Data futurities: Negotiating statistical citizenship and ethnoracial visibility in Portugal's 2021 population census

Abstract

This article examines the controversies surrounding the 2021 Portuguese census, centring on debates over proposed but ultimately excluded ethnoracial questions amidst a broader shift towards digital and register-based census-making. We argue that ‘data futurities’ – in and of censuses – are produced through reform narratives that promise modernisation while deferring politically fraught forms of visibility. Treating censuses as performative technologies that organise national imaginaries across time, we analyse how changing data practices reconfigure participation, visibility, and statistical citizenship. Drawing on archival materials, reborn digital media sources, and official documentation, we trace how Afro-descendant and Portuguese Roma communities mobilised competing claims over recognition, data ownership, and participation, and how these demands were channelled through technical justifications and institutional risk management. We show that the expansion of administrative data does not resolve demands for ethnoracial visibility; it displaces classificatory decisions to other instruments and infrastructures, reshaping the terms of belonging. These findings suggest that the stakes of ‘future’ censuses lie as much in inclusive nation-making via statistical citizenship as in technological innovation.

Introduction

The collection and organisation of population data through censuses has a long history, with early forms of enumeration dating back to ancient times (Thorvaldsen, 2018). The ‘traditional’ population census – based on questionnaires administered periodically across a national territory – emerged in the 19th century alongside the consolidation of nation-states and the standardisation of statistical practices. From their inception, censuses have functioned as instruments of state-building, classifying populations and rendering society legible for administrative and political purposes. Although census-making has evolved through incremental technical changes, these shifts have been consequential, reconfiguring methods, meanings, and political stakes while often being framed as continuity (Vitali-Rosati, 2018: 34).

In this article, we take the 2021 Portuguese population census, marked by intense public debate over proposed ethnoracial questions and their eventual exclusion, as a starting point to examine how populations are known, counted, and governed under conditions of digital transformation. We develop a grounded account of the infrastructural and conceptual changes shaping contemporary census-making in Portugal, particularly the turn towards digital and register-based data production. We argue that censuses function as performative technologies that link past, present, and future projects of nation-making, and we introduce the concept of data futurities to analyse how futures are anticipated, managed, and deferred through census reform, including through promises of modernisation and efficiency, institutional risk management, and the displacement of politically contested forms of visibility.

The build-up to the 2021 census was shaped by two intersecting dynamics. On the one hand, the digitalisation of census practices and growing reliance on administrative data raised concerns about changing infrastructures of data production, accountability, and participation. On the other, social mobilisations challenged emergent datafication practices by questioning not only whether ethnoracial data should be collected, but also who should participate in defining the criteria, categories, and governance of such data. Disputes over official ethnoracial data thus became a focal point for broader struggles over statistical citizenship: who is seen by the state, on what terms, and through which data systems.

Focusing on absences as much as on explicit controversies, we ask: how do imaginaries of future censuses build upon, rework, or disrupt existing statistical practices? How do debates over ethnoracial classification articulate competing claims about recognition, participation, and data governance? And what does the exclusion of ethnoracial questions from the 2021 census reveal about how futures are managed through postponement, displacement to other instruments, and institutional framings of risk?

The article proceeds as follows. We first develop a conceptual framework around statistical citizenship and ethnoracial population data. We then outline our methodological approach before tracing the history of Portuguese censuses and key datafication policies. Next, we analyse documents produced by the Working Group (WG) on Census 2021 – Ethnoracial Questions, examining how Afro-descendant and Portuguese Roma actors articulated competing demands around inclusion, participation, and data governance, and how these were mediated through technical justifications and institutional risk management. We conclude by showing how data futurities help rethink the past and future of Portuguese census-making, particularly where citizenship is rooted in struggles over visibility and participation.

Census taking as frontiers of data futurity

Over recent decades, statisticians have undertaken experiments seeking to reorganise census-taking driven by pressures to publish statistics more frequently, reduce costs, and modernise data infrastructures (Ruppert and Scheel, 2021: 2). We treat these reform agendas as future-oriented projects: they mobilise promises of efficiency and quality, foreground risks (concerns around privacy and legitimacy), and reorganise what becomes possible to count.

Across Europe, hybrid and register-based census infrastructures are key sites where the future of census-making is negotiated through standards and governance, as seen in Spain, North Macedonia, and Italy (Eurostat, 2024: 100–104). Spain completed its first fully register-based census in 2021 after investing in the Padrón Municipal and testing. North Macedonia adopted a combined-method census the same year, using registers, field, and web responses after decades without a successful count. Italy is developing an Integrated System of Statistical Registers linking various data sources. These cases show how such infrastructures shape census policies through technical standards, governance, and public debate, situating Portugal within wider European trends.

Within this process, statistical data function as technologies of governance inseparable from social practices (Foucault, 2007): statistics do not merely describe populations; they help constitute them through categories, norms, and measurable differences, shaping who can be known, governed, and recognised. Scholarship on the politics of enumeration shows that censuses are not merely technical exercises but key infrastructures of democratic representation and statecraft (Bartl et al., 2024; Egeler et al., 2013). Statistical and political representation are deeply intertwined, as being counted often becomes a condition for recognition as a subject of rights and governance. It is within this analytical frame that Hannah (2001: 516) defines statistical citizenship as a strategic engagement in shaping the statistical representations that classify individuals as political actors, social policy subjects, or consumers. Statistical citizenship thus involves managing one's personal information politically in everyday life and cultivating a critical awareness of censuses as instruments that participate in nation-making and futurity. In this sense, statistical citizenship can be understood as one of the many ways individuals enact active, participatory citizenship.

