant countries, in part because such denial of access is relatively rare, but primarily because, at this point in their career at least, they are mixed-methods researchers who think of themselves as political scientists who happened to do fieldwork in one or two specific countries. Our Morocco researcher found doing research in Tunisia to make a refreshing change and is also thinking about broadening her expertise to West Africa. Our Kazakhstan researcher thinks of herself as a country expert first and foremost, but has also written on Central Asia more generally, and considers future research on Russia. Our Iran researcher, while he has invested profoundly in learning Farsi and understanding Iran, has partly shifted his research agenda, toward studying the Iranian diaspora, on the one hand, and a broader comparative focus on media in authoritarian contexts on the other hand. While the primary motivation for broadening our research agendas has not been to mitigate against the risk of the authoritarian state obstructing our research, it does make it easier to navigate the dilemmas regarding publicity and access. Our China researcher is more exclusively invested in understanding the universe that is China, albeit comparatively. Moreover, she is and wants to remain a Chinese citizen, so for her the stakes in navigating what to write, and where to write it, are higher, as they are likely to be for any national investigating their own country. In sum, we all think of how we couch our criticisms of authoritarian regimes and how publicly we do so, in relation to future access as a trade-off, but the choices we make depend on our specific professional and personal relation to the field.

There is an inherent tension in doing, but especially in publishing, research on authoritarianism. Ahram and Goode (2016, 838) describe authoritarian regimes as ‘engines of agnotology’, by which they mean that these regimes have an interest in maintaining ignorance and uncertainty about many aspects of how they function. Hence, publication can raise problems for our future access, but more importantly, potential harm to sources. We add our voice to the chorus of scholars who have argued that a concern for transparency in research cannot be translated into a requirement to make transcripts or field notes public, even in anonymized version. Nor should they be stored in potentially unsafe digital repositories. Our responsibility to do no harm to respondents is simply paramount. But we have also tried to go beyond only rejecting inappropriate transparency requirements. In this chapter and in this book, we have tried to increase transparency about how we do research: by explaining in detail how we have navigated the methodological and ethical trade-offs that follow from doing research in the authoritarian field, and what general learnings we think may be gleaned from our common experiences.