Does the Name Still Matter? A Longitudinal Correspondence Study of Hiring Discrimination by Ethnicity and Gender in a Nordic Labour Market Authors: Akhlaq Ahmad Published: 2026 ABSTRACT ABSTRACTThis article examines changes in labour‐market discrimination against second‐generation immigrants in Finland over an 8‐year period, focusing on five groups. It draws on two large‐scale field experiments conducted in 2016 and 2024, in which 10,000 matched applications were sent to 2000 vacancies across five sectors. The findings show a nuanced shift in employer attitudes, reflecting both progress and persistence in discrimination. Discrimination against non‐European applicants has declined significantly, while bias towards European‐origin candidates shows no improvement. Non‐European men have seen a greater reduction in employer reluctance, while male disadvantage among European‐origin candidates has remained stable. Despite these improvements, results from both periods highlight enduring structural biases, with immigrant applicants continuing to receive fewer callbacks than their Finnish counterparts. The findings underscore that achieving a fair labour market remains a long‐term generational challenge rather than a quick policy fix, and provide support for ethnic hierarchies and social distance thesis, and the subordinate male target hypothesis.Keywords: correspondence experiment, discrimination, ethnicity, Finland, second‐generation immigrants Introduction The degree to which immigrants are integrated into the labour market is often regarded as a central measure of their broader inclusion in society. Given its importance, creating fair and equal access to employment has become a key priority in policy agendas across many European countries. Nevertheless, immigrants continue to face notable obstacles in securing employment opportunities, despite extensive policy efforts. These challenges are frequently attributed to so‐called ethnic penalties, referring to disadvantages associated with ethnic background, which manifest in restricted mobility across occupational hierarchies, lower income levels, greater exposure to precarious job contracts and reduced workplace satisfaction and well‐being (Paradies et al. 2015; Triana et al. 2010). The impact has been reported to be most pronounced among visible minorities and those from non‐European backgrounds (e.g., Branker 2017; Weichselbaumer 2017). More significantly, these disadvantages persist even among second‐generation immigrants—the focus of this study—who possess education and work experience acquired within the host country (Ahmad 2022; Weichselbaumer 2015; Zschirnt and Ruedin 2016). The weaker labour‐market outcomes among some second‐generation groups in many European and Nordic countries may suggest that structural barriers, rather than individual deficits, could be playing a significant role in shaping employment prospects.1 While the performance of first‐generation immigrants may be hindered by language barriers, lack of locally recognised qualifications, limited country‐specific cultural capital and insufficient access to resource‐rich social networks, these disadvantages are not expected to apply to the second generation. As such, their outcomes serve as a more accurate benchmark of immigrants’ long‐term integration. Discrimination has widely been recognised as a key contributing factor to their differential access to opportunities, a conclusion supported by numerous field experimental studies, including multi‐group correspondence designs comparing several ethnic groups within the same experiment (Ahmad 2020; Baert et al. 2017; Fibbi et al. 2022; Granberg et al. 2020; Pager et al. 2009; Pierné 2013; Thijssen et al. 2021; Vernby and Dancygier 2019; Zschirnt 2020;).2 However, while the existence of discrimination in hiring is well documented, the reasons for its persistence remain less well understood. The continued disadvantages associated with ethnic background raise a pressing question: why, despite sustained policy efforts, have interventions still not succeeded in eliminating labour‐market discrimination? To explain the persistence and complexity of discrimination in contemporary labour markets, conceptual frameworks spanning structural accounts to cognitive mechanisms offer distinct lenses on how exclusionary practices emerge and endure. Because the analysis that follows relies on a correspondence design, the empirical leverage lies in identifying differential treatment at the first stage of recruitment, which allows us to assess which theoretical accounts the observed patterns are most consistent with, even if the underlying organisational or societal mechanisms cannot be directly identified. While the introduction reviews a broad set of perspectives on discrimination, the empirical analyses focus most directly on a subset of these, namely the pure discrimination model, the ethnic hierarchies and social distance thesis and the subordinate male target hypothesis (SMTH) and double burden hypothesis (DBH) frameworks. Sociological accounts, particularly those grounded in conflict theory, interpret discrimination as a means by which dominant groups preserve their privileged access to scarce resources such as jobs and status (Tilly 1998; Tomaskovic‐Devey 1993). From this perspective, exclusion is strategic: it functions to sustain existing hierarchies and protect in‐group advantage (Reskin 2000). Social psychological approaches, by comparison, locate discrimination in mental processes of social categorisation. These theories argue that individuals tend to divide others into in‐groups and out‐groups as a cognitive shortcut, which, although useful for managing social complexity, can lead to biased perceptions and exclusionary behaviours (Fiske 1998; Fiske et al. 1999). Another framework called social norm theory explains how individual behaviour is shaped by shared social expectations, what people typically do (descriptive norms) and what is socially approved (injunctive norms) (Cialdini et al. 1991). Because individuals often conform to perceived norms to gain acceptance or avoid sanctions, discriminatory practices in the labour market may persist not only due to overt prejudice but also because exclusion becomes socially normalised. Employers may unconsciously follow established patterns of in‐group favouritism or out‐group avoidance, particularly when such behaviour is seen as typical within their social or professional environment. Thus, even in the presence of anti‐discrimination laws, bias may persist through status beliefs that become embedded in institutional routines and hiring practices, reinforced by widely shared normative expectations (see Bicchieri 2005; Ridgeway and Correll 2006). Economic research models discrimination through several mechanisms, including preference‐based ‘taste’ and information‐based statistical discrimination, alongside more recent accounts that emphasise implicit bias, labour‐market frictions and employer market power. The taste‐based discrimination model, originally formulated by Becker (1957), posits that some employers or clients hold prejudices and are willing to incur costs to avoid working with members of certain groups (Cain 1986; Kirschenman and Neckerman 1991). In contrast, the model of statistical discrimination assumes no explicit prejudice but rather limited information: employers use observable group characteristics, such as race or ethnicity, as proxies for unobservable traits like productivity (Arrow 1973; Phelps 1972). Although both models acknowledge differential treatment, they differ in motivation, prejudice versus uncertainty, and share a belief that market competition will eventually reduce discrimination. More recent economic contributions, however, also highlight implicit discrimination and the situational conditions under which implicit attitudes may shape decisions, including inattentiveness, time pressure, stress and cognitive load as well as ambiguity (Bertrand et al. 2005). Related work also shows how discriminatory outcomes can arise under labour‐market frictions and monopsonistic wage setting, where employers may be able to exploit immigrants’ lower labour‐supply elasticities (Hirsch and Jahn 2015). Overviews of contemporary economic research further emphasise the conceptual and empirical evolution of the field, including renewed debates on taste‐based versus statistical mechanisms and the centrality of experimental evidence (Guryan and Charles 2013; Lang and Kahn‐Lang Spitzer 2020; Neumark 2018). These developments are closely tied to the expansion of correspondence and audit methods within economics, which can credibly identify differential treatment in hiring, while also clarifying what experimental designs can and cannot identify about mechanisms and later outcomes in the employment process. More recent frameworks move beyond individual motives to emphasise the institutional and cultural conditions that shape discriminatory practices. Political economy approaches argue that discrimination stems not from isolated bias but from the systemic functioning of social institutions, which produces unequal outcomes through their rules, norms and structures (Sampson 2008). From this view, discrimination is institutionalised and operates independently of individual intention. Structural accounts similarly conceptualise racism as embedded in racialised social systems that structure social relations and shape groups’ life chances through unequal rewards and penalties, rather than as a matter of individual prejudice alone (Bonilla‐Silva 1997). Building on this perspective at the organisational level, Ray (2019) argues that organisations are themselves racial structures, such that routines, hierarchies, and the decoupling of formal rules from practice can legitimate and reproduce unequal outcomes. Similarly, the aesthetic labour framework introduces a cultural dimension, suggesting that hiring decisions, particularly in customer‐facing occupations, can be shaped by embodied traits such as appearance, accent and mannerisms, which are aligned with perceived market or client preferences (Nickson et al. 2012; Nickson et al. 2001; Warhurst et al. 2000). For example, in Canada, Lorente (2018) illustrates how maid agencies frequently use accents to perpetuate colonial stereotypes about migrant workers’ traits. While European accents tend to be linked with professionalism, Filipino accents are often interpreted as signs of limited competence, intelligence and confidence, leading to their exclusion from roles that demand effective communication (see also Creese and Kambere 2003; Ramjattan 2022). In addition to these theoretical explanations, research on ethnic hierarchies and social distance shows that discrimination is not uniform across out‐groups. Instead, groups are often ranked by perceived social desirability, with a broad consensus among in‐group members about these hierarchies (Hagendoorn 1995; Verkuyten and Kinket 2000). Numerous studies across countries support this pattern (e.g., Bobo and Zubrinsky 1996; Snellman and Ekehammar 2005). However, evidence on ethnic hierarchies in the labour market remains limited. Existing studies (Andriessen et al. 2012; Baert et al. 2017; Vernby and Dancygier 2019; Weichselbaumer 2017) suggest that such hierarchies shape employment opportunities and non‐native applicants’ chances of receiving a callback decline with greater socio‐cultural and ethnic distance from the majority group. Taken together, the frameworks discussed above suggest that discrimination is not solely economic or psychological in nature but also institutional, cultural and deeply relational. Meta‐analytic evidence shows that hiring discrimination remains widespread, but varies across national contexts, minority groups and time periods (Quillian et al. 2019; Quillian and Lee 2023; Zschirnt and Ruedin 2016). What remains less clear is whether, within the same country and research design, employer responses