Transitional justice in Portugal

Portugal endured nearly five decades of authoritarian rule, with the Estado Novo (the New State) institutionalized in 1933 under the leadership of António de Oliveira Salazar, following the principles of European fascism (^{Pinto and Kallis, 2014}). Preceding this, a military dictatorship occurred between 1926 and 1933, during which Salazar, serving as the Minister of Finance since 1928, held a prominent role (^{Accornero, 2018}).

To suppress the opposition, Salazar’s regime used various repressive methods, curtailed individual rights and freedoms, imposed censorship, and implemented surveillance measures (Raimundo and ^{Pinto, 2014}). The regime established prisons specifically for political offenders and created the Surveillance and State Defense Police (PVDE), which was renamed PIDE in 1945. PVDE was primarily tasked with controlling political crime and was heavily influenced by the OVRA, the Italian Fascist Secret Police (^{Pimentel, 2004}). Repression became a hallmark of the New State, closely following the models of other European fascisms (^{Accornero, 2018}).

In 1974, a bloodless coup led by middle-ranking military officers brought an end to Salazar’s regime (^{Lobo et al., 2016}) and ended a 13-year-long colonial war in Angola, Mozambique, and Guinea-Bissau (Raimundo and ^{Pinto, 2014}), signaling the onset of democratic transitions in Southern Europe ( ^{Diamond et al., 1997}).¹

After that, a large set of tj mechanisms were implemented, including trials, purges, truth-revelation, professional, financial, and symbolic reparation for victims of repression,² and memorialization (^{Pinto, 2006}; ^{Raimundo, 2015}; ^{Rezola, 2019}). The first phase lasted two years, from 1974 to 1976, and it was mainly characterized by perpetrator-oriented measures to punish those who committed crimes and collaborated with the dictatorship. This process included military trials against the political police, vetting measures against individuals who allegedly collaborated with the regime, and the forced exile of the previous political elite (^{Pinto, 2006}). This period was characterized by a State crisis, influential social movements, and military intervention, that influenced societal perspectives on punishing individuals associated with the previous regime (^{Pinto and Morlino, 2011}). In a second phase, beginning in 1976, occasional initiatives to recognize and preserve the memory and compensate the victims of the dictatorship were discussed and approved. However, it was not until 1997 that government legislation allowed members of the opposition to the dictatorship to claim compensation through social security and retirement pension entitlements for the years they spent in clandestinity or exile (^{Pinto, 2011)}.

Although Portugal did not include truth-seeking mechanisms in its transitional justice framework (^{Pinto, 2010}), it is noteworthy to highlight the existence of a State entity, established in 1977 and dissolved in 1991, that closely resembled a traditional ‘truth commission’ (^{Pinto and Morlino, 2011)}: the Commission of the Black Book on the Fascist Regime. This commission aimed to investigate the abuses committed during the dictatorship, which ended with the military coup of April 25, 1974 (^{Morais, 2022}). According to the commission’s estimates, approximately 30 000 individuals were imprisoned, 32 died in the Tarrafal labor camp, and about 20 members of the opposition were killed directly by the political police (^{Raimundo and Pinto, 2014}).

tj in Portugal has been a long and occasionally contentious process, especially regarding the trials and purges of the early years, and remains wide open for further exploration, particularly in understanding the individual-level impact of the tj mechanisms within the target population. This research note aims to present the methodological challenges of a study designed to understand the perspectives of individuals at the center of this process - those who experienced some form of repression for their political ideals during the Estado Novo. Those among the victims who were alive during the timeframe of our project (2016) were socialized under the authoritarian regime, particularly in the 1950s and 1960s. The study explored differences in attitudes, perceptions, and behaviors among victims of the Portuguese right-wing dictatorship. It also explored the underlying mechanisms of these attitudes, with a specific focus on ideology and political affiliation, for which clear evidence is lacking. It is also important to point out that our dataset is the only survey of direct victims of repression in Portugal offering a significant contribution. Furthermore, these data allow us to test existing theories in the field and compare the Portuguese case with others where the role of ideology and political affiliation has been explored (^{Aguilar, Balcells and Cebolla-Boado, 2011}; ^{Bratton, 2011}).

A mixed-methods research design to enhance the study of Transitional Justice in Portugal

To study the attitudes of the victims of the Portuguese authoritarian regime, we conducted a two-phase exploratory sequential mixed-methods design (^{Ivankova, Creswell and Stick, 2007}), consisting of: a qualitative phase and a quantitative phase, with the first phase iteratively guiding the second phase of the study.

