Denominational Differentiation and Religiosity Among the Hungarian Minority of Transylvania: Evidence from the European Values Study

Abstract

The Hungarian minority of Transylvania comprises four historically received denominations—Roman Catholic, Reformed, Unitarian, and Lutheran—whose institutional profiles differ markedly despite their shared function as carriers of minority cultural identity. Using the European Values Study 2017 Romanian Hungarian minority oversample (GESIS ZA7550; $N=1106$), this article presents the first regression-based analysis of intra-community denominational variation in religiosity in this dataset. Four binary logistic regression models test whether denomination independently predicts church attendance, confidence in church, subjective importance of religion, and self-described religiosity type (institutional versus personalised), net of sociodemographic controls. Catholics attend services significantly more frequently than Reformed members, while Reformed members express higher confidence in their church—a practice–trust reversal explicable by the distinction between canonical obligation and ethnic embeddedness. Subjective religious importance does not vary by denomination, consistent with an identity-protection mechanism operating uniformly across confessions. Denomination does not independently predict institutional versus personalised religiosity type once sociodemographic controls are applied, with age emerging as the dominant axis of variation on this dimension. The findings engage with Davie’s believing/belonging/behaving framework and the debate on whether denominational cleavage or the secular–religious divide constitutes the primary axis of religious differentiation in contemporary Europe.

1. Introduction

The Hungarian minority of Transylvania occupies a singular position in the comparative sociology of religion and ethnicity in Central and Eastern Europe. As a community of approximately one million people living as an ethnic and religious minority within an Orthodox-majority state, it has long attracted scholarly attention for the unusually strong connection between religious affiliation and ethnic identity (Kiss 2018; Kiss et al. 2022). Less noticed, however, is a feature which makes this community analytically exceptional: unlike most minority communities in the region, Transylvanian Hungarians are internally divided across four historically established denominational traditions—Roman Catholic, Reformed (Calvinist), Unitarian, and Lutheran—whose institutional characters, canonical expectations, and relationships to Hungarian ethnic identity differ in theoretically significant ways.

Existing quantitative research has examined the community’s religiosity primarily through a comparative lens, situating Transylvanian Hungarians between Romanians and Hungarians in Hungary on aggregate measures of religious commitment (Kiss et al. 2022; Zsigmond and Attila 2022). This approach has established the community’s distinctively elevated religiosity—consistent with a minority identity-protection dynamic—but it has necessarily treated the community as a homogeneous bloc, leaving open the question of whether denomination shapes the form of religious expression within it. The present article addresses that question directly, using the 2017 European Values Study Romanian Hungarian minority dedicated oversample (GESIS ZA7550; DOI: 10.4232/1.13562) to estimate the independent effects of denomination on four religiosity outcomes, net of sociodemographic controls.

The analytical argument advanced here is not self-evident, and it is worth stating at the outset what it is competing against. A prominent strand of European comparative sociology of religion maintains that denomination carries little independent explanatory power once the fundamental cleavage between the practising believer and the secular individual is controlled. Bechert (2018), in a cross-national EVS-based analysis of religiosity profiles, and Stegmueller et al. (2012), examining welfare preferences across Western Europe, converge on the conclusion that the secular–religious divide, rather than confessional affiliation, constitutes the dominant axis of religious variation. If this finding generalises to minority communities, denomination would be expected to leave no residual imprint on individual religiosity once the intensity of religious commitment is controlled. The present article takes this null hypothesis seriously while arguing that its scope is limited by the specific configuration of the Transylvanian Hungarian case. Where religious affiliation functions simultaneously as a spiritual commitment, an ethnic boundary marker, and a vehicle of minority identity-protection, denominational distinctions may operate selectively—shaping the form of religious engagement—its behavioural expression and institutional orientation—while an identity-protection mechanism operating at the community level suppresses denominational variation on the intensity of religious commitment. Testing whether this qualified claim holds requires a framework that treats the multiple dimensions of religiosity as distinct outcomes rather than interchangeable indicators of a single latent variable.

Three specific contributions are claimed. First, this article presents the first regression-based analysis of intra-community denominational variation within the ZA7550 dataset, filling a gap left by prior descriptive and comparative work which, however valuable, has necessarily treated the Transylvanian Hungarian community as a homogeneous religious bloc in order to situate it between the Romanian majority and the Hungarian national population. Second, it introduces the country-specific ZA7550 religiosity-type variable (RO_hu_cs25), which distinguishes institutional from personalised religious self-description, as a dependent variable in multivariate regression analysis for the first time; prior published uses of this variable have either employed it as a predictor in models of non-religious outcomes (Veres 2024) or have not engaged with it in regression form at all. Third, by demonstrating that denomination independently predicts some religiosity dimensions but not others, the analysis offers a replicable template for disentangling the effects of denominational structure from those of minority status wherever comparable oversample data are available in the European Values Study or related cross-national archives.

The article proceeds as follows. Section 2 situates the analysis in the historical and sociological context of the Transylvanian Hungarian community, characterises each of the four historical denominations, and reviews the existing empirical literature on religiosity within this population. Section 3 develops the theoretical framework—drawing on Davie’s (1994) believing, belonging, and behaving model, alongside the debate on denominational effects initiated by Bechert (2018) and Stegmueller et al. (2012)—and derives four testable hypotheses. Section 4 describes the data source, variables, and analytical strategy. Section 5 presents the results of the four logistic regression models. Section 6 discusses the findings in relation to the hypotheses and the broader theoretical debate, and Section 7 concludes with a summary of contributions, limitations, and directions for further research.

2. Background: Religion, Ethnicity, and Denomination in Transylvania

2.1. The Ethnic–Religious Nexus in Post-Communist Romania

The intertwining of religious and national identity in Romania is neither an accident of history nor a residue of pre-modern thought awaiting eventual dissolution by the forces of secularisation. It is, rather, the durable product of a nation-building project which, across the nineteenth century and well into the twentieth, made Orthodoxy the confessional marker of Romanian nationhood with a thoroughness that few European cases can match. The theologian’s blunt formula—“if you are born Romanian, you are born Orthodox”—may not constitute a sociological proposition, yet it encapsulates a cultural equation that carries real weight in contemporary Romanian identity discourse (Guglielmi and Piacentini 2024; Müller and Rosta 2025).

Analysis of the 2017 Pew Research Center survey data for Romania confirms the persistence of this equation: no fewer than 74% of Romanian respondents considered being Christian an important component of truly being Romanian, a figure that places the country among the highest in all of Europe on this measure (Pew Research Center 2017). Modernity, EU membership, and the emergence of a secular civil society have complicated but not dissolved the Orthodox-national nexus, as Guglielmi and Piacentini (2024) and, more recently, Müller and Rosta (2025) have confirmed in their comparative studies of Central and Eastern European countries, finding that ethno-religious conceptions of nationhood remain distinctively entrenched in this part of the continent, with nationalist forms of national identity particularly prevalent in Orthodox-majority contexts.