If statistical citizenship concerns how people engage with governing representations, it also draws attention to how those representations are produced. Censuses are not neutral instruments but performative practices shaped by datafication – the translation of social life into quantifiable traces (Mayer-Schönberger and Cukier, 2013; Mejias and Couldry, 2019) – a process that raises ethical and political concerns (Iliadis and Russo, 2016; Zuboff, 2015; 2018). In census-making, datafication materialises through practices such as defining, standardising, inferring, and harmonising populations and categories (Ruppert and Scheel, 2021).

Censuses, in this landscape, determine how populations are known and acted upon, shaping how a population (re)imagines itself as a distinct national community (Anderson, 2008). These categories help legitimise state action (Ruppert, 2008) yet remain historically contingent and politically negotiated. As Simon et al. (2015: 2) note, demographic classifications embed subjective judgements within ostensibly objective systems, ultimately serving as the foundation for policy formulation and contributing to the shaping of national identity.

As census-making becomes embedded in digital infrastructures, classification extends beyond enumeration into broader processes of data production and circulation. Through integration of datasets, automated systems, and human interactions, censuses engage in what Van Dijck et al. (2018: 3) call ‘making sociability technical’, where technology shapes social life by formalising, managing, and influencing it. These dynamics reshape how populations are understood and governed, making social relations legible and actionable through data. This is especially significant when classifications involve ethnoracial difference, as visibility, recognition, and vulnerability intertwine with technical and institutional counting practices.

Race, as Nobles (2000: 2) argues, is not a fixed attribute to be measured but a discourse produced through institutional practices such as the census. Ethnoracial categories acquire social force through recognition, yet derive particular authority from their inscription in state statistics (Loveman, 2014). As Benjamin (2019) shows, claims of technical neutrality lend these classifications legitimacy while obscuring the judgements and futures they encode. Datafication further narrows possibilities for contestation by prioritising infrastructural stability over lived circumstances (Ustek-Spilda and Alastalo, 2020). Together, this scholarship highlights that ethnoracial census categories are political instruments that organise visibility and invisibility and shape how national belonging is made and contested.

To fully grasp how classification practices influence citizenship and belonging, it is important to consider not only how populations are represented today but also how data practices organise relations between past, present, and future. Census-taking is inherently temporal: it draws on past classifications and memories, activates decisions in the present, and projects particular futures through its categories and infrastructures. Temporality is, therefore, not external to the census but enacted through it, as counting practices stabilise certain histories while opening or foreclosing future possibilities. In this sense, time operates as relational and situated, unfolding multi-scalar articulations rather than a linear sequence (Bear, 2016).

We approach temporality in census-making along two intertwining dimensions. First, censuses generate temporal vectors through time series, classification evolutions, and encoding protocols (Desrosières, 1998), shaping how populations are stabilised, compared, and reproduced over time (Ruppert and Scheel, 2021). Second, temporality matters insofar as censuses carry social imaginaries about the futures of numeric infrastructure as these are conceived, negotiated, and developed over time by communities of statistical producers and users. These imaginaries, in turn, influence how citizenship is experienced and governed.

To specify how ‘the future’ matters here, we build on scholarship that treats future visions not as mere rhetoric but as performative: they provide orientation in the present and help to construct the futures they invoke. Mager and Katzenbach (2021) argue that future imaginaries become analytically distinctive when they are collectively held and institutionally stabilised, and that in digital domains, such imaginaries are typically multiple, contested, and often commodified, rather than singular state-led trajectories. In census-making, this means tracking how competing actors seek to stabilise particular futures of population knowledge – through standards, governance rationales, and the material design of data infrastructures.

We call these competing yet complementary temporal dimensions of census making, data futurities. We use ‘data futurities’ as a mid-level concept that links reform discourse to infrastructural inscription. By this, we mean the ways in which statistical practices not only record populations but also configure expectations, orientations, and constraints around how societies will be known and governed in the future. We distinguish two dimensions.

First, the future in the census refers to the imaginaries through which expectations, promises, and fears orient (Bryant and Knight, 2019) the design, discussion, and implementation of census reforms. These imaginaries may appear as narratives of modernisation and inevitability, as anticipations of new forms of visibility or inclusion, or as institutional risk framings that justify postponement and caution. We examine this dimension through the longitudinal history of the (non-)collection of ethnoracial data in Portugal, showing how absences, inconsistencies, and deferrals have contributed to the sedimentations of specific understandings of citizenship and nation-making.

Second, the future of the census refers to the material infrastructures – such as administrative registers, interoperability standards, digital workflows, and auditing regimes – that condition which forms of enumeration remain possible, legitimate, or contested over time. Analytically, this distinction allows us to show how struggles over ethnoracial classification in Portugal are shaped both by competing future imaginaries (recognition, protection, participation) and by infrastructural shifts that can relocate classificatory decisions away from the census questionnaire and into other data systems, reshaping the terms of statistical citizenship and belonging.

Methods and materials

Our analysis draws on publicly available sources, including archival documents on Portuguese census practices and classification regimes; a reborn-digital media archive (Brügger, 2018) tracing public debates on the proposed inclusion – and eventual exclusion – of ethnoracial questions in the 2021 census; and official publications produced by members of the WG on Ethnoracial Questions and by statisticians involved in Portugal's transition towards a register-based census. We were not members of the WG, did not participate in its meetings, and therefore write as researchers examining official statistics as a political and infrastructural field rather than as practitioners within a national statistical office.

We view these materials not as ‘facts’ but as traces of practices shaping census knowledge: how actors justify, challenge, and negotiate categories and data infrastructure. Instead of analysing microdata, we focus on how institutional actors and publics defend classifications, invoke risks, and determine participation. This approach treats documents as social artefacts created for specific audiences, moments, and impacts, not neutral reflections of reality (Riles 2006; Ruppert and Scheel, 2021).