For the qualitative phase, we chose to conduct face-to-face (f2f) focus groups. Compared to one-on-one interviews, they are a more appropriate approach given the nature of the study and the characteristics of the participants. First, these are considered to be a more appropriate approach for conducting research that involves potentially sensitive topics (^{Mauthner, 1997}; ^{Wutich et al. 2010}). In fact, the topic of our research may evoke strong feelings and opinions as well as painful memories for the participants. Therefore, subjects may be more willing to share their views and experiences of victimhood if they are surrounded by others who are similar to them (^{Morgan, 1996}). Group settings favor the formation of a comfortable environment and the perception of being among ‘equals’ (Acocella and Cataldi, 2021). Second, given we are asking the participants to remember and share past events that occurred over 40 years ago, the dynamics of a group of participants may be useful in reviving and activating the memory of forgotten details and aspects not previously considered (Hennink, 2014). Third, a group setting can also feel less invasive or threatening for participants, particularly when the researcher is unfamiliar to them prior to the study (^{Farquhar, 1999}). While this does not entirely apply to our case - given our research experience -, we recognize that we do not share the participants’ lived experiences, having not endured the same realities. However, group settings may not always be ideal when addressing sensitive memories, as they can heighten discomfort or vulnerability. To mitigate this, we opted for small focus groups, allowing for a more controlled dynamic and fostering deeper, more meaningful discussions (Acocella and Cataldi, 2021).

The exploratory f2f focus groups informed the second phase of the study - the quantitative stage, i. e. the development and implementation of a survey instrument. While surveys are not necessarily recommended in every instance of research on tj, numerous important surveys have been conducted effectively and safely in the field (^{Backer and Kulkarni, 2016}). Several studies on tj have relied on surveys to collect data on victims’ attitudes toward retribution, reparation, and memorialization policies through the use of representative face-to-face surveys (^{Aguilar, Balcells and Cebolla-Boado, 2011}; ^{Nussio, Retteberg and Ugarriza, 2015}; ^{Hall et al., 2018}), representative telephone surveys (e. g., ^{Gebauer, 2010}), or experiments (e. g., ^{Gibson and Gouws, 1999}; ^{Rimé et al., 2011}). These sampling techniques have a number of requirements: that a significant number of victims are still present in the population; that the researcher has sufficient knowledge of the characteristics of the population; ideally, that they are conducted where the target population is less geographically dispersed; and that a significant amount of financial resources are available. As explained before, many of these conditions are not easily met, especially in the case of studies focused on twentieth-century dictatorships, such as ours.

The mixed-methods design, outlined in Figure 1, played a crucial role in achieving the goals of our study. This was particularly important due to the limited information available on the conceptualization of the phenomenon, the measurement of potential key variables in the specific context of the Portuguese case, the number of years that elapsed since the events of interest, and the sensitivity of the topic under study. Considering the scarcity of empirical research in the field of tj (especially in Portugal), there were no immediate resources available when operationalizing the key concepts of the phenomena. This design has proven useful in improving the quality of the survey instruments and, consequently, the survey data. It helps to develop a questionnaire that is comprehensive and fit for purpose, avoids frivolity or ambiguity, prevents boredom, and minimizes comprehension difficulties (^{Galliot and Graham, 2016}). Such careful design is particularly important in studies such as ours, which aim to explore uncharted territory in the Portuguese context and go beyond previous literature. The focus groups were especially helpful in tailoring our survey to the target population, particularly in refining the wording of our survey - an essential aspect of survey design that has an immense impact on the quality of the data collected (^{Puchta and Potter, 2004}). This is an issue that we will explore further in following sections.

Figure 1 Research design: an exploratory sequential mixed-methods design. Note: Authors’ elaboration. 

Leveraging civil society organizations for sampling and recruitment

As with other studies interested in retrospective research, the first challenge was sampling and recruitment. How do we find the target population if we do not know the population universe? How to build a sampling frame when there are no official, up-to-date figures on the size of the target population or its socio-demographic data?

The first step was to identify as precisely as possible the characteristics of the target audience. Since Portugal experienced an authoritarian regime without internal conflict, the boundaries between victims and perpetrators are reasonably well defined. Following previous works (^{Nussio, Retteberg and Ugarriza, 2015}), we relied on national laws passed since 1974 to recognize and repair individuals who experienced some form of repression during the authoritarian regime.³ The only exception was individuals who were ‘tortured’, as there is no specific law to recognize or repair those who were tortured during the dictatorship as political prisoners. Excluding these individuals from our study of victims of repression would have been problematic, as it could be argued that torture represents the most extreme level of violence that regime opponents could experience in Portugal. Therefore, our target audience included individuals who were arrested and/or imprisoned, went into exile, spent time underground or engaged in clandestine activity, were dismissed from their jobs, were expelled from high school or university, and/or were deserters or draft evaders for their political and ideological reasons.