What makes the Romanian case particularly instructive for the study of religious minorities is the mirror logic it generates. Where the majority uses its confession as a national badge, the religious traditions of minority communities become, by the same logic, markers of ethnic distinctiveness—not merely spiritual homes but institutional embodiments of a cultural identity under permanent symbolic pressure. This dynamic takes a concrete and observable form in Transylvania, where the denominational landscape is deeply ethnically coded: the Romanian Orthodox Church is the majority church, Roman Catholicism and the Reformed faith are implicitly Hungarian, and the Lutheran tradition is associated, in the popular imagination, with the Transylvanian Saxon community—the very community whose dramatic emigration to West Germany in the years surrounding 1989 has reduced its local presence to near-invisibility. In mixed-ethnicity localities, confessional and ethnic boundaries reinforce one another, communities worshipping separately, schooling their children separately, and consuming distinct media worlds—a parallel society structure in which the historical churches of the Hungarian minority serve as the primary institutional carriers of a bounded identity (Kiss 2018).

Kiss et al. (2022) have provided the most systematic empirical demonstration of the identity-protective function of religion for the Transylvanian Hungarian minority, showing that the community displays levels of religiosity substantially higher than those of Hungarians residing in Hungary itself, while remaining closer to the Romanian majority in this respect. As Veres (2015) has documented across four states hosting Hungarian minorities, the religious and ethnic dimensions of group identity are deeply intertwined, with confessional belonging functioning as a proxy for national attachment where direct political expression of minority solidarity is constrained. The implication is theoretically suggestive: minority conditions generate a form of religious surplus, in which the church is valued not only for its spiritual functions but for its irreplaceable role in maintaining a collective identity that the wider social environment places under recurrent strain. Yet if the identity-protection mechanism operates at the community level—making Transylvanian Hungarians more religious than their kin-state co-nationals as a bloc—the question of whether this mechanism operates with equal intensity across denominational lines within the community, and whether it produces the same forms of religious expression in all three historical confessions, has not yet been posed. That is precisely the question this article addresses.

2.2. The Four Historical Denominations of the Transylvanian Hungarian Community

The influence and presence of churches in the life of Transylvanian Hungarians extends well beyond the strictly religious domain. As Kiss (2018) has argued, the legitimacy of the Hungarian churches is increased by the widespread conviction—shared even by many non-religious community members—that they are indispensable institutional tools of minority self-organisation, lending them an impetus to take active roles in education, cultural life, and social care that has no parallel among the majority population’s denominational institutions (Kiss 2020).

The confessional pluralism of the Transylvanian Hungarian community is not a modern curiosity but the legacy of a singular historical episode. It was at the Diet of 1595, building upon a sequence of tolerance declarations beginning with the Diet of Torda in 1568, that the Principality of Transylvania formally recognised four religiones receptae—received religions enjoying full legal standing: the Roman Catholic Church, the Lutheran (Evangelical, Augsburg Confession) Church, the Reformed (Calvinist, Helvetic Confession) Church, and the Unitarian Church (Nagy 2023; Wien 2022). Eastern Orthodoxy, the faith of the Romanian-speaking population, was tolerated but not received—a distinction with profound political consequences that would reverberate well beyond the principality’s dissolution. This act of early institutionalised pluralism, unparalleled in its era across Christian Europe, created a denominationally stratified Hungarian community which has persisted, across the upheavals of Ottoman occupation, Habsburg consolidation, Trianon, communist repression, and post-communist transition, to the present day.

Of the four historical churches, the Reformed (Calvinist) Church occupies by far the most prominent position among the Hungarian minority today, both numerically and institutionally. With some 495,000 members recorded in the 2021 Romanian census—a decline from 600,000 in 2011, reflecting the accelerating emigration and demographic contraction of the Hungarian community more broadly (Kiss 2023)—the Reformed Church is almost entirely ethnically Hungarian in membership, its synods conducted in Hungarian, its geographic concentration heaviest in the Szeklerland counties of Harghita and Covasna, and its institutional network encompassing schools, publishing houses, and social welfare organisations that together constitute a parallel civil society for a minority community which has learned, across centuries, not to rely on the state for the reproduction of its cultural life. The relationship between the Reformed Church and the Democratic Alliance of Hungarians in Romania (RMDSZ), the principal political representative of the Hungarian minority, is close and historically rooted, the church functioning in effect as an institutional anchor for a form of minority solidarity that requires no explicit political articulation to be socially operative (Veres 2015). Hungary’s government has also provided documented financial support to all four historical churches through the Bethlen Gábor Foundation, underscoring the degree to which the minority churches function as nodes in a transnational Hungarian national community (Chiruta 2022).

The Roman Catholic Church presents a structurally different case. Accounting for roughly 40% of ethnic Hungarians in Transylvania (Kiss 2023), making the Latin rite the numerically second denomination within the community, the Catholic Church’s institutional character is not defined by its Hungarian membership in the way the Reformed Church is. Headquartered in Rome, globally hierarchical, and sharing its confession with a portion of the Romanian majority population—including the Csángó communities of Moldavia, historically Hungarian-speaking Catholics whose ethnic identification has long been a matter of political contestation—the Catholic Church cannot be straightforwardly ethnicised as a Hungarian institution in the manner that the Reformed Church can. It is Hungarian in the sense that most Transylvanian Catholics are Hungarian; it is not Hungarian in the sense of being institutionally organised around, or symbolically identified with, Hungarian minority identity as its primary purpose. This distinction, as the empirical analysis below will demonstrate, carries observable consequences for how Catholic Hungarians relate to their church as an institution, particularly with respect to confidence and trust, even as their levels of subjective religious commitment are entirely comparable to those of their Reformed counterparts.

The Unitarian Church deserves particular attention despite its modest numbers—approximately 48,000 members in 2021, a decline from 57,000 a decade earlier (Kiss 2023). Unique to Transylvania in the broader European context, and tracing its organised existence to the very Diet of Torda whose tolerance declarations created the framework for the religiones receptae, the Unitarian tradition has historically been associated with the intellectually oriented stratum of Transylvanian Hungarian culture, its theological antitrinitarianism demanding a certain willingness to deviate from confessional convention that has given the tradition a distinctive sociological profile. Its geographic centre is Kolozsvár/Cluj-Napoca, where the Unitarian Church maintains its bishopric and a university-level theological faculty. Institutionally small and liturgically less demanding than either the Reformed or the Catholic tradition, the Unitarian Church nonetheless partakes fully in the ethnic boundary-marking function common to all three numerically significant denominations: it is recognisably and entirely a Hungarian church.