Following Kopper and Knox (2024), we attend to the affective and imaginative dimensions of datafication – how hopes, anxieties, and anticipatory narratives become folded into seemingly technical choices. Drawing on Bonilla-Silva and Zuberi's (2008) critique of ‘white methods’, we remain attentive to how ostensibly neutral statistical practices can reproduce dominant epistemologies and structural inequalities. Our aim is to analyse how census infrastructures and institutional rationales shape contested data futures, acknowledging the limits of a public-document corpus that captures what actors chose or could make visible, potentially under-representing debates and experiences outside official records.

The past of Portuguese censuses

Census-taking in Portugal dates to the 13th century, but became standardised in the mid-19th century under international statistical influences that emphasised uniform classification and individual-level data (Thorvaldsen, 2018). The first modern census was held in 1864, with regular decennial censuses starting in 1890. Until the 1970s, censuses encompassed both the metropolis and the colonies, operating within dual imperial logics: while enumeration at home aimed at national unification, colonial censuses relied on division, producing differentiated regimes of citizenship (de Castro Oliveira, 2012). Racial lines heavily shaped classification within Portuguese colonial rule. In Angola, census and population classification differentiated between the civilised ‘whites’ and the ‘blacks’ or ‘mestizos’, who had to demonstrate civilisational advancement through tests; successful ones could be promoted to assimilated status. No equivalent existed in the metropolis (Matos, 2013: 65). The 1890 census, following French models, categorised the population by sex, nationality, marital status, and education (Ministério das Obras Públicas, Comércio e Indústria, , 1896: VII).

Despite the absence of explicit racial distinctions in metropolitan censuses, ethnic and cultural diversity was not absent from the territory. Migrant presence in the metropolis remained limited (de Castro Oliveira, 2012), but Roma communities have been present in Portugal since at least the 15th century. Yet institutionalised forms of exclusion have long shaped their social positions. As Mendes and Magano (2022: 57) argue, Portuguese Roma (Ciganos ¹ ) have been constructed as the ‘internal stranger’, a position historically produced through persecution, forced assimilation, and persistent stigmatisation. In Portugal, the persecution of Roma people and the categorisation of Black populations – often through notions of ‘non-civilisation’ in colonial contexts – developed alongside the construction of the modern nation-state and the consolidation of the colonial empire (Araújo, 2019).

These dynamics align with broader Iberian patterns in which national identity has been reconstructed through narratives of modernity and European belonging (Rodríguez Maeso et al., 2025). As Reis and da Silva (2025) argue, the official denial of race as a relevant category in Portugal can sustain forms of racism by preventing recognition of how institutional practices marginalise non-white populations. Thus, rather than a narrative of failed integration, the history of Roma communities in Portugal exemplifies how state policies, classificatory regimes, and social imaginaries have produced forced assimilation (Mendes and Magano, 2022) and enduring boundaries of difference.

As the long-standing dynamics of differentiation shaped Portuguese society, census-taking also underwent important institutional transformations. A key turning point came in 1940, when INE (Statistics Portugal), established in 1935, published its first historical series. At the time, the institute's creation was described as ‘a new phase in the history of national statistics’, with the 1940 census ‘inaugurating a new series of Portuguese censuses grounded in emerging international standards’ (INE, 1940: 5). This shift marked a move towards greater centralisation, authority, and autonomy, principles that would guide the development of Portuguese official statistics in the decades that followed. At the time, INE acknowledged the existence of colonial censuses but treated them as administratively fragmented, given the autonomy of different colonies. Nonetheless, racial data continued to be collected in colonial contexts, including Cape Verde, Guinea, Angola, and Mozambique. Only in the second half of the 20th century did international census standards begin to shift, particularly regarding the treatment of race and ethnicity (Simon et al., 2015).

Approaches to ethnoracial data collection diverge markedly across countries with colonial histories. In the United Kingdom, the census progressively expanded ethnic questions from 1991 onward, moving from non-counting in the name of multiculturalism to counting in order to justify positive action (Thompson, 2015). By contrast, France prohibits questions on race and ethnicity, relying instead on nationality and country of birth to distinguish citizens, foreigners, and immigrants (Simon, 2015). Spain similarly avoids ethnoracial classification (Rodríguez Maeso et al., 2025). These contrasts underscore that census categories evolve along broader social, political, and demographic transformations – and that refusals to classify are themselves political and historically situated.

Portugal was no exception. The latter part of the 20th century saw significant transformations in the structure of Portuguese society, reshaping how official statistics addressed migration and citizenship. As migration from former colonies increased, demographic composition shifted, and new debates about citizenship and rights emerged. In the early 1950s, the state introduced nationality criteria aimed at promoting racial and cultural homogenisation and facilitating assimilation (de Castro Oliveira, 2012: 136). From the 1980s onward, immigration policy oscillated between restrictive controls and periodic regularisation measures, reflecting the growing perception of immigration as a social and political challenge as Portugal emerged as a destination country.

A closer look at how the census classified ‘foreigners’ during this period is instructive. In 1960, the census collected detailed information on resident foreigners, including nationality, length of stay, occupation, and socio-economic conditions (INE, 1960). By 1980, however, this detail was substantially reduced: only basic variables such as age, nationality, and profession were retained (INE, 1981b: 157). Although tables cross-classifying foreign residents by occupation, age, and economic activity were produced, they were never published, with INE citing financial constraints and limited perceived relevance (INE, 1981a: 7).