The second phase involved building a sample pool and recruiting individuals for our two-part study. We used convenience and snowball sampling to recruit participants for both the focus groups and web surveys. Given the absence of a comprehensive and accessible sampling frame, namely information on individual names, whereabouts, or official up-to-date figures for this subpopulation in Portugal, we relied on key associations closely linked to our target research participants in order to reach potential participants. This approach is in line with previous studies in tj, such as ^{Backer and Kulkarni (2016}).

We selected three organizations that were active at the time of the project: the ‘Do Not Erase the Memory’ (Não apaguem a memória, NAM), the ‘Union of Portuguese Antifascist Resistance’ (União de Resistentes Antifascistas Portugueses, URAP), and the ‘April 25^th Association’ (Associação 25 de Abril). Despite their different nature and organizational structures, we chose them because they: a) are committed to preserving the memory of antifascist resistance; b) uphold civic and democratic values; c) have developed important activities over the years, such as lobbying political authorities on the issues of reparation of victims and memorialization policies; d) and also organize events to commemorate the coup of April 25^th. These factors made them an ideal gateway to reach our target population.

After the democratization process, several grassroots organizations were created to advocate for support and justice for the victims of the dictatorship, but many did not survive the transition (^{Pinto and Raimundo, 2014}). Therefore, we specifically selected the three organizations that have remained active in recent years in preserving the memory of the resistance to the dictatorship and the freedom achieved on April 25, 1974.

Establishing a trusting relationship with these organizations through multiple contacts and open dialogue was critical to the success of the recruitment process. After initial contacts and discussions with the representatives of these organizations, during which we explained the scope of our research, we asked them to forward our email invitation to their members.⁴ As a result, individuals who were interested in our study reached out to us, allowing us to build a pool of participants who met our criteria. From this pool, we randomly selected individuals to invite to the focus groups (phase one of the study) and extended invitations to all of them to participate in the web surveys that followed (phase two of the study).

We also asked each person if they knew people who met the research criteria and would refer them to the project (snowball sampling). When the original contacts used to start chains were exhausted, we started a new wave aiming to find new participants and new referral chains. We also used locators to find participants. Locators are individuals who possess specialized knowledge or insights about a particular area of life due to their past or present experiences or circumstances, which grant them privileged access to information or perspectives that others may not have (^{Biernacki and Waldorf, 1981}). This strategy proved to be highly effective, as locators can more easily establish referral chains because they already know potential respondents or are more likely to have others refer potential participants to them. In our case, an exemplary locator was a research colleague whose past and current work was relevant to our study and the target population.

The underlying assumption of snowball sampling is that there is a link between the initial known subjects and others in the same target population. If this assumption is correct, a chain of acquaintances can be created from the primary contacts. The main value of snowball sampling is its usefulness when a certain level of trust is required to initiate the process of recruiting subjects for the study. However, findings from data collected through snowball sampling are not easily extrapolated to the target population, which raises the issue of limited validity of the data. Traditionally, the non-random nature of social network connections has led to such samples being considered as convenience samples from which unbiased estimation is not possible (^{Berg, 1988}).

An alternative method that challenges these views is respondent-driven sampling (RDS and web-based RDS), which is an extension of network or ‘snowball’ sampling that uses data on who recruited whom and the extent of network connections as the basis for calculating relative inclusion probabilities, population indicators of minimal bias, and the variability of these indicators (^{Wejnert and Heckathorn, 2008}; ^{Salganik and Heckathorn, 2004}). This approach challenges the notion that unbiased estimation from convenience samples is impossible (Wejnert and Heckathorn, 2008). Despite the widespread use of RDS in different fields and its potential for the tj field to reach hidden and hard-to-reach populations, the assumptions underlying RDS theory may not always hold, and recruitment may not generate large numbers of subjects (^{Jonsson, 2019}; ^{Léon et al., 2016}). Furthermore, sometimes the population of interest is not sufficiently socially networked, which may be the case for our target population. In our case, we do not have enough information to know whether these individuals are sufficiently networked with each other to sustain recruitment chains or if they are able to recruit each other. Their affiliation with these organizations does not inform us on the extent to which they are or are not networked with each other. Regardless of whether the target population is interconnected, we know that these individuals are connected to the organizations we considered and contacted. Therefore, we relied primarily on these organizations as the direct point of contact to recruit participants. Nevertheless, RDS may be useful in other cases.