The Lutheran Church of Transylvania was historically the church of the Transylvanian Saxon community; following the mass emigration of Saxons to West Germany in the late communist and early post-communist period, its presence among the Hungarian minority has been reduced to statistical insignificance (Kiss 2023). In the ZA7550 dataset, only six respondents identify as Lutheran. For this reason, the present analysis merges Lutheran respondents with Unitarians into a single “Other Historical Protestant” category, not because the Lutheran church lacks historical standing—its status as one of the four received religions is unambiguous—but because its present-day presence within the Transylvanian Hungarian community is negligible in scale, while its institutional character as a non-Catholic, ethnically bounded historical church is sufficiently similar to the Unitarian tradition to make the merger analytically defensible.

It is worth noting, in this connection, the small but theoretically remarkable presence of neoprotestant denominations within the Hungarian minority. The 2021 Romanian census recorded a modest growth of neoprotestant affiliation even within the Hungarian community (Institutul Naţional de Statistică 2023; Rotaru et al. 2024)—a trend consistent with the broader expansion of Baptist and Adventist congregations that has reshaped the Romanian religious landscape since 1989 (Kiss 2023). These groups are too small in the present sample for multivariate analysis, but their descriptively distinctive profiles—markedly higher attendance and institutional religiosity than any of the historical denominations—will be noted in the results (Kiss 2020).

2.3. State of the Literature

The empirical sociology of religion among the Transylvanian Hungarian minority has advanced considerably in the past decade, driven in large part by the availability of the dedicated EVS 2017 dataset (GESIS ZA7550, DOI: 10.4232/1.13562), a dataset which provides individual-level data on religious attitudes and behaviour for 1106 members of the Hungarian minority community, collected in 2019–2020 and constituting the most comprehensive survey instrument yet applied to this population. Prior to its availability, research in this area relied predominantly on secondary analysis of standard national surveys in which the Hungarian minority appeared as a small and statistically unreliable subgroup, or on qualitative and historical approaches which, however valuable, could not speak to the distribution of religious attitudes at the individual level. An earlier contribution by Kiss (2010), drawing on EVS 2000 and 2008 data, provided the first descriptive account of denominational variation in religiosity among Transylvanian Hungarians, noting that Catholics scored highest on subjective religious importance and institutional self-identification while Unitarians scored lowest—a pattern that the present analysis subjects, for the first time, to multivariate regression controls. A comprehensive treatment of the subject is offered by Kiss (2018), whose chapter draws on a 2009 Romanian Institute for Research on National Minorities representative survey (N = 3991) to document attendance and religiosity type by denomination, finding Catholics to attend at notably higher rates than Reformed respondents (40.8% vs. 27.1% weekly attendance) and Catholics more likely to identify as religiously institutional. These descriptive findings, consolidated in Kiss’s (2020) monograph-length survey-based account of the community’s religious life, are the direct empirical precursors that the present regression analysis extends and formalises.

The existing published analyses of ZA7550 are instructive both for what they have established and for what they have left open. Kiss et al. (2022), in a contribution to a volume honouring the EVS’s long-standing comparative tradition, provide the most thorough descriptive examination of the dataset to date, positioning the Transylvanian Hungarian community between the Romanian majority and the Hungarian national population across multiple dimensions of religiosity, and demonstrating that the minority’s elevated religious engagement is consistent with an identity-protection interpretation. Zsigmond and Attila (2022), writing in a Hungarian-language festschrift context, draw on the same data to document the principal trends in the community’s religious life in aggregate terms. In the broader Central and Eastern European (CEE) context, Guglielmi and Piacentini (2024) have examined national and religious identity linkages across the region using the integrated EVS 2017 dataset, though their analysis necessarily treats minority communities as absorbed into national samples and cannot speak to the internal dynamics of any specific minority group.

More recently, Müller and Rosta (2025) have revisited the same question using Pew 2017 data across sixteen CEE countries, finding that the institutional–personalised religiosity distinction carries consequences for national identity orientation—a finding which underscores the timeliness of treating that distinction as a dependent variable in its own right, as the present article does.

A partial exception to the descriptive pattern is provided by Veres (2024), who, in a regression-based study of socioeconomic stratification among minority Hungarian youth using a different dataset, includes the self-description of religiosity type as a predictor variable and finds a theoretically significant result: among Transylvanian Hungarians specifically, identifying as religious according to church teachings—as distinct from being religious in one’s own way—significantly reduces the likelihood of belonging to the socioeconomically deprived group, an effect the author attributes to the organisational social capital embedded in institutionalised religious participation. This finding, though not concerned with religiosity as a dependent variable and operating on a youth sample rather than the full adult population, establishes that the institutional–personalised religiosity distinction carries differential predictive power for Transylvanian Hungarians, and thereby lends indirect support to the analytic choice made in the present article to treat the relevant ZA7550 country-specific variable as a theoretically substantive outcome.

What no prior contribution has attempted is such an intra-community regression analysis—an inquiry into whether the Catholic, Reformed, and Unitarian traditions predict individual-level religiosity outcomes differently once sociodemographic controls are applied. The treatment of the community as a homogeneous religious bloc is understandable given that the primary analytical interest has been comparative—the minority against the majority, or against the kin-state—but it forecloses a question which the data are, as this article will demonstrate, well positioned to answer. A second gap concerns the country-specific religiosity-type variable, which has appeared in only one published regression context (Veres 2024), and then only as a predictor in a non-religious outcome model rather than as a dependent variable in its own right. The claim advanced here is specific and verifiable: that denomination shapes the form of religious engagement among Transylvanian Hungarians—its behavioural expression, its institutional orientation, its mode of self-understanding—in ways that are not reducible to the secular–religious cleavage identified by Bechert (2018) and Stegmueller et al. (2012) as the dominant axis of religious variation in European comparative research. The theoretical elaboration of that claim is taken up in the following section.

3. Theoretical Framework and Hypotheses

3.1. Believing, Belonging, and Behaving: The Three-Dimensional Model

The sociology of religion has long wrestled with the question of whether denomination constitutes a meaningful independent variable in the explanation of individual religiosity, or whether it is merely a proxy for the more fundamental cleavage between the religious and the secular. A prominent strand of the literature, forcefully represented by Stegmueller et al. (2012), concludes that denominational affiliation contributes little to the explanation of economic or social attitudes once the intensity of religious practice is controlled, a finding extended by Bechert (2018) in a comparative EVS-based analysis of religiosity profiles across European countries. In this view, the relevant boundary runs between the practising believer and the secular individual, not between the Catholic and the Protestant. The present article does not so much contest this claim as qualify it, and qualification requires a theoretical framework attentive to the multiple, partially independent dimensions of religious life—dimensions which, it will be argued, respond differently to denominational context.