Between the 1990s and early 2000s, migrants’ rights expanded, influenced by European directives promoting anti-discrimination and political participation of foreign residents (de Castro Oliveira, 2012: 154). These legal changes mirrored census practices. The 1991 census added detailed data on nationality, birthplace, residence, and migration flows, allowing multiple nationalities (INE, 1995). The 2001 census kept nationality and birthplace data but improved accessibility by translating the questionnaire into Russian, reflecting new migration waves (INE, 2003). The 2011 census integrated traditional methods with digital collection via e-Censos, which accounted for half of the responses. It translated questionnaires into Russian, Romanian, Chinese, and English to include diverse migrants (INE, 2013). Despite procedural innovations and demographic diversity, ethnoracial data remained excluded, continuing Portugal's reliance on migration-based proxies for difference.

Góis and Marques (2018) view this 30-year period as crucial for the statistical visibility of citizenship in Portugal. That visibility is, nonetheless, conditional, arising from ‘a process of extraordinary regularisation demonstrating the reactive logic of Portuguese migration policies and their subjugation to independent logics originating in the national labour market’ (Góis and Marques, 2018: 131). We emphasise that such conditional visibility affects primarily those categorised as ‘migrants’. Yet by this period, many immigrants had acquired Portuguese citizenship, and many of their descendants had been born on Portuguese soil. ² Roma communities, by contrast, are composed primarily of Portuguese citizens and cannot be understood through a migration lens, even as they remain subject to durable forms of institutional marginalisation.

Moreover, as Scheel and Ustek-Spilda (2019) argue, migration management relies on a politics of expertise and ignorance that renders migration governable by obscuring the limits of quantification. Through practices such as omission, compression, and sanitisation, migration statistics sustain authoritative narratives even when data are partial or inconsistent. As a result, increased statistical visibility of migration does not translate into ethnoracial visibility, nor does it address deeper classificatory absences in Portuguese census infrastructures. These dynamics are central to our argument about data futurities. The continued use of migration proxies, the repeated deferral of ethnoracial classification, and the gradual expansion of digital and administrative infrastructures do not merely describe what is counted; they shape expectations about what can be counted and governed. Over time, census categories and infrastructures sediment particular horizons of recognition and participation, relocating classificatory decisions and narrowing the space for contestation.

In 2018, tensions over visibility, classification, and census limits led to the formation of a WG, sparked by public mobilisation, to discuss including ethnoracial categories in the 2021 Portuguese census.

The future in the censuses

As census classifications increasingly influence how futures of citizenship and belonging are envisioned, they have become central to institutional and public debates. Scholars show that census practices arise from historically contingent assemblages where the state is one actor amongst many (Bartl et al., 2024: 7). Categories and systems are not merely imposed but are enacted, contested, reworked, or resisted through everyday practices and institutional frictions (Ruppert et al., 2018; Ustek-Spilda and Alastalo, 2020). It is against this backdrop of contested classification, expanding sociotechnical infrastructures, and growing demands for participation that the Portuguese WG Census ‘Ethnoracial Issues’ 2021 was established. Despite repeated adjustments to census methods, Portugal has maintained persistent ambiguity about the need and potential consequences of collecting ethnoracial data, marked by disagreements over categories, uncertainty over governance and safeguards, and uneven political commitments to diversity and equality agendas.

Census-making is inherently future-oriented: preparations begin years before enumeration and involve projecting which categories, instruments, and methodological choices will remain meaningful and legitimate over time. The pre-enumeration period, therefore, becomes a critical site of negotiation amongst expert users, institutional actors, and organised publics. In this sense, the creation of a WG in 2018 to discuss the potential inclusion of ethnoracial data is not in itself unexpected. What was unprecedented, however, was the scale and visibility of the mobilisation it generated, signalling that struggles over census categories had become struggles over future recognition, participation and statistical citizenship.

The WG for the 2021 Census on ‘Ethnoracial Issues’ was created by Order No. 7363/2018, referencing UN recommendations to gather ethnic origin data aligned with Portugal's laws. Its goal was to recommend how the census, including its questionnaire, could describe Portugal's ethnoracial composition. Starting in August 2018, the group met over 18 months and submitted a report to the Secretary of State for Citizenship and Equality. The WG should be understood as the outcome of a broader political and social process rather than an isolated governmental initiative. For a long time, public debate on racism in Portugal was largely confined to activist and academic circles. This began to shift in the mid-2010s amidst public denunciations of racist police violence and intensified social and political mobilisation, including a public letter signed by Afro-descendent organisations and addressed to the UN Committee on the Elimination of Racial Discrimination. As Marta Araújo (2019), one of the scholarly members of the WG, noted, demands to make structural inequalities statistically visible increasingly framed disaggregated ethnoracial data as a tool for enhancing transparency and public accountability within the Portuguese state.

The composition of the WG included members from different sectors: INE officials; representatives linked to migration and equality governance (including the High Commission for Immigration and the Observatory of Migration); academics from Portuguese research institutes and universities; and representatives of NGOs and social movements. ³ Participation was unpaid and entailed a lengthy agenda of meetings to assess the state of data production and deliberate over possible classificatory models.

Discussions encompassed the feasibility of data collection; alternative models for categorising origins and/or ethnoracial belonging that could be integrated into censuses; the design and presentation of focus group results with different communities (including Afro-descendants, Roma, and immigrants) discussing the relevance and risks of these models; presentations of studies and surveys on collecting ethnoracial data; internal deliberations; and the drafting of recommendations. Ultimately, despite majority (though not unanimous) support within the WG for including ethnoracial questions in the census, INE – invoking principles of technical and professional independence – opted not to include these variables in the 2021 Census.

Focusing on Afro-descendant collectives and Portuguese Roma associations, this section analyses how different community actors articulated competing claims about ethnoracial visibility, participation, and data control in the 2021 census debates, revealing divergent expectations about what future census-making should enable or constrain.