Qualitative strand: focus groups, procedures and challenges

The second challenge in our study was to develop a robust instrument for collecting individual-level data. How do we create such an instrument that is comprehensive and sensitive, given the complexity of the experiences and attitudes we aim to capture, when there is no immediate guidance on how to operationalize key concepts that take into account the cultural, political, and historical context of events in Portugal?

We relied on exploratory face-to-face (f2f) focus groups, with a small group of six to nine individuals in three separate groups (n = 21),⁵ to explore the phenomenon in depth and develop our measurement instrument. They were conducted at the Institute of Social Sciences of the University of Lisbon, in February 2016. We created an interview guide consisting of open-ended questions that broadly explored attitudes towards forgiveness, truth-seeking, historical memory, prosecution and reparations for victims. These open-ended questions deliberately allow the participants as much latitude as possible in their responses (^{Krueger and Casey, 2009}). While the questions in the interview guide provided the main blueprint for the flow of the interview, the moderator had several opportunities to explore the emerging themes and participants’ responses. The f2f focus group was preceded by warm-up activities and an explanation of some ground rules for the discussion. The three video and audio recordings of the group interviews were transcribed verbatim.

Each focus group was conducted by a team consisting of a moderator and an assistant moderator. The moderator facilitated the discussion while the assistant took notes and recorded the session. All participants signed a consent form. As we expected the research topic to be interesting enough for this specific population, there was no need to use any kind of incentive to participate.

The F2F focus groups were designed to support the development of our measurement instrument - the web survey. The open-ended questions were designed to address several themes related to victims’ experiences and perceptions of tj mechanisms in Portugal. Participants were given time and a non-judgmental setting to articulate and discuss their views and perspectives on the themes defined by the research team: a) personal experiences of repression; b) practices and experiences of association and mobilization; c) experiences of compensation and recognition; d) attitudes towards justice and punishment (e. g., attitudes towards punitive measures implemented in Portugal); and e) truth and memory (e. g., knowledge about museums and exhibitions on the topic). For each of these themes, cards were introduced to facilitate discussion. For instance, a card with different experiences of repression was introduced, as well as a card with a list of national laws passed since 1974 to recognize and repair individuals who experienced some form of repression. This proved to be an effective method to stimulate participants’ memories, encouraging richer discussions and refining the future operationalization of survey questions.

Many participants were willing to openly discuss sensitive topics, such as their experiences of repression during the dictatorship, including cases of torture. F2F focus groups provided a dynamic space for individuals to freely express their views, opinions, and memories in a safe environment (^{Dodson, Piatelli, and Schmalzbauer, 2007}). This research technique proved to be especially useful in helping participants collectively remember and share events that occurred over 40 years ago. They played a crucial role in developing a more robust measurement instrument, refining the survey questions, responses, scales, and language to ensure the questions took into account the subtleties of the topic. Because the topic can be sensitive for some participants and evoke painful memories or feelings (^{Simic, 2017}), relying on in-depth and face-to-face conversations in a safe environment through a group setting helped us immensely to understand the impact of language. For example, determining what word(s) should be used to refer to what is commonly defined as ‘victims.’ Through these discussions, we realized that those who opposed the authoritarian regime do not perceive (or define) themselves as victims (some even expressed an aversion to the term), but as active opponents of the regime, and consequently they tend not to demand reparations. As a result of these findings, we refrained from using this term in the survey. Instead, we introduced a multiple-choice question to determine how respondents define themselves based on the experiences of repression they reported. In this question, the options, which were informed by the focus groups, included ‘victim,’ ‘resistant,’ ‘hero,’ ‘opponent,’ and ‘I don’t know.’ Notably, only 1% of survey respondents chose the term ‘victim.’

Quantitative strand: web-surveys, procedures and challenges

In the second phase of our study, we invited 360 individuals to participate in the web survey. We sent personalized email invitations to our sample with the link and instructions to complete our survey (^{Couper, 2008}). A thank you note was sent to the respondents within one day of the survey completion. Because we expected e-mail use to be highly variable in our target population, the sampling period was left open long enough for occasional e-mail users to check their inboxes and respond.

The research recruitment criteria (see above) were explicitly described on both pages of the web survey (introduction and first page). Nevertheless, some of the participants (referred by others), who in one way or another considered themselves to be victims of the dictatorship, did not meet the research criteria and left the survey immediately after the introduction (n = 139) or after the first page of the survey (n = 69). On the first page, participants were asked to select the situation that applied to them, i. e., the nature of repression they experienced during the Portuguese dictatorship. Since these experiences are not mutually exclusive, in other words, it is likely that some participants have suffered multiple experiences of repression, this was a multiple-choice question. Furthermore, we predicted that some of the referrals would not meet the research criteria even though they felt they were victims in some way. Therefore, we added an option on the first page to cover these cases: ‘I was harmed in my personal and/or professional life, but I do not fit any of the situations described above.’