shift over time, and whether any change is evenly shared across groups or concentrated among those facing the sharpest baseline penalties. To address this gap, this article examines whether name‐based hiring penalties changed between 2016 and 2024 in Finland by replicating, with an identical sample size, a large‐scale correspondence study across the same immigrant groups and job sectors, with a focus on occupations requiring lower to intermediate levels of formal qualifications, including jobs such as cooks, waiters, shop assistants, office clerks and receptionists. The study offers a concrete picture of whether the significant disadvantages that second‐generation immigrant applicants experienced in 2016 have remained stable, improved or declined by 2024 and whether entrenched ethnic hierarchies prevailing in the labour market continue to reward them differentially based on their origin. The article especially explores whether the gendered patterns in access to the labour market have shifted over this eight‐year period. This study contributes to the literature in a number of ways. First, despite extensive evidence of name‐based penalties in hiring, there is much less scholarship on whether, and for whom, these penalties change as the institutional and discursive environment evolves. Replicating a large‐scale correspondence design in the same national context, among the same groups and across the same occupational sectors eight years apart allows us to move beyond static estimates and assess temporal dynamics in discriminatory behaviour. Because such longitudinal correspondence studies within a single country remain rare, this design also contributes to the wider comparative literature by offering a benchmark for understanding how ethnic penalties evolve over time in a Nordic welfare state. Second, Finland offers a theoretically informative setting: it combines extensive equality discourse and integration initiatives, creating a useful case for testing whether institutional efforts translate into measurable changes in employer behaviour or whether ethnic hierarchies and stereotypes remain durable. Third, while most correspondence studies establish the existence of discrimination, relatively few explicitly engage with competing theoretical explanations such as the pure discrimination model, ethnic hierarchies and social distance or the SMTH versus DBH frameworks. By replicating the design over time, this study not only documents whether discrimination persists but also provides evidence on whether temporal and group‐specific patterns align with the expectations of these accounts, even if the underlying mechanisms cannot be directly identified. Because this study examines change over time, national context is likely to be analytically important: institutional reforms, labour‐market conditions and public discourse may shape both employer decision making and the opportunities for discrimination to persist. The next section, therefore, outlines the Finnish setting, focusing on developments relevant to interpreting change between the 2016 and 2024 period, before the article discusses the research questions, hypotheses and experimental design. Labour Market and Immigrant Integration: The Finnish Setting Finland, whose migration history has been marked more by emigration than immigration, has experienced a mix of institutional, corporate and civic efforts aimed at promoting diversity and inclusion in the labour market. Measures such as pilot projects in anonymous recruitment, inclusive leadership training and municipal‐level diversity, equity and inclusion (DEI) programmes have sought to address long‐standing ethnic disparities. As a result, the value of diverse workplaces has been increasingly acknowledged, and discourse in the mainstream media has given growing visibility to themes of equality. Apart from ethical and legal dimensions, the justification for these broad‐scale efforts is well grounded: Finland is confronting an acute demographic challenge due to a sharply ageing population. Thus, there is a greater need for foreign labour to help sustain the country's welfare system and productivity. Yet, these developments may contrast with policy shifts that have tightened immigration controls, restricted labour mobility and weakened the residency security of immigrants and foreign workers. In recent years, a range of restrictive measures have been introduced, including stricter family reunification rules as well as new financial requirements and age thresholds. Eligibility criteria for Finnish nationality have also been made more demanding, with the residency requirement extended from 5 to 8 years, higher income thresholds, advanced Finnish‐language proficiency and expanded background checks. Cuts to integration services have further reduced funding for language instruction and employment counselling. Proposals have also been made to lower the intake of quota refugees from Muslim‐majority countries while prioritising Christian refugees. Public institutions such as the Non‐Discrimination Ombudsman have criticised these proposals as potentially violating equality principles. On the labour market front, despite a national narrative favouring labour‐market integration and Finland's broad anti‐discrimination legislation, the realities for immigrants tell a different story. According to a survey conducted by the Academic Engineers and Architects in Finland (TEK) and the Union of Professional Engineers in Finland, 41% of male respondents and 43% of female respondents reported discrimination in the job market.3 Other surveys show foreign experts are increasingly disillusioned, with many planning to leave Finland due to limited opportunities and a hostile social climate. Finland's global reputation among immigrants has also plummeted, falling from 16th to 51st in the Expat Insider ranking between 2023 and 2024.4 While the economy increasingly depends on immigrant labour, integration outcomes remain uneven and