Advocacy for collecting ethnoracial data was consistently evident in the positioning of Afro-descendant collectives engaged in the WG's debates. Their interventions emphasised that collecting and analysing such data was essential to support affirmative public policy capable of addressing structural racism in Portugal. Community representatives repeatedly stressed that anti-racist policy agendas should not be reduced to migration frameworks, since Afro-descendants should be recognised as ‘Portuguese citizens’ entitled to active participation. In a public open letter to O Público in February 2018 (included in the WG's annexes), the collectives argued that: ‘Data collection can be a tool in the service of ethnoracial equality, but only if it results from the active participation of those who have no voice and no statistical footprint’ (GT Censos, 2019: 166).

In an opinion article published in Público prior to INE's final decision, WG members Roldão et al. (2019) argued that including ethnoracial data in the 2021 census would represent an important step in addressing racism in Portugal. They framed census inclusion as an opportunity to regulate and make transparent existing practices of ethnoracial identification already occurring across public institutions such as policing, healthcare, and education. They also stressed the need for broad public awareness efforts and for institutions dedicated to combating racism and xenophobia beyond migration governance, while emphasising that racialised communities and anti-racist organisations should participate meaningfully in the process, including in decision-making roles.

A central question raised in these debates concerned how those described as having ‘no voice and no statistical footprint’ could participate in defining future census practices. One scenario discussed in the WG involved fostering public understanding – within a population largely unfamiliar with ethnoracial self-classification – of the meaning and implications of open-ended or self-identification questions. While such learning could be cultivated over time through schools and public communication, in this imagined future, the pedagogical role would also be carried by census takers administering questionnaires through household visits.

In focus group sessions with Afro-descendant communities in Lisbon, presented at the WG, participants emphasised the need for more census takers trained to address these issues and for greater ethnoracial diversity amongst interviewers and coordinators. It was argued that diversity amongst census staff could enhance participation, strengthen legitimacy, and reduce miscommunication during enumeration amongst marginalised communities. Participants also recommended involving technicians from relevant institutions and collectives of racialised individuals in analysis and dissemination, ensuring that teams responsible for interpreting results reflect the population's ethnoracial diversity (GT Censos, 2019: 87).

Here, census takers are imagined as key mediators – actors who do not simply transmit fixed meanings but transform, translate, distort, and modify the meanings or elements they convey (Latour, 2016: 65). As mediators, census takers can help render categories intelligible, shape perceptions, and facilitate (or hinder) trust (da Cruz Freitas, 2018; Kopper and Duarte, 2024). If census takers were chosen with diversity in mind and trained in a methodology that takes into account the intersectionality of race and ethnicity, the argument went, data collection could be transformed into an educational activity – one in which the Portuguese population would ‘learn’ to self-represent themselves along the lines of race and ethnicity. In this future-oriented vision, census takers would act as conceptual mediators, contributing to the standardisation of categories and their use across fault lines of race and ethnicity amongst an informed, educationally engaged, and informationally aware citizenry. Nevertheless, this imaginary raises a critical question for Portugal census futurities: what happens to these pedagogical and mediating functions as census-making shifts towards register-based models that reduce or eliminate the role of census takers and relocate classificatory decisions into administrative infrastructures?

Portuguese Roma associations and community leaders positioned themselves on the other end of the spectrum, underscoring risks of racialisation, misuse, and loss of control. In an official statement published in 2019, they warned that such data could be misused – particularly in contexts of rising far-right mobilisation – and could deepen stigma and persecution rather than enable anti-racist policies. As they argued: ‘the collection of “ethnoracial” data by the state (…) and its ownership by the various state bodies, can lead to increased discrimination and racism, as well as the persecution of the Roma population, who are socially and geographically fully characterised and “registered”’ (GT Censos, 2019: 171).

This concern was echoed in public interventions by Roma activists during the same period. In an interview with Expresso, Bruno Gonçalves, member of the association Letras Nómadas, explained that the community's opposition stemmed from profound distrust of data protection mechanisms and from a historical experience of being treated as outsiders despite centuries of presence in Portugal (Martins, 2019). From this standpoint, fears surrounding ethnoracial data were not abstract; they were grounded in vulnerability and in the anticipation that such data might structure future forms of surveillance and governance rather than operate in Roma communities’ favour.

When assessing proposed survey models, WG participants often framed ethnoracial data collection as double-edged. On the one hand, they cited potential benefits such as non-selective and non-negative visibility and the ability to evaluate public measures targeted at Roma communities. On the other hand, they warned that a mandate to publish ethnoracial data could intensify and reinforce stereotypes (for instance, by associating Roma communities with social subsidies); further, ‘if it were mandatory, it would not allow ethnic identity to be concealed (“ethnic clandestinity”)’ (GT Censos, 2019, 92).

‘Ethnic clandestinity’ is used by some Roma activists, such as Bruno Gonçalves, to describe individuals who do not publicly acknowledge their ethnic identity in everyday life to avoid prejudice and expand opportunities (Pereira, 2019). Consequently, despite the varied economic and social conditions within this community, the prevailing public image associates them with poverty, exclusion, and nomadism. This reflects an ethnocentric and romanticised interpretation that conceals individuals from different social classes living at various stages of sedentariness, who are often referred to as ciganos invisíveis (‘invisible Portuguese Roma’) (Mendes et al., 2019: 81–82).

These tensions foreground how debates over ethnoracial data are also debates over control over the future use of data. On this view, the key question is not only whether ethnoracial data should be produced, but under what conditions communities can shape how such data are collected, secured, interpreted, and mobilised. Here, claims about data ownership are less about commodified property rights than about future-oriented agency: who can authorise uses, contest interpretations, and prevent harmful applications over time.