There is ample variation in how researchers measure individual perceptions and evaluations of previously implemented mechanisms. Some examples include: a list of possible answers for respondents to choose from (^{Pham et al., 2007}; 2010); multiple-choice questions that ask respondents to choose between two different statements on a similar topic;⁶ Likert scale questions (^{Nussio, Retteberg and Ugarriza, 2015}); or qualitative scales to measure competing views of attitudes toward dealing with the past.⁷ Although different strategies have been used, as shown above, they all follow a holistic approach, often leading to results in the construction of indices of support that combine multiple variables (Aguilar, Balcells and Cebolla-Boado, 2010) or to the analysis of the different components of tj individually to identify what determines support for punishment or forgiveness and retribution or truth (^{Samii, 2013}). Inspired by the literature and building on the data from the focus groups, we developed ten indicators of attitudes and perceptions toward tj, operationalized by a five-point agree/disagree scale. A few examples of the statements considered include: ‘to what extent do you agree or disagree with the following statements?’ ‘Victims of the dictatorship have not yet been recognized’ and ‘There has not been enough justice for members of the State police.’ We have also collected data on individual characteristics such as age, gender, ethnicity/race, marital status, educational level, location, socioeconomic status, religion, political affinity, and organizational affiliation.

The survey was conducted between March and December of 2016, with multiple reminders sent during this period to encourage participation. We achieved a completion rate⁸ of 41.6%, meaning that of all 360 initial personal invitations to participate, 150 respondents completed the survey. In the sample of 150 individuals, the majority of the participants are male (77.6%), have a high level of education (79.2%), are living in Lisbon at that time (62.0%), and are, on average, 73.3 years old. These descriptive statistics are a good starting point to reflect on the quality and validity of the data collected through web surveys and a convenience sample.

These statistics show how biased our sample is toward individuals who live in Lisbon and have a high level of education. They reflect, of course, the non-probabilistic⁹ technique used to build the sample of participants. In the absence of a census or any other database to build a sampling frame of the target population, the solution was to resort to key civil society organizations, hoping to reach their members. Although this strategy proved to be a great source of participants, it excludes a priori those who are not connected to any of the associations considered and/or those who are not accessible electronically or able to participate in a web survey.

In order to meet our criteria, our potential respondents would have to be born in 1956 or earlier, which would make them 60 years old or older at the time they were invited to participate in our survey in 2016. Although internet access is increasing in Portugal, it is most likely that our target population consists of individuals who may not use ICT in their daily lives and may not be easily accessible online or able to participate in an online survey. Considering that research (e. g. ^{Brandtzæga, Heima and Karahasanović, 2010}) has shown that Internet users differ from non-users in terms of socio-demographic characteristics (Internet users are younger, more educated, have a higher socio-economic status, and come from urban areas) and in lifestyle, behavior, and attitudes, especially those related to ICT, this represents a significant bias in our sample.

Upon further reflection, we confirmed the accuracy of this observation. After all, our sample consisted predominantly of highly educated individuals (79.2%), more than half of whom had been involved in student associations in the past (57%), and almost half of whom experienced forced exile during the dictatorship (45%). It is, therefore, reasonable to conclude that many respondents were likely to have been part of the student movement in the 1960s and 1970s. These individuals not only had a higher level of education, but also different experiences and a distinct political profile. At the end of the regime, students represented the social category most affected by the repression, which is even more significant considering that this was a relatively small group of people in Portuguese society (^{Accornero, 2018}). For instance, in contrast to the working class, which was mainly persecuted in the 1930s and 1940s and was less represented in our sample, the individuals predominantly captured in our study show notable differences.

Despite these limitations, our sample shows significant diversity in terms of their experiences. Two-thirds of the participants were arrested or imprisoned at some point during the dictatorship (64%). Additionally, a substantial proportion experienced forced exile (43%), while a third were tortured (36%) or deserted from the armed forces (32%). It is important to note that these forms of repression are not mutually exclusive; in fact, most individuals reported experiencing multiple forms of repression. Specifically, 74% of the participants reported experiencing more than one form of repression. These figures underscore the lasting effects of one of the longest periods of undemocratic rule in the twentieth century. Despite this diversity, it is important to note that the survey represents only a snapshot of those who suffered during the Estado Novo and is by no means representative of the entire universe of this population.