The most influential attempt to disaggregate religiosity into analytically distinct components remains Davie’s (1994) diagnosis of “believing without belonging,” elaborated in the British context but since generalised into a broader European sociology of religion. Davie’s (1994) key insight was that affiliation, practice, belief, and belonging do not necessarily travel together: modernity does not produce a uniform retreat from religion but rather a restructuring of its forms, with institutional participation declining faster than subjective conviction. Voas (2009) subsequently introduced the concept of “fuzzy fidelity” to capture the growing proportion of Europeans who occupy an indeterminate middle ground—neither devout nor secular, retaining a cultural religiosity without sustained commitment to any particular institutional form. Taken together, these formulations invite a research design that treats attendance, institutional trust, and subjective religiosity as separate outcomes rather than interchangeable indicators of a single underlying variable—and it is precisely this disaggregation which opens up the question of whether denomination predicts them differently.

3.2. Denominational Structure and Practice Expectations

There is good theoretical reason to expect that denomination does predict practice and institutional trust differently, at least in the specific configuration under examination. The Roman Catholic Church and the Reformed (Calvinist) Church impose structurally different demands upon their members with respect to attendance. Canon law obliges Catholics to attend Sunday Mass, a normative expectation enforced through generations of socialisation and, historically, through sacramental discipline; the Reformed tradition, by contrast, while strongly encouraging regular worship, lacks an equivalent juridical obligation (McQuillan 2004; Peri-Rotem 2016). This difference in institutional prescription, independent of any difference in subjective religiosity, generates a testable prediction: controlling for individual-level characteristics, Catholics should attend religious services more frequently than Reformed members—not necessarily because they are more religious in any deeper sense, but because their church makes a more explicit behavioural demand. Descriptive evidence from a 2009 survey representative of Transylvanian Hungarians already showed a nearly 14-percentage-point gap in weekly attendance between Catholics and Reformed church members (40.8% vs. 27.1%), suggesting that canonical attendance obligations have a real behavioural correlate in this community (Kiss 2018).

The institutional trust dimension, however, follows a different logic entirely. The Reformed Church of Romania occupies a position in the life of the Transylvanian Hungarian minority that no amount of canonical obligation can replicate: it is, in the most literal institutional sense, the Hungarian church—its synods conducted in Hungarian, its parishes coterminous with Hungarian-speaking communities, its history inseparable from the political and cultural survival of the minority in an Orthodox-majority state. Kiss et al. (2022) have demonstrated that the elevated religiosity of Transylvanian Hungarians relative to their co-nationals in Hungary is plausibly attributable to the identity-protection function of religion under minority conditions, a mechanism operating across denominational boundaries. Abbasov (2023), examining the determinants of public trust in religious institutions across national contexts, demonstrates empirically that the strength of national identity predicts confidence in churches—a finding grounded in cases where, as in Transylvania, a particular denomination is fused with ethnic self-understanding. The present article applies this logic at the within-community denominational level: it is not national identity as such that varies across Reformed and Catholic Transylvanian Hungarians, but the degree to which each denomination is experienced as institutionally one’s own. Confidence in one’s particular church as an institution is likely to reflect the degree to which that institution is experienced as ethnically one’s own—a dimension on which the subjective salience of religion, being uniformly elevated across confessions by minority conditions, provides no purchase. The Reformed Church’s advantage in institutional trust therefore reflects not its members’ greater personal piety but its structural position as the primary bearer of Hungarian minority identity in Transylvania. The Roman Catholic Church, whatever its local strength, remains part of a universal hierarchical structure headquartered in Rome, sharing its confession with significant portions of the Romanian majority population; it cannot carry the same specific ethnic charge.

A similar logic, extended further, would lead one to expect the Unitarian and Lutheran churches—both historical Protestant institutions in the Transylvanian sense, both ethnically bounded and modest in institutional scale—to pattern more closely with the Reformed tradition than with the Catholic one on the trust dimension, even as their small size and less central institutional role may produce slightly attenuated effects.

3.3. The Institutional vs. Personalised Religiosity Distinction

A third theoretical strand informs the analysis, one which the data make unusually tractable. The country-specific module of ZA7550 includes a variable absent from the standard EVS country files, asking respondents which of several statements best describes their religiosity: religious according to the teachings of the church, religious in their own way, undecided, or non-religious. This operationalises with rare directness the distinction between institutional and personalised religiosity that Davie (1994) and Voas (2009) place at the centre of their arguments. The ecclesiological contrast between Catholicism and the Reformed tradition maps neatly onto this dimension: the hierarchical, magisterial Catholic Church transmits a normative theological content to which the faithful are expected to conform, while the congregationalist Reformed tradition places greater emphasis on the individual believer’s personal relationship with scripture and community, without an equivalent teaching authority mediating that relationship. One would consequently expect Catholics to identify more frequently as religious “according to the teachings of the church,” while Reformed respondents gravitate toward the personalised formulation—and this, it should be noted, entirely independently of any difference in the strength of their religious conviction. Prior descriptive evidence supports this expectation: Kiss (2018) found a 10-percentage-point difference between Catholics (48.3%) and Reformed respondents (37.8%) in institutional self-identification using 2009 data, the precise denominational gap that H4 predicts will persist net of sociodemographic controls in the ZA7550 data collected in 2019–2020. Recent empirical support for the analytical salience of this distinction in the Transylvanian Hungarian context is provided by Veres (2024), whose regression analysis demonstrates that the two self-descriptions carry different predictive weights for socioeconomic outcomes within this population. Moreover, Müller and Rosta (2025) demonstrate, using Pew 2017 data across sixteen CEE countries, that institutional religiosity correlates with more exclusivist orientations toward national identity, suggesting that how one is religious—not only how intensely—carries sociological weight independent of denomination.

3.4. Hypotheses

From the theoretical considerations developed above, four hypotheses are derived. The Reformed (Calvinist) Church is set as the reference category throughout, as the numerically dominant and theoretically central institution within the Hungarian minority community.

Hypothesis 1

(Practice). Controlling for sociodemographic variables, Roman Catholic respondents will attend religious services significantly more frequently than Reformed members, reflecting the canonical obligation of Sunday Mass attendance; the Other Historical Protestant category will not differ significantly from the Reformed reference group.

Hypothesis 2

(Institutional trust). Controlling for sociodemographic variables, Reformed respondents will express significantly higher confidence in their church than Catholic respondents, reflecting the ethnic embeddedness of the Reformed Church as the primary institutional carrier of Hungarian minority identity in Transylvania.

Hypothesis 3

(Subjective salience). Denomination will exert no significant independent effect on the subjective importance of religion—the identity-protection function of religion under minority conditions operates across all three historical denominations with equal force, producing uniformly high subjective salience regardless of confessional affiliation.

Hypothesis 4

(Religiosity type). Catholic respondents will be significantly more likely to describe themselves as religious according to the teachings of their church rather than in a personalised way, reflecting the Catholic institutional tradition’s stronger normative claim over its members’ mode of religious self-understanding.