For Hummel et al. (2020: 568), data ownership concerns self-determination: it entails securing the prerequisites needed to preserve self-determination during digitalisation and data analysis processes. Despite the existence of multiple (and sometimes contradictory) interpretations, a shared thread suggests a concern with controlling how data is used. In the Portuguese case, such concerns can be read as demands for collective capacity to decide whether ethnoracial data should be collected at all, how it should be safeguarded, who may access it, how it will be analysed, and whether it can support community-led initiatives rather than only state-led governance.

The authors also stress that data ownership is not reducible to technical safeguards; it is a matter of social and political negotiation (2020: 566). Claims to data ownership, therefore, express demands for both material rights and recognition as legitimate actors within data-driven processes, including the ability to shape institutional accountability. In this sense, informational self-determination depends not only on privacy protections but on meaningful participation in how data infrastructures are designed, governed, and audited, impacting people's lives.

These concerns align with broader critiques of data-driven governance. Hintz (2022) argues that black-boxed data processes limit opportunities for contestation and undermine participation, positioning citizens as recipients of data-based decisions rather than as co-creators of the systems that govern them. Read alongside Roma resistance to ethnoracial data collection, data ownership becomes a claim about futurity: who gets to shape the trajectories through which population data are transformed into policy, surveillance, or redistribution – especially as census-making is reorganised through administrative registers and automated workflows.

Seen through the lens of data futurities, the WG process makes visible two intertwined dynamics. First, competing imaginaries of the future in the census were articulated: for Afro-descendant actors, future-oriented equality depended on recognition through counting; for Roma actors, future-oriented safety depended on limits, safeguards, and the possibility of non-disclosure. Second, these imaginaries were negotiated in anticipation of the future of the census, with its impending shift towards register-based infrastructures, away from participatory enumeration and into dispersed administrative systems, where contestation may be harder, and where control over downstream uses becomes more diffuse.

As we have seen, the WG issued a majority, but not unanimous, recommendation in favour of including ethnoracial categories in the 2021 Census; the final decision rested with INE. In its official position, INE rejected the inclusion of ethnoracial questions, arguing that such data could not be collected with sufficient quality and that their inclusion posed risks to the census as a whole (INE, 2019). This decision did not occur in an institutional vacuum. At the EU level, the European Commission has issued non-binding guidance encouraging the voluntary collection and use of equality data based on racial or ethnic origin to address discrimination, under strict data-protection safeguards (European Commission, 2021). At the same time, this guidance coexists with the absence of harmonised census standards on ethnoracial data within the European Statistical System. The Portuguese case thus illustrates how national statistical authorities navigate competing institutional logics: emerging European equality-data agendas on the one hand, and concerns over data ownership, governance, and public trust on the other. In this context, the rejection of ethnoracial variables can also be understood as a recalibration of census legitimacy amidst datafication, heightened scrutiny, and anticipated futures of register-based counting.

It is also important to note that neither the WG's recommendation nor INE's final decision satisfied the communities involved. Representatives of Roma associations opposed the recommendation, arguing that questions framed around ethnoracial affiliation, rather than nationality, risked reinforcing stigma (Henriques, 2019a). Conversely, Afro-descendant representatives expressed deep dissatisfaction with INE's decision to exclude such questions. Researcher Cristina Roldão, for example, stated publicly that the decision perpetuated Portugal's failure to recognise itself as a multicoloured society and to acknowledge the persistence of individual, institutional, and structural racism (Henriques, 2019b).

Against this backdrop, the adoption of the National Plan to Combat Racism and Discrimination (2021–2025) marked a symbolic shift by formally recognising structural racism as a legacy of slavery and colonialism (Reis and da Silva, 2025: 537). Simultaneously, INE launched a pilot survey on living conditions, origins, and trajectories of the resident population (ICOT/INE, 2023). Rather than solving the controversy, this initiative reframed ethnoracial visibility as a methodological and explicitly future-oriented issue, deferring high-stakes classification to pilot-based experimentation beyond the census. Whether and to what extent this move represents an inflexion in Portugal's ethnoracial data regime – and the data futurities enacted through the production of such numbers – remains an open question.

Despite differing views on collecting ethnoracial data, a common theme links Afro-descendant and Portuguese Roma arguments: the demand for meaningful participation in statistical representation and governance. As Portugal shifts towards administrative registers and automated data, questions about participation, accountability, and collective control in a datafied regime remain unresolved. From a data futurities perspective, the key issues are not only what will be counted but also where classificatory decisions are made, who can contest them, and how communities can influence how population data becomes policy, stigma, or rights.

The future of the census

Population census planning is an ever-evolving process in which statisticians continually refine categories, methodologies, collection protocols, budgets, timelines, and dissemination strategies. By the 1970s, most European countries had consolidated traditional enumeration models, but growing pressures to produce statistics more frequently, reduce costs, and modernise data infrastructures increasingly pushed various statistical authorities to explore alternatives to conventional census operations (d’Alva and Paraná 2024; Ruppert and Scheel 2021). Coleman (2012: 336) identifies several drivers behind this shift, including the high costs of field-based enumeration, particularly the need to employ a large temporary workforce (such as census takers and managers), and growing difficulties in securing public participation. Taken together, these developments reflect not only methodological transformations but also shifts in institutional practice and political culture (including the public's views of censuses) that help explain the growing appeal of register-based models.