significant disparities persist between immigrant groups and the native Finnish population across key employment indicators. In 2024, the unemployment rate among foreign job seekers stood at 17.8%, more than double the 7.2% recorded for Finnish citizens (Employment Bulletin 2024). These gaps are not uniform: individuals from Asian and African backgrounds consistently face higher unemployment rates than those from European countries. Particularly striking are the 2022 figures, showing unemployment rates of 17% among immigrants from Russia and the former Soviet Union, 31% for Somalis and 38% for Iraqis (Employment Bulletin 2024). Additionally, immigrants are overrepresented in low‐paid occupations and are significantly more likely to work in positions below their qualifications than native Finns (Myrskylä and Pyykkönen 2014). They are also more commonly employed in temporary or part‐time roles with less favourable conditions, particularly in the retail, hospitality and service sectors (Ojala et al. 2024; Saukkonen 2017). They also continue to earn less than their Finnish counterparts. More importantly, young adults born in Finland to immigrant parents are less likely to be employed and often face name‐based discrimination during recruitment (Ahmad 2020). Employers also reportedly assume non‐Finnish names indicate meagre Finnish skills, even when applicants have fluent proficiency.5 Against this backdrop, the article next specifies the research questions and hypotheses that guide the comparison of the 2016 and 2024 correspondence studies. These focus on overall change in hiring penalties, the persistence of ethnic hierarchies and potential shifts in gendered patterns of callbacks. Research Questions and Hypotheses This article compares the findings of two large‐scale correspondence tests conducted in 2016 and 2024, in which fictitious, equally qualified applicants with Finnish, English, Iraqi, Russian and Somali backgrounds, who differed essentially only in their names, submitted a total of 10,000 applications to publicly advertised positions across Finland. It presents a concrete picture of how the reception of these groups has shifted over the 8‐year period in the Finnish labour market. It explores whether the significant disadvantages experienced in 2016 by young second‐generation immigrant candidates have receded or whether entrenched ethnic hierarchies continue to influence employers’ hiring decisions and sustain origin‐based disparities in access to employment. Given that both studies examined the same immigrant groups, economic sectors and employed comparable sample sizes, they provide a unique opportunity and a solid basis for analysing changes in recruitment trends among these groups over time. To systematically examine patterns of employer discrimination and their evolution between 2016 and 2024, the article tests several hypotheses. These hypotheses are framed as testable expectations about change over time and heterogeneity by origin group and gender, rather than as direct tests of underlying mechanisms. We begin with expectations about whether discrimination has changed between 2016 and 2024, drawing on developments in Finland. Over the past 8 years, Finland has introduced a range of measures to promote equality in the labour market, including the Ministry of Economic Affairs and Employment's Working Life Diversity Programme, EU‐supported integration efforts, corporate practices such as anonymous recruitment and inclusive leadership training and various DEI initiatives. In addition, discourse in the mass media has increasingly focused on equality issues and Finland's future need for migrant labour in light of its rapidly ageing population. These developments suggest a broader institutional and societal motivation to equitable employment chances across diverse groups. Accordingly, Hypothesis 1a posits that the level of discrimination in 2024 will be substantially lower than that observed in the 2016 baseline experimental study. At the same time, evidence from field experiments indicates that hiring discrimination often remains stable over time, with one meta‐analysis reporting no aggregate decline across decades of experimental evidence (Quillian et al. 2017). More recent meta‐analyses further underline that change, where it occurs, may be uneven across national contexts and groups (Quillian et al. 2019; Quillian and Lee 2023). In this light, and consistent with descriptive evidence of persistent labour‐market gaps by origin in Finland (Statistics Finland 2024; OECD 2018), discriminatory practices may be resilient to institutional change. In Becker's (1957) terms, such persistence is compatible with preference‐based avoidance by employers, coworkers or clients, but it may also reflect durable group stereotypes, uncertainty‐driven inference, norm compliance or routinised screening practices that reproduce unequal treatment at the first stage of recruitment even when applicant qualifications are held constant. Therefore, Hypothesis 1b states that discrimination will not decline markedly between 2016 and 2024 and that immigrant‐origin applicants will continue to face unequal treatment despite being matched on age, education and work experience acquired in Finland. The ethnic hierarchies and social distance thesis argues that societies rank ethnic groups based on their proximity to the dominant group along multiple dimensions. Those perceived as more culturally, racially or religiously distant are more likely to face greater exclusion (Ford 2011; Sides and Citrin 2007; Sniderman et al. 2004). In line with this framework, Hypothesis 2 postulates that if ethnic discrimination persists in 2024, it will be more pronounced against non‐European than European‐origin job candidates, who are perceived as culturally closer to the Finnish majority population. Evidence from the 2016 correspondence study revealed that immigrant men received significantly fewer callbacks than women of the same ethnic background (Ahmad 2020). The findings were consistent with the SMTH, which argues that minority men are more likely to be perceived as threatening or socially deviant and are therefore disproportionately subject to labour‐market discrimination (see Berdahl and Moore 2006; Sidanius and Veniegas 2000). They were not consistent with the DBH, which holds that, compared to immigrant men, immigrant women may face more discrimination because of their combined status as immigrants and women. The DBH resonates with intersectional scholarship, which emphasises that disadvantages may be produced through intersecting social positions and cannot always be understood by examining gender and immigrant status separately (Crenshaw 1989; Crenshaw, 1991). At the same time, intersectional disadvantage need not be additive in a simple sense. Pedulla (2014), for example, shows that negatively stereotyped categories can combine in complex ways in hiring evaluations, such that one stigmatised identity may partly offset stereotypes attached to another. This insight is conceptually relevant here because it helps situate the contrast between the DBH and the SMTH within a broader literature that does not assume cumulative disadvantage in every case. Since the first correspondence experiment in 2016, Finland has witnessed a growing societal and policy emphasis on gender equality. Against this backdrop, Hypothesis 3 proposes that employer attitudes towards male immigrant applicants have become more equitable, resulting in a narrowing of the gender gap in 2024. However, if equally qualified male applicants continue to receive significantly fewer callbacks than their female counterparts, this would remain consistent with the SMTH framework. Data, Research Design and Methods This article's data is drawn from a larger project that used the correspondence methodology, widely applied across disciplines, to measure differential treatment in hiring in the Finnish labour market. In this project, two field experiments were carried out in 2016 and 2024. In correspondence tests, closely matched applications are sent to advertised positions from fictitious applicants who differ only in a specific characteristic (e.g. name, ethnicity, gender), allowing researchers to isolate the impact of that variable while holding all else constant. As detailed in Ahmad (2025), five applicants of Finnish, English, Russian, Iraqi and Somali origin submitted equivalent job applications to vacancies posted on Finland's national employment platform (te‐palvelut.fi). Applicants were matched in age, education, work experience and vocational diploma but differed only in one respect: name. This name‐based signalling strategy builds on a longer correspondence‐study tradition using ethnic name signals (Riach and Rich 2002), while also following the influential hiring‐discrimination study by Bertrand and Mullainathan (2004), which demonstrated how racially distinctive names can be used to identify differential treatment while holding applicant credentials constant. Each field experiment consisted of applying to 1,000 job openings. Thus, across the two experiments, a total of 2,000 jobs were applied for, and altogether 10,000 job applications were submitted to various firms. The large size of the dataset helps ensure that the observed patterns reflect real relationships rather than random correlations. Central to correspondence testing is the use of names that clearly signal ethnic origin. Applicant names were displayed prominently on the CVs. Each cover letter stated the applicant's mother tongue—such as Arabic, English, Russian or Somali—to further highlight ethnic background. Each application included a tailored cover letter and CV—stylistically varied but substantively equivalent—written in fluent Finnish and conveying enthusiasm, competence and adaptability. The order of assigning letters to the applicants was changed each time they responded to a new position. All immigrant applicants clearly stated that they had obtained their education and work experience in Finland to avoid employer bias related to foreign qualifications. Cover letters also briefly articulated why the candidate was well‐suited to the job, emphasising personal attributes such as teamwork, communication skills and work ethic. Each application matched the vacancy, educational attainment, vocational credentials and the amount and type of work experience presented on the CVs, in line with the requirements of each job posting. Where a posting required a particular qualification or prior experience, this was included in the CVs, and where no such requirements were stated, applications did not necessarily list additional credentials beyond those needed to present a credible profile. Importantly, for every vacancy, all five applicants were constructed to be equivalently qualified for that specific position and differed only in name‐based origin signalling. CVs were created using a CV generator software and included detailed information on education, work history, language proficiency (Finnish and English), IT skills, and the institutions where qualifications were obtained. It was clear from the letters and CVs that all immigrant applicants had either been born in Finland or had arrived at a very young age. Applicants were aged 24 to 28, with small variations introduced to prevent detection. This age range was selected to avoid conflating ethnic discrimination with age‐related bias, which has been well documented in the labour economics literature (e.g., Wilson et al. 2019; Ye and North 2025). Jobs targeted fell into five occupational categories: restaurant and catering, retail, cleaning, clerical and customer service, which together represent an important point of labour‐market entry for young applicants. These included roles such as cooks, waiters, shop assistants, office clerks and receptionists. Applications were submitted via email, the standard method in Finland. The order of submitting applications was changed