Taken together, these hypotheses imply that denomination shapes the form of religious engagement—its behavioural expression, its institutional orientation, its mode of self-understanding—while leaving largely untouched the intensity of religious identity that minority conditions generate uniformly across confessional lines.

4. Data and Method

4.1. Data Source

The analysis draws on the European Values Study 2017 Romanian Hungarian dedicated minority oversample (EVS 2020a), archived at GESIS as study ZA7550 (DOI: 10.4232/1.13562). This dataset was collected between November 2019 and March 2020 by TT Research & Communications, Cluj-Napoca, under the direction of Tamás Kiss and István G. Székely of the Romanian Institute for Research on National Minorities (ISPMN). The target population comprises adults aged 18 and over residing in private households in areas of Romania with a significant Hungarian minority presence. The full dataset contains $N=1106$ respondents. The analytic sample for the present study is restricted to respondents identifying with the Roman Catholic ($n=419$), Calvinist/Reformed ($n=560$), or Unitarian/Lutheran ($n=47$) denominations, yielding a working sample of $N=1026$, representing 93% of the full dataset. According to the 2021 Romanian census, Reformed, Catholic, and Unitarian members together account for over 90% of all ethnic Hungarians in Romania (Kiss 2023), confirming that the analytic sample covers the vast majority of the target population.

4.2. Variables

Four binary dependent variables were constructed from the ZA7550 dataset. Church attendance (v54) was recoded as 1 for weekly or more frequent attendance and 0 otherwise. Confidence in the church (v115) was recoded as 1 for respondents expressing a great deal or quite a lot of confidence and 0 for those expressing not very much or none at all. The importance of religion (v6) was recoded as 1 for very or rather important and 0 otherwise. Religiosity type (RO_hu_cs25), a country-specific variable unique to ZA7550, was recoded as 1 for respondents describing themselves as religious according to the teachings of their church and 0 for those describing themselves as religious in their own way; respondents identifying as undecided or non-religious, and a small number non-responses (totally: $n=59$) were excluded from Model 4, as the institutional–personalised distinction is analytically meaningful only within the religious population1.

The decision to operationalise all four dependent variables as binary outcomes reflects two considerations. First, the original scales are ordinal with few categories, and binary logistic regression makes fewer distributional assumptions while producing directly interpretable odds ratios. Second, the specific cutoff points reflect theoretically meaningful thresholds: weekly attendance captures the normative minimum in canonical Catholic practice; high confidence separates active institutional trust from its absence; very or quite important separates those for whom religion is a positive value from those who are indifferent or opposed; and the institutional versus personalised dichotomy follows directly from the ZA7550 response categories. The trade-off is a reduction in variance relative to the full ordinal scales; sensitivity analyses using ordered logistic regression confirm that the direction and significance of denomination contrasts are robust to this choice.

The key independent variable in the analysis is the denomination (v52_RO_hu), a ten-category variable collapsed into three analytical groups: Reformed/Calvinist (reference category, $n=560$), Roman Catholic ($n=419$), and Other Historical Protestant—comprising Unitarian ($n=41$) and Lutheran ($n=6$) respondents merged due to the near-absence of Lutherans in the community. Given the small size of this combined category ($n=47$), statistical power for OHP contrasts is limited; the wide confidence intervals reported in Table 2 for this group should be interpreted as exploratory rather than confirmatory. Neoprotestant respondents (Baptist, Adventist, Jehovah’s Witnesses; $n=29$) and respondents belonging to other or no denomination are excluded from the regression models and treated descriptively only.

Five control variables were taken into account. The first is gender (v225), as women consistently show higher religious participation across European contexts and the Transylvanian Hungarian community is no exception. Age group (age_r), with six categories (15–24, 25–34, 35–44, 45–54, 55–64, and 65 and over), is another very strong sociodemographic predictor of religiosity. The lowest category follows the standard EVS age_r coding; given the 18-and-over target population of ZA7550, this category contains respondents aged 18–24 only. Education (v243_ISCED_1) was recoded into three broad categories following the ISCED classification: low (less than primary, primary, and lower secondary; ISCED 0–2), medium (upper secondary and post-secondary non-tertiary; ISCED 3–4; reference category), and high (short-cycle tertiary, bachelor’s, and master’s or equivalent; ISCED 5–7). The original eight-category variable produced an unstable reference category ($n\approx 6$, 0.6% of the analytic sample), resulting in very wide confidence intervals for all education contrasts; the three-category specification eliminates this instability while preserving the theoretically relevant distinction between lower, medium, and higher educational attainment; a sensitivity analysis using the original eight-category specification produces substantively identical denomination contrasts. As rural populations in Transylvania maintain higher rates of religious practice and institutional participation, urban (1) or rural (0) settlement type (RO_hu_cs31) was also used as a control variable. Finally, considering that partnership status is associated with higher institutional religious participation through the lifecycle, a binary control variable for marital status was constructed, coded 1 for married or in stable partnership and 0 for all other statuses (divorced, widowed, single).

The analytic sample comprises 1026 respondents identifying as members of one of the four historically received denominations of the Transylvanian Hungarian minority. Reformed Calvinist respondents constitute the largest group (54.6%), followed by Roman Catholics (40.8%) and Other Historical Protestants, Unitarians and Lutherans (4.6%), merged together due to the low case count. The sample is approximately gender-balanced (51.8% female). The age distribution is skewed toward older cohorts, with respondents aged 65 and over constituting the largest single age group (28.7%), while those aged 15–24 represent the smallest (11.6%). In terms of educational attainment, the largest share holds upper secondary qualifications (41.7%), followed by lower secondary (28.7%) and bachelor’s or equivalent degrees (9.5%). A minority of respondents live in urban areas (43.6%), and the majority are married (59.2%).

Missing data are minimal for three of the four outcome variables (attendance: $n=2$; importance of religion: $n=4$; confidence in church: $n=5$). Missingness is higher for the institutional religiosity variable (cs25: $n=59$), reflecting the deliberate exclusion of respondents identifying as non-religious or undecided, as Model 4 is restricted to the religious population.

4.3. Analytical Strategy

Survey design is accounted for using the post-stratification weight variable RO_hu_WEIGHT supplied with ZA7550, which aligns the sample to the Hungarian minority population on region (four regions), settlement type (rural/urban), gender, and age group, based on a demographic projection from the 2011 Romanian census (EVS 2020b). As the publicly released anonymised data file does not include primary sampling unit (PSU) identifiers, the sampling design is specified as ids=~1 following standard practice for the public-access EVS data files. Standard errors may be modestly underestimated relative to full design-based inference, given the three-stage cluster structure of the original sample; this limitation is acknowledged.