Register-based censuses are not new. In the early 1980s, Denmark implemented a population census based entirely on administrative data. In UNECE terms, a register-based census is an approach in which no population data are collected directly from respondents; instead, traditional enumeration is replaced by the use of administrative data held in various registers (such as population registers, building/address registers, and social security registers), linked through matching procedures, typically via personal identification numbers (UNECE, 2018: 2). While this model is often presented as cost-efficient because it reduces fieldwork and temporary labour, it also entails substantial and ongoing investments in data storage, integration, and processing infrastructures. Moreover, its effectiveness depends on the quality and coverage of administrative registers and on governance arrangements that enable linkage and access – conditions that may reproduce existing gaps, reduce opportunities for public participation, and shift statistical authority away from respondents towards administrative systems.

For a few years now, INE has researched how to implement such a system in Portugal, aiming to complete a full transition for the 2031 census. The Portuguese strategy for the 2021 Census envisaged a combined approach in which administrative data would be used to provide information on specific census topics, such as place of residence, country of birth, and country of citizenship. The aim was to align INE with broader trends in EU and UNECE countries towards a streamlined census method that could uphold high-quality standards while reducing some operational burdens associated with field enumeration (Lagarto et al., 2017: 749).

Portugal's extended reliance on traditional census models reflects structural limitations in its administrative data infrastructure rather than institutional or methodological resistance. As explained by INE statisticians (Araújo et al., 2024), the country lacks two key preconditions for a fully register-based census: an administrative population register and a unique personal identification number enabling consistent linkage across databases. Although multiple administrative sources exist (such as civil identification, tax, and social security records), these systems were developed for distinct purposes, use different identifiers, and present coverage and accuracy limitations, including systemic overestimation of the resident population. In this context, the traditional census has continued to function as a repository of past classificatory practices and as a benchmark through which future census infrastructures are validated, calibrated, tested and legitimised. Both the 2011 and the 2021 Censuses thus marked a transitional moment, positioned between inherited models of enumeration and prospective register-based designs, enabling cautious experimentation while safeguarding data quality and methodological reliability.

Comparative experiments combining administrative sources with 2011 census data suggested a strong potential for register-based transition, given the high levels of consistency observed between datasets. For example, matching individual administrative records with seven demographic variables produced a 90% equality rate (Lagarto et al., 2017, 753). Still, statisticians highlighted persistent structural constraints, notably limited control over administrative data held by other institutions and gaps in key variables. Preliminary assessments identified restricted access to certain registers, incomplete coverage of household and family relations and limits to producing data at detailed geographical scales as major obstacles to a full transition in 2021 (INE, 2020; Lagarto et al., 2017: 752). These constraints underscored the continued reliance on traditional enumeration as a stabilising reference point during the transition period.

More recent analyses from INE further position the 2021 Census as a benchmarking exercise. Drawing on comparisons between the Resident Population Database and the 2021 Census, Araújo et al. (2024) show a high degree of convergence between the two sources: 92% of individual records matched. and similarity rates exceeded 97% for demographic variables such as age, sex, citizenship, and marital status. Still, the analysis identified persistent coverage biases, particularly affecting non-nationals and non-economically active populations, linked to limitations in matching attributes, reference dates, and the application of signs of life criteria (administrative indicators used to infer whether a person is actively residing in the country based on traces of interaction with state systems). This fact underscores how register-based censuses may risk deepening citizenship-based asymmetries in statistical visibility, particularly as enumeration gives way to silent forms of administrative data collection.

Data limitations present a significant challenge, as administrative data alters the usual supply-and-demand dynamics of statistical production. In thinking about the future of censuses in Portugal – and particularly in light of community demands regarding the collection of ethnoracial data – a central concern is what Coleman calls ‘silent data-gathering’: administrative sources provide numerous opportunities for silent data collection, with minimal citizen involvement or awareness, potentially resulting in data that is more complete and comprehensive than any census could achieve (Coleman, 2012: 345).

Such ‘silent’ data collection stands in tension with calls for active participation in future census-making that we observed in the WG debates. Regardless of whether future Portuguese censuses collect and analyse ethnoracial data, the ‘silent’ collection of demographic information may hinder public participation in both data collection and analysis, as urged by Afro-descendant actors. It may also undermine claims for greater control over data ownership and even the capacity to refuse data collection, as advocated by Portuguese Roma representatives.

This tension becomes even more pronounced under register-based census systems. As Grommé and Scheel (2021) show, the shift from questionnaire-based censuses to register-based statistics transforms how individuals are assigned to statistical categories. While traditional censuses rely on self-identification (as proposed by the WG in Portugal), register-based approaches allocate individuals to categories through administrative data practices such as inferring or assigning. In this model, classification is grounded less in citizens’ self-definition and more in institutional interpretations of existing records, often spanning generations. These categorisations do not merely describe populations; they also enact broader notions of national belonging and citizenship, particularly by differentiating migrants and minorities. In the Portuguese case, where demands for ethnoracial visibility, data control, and the right to refuse enumeration have been central to public controversies and future imaginaries, this shift risks producing census futurities in which participation is further displaced, citizenship is enacted through opaque infrastructures, and decisions about belonging are progressively removed from public debate and collective negotiation.

In terms of participation in census production, Kopper (2023: 5) remarks that the population census questionnaire can be viewed as a co-produced and continually evolving historical artefact, shaped concurrently by national histories of identification and population-making, as well as by technical conventions and decisions made at the level of transnational organisations. In this sense, the questionnaire becomes a key site where past classificatory logics and future expectations converge, as planners endeavour to minimise and anticipate informational needs and stabilise categories oriented towards future governance.

In register-based censuses, however, the role of these actors – conceptualised here as mediators in the Latourian sense – would be significantly altered. While workloads may be reduced and analytical outputs accelerated through access to larger (and sometimes underutilised) administrative databases, active participation in census-making becomes displaced in time and space. In such scenarios, ethnoracial data would be collected by administrative institutions with their own mandates through technicians who might lack the specific training or situational mediation required to negotiate meanings with respondents, reshaping not only census practices but also the future conditions under which statistical citizenship is enacted.