for each job to prevent any applicant from being favoured based on submission timing. Responses were categorised as positive (interview invitations) or negative (rejections or no reply). While the issue of whether a negative response or no response from an employer constitutes equal treatment or non‐observation has been debated in the experimental literature (Riach and Rich 2002), this classification offers a useful methodological framework. The analyses in this article rely on a combination of logistic regression, binary comparisons and chi‐square tests to examine the statistical significance of variations in callback rates across five applicant groups. Binary and chi‐square tests are used for the analysis shown in Panel A of Table 1, and logistic regression models are employed in Panel B of Tables 1 and 2. While the chi‐square test evaluates whether an association exists between two categorical variables (in this case, whether a callback was received and the study year 2016 versus 2024) by comparing observed and expected frequencies under the assumption of independence, logistic regression, in contrast, models the likelihood of a binary outcome, such as receiving a callback, based on multiple explanatory variables. In logistic regression, the coefficients represent adjusted estimates of these changes, considering variability and the influence of additional covariates. Each coefficient is interpreted relative to a reference category. In this study, the Finnish group serves as the reference when comparing the differences between ethnic backgrounds of the applicants. A higher coefficient for any other group indicates a greater likelihood of receiving callbacks compared to the Finnish group. Statistical significance is assessed using a t‐statistic, which, in large samples, follows a normal distribution under the null hypothesis. In addition to regression estimates, the analysis employs two descriptive measures that are widely used in audit and correspondence studies to summarise differential treatment in hiring: the net discrimination rate and the relative callback rate (Bertrand and Mullainathan 2004; Neumark 2018; Riach and Rich 2002). The net discrimination rate (NDR) summarises the imbalance in positive responses between majority and minority applicants, while the relative callback rate (RCR) indicates how many more applications minority job seekers must typically submit to obtain a comparable number of interview opportunities. The formulae for NDR and RCR used in this study are provided in the notes under Table 1. TABLE 1A comparison of the net discrimination rates for different ethnicities in 2016 and 2024. F/EF/RF/IF/S201620242016202420162024201620241. No response/both rejected5615455725495875555985642. Both invited220264190186111173871483. Applicant1 invited, Applicant2 not1701522002302792433032684. Applicant2 invited, Applicant1 not49393835232912205. Net discrimination rate28 %25 %38 %43 %62 %48 %72 %57 %6. Relative callback rate1.451.371.711.882.912.063.942.48Statistical tests for net discrimination rate change 2016–2024Combined rate0.2650.4050.550.645 z −1.5202.278−6.293−7.009 p 0.1290.0230.0000.000 Note: Net discrimination rate = (Row 3 ‐ Row 4)/(Row 2 + Row 3 + Row 4); Relative callback rate = (Row 2 + Row 3)/(Row 2 + Row 4); F = Finnish, E = English, R = Russian, I = Iraqi, S = Somali; For example, F/E means Applicant1 is Finnish, Applicant2 is English. A comparison of the net discrimination rates for different ethnicities in 2016 and 2024. Note: Net discrimination rate = (Row 3 ‐ Row 4)/(Row 2 + Row 3 + Row 4); Relative callback rate = (Row 2 + Row 3)/(Row 2 + Row 4); F = Finnish, E = English, R = Russian, I = Iraqi, S = Somali; For example, F/E means Applicant1 is Finnish, Applicant2 is English. TABLE 2Logistic regression analysis of callback odds by ethnicity and dataset year.95% Confidence IntervalBLowerUpper t‐statisticSig.Exp(B)(Intercept)−0.447−0.574−0.320−6.8980.0000.639Dataset 2016 (ref.)Dataset 20240.108−0.0710.2871.1850.2361.114Finnish (ref.)Somali−1.761−1.965−1.557−16.9000.0000.172Iraqi−1.419−1.600−1.237−15.3260.0000.242Russian−0.772−0.912−0.633−10.8580.0000.462English−0.552−0.682−0.423−8.3550.0000.576Dataset 2024 * Somali0.5000.2410.7603.7790.0001.649Dataset 2024 * Iraqi0.3840.1490.6193.2070.0011.468Dataset 2024 * Russian−0.148−0.3490.053−1.4480.1480.862Dataset 2024 * English0.059−0.1150.2320.6610.5091.060 Note: Dependent Variable: Response to the application submitted (reference category = No callback); Model: (Intercept), Dataset, Ethnicity, Dataset * Ethnicity. The logistic regression model accounts for the clustering of the data by employer. For this purpose, a complex sample module of the SPSS was used. Logistic regression analysis of callback odds by ethnicity and dataset year. Note: Dependent Variable: Response to the application submitted (reference category = No callback); Model: (Intercept), Dataset, Ethnicity, Dataset * Ethnicity. The logistic regression model accounts for the clustering of the data by employer. For this purpose, a complex sample module of the SPSS was used. The dataset comprises responses to five applications submitted for each individual job. Since employers may vary in how they respond to applicants, this introduces the possibility of response patterns that differ across employers. Such variation can lead to intra‐employer correlations, thereby undermining the core assumption of independent observations required in regression analysis. These correlations can distort standard error estimates and thus compromise statistical inference. To correct for this, a complex sample logistic regression procedure was used. Within the SPSS programme, a complex sampling plan was specified, assigning equal weights to cases and including a clustering variable to reflect employer‐level responses. This approach allowed the analysis to account for within‐cluster dependencies and provided more reliable estimates of standard errors. Submitting five applications to the