Four binary logistic regression models were estimated using the svyglm() function from the survey package (Lumley 2010) in R (R Core Team 2026), with a quasibinomial variance family and logit link. Each model uses an identical right-hand-side specification: denomination (Reformed as reference category) plus five sociodemographic controls (gender, age group, education, urban/rural settlement, and marital status). The four dependent variables—weekly or more frequent church attendance, high institutional confidence in the church, high subjective importance of religion, and institutional religiosity type—are each modelled separately, as they are treated as theoretically distinct dimensions of religiosity rather than interchangeable indicators of a single latent variable. Model 4 is estimated on a restricted subsample of religious respondents only ($n=967$), as the institutional–personalised distinction is analytically meaningful only within the religious population; undecided and non-religious respondents are excluded. Results are reported as odds ratios with 95% Wald confidence intervals. Model fit is reported as McFadden’s pseudo-$R^2$, computed as $1-(D_{\mathrm{model}}/D_{\mathrm{null}})$ where D denotes the quasi-deviance of the full model relative to an intercept-only null model estimated on the same survey design; this measure is preferred over Nagelkerke’s $R^2$ because the quasibinomial family does not produce a proper log-likelihood from which the latter can be computed. Robustness was assessed by re-estimating all models without survey weights; results were substantively identical in direction and significance across all denomination contrasts.

During the preparation of this manuscript, the author used Claude (Anthropic, model Sonnet 4.6) for the following purposes: drafting, editing, and improving manuscript text, including background prose in Section 2 and Section 3; conducting literature searches to identify relevant published sources; formulating and debugging R code for the survey-weighted logistic regression models; and discussing the interpretation of regression results. All AI-assisted content, including any interpretive suggestions, was critically reviewed, verified, and validated by the author, who takes full responsibility for the analysis, conclusions, and the content of this publication.

5. Results

5.1. Descriptive Overview

Table 1. Religiosity indicators by denomination (%), ZA7550 analytic sample.

Indicator	Calvinist	R. Catholic	Other Hist. Prot.
Weekly+ attendance (v54)	28.4 (559)	40.2 (418)	23.4 (47)
Religion very/quite important (v6)	87.3 (559)	88.8 (418)	80.0 (45)
High confidence in church (v115)	83.0 (559)	76.1 (415)	76.6 (47)
Institutional religiosity (cs25)	47.0 (526)	42.8 (397)	45.5 (44)

Note: Values are unweighted percentages; valid n in parentheses. All four outcomes are recoded as binary for the multivariate models; see Section 4.2. The full analytic sample comprises Reformed $n=560$, Catholic $n=419$, and Other Historical Protestant $n=47$; valid-case n per outcome differs by up to 5 due to item non-response (see Section 4.2 for details). For institutional religiosity (cs25), undecided and non-religious respondents are additionally excluded, as Model 4 contrasts institutional versus personalised religiosity within the religious population only (Reformed $n=526$, Catholic $n=397$, Other Historical Protestant $n=44$).

presents key religiosity indicators by denomination for the analytic sample. The denominational profiles reveal a striking practice–trust reversal: Roman Catholics attend religious services more frequently than Reformed members (40.2% vs. 28.4% weekly or more), while Reformed members express higher confidence in their church than Catholics (83.0% vs. 76.1% great deal or quite a lot). The subjective importance of religion is uniformly high across all three groups, ranging narrowly from 80.0% (Other Historical Protestant) to 88.8% (Catholic).2

The attendance and confidence patterns are consistent with those documented descriptively by Kiss (2018) using 2009 data and subsequently confirmed in Kiss (2020). The institutional religiosity type distribution, however, shows a notable departure from the 2009 figures: where Kiss (2018) found Catholics more likely than Reformed respondents to self-identify as institutionally religious (48.3% vs. 37.8%), the present data show a modest reversal (Reformed: 47.0%; Catholic: 42.8%), suggesting that the denominational gap on this dimension has not only narrowed but changed direction over the intervening decade—a pattern consistent with the non-significant H4 regression result discussed in Section 5.6.

Neoprotestant respondents (Baptist, Adventist, and Jehovah’s Witnesses; $n=29$), excluded from the multivariate models due to insufficient cell sizes, display a markedly more intense profile: 79% attend weekly or more and 90% describe themselves as religious according to church teachings—a pattern consistent with the behavioural expectations generated by high-demand religious organisations (Stark and Finke 2000).

5.2. Multivariate Models

Table 2. Binary logistic regression models predicting religiosity outcomes: odds ratios (95% CI), ZA7550 analytic sample.

	M1 Attendance	M2 Conf. Church	M3 Importance	M4 Relig. Type
Denomination (ref.: Reformed/Calvinist)
Catholic	1.895 *** [1.408–2.552]	0.719 † [0.509–1.015]	1.044 [0.678–1.607]	0.867 [0.651–1.153]
Other Hist. Prot.	0.770 [0.356–1.666]	0.683 [0.312–1.495]	1.041 [0.427–2.539]	0.769 [0.397–1.490]
Gender
Female	2.392 *** [1.788–3.200]	1.233 [0.876–1.735]	1.788 ** [1.168–2.736]	1.431 * [1.083–1.892]
Age group (ref.: 15–24 years)
25–34 years	0.827 [0.438–1.558]	0.702 [0.376–1.309]	0.890 [0.421–1.884]	1.232 [0.670–2.262]
35–44 years	1.240 [0.660–2.329]	0.856 [0.445–1.646]	1.093 [0.477–2.507]	1.681 † [0.930–3.038]
45–54 years	1.350 [0.746–2.443]	0.734 [0.390–1.381]	0.802 [0.370–1.737]	2.018 * [1.135–3.587]
55–64 years	2.336 ** [1.289–4.235]	1.355 [0.673–2.727]	1.017 [0.452–2.287]	2.608 ** [1.439–4.726]
65 and over	2.664 *** [1.556–4.560]	1.633 [0.896–2.975]	1.614 [0.793–3.286]	3.267 *** [1.922–5.553]
Education (ref.: Medium—upper secondary and post-secondary non-tertiary)
Low (ISCED 0–2)	1.065 [0.760–1.493]	0.843 [0.564–1.262]	0.886 [0.523–1.501]	1.293 [0.938–1.782]
High (ISCED 5–7)	1.231 [0.788–1.922]	0.686 [0.432–1.091]	0.539 * [0.316–0.920]	0.958 [0.631–1.454]
Structural controls
Urban	0.837 [0.621–1.128]	0.608 ** [0.431–0.859]	0.264 *** [0.171–0.408]	0.794 [0.596–1.058]
Married	0.963 [0.702–1.322]	1.553 * [1.073–2.247]	1.499 † [0.946–2.374]	1.244 [0.917–1.687]
N	1024	1021	1022	967
$R_{\mathrm{McF}}^2$	0.080	0.042	0.088	0.050