This point is consequential. The shift towards a register-based census reshapes where and how decisions about inclusion, exclusion, and categorisation are made. As Ustek-Spilda and Alastalo (2020) show, in automated and register-based systems, such decisions are frontloaded at the moment of data entry and embedded in administrative procedures rather than negotiated through interaction or revisited through analysis. Because administrative procedures are designed to serve institutional logics (rather than primarily statistical or social ones), they may reflect procedures more than lived circumstances, often reproducing partial, fragmented, or misaligned representations, especially for marginalised groups. Here, statisticians do not disappear; they become back-office policy actors, exercising discretion through technical decisions that adapt abstract guidelines to national infrastructures (Ustek-Spilda, 2020).

For Portugal, where current register-based experiments already reveal coverage biases affecting non-nationals, this raises critical questions for future census-making. In the absence of mechanisms for self-identification, mediation, or refusal, administrative categorisation risks consolidating forms of belonging and exclusion that become difficult to scrutinise publicly. Such developments place the conditions of statistical citizenship at stake. Rather than operating as sites of negotiated citizenship, censuses risk becoming infrastructures where political decisions about belonging are made silently and in advance, reshaping the futurities imagined by communities demanding visibility, participation and control.

Following Hannah (2001: 526), statistical citizenship hinges on gaining deliberate influence over how individuals are stratified and clustered within statistical systems. Crucially, Hannah suggests that the act of data collection can itself be politically consequential: when individuals are made aware of how their information is gathered and used, they may be empowered to intervene in its management. Thus, participation in statistical processes carries the potential to destabilise the historical hierarchical politics of statistical knowledge. Statistical citizenship, therefore, concerns not only visibility in data, but also the possibility of reshaping the power relations embedded in data by insisting on active, participatory citizenship.

Here, it is useful to distinguish being an actively participating citizen from being an active citizen. Active citizenship can imply an individualistic liberal-democratic vision in which collective decision-making is channelled through scripted forms of political participation. In such contexts, actors operate on existing political stages through standardised practices. They can propose policy changes, but there is no guarantee these issues will lead to effective changes, or even that state authorities will formally consider them (Saward et al., 2013: 228). Certainly, the public's influence on government decisions, even if consultative and limited, may nonetheless enhance policy adequacy. However, for effective social control, active participation must be ensured at all stages, from decision-making and planning to implementation, monitoring, and evaluation, as statistical citizenship implies.

In this model, avenues for participation should be deliberative, oriented towards more radical democratic practice; the population contributes to policy formulation, and the state integrates these decisions into its actions, communicates technical constraints, and manages resources transparently. Therefore, ensuring active participation requires integrating top-down and bottom-up infrastructures in the production of official statistical data, commonly seen as a pillar of democracy. While register-based censuses promise efficiency and continuity, they also complicate this balance by potentially distancing communities from the process through which data about them are produced and mobilised.

To reflect on the future of ethnoracial data in Portugal, the metaphor of the ‘entrance door and the exit door’ (Cruz Freitas, 2018) is instructive. The entrance signals transition, possibility, and the opening of a new phase; the exit – often experienced as refusal, denial, or failure – also marks transformation and the beginning of new challenges. The debate on ethnoracial data entered the census arena after a long history of colonial classification, racial silence, and community mobilisation, generating expectations of change through negotiated inclusion. The final decision not to include ethnoracial data in the Portuguese census can be read as an exit – one that left many community demands only partially addressed. Yet this exit coincides with another structural transition: the gradual move towards a register-based census. In this emerging configuration, decisions about classification, assignment, inclusion, and belonging risk being displaced from participatory encounters to administrative infrastructures with limited public visibility.

Seen in this light, the future of ethnoracial data in Portugal will be shaped not only by whether such variables are included, but by where and how decisions about population categories are made. A central question thus emerges: how can register-based census infrastructures sustain statistical citizenship rather than foreclose it? As Lane (2016) reminds us, citizenship concerns who is included, who is excluded, and who has a voice in shaping those terms. The exit door of Census 2021 may therefore open a new challenge: ensuring that future census models do not turn citizenship into a silent administrative outcome, but remain open to democratic negotiation, recognition, and collective claims to data.

Conclusion

Population censuses are often portrayed as static snapshots of society. Our analysis instead shows that censuses operate as dynamic infrastructures, connecting past classifications, present practices, and imagined futures. Through the lens of data futurities, we highlighted how census-making simultaneously embeds anticipatory projections about social order (the future in the census) and reshapes the material and institutional conditions of enumeration (the future of the census).

Focusing on Portugal's 2021 census, we examined how shifts towards register-based models intersect with long-standing demands for ethnoracial visibility and participation. We argued that technological changes, while often framed as progress, do not automatically expand democratic possibilities. Rather, it may reconfigure statistical citizenship by relocating decisions about categorisation, recognition, and accountability into administrative and technical domains, potentially narrowing opportunities for negotiation, self-identification, and refusal.

Our findings also point to important directions for further research. This article does not offer a detailed legal analysis of data protection and statistical governance, nor does it examine other administrative sites where ethnoracial data may be produced or inferred beyond the census. In addition, while we reference the pilot Survey on Living Conditions, Origins, and Trajectories (ICOT), its design, reception, and capacity to respond to community demands require systematic evaluation. Future research should also examine the uneven conditions under which different groups can influence statistical infrastructures and practices of participation within register-based census-making. Taken together, these gaps underscore that the future of census-making in Portugal hinges not only on technical change, but on how evolving data infrastructures reshape the effective – and affective – conditions under which diverse communities are included in the production, governance, and consequences of official statistics.