same vacancy strengthens within‐job comparability by holding constant vacancy‐specific characteristics, and it allows callback chances to be compared across the different origin groups within the same advertised job. However, recent methodological work notes that matched designs with multiple applications to the same employer can raise concerns about within‐vacancy dependence and potential interference or ‘search externalities’ in employers’ evaluation of an applicant pool (Hollister et al. 2025; Phillips 2016; Rivera and Tilcsik 2017). These concerns are addressed by accounting for employer‐level clustering in the analyses and by designing the applications in a way that ensured they did not appear identical. Results Do Second‐Generation Immigrants Face Less Discrimination in 2024? We begin by examining whether employer reluctance to recruit second‐generation immigrant jobseekers, whose credentials matched those of Finnish candidates in terms of age, education, vocational qualifications and work experience, has declined in 2024 relative to 2016. Table 1 provides the initial descriptive assessment of hypothesis 1a, which posited that, given the extensive public debate in the mass media on equality issues and the numerous institutional efforts to promote immigrant integration in the Finnish labour market over the past 8 years, the level of ethnic discrimination in 2024 would be substantially lower than that reported in the 2016 baseline correspondence study. Table 1 presents net discrimination rates for various groups. As is evident, in 2016, the highest levels of net discrimination were observed among Somali and Iraqi applicants, with rates of 72% and 62%, respectively. This means that in more than six or seven out of every ten matched application pairs, the Finnish applicant received a callback while the Somali or Iraqi applicant did not. By 2024, these figures have decreased to 57% for Somalis and 48% for Iraqis, indicating notable improvement in their relative treatment by employers, even though these rates remain notably high. In contrast, English and Russian applicants experienced little to no meaningful change over time. Net discrimination against English‐named applicants decreased only marginally, from 28% to 25%, while Russian applicants saw a slight increase from 38% to 43%. This increase may also reflect contextual influences, notably the post‐2022 war in Ukraine, rather than a gradual temporal trend in employer aversion towards Russian‐named applicants. Another way to quantify ethnic discrimination more concretely is through the use of the relative callback rate, which appears in row 6 of the table. While net discrimination rates capture patterns of differential employer behaviour, the relative callback rate provides a more concrete understanding of how much harder minority job seekers must try to secure the same number of interview opportunities as native applicants. In 2016, Somali candidates had to submit nearly four times as many job applications as their Finnish counterparts to receive a similar number of callbacks (3.94). By 2024, this burden had eased somewhat, though they still needed to send more than twice as many applications (2.48). Iraqi applicants also show a comparable improvement, with their relative callback rate decreasing from 2.91 to 2.06. In contrast, English and Russian applicants see only marginal changes in their situation. Thus, while hypothesis 1a anticipated a substantial decline in overall discrimination due to policy reforms and public discourse, the findings reveal only partial confirmation. At the same time, the limited change for English and Russian applicants, alongside the continued high net discrimination rates for Somali and Iraqi applicants, is consistent with Hypothesis 1b. In this sense, the results suggest that a ‘taste for discrimination’ may persist among some recruiters or clients despite expanded equality initiatives and public debate. To determine whether the observed changes in net discrimination rates are statistically significant, z‐tests were conducted on the differences in net discrimination rates between 2016 and 2024 for each ethnic group. The null hypothesis in each case stated that there is no change in the net discrimination rate between the two time points. As we can see, the null hypothesis can be rejected for Somali and Iraqi applicants (p = 0.000 for both), indicating significant reductions in discriminatory outcomes. For Russian applicants, the null hypothesis is also rejected (p = 0.023), whereas for English applicants it cannot be rejected (p = 0.129). These results thus reflect a heterogeneous pattern of change, with improvement limited to the most disadvantaged groups. To further assess the robustness of the descriptive findings reported in Table 1 and to control for potential confounders, logistic regression was conducted with interaction terms between ethnicity and dataset year, as presented in Table 2. The dependent variable was whether the applicant received a callback, with “no callback” as the reference category. The model includes main effects for ethnicity and dataset year, as well as interaction terms to test changes in treatment across time. It should be noted that, in models with interaction terms, the main effects refer to the situation in which the interacting variable is fixed at its reference category. In this case, the ethnicity coefficients refer to differences between ethnic groups in the 2016 dataset, while the interaction terms indicate how those differences changed in 2024 relative to 2016. As we can note, in 2016, all four immigrant groups had significantly lower odds of receiving a callback than Finnish applicants. Specifically, Somali (Exp(B) = 0.172), Iraqi (0.242), Russian (0.462), and English (0.576) applicants are all significantly disadvantaged (p = 0.000 for each group), which confirms substantial ethnic penalties in hiring. To examine whether these penalties changed over time, the model included