^† $p<0.10$; * $p<0.05$; ** $p<0.01$; *** $p<0.001$ (Wald 95% CI in brackets). Weighted binary logistic regression (svyglm, quasibinomial family). Reference categories: denomination = Reformed (Calvinist); age group = 15–24 years; education = Medium (upper secondary and post-secondary non-tertiary; ISCED 3–4). Low education: less than primary, primary, and lower secondary (ISCED 0–2); High education: short-cycle tertiary, bachelor’s, and master’s or equivalent (ISCED 5–7). Model 4 is restricted to religious respondents only ($n=967$); see Section 4.2. $R_{\mathrm{McF}}^2$: McFadden’s pseudo-$R^2$, computed as $1-(D_{\mathrm{model}}/D_{\mathrm{null}})$ where D denotes the quasi-deviance relative to an intercept-only null model estimated on the same survey design. Italic labels denote predictor category groups.

presents the results of four weighted binary logistic regression models estimated on the analytic sample, each testing a distinct dimension of religiosity as the dependent variable. Model 1 examines weekly or more frequent church attendance (H1), Model 2 institutional confidence in the church (H2), Model 3 the subjective importance of religion (H3), and Model 4 institutional versus personalised religiosity type among religious respondents (H4). All models include identical control variables: gender, age group, educational attainment, urban/rural settlement, and marital status. Denomination is entered as a categorical variable with Reformed (Calvinist) as the reference category, so all denomination coefficients represent contrasts against the Reformed baseline.

5.3. Model 1: Church Attendance (H1)

H1 is strongly confirmed. Controlling for gender, age, education, settlement type, and marital status, Roman Catholic respondents attend religious services significantly more frequently than their Reformed counterparts (OR = 1.895, 95% CI [1.408–2.552], $p<0.001$). The odds of weekly or more frequent attendance are nearly twice as high among Catholics as among Reformed members, a magnitude consistent with the nearly 14-percentage-point gap (40.8% vs. 27.1%) documented descriptively by Kiss (2018) using 2009 data and replicated in Table 1. Other Historical Protestants do not differ significantly from the Reformed reference group (OR = 0.770, $p=0.506$). Among the controls, gender and age emerge as the strongest predictors: women have nearly 2.4 times the odds of weekly attendance as men (OR = 2.392, $p<0.001$), and attendance rises monotonically with age, reaching significance in the two oldest cohorts (55–64 years: OR = 2.336, $p=0.005$; 65 and over: OR = 2.664, $p<0.001$). Education, urban residence, and marital status are not significant predictors of attendance in this sample.

5.4. Model 2: Confidence in Church (H2)

H2 is not confirmed in its strict form but the results are directionally consistent with the hypothesis and approach conventional significance. Catholics express lower confidence in the church than Reformed members (OR = 0.719, 95% CI [0.509–1.015], $p=0.061$), a borderline result that aligns with the predicted practice–trust reversal: the same denomination that attends more frequently reports less institutional confidence. The Other Historical Protestant group does not differ significantly from the Reformed baseline (OR = 0.683, $p=0.340$). Notably, the robustness check confirmed that this effect is more pronounced in the unweighted model (OR = 0.651, $p=0.009$), suggesting that the borderline result in the weighted model reflects the correction for demographic imbalance rather than a weak substantive effect. Among controls, urban respondents show significantly lower confidence than rural ones (OR = 0.608, $p=0.005$), and married respondents show higher confidence than unmarried ones (OR = 1.553, $p=0.020$). Education, gender, and age do not reach significance in this model.

5.5. Model 3: Importance of Religion (H3)

H3 is confirmed. Neither Catholic nor Other Historical Protestant respondents differ significantly from Reformed members in the subjective importance they assign to religion (Catholic: OR = 1.044, $p=0.845$; OHP: OR = 1.041, $p=0.929$), and both odds ratios are negligibly close to 1. This null result is substantively meaningful: it supports the identity-protection interpretation, in which the minority condition generates a uniformly elevated subjective salience of religion across all three historical confessions, leaving no residual variation for denomination to explain. Women assign significantly higher importance to religion than men (OR = 1.788, $p=0.008$). The most striking finding in this model concerns urban residence: respondents living in urban areas have only 0.26 times the odds of describing religion as very or quite important compared to rural respondents (OR = 0.264, 95% CI [0.171–0.408], $p<0.001$). This is the strongest single control effect across all four models and suggests that the secularising effect of urban settlement operates with particular force on the dimension of subjective religious salience, independently of denominational affiliation. Education also emerges as a significant predictor: respondents with high educational attainment (short-cycle tertiary, bachelor’s, or master’s; ISCED 5–7) have significantly lower odds of describing religion as very or quite important compared to those with medium education (OR = 0.539, 95% CI [0.316–0.920], $p=0.024$), a finding consistent with the well-documented secularising association of tertiary education in the European context, which the more stable three-category specification makes visible.

5.6. Model 4: Institutional Religiosity Type (H4)

H4 is not confirmed. Among religious respondents, Catholics are no more likely than Reformed members to describe themselves as religious according to the teachings of their church rather than in their own way (OR = 0.867, 95% CI [0.651–1.153], $p=0.326$). The direction of the coefficient is in fact slightly negative—Catholics are marginally less likely than Reformed members to select the institutional self-description, though the effect is far from significant. Other Historical Protestants similarly do not differ from the Reformed baseline (OR = 0.769, $p=0.436$). The absence of a denominational effect on religiosity type is notable given the prior descriptive evidence from Kiss (2018), who found a 10-percentage-point gap between Catholics (48.3%) and Reformed respondents (37.8%) on this measure in 2009; the present regression results suggest that this gap does not survive sociodemographic controls in the ZA7550 data collected in 2019–2020. It is worth noting that the same 2009 descriptive data show Unitarian respondents at 47.9% institutional self-identification—nearly identical to Catholics and substantially higher than Reformed—a pattern which, if replicated in the current ZA7550 data, would further undermine any simple Protestant–Catholic cleavage on this dimension and point instead toward within-Protestant variation as the more salient axis. Age is the dominant predictor in Model 4: the odds of institutional self-identification rise monotonically and steeply with age, from a non-significant contrast in the youngest cohorts to an OR of 3.267 in the oldest group (95% CI [1.922–5.553], $p<0.001$), suggesting that the institutional–personalised distinction maps more strongly onto a generational than a denominational cleavage within this community. Among the controls, women are significantly more likely than men to identify as institutionally religious (OR = 1.431, $p=0.012$), a result consistent with the broader literature on gender and institutional religious participation.

6. Discussion

6.1. The Practice–Trust Reversal: Canonical Obligation and Ethnic Embeddedness

The most theoretically significant finding is the simultaneous confirmation of H1 and the directional result of H2: Catholics attend more, Reformed trust more. This practice–trust reversal is not paradoxical once the two dimensions are disentangled. Attendance is, in the Catholic case, partly a normative compliance with canonical obligation rather than a pure expression of subjective commitment—a mechanism entirely consistent with the non-significant denomination effect in Model 3, which shows that the two groups assign identical subjective importance to religion. The Reformed Church’s advantage in institutional confidence, by contrast, reflects its structural position as the primary ethnic institution of the Hungarian minority: it is trusted not only as a religious authority but as an irreplaceable vehicle of collective identity under minority conditions, a function which the Roman Catholic Church—hierarchically headquartered in Rome and shared with portions of the Romanian majority—cannot perform in the same way (Abbasov 2023; Kiss 2018). The ecclesiological dimension of H4, while not confirmed in the regression, is theoretically consistent with this picture: that denomination does not independently predict the institutional–personalised self-description once sociodemographic controls are applied suggests that the Catholic–Reformed difference observed descriptively in 2009 (Kiss 2018) has narrowed or dissolved in the intervening decade, with age emerging as the dominant axis of variation. The 2009 descriptive data also showed Unitarian respondents identifying as institutionally religious at rates closer to Catholics than to Reformed members—a pattern which, if present in the current data, would further complicate any simple Catholic–Protestant ecclesiological reading, though the merger of Unitarians and Lutherans into the Other Historical Protestant category precludes a definitive test.

The emergence of age as the dominant predictor of institutional versus personalised religiosity type, displacing the Catholic–Reformed denominational cleavage visible in 2009 descriptive data, invites at least two interpretive hypotheses that future research should test against longitudinal data. The first is a secularisation hypothesis: younger cohorts of Transylvanian Hungarians, exposed to greater urban mobility, higher education rates, and Europeanised cultural norms, may be converging toward a more individualised religiosity irrespective of their denominational background, consistent with Voas’s (2009) account of fuzzy fidelity spreading across previously resistant religious communities. Evidence that personalised rather than institutional religious self-identification is becoming more characteristic of Hungarian minority youth is provided directly by Kiss and Rosta (2023), whose analysis of the 2020 Hungarian minority youth survey across four Carpathian Basin countries finds that individual religiosity predominates among young people over close ties to the historical churches—a pattern they document specifically in the dimension of religious self-categorisation, the youth-survey analogue of the ZA7550 RO_hu_cs25 variable analysed in the present study. Notably, Transylvania is confirmed as one of the two most religious Hungarian minority regions on all examined dimensions, which is consistent with the identity-protection interpretation but does not preclude the convergence in how younger cohorts express that religiosity across denominational lines. A further possibility specific to the Catholic side of the reversal is the declining salience of clerical authority among younger Hungarian Catholics following post-communist institutional liberalisation: if the teachings of the church have become a less salient frame of self-identification for younger Catholic Transylvanian Hungarians, the descriptive advantage of Catholics on institutional self-identification documented in 2009 data would erode even in the absence of any change in Reformed self-understanding (Davie 1994). The second hypothesis is a minority-status transformation hypothesis: as younger generations have grown up under EU membership and with normalised cross-border mobility to Hungary, the identity-protection function of religion may be becoming less denominationally differentiated and more uniformly cultural, producing a convergence in how members of different confessions relate to their church rather than simply whether they remain attached. Distinguishing between these explanations would require panel data or cohort analysis across EVS waves, which the present cross-sectional design cannot provide.

6.2. Denomination-Agnostic Salience and the Identity-Protection Mechanism

The confirmation of H3—the null result on subjective religious importance—is the finding most directly interpretable within the identity-protection framework (Kiss et al. 2022; Veres 2015). When minority conditions render the church indispensable as an institutional carrier of ethnic survival, the subjective salience of religion is elevated across all confessional lines, leaving denomination without explanatory purchase on this dimension. The urban effect in Model 3, by contrast, suggests that the secularising pressure of urban settlement does penetrate the community’s subjective religiosity even where denominational affiliation does not differentiate it—an important qualification to any blanket characterisation of Transylvanian Hungarians as uniformly resistant to secularisation.

6.3. Contributions and Broader Implications

The findings engage directly with the debate initiated by Stegmueller et al. (2012) and Bechert (2018) on whether denomination constitutes an independent explanatory variable once the secular–religious divide is controlled. The present results offer a partial qualification of their scepticism: denomination does independently predict attendance and trends toward significance on institutional confidence, but exerts no effect on subjective salience or religiosity type. The relevant boundary is not simply secular versus religious, but cuts across dimensions—denomination shapes the form of religious engagement while the minority condition shapes its intensity. This finding has implications beyond the Transylvanian case. Minority religious communities in CEE represent a distinct configuration in which the usual predictors of denominational variation—doctrinal content, institutional demand, canonical obligation—are overlaid by an identity-protection function that operates uniformly across confessional lines. Research treating such communities as homogeneous blocs risks confounding these two sources of variation; the regression framework applied here offers a replicable strategy for disentangling them wherever comparable oversample data are available.

7. Conclusions

This article has presented the first regression-based analysis of intra-community denominational variation in religiosity among Transylvanian Hungarians, using the 2017 European Values Study Romanian Hungarian minority dedicated oversample (GESIS ZA7550, fieldwork 2019–2020). Four binary logistic regression models tested whether denomination independently predicts church attendance, institutional confidence in the church, subjective importance of religion, and institutional versus personalised religiosity type, net of sociodemographic controls.

The results confirm a practice–trust reversal that descriptive research had previously noted but not subjected to multivariate controls: Catholics attend religious services significantly more frequently than Reformed members, consistent with the canonical obligation of Sunday Mass, while Reformed members display higher institutional confidence in their church, consistent with its unique role as the primary ethnic institution of the Hungarian minority community. Subjective religious salience does not vary by denomination, confirming that the identity-protection mechanism documented in prior comparative research operates with equal force across all three historical confessions. Denomination does not predict the institutional–personalised religiosity distinction once controls are applied, with age emerging as the dominant axis of variation on this dimension.

Three limitations should be acknowledged. First, the absence of PSU identifiers in the public use file means that standard errors may be modestly underestimated, and findings should be interpreted accordingly. Second, the small size of the Other Historical Protestant category ($n=47$) limits statistical power for contrasts involving Unitarians and Lutherans, who cannot be analysed separately. Third, the cross-sectional design precludes causal inference; the observed associations are consistent with the theoretical mechanisms proposed but do not establish them definitively.

These limitations notwithstanding, the article makes three contributions. It demonstrates that denomination retains independent explanatory power for behavioural practice and trends toward significance for institutional trust even within a community where the identity-protection mechanism suppresses denominational variation in subjective salience. It introduces the ZA7550 country-specific religiosity-type variable as a dependent variable in regression analysis for the first time. And it offers a replicable analytical framework for distinguishing the effects of denominational structure from those of minority status in communities where the two are systematically confounded—a framework applicable wherever comparable minority oversample data exist in the EVS or related cross-national archives.