interaction terms between dataset year (2024) and ethnicity. The null hypothesis in this context posits that there are no differences in the odds of receiving a callback between 2016 and 2024 for any ethnic group, that is, the interaction coefficients equal zero. The results show significant increases in the odds of receiving a callback for Somali (p = 0.000, Exp(B) = 1.649) and Iraqi applicants (p = 0.000, Exp(B) = 1.468), indicating measurable improvement in their treatment by employers over time. In contrast, the interaction terms for Russian (p = 0.148, Exp(B) = 0.862) and English applicants (p = 0.509, Exp(B) = 1.060) are not significant, supporting the conclusion that their relative positions have not meaningfully improved. These findings thus reinforce those from the descriptive analysis in Table 1 but do so through a more rigorous inferential framework. In brief, the 2016–2024 comparison indicates a statistically meaningful narrowing in callback gaps for Somali and Iraqi applicants, whereas English and Russian applicants show no comparable improvement across the two study periods. For Russian‐named applicants in particular, the interaction effect is not statistically significant, so no change over time can be inferred. However, the low callback rates for this group may plausibly reflect broader geopolitical developments, namely, heightened tensions following the war in Ukraine, which may have intensified negative public sentiment towards this community. This contextual explanation does not alter the empirical finding of no statistically significant change, but situates the limited employer response to Russian applicants within the broader societal climate. Is There Greater Discrimination against Non‐European than European Applicants? Next, we test hypothesis 2, which states that if ethnic discrimination persists in 2024, it will be more pronounced against non‐European compared to European‐origin candidates, who are perceived as culturally closer to the Finnish majority population. This was done by calculating the differences in net discrimination rates and using the z‐tests (not shown here) to determine whether the observed differences would lead to either rejecting or confirming the hypothesis. In 2016, all differences between European and non‐European groups were statistically significant in favour of the European groups. The smallest European‐non‐European difference was between the net discrimination rate of Russian (38%) and Iraqi applicants (62%), which was statistically significant (p = 0.000). By 2024, the net discrimination rate for Russian applicants has risen slightly to 43%, while the rate for Iraqi applicants has declined to 48%. The resulting difference of five percentage points indicates that the gap in hiring discrimination between the two groups has narrowed over time. However, the difference is still significant. However, other comparisons between European and non‐European groups continue to reveal robust and statistically significant disparities. English applicants, with a net discrimination rate of 25%, experience significantly less discrimination than Iraqi applicants at 48% (p = 0.000). Similarly, Russian applicants (43%) face significantly less discrimination than Somali applicants (57%) (p = 0.000), and English applicants (25%) are significantly favoured over Somali applicants (57%) as well (p = 0.000). Overall, these comparisons show statistically significant European‐non‐European disparities in 2024. They therefore provide clear support for Hypothesis 2, with non‐European applicants facing higher net discrimination rates than European‐origin applicants. Has the Gender Gap in Callbacks Narrowed? In this section, we test hypothesis 3, which posits that employer discrimination against male applicants will diminish in 2024 as a result of the increased societal and policy emphasis on gender equality in Finland over the past eight years. Before turning to the interaction models, Table 3 first describes callback rates by gender within each ethnic grouping and study year to establish the basic pattern. As we can note, the non‐European group shows a clear gender gap in both years, with women receiving more callbacks than men. For Somali applicants, male callback rates increased from 6.8% in 2016 to 15.8% in 2024, while female rates rose from 13.0% to 17.8%. For Iraqi applicants, male callback rates increased from 9.2% to 17.8%, whereas female rates increased from 17.6% to 22.6%. Substantively, these figures suggest that the male penalty that characterised the 2016 baseline diminished in magnitude in 2024, even though a gender gap remains. TABLE 3Callback rates by applicant ethnicity, gender and dataset year, %. Applicant gender EthnicityDataset yearMaleFemaleTotalSig.Finnish201633.844.239.0 *** 202437.445.841.6 ** Total 35.6 45.0 40.3 *** English201624.629.226.9 202426.434.230.3 ** Total 25.5 31.7 28.6 ** Russian201619.226.422.8 ** 202420.024.222.1 Total 19.6 25.3 22.5 ** Iraqi20169.217.613.4 *** 202417.822.620.2 Total 13.5 20.1 16.8 *** Somali20166.813.09.9 ** 202415.817.816.8 Total 11.3 15.4 13.4 ** Total201618.726.122.4 *** 202423.528.926.2 *** Total 21.1 27.5 24.3 *** Note: Statistical significance refers to the chi‐square test for gender differences in callback rates within each ethnic group and dataset year: *** p < 0.001, ** p < 0.01, * p < 0.05. In each ethnic group, there are 500 male and 500 female applicants in both 2016 and 2024. Callback rates by applicant ethnicity, gender and dataset year, %. Note: Statistical significance refers to the chi‐square test for gender differences in callback rates within each ethnic group and dataset year: *** p < 0.001, ** p < 0.01, * p < 0.05. In each ethnic group, there are 500 male and 500 female applicants in both 2016 and 2024. To test whether the gender gap changed significantly between 2016 and 2024, we now estimate logistic regression models with a gender by year interaction in Table 4. In this analysis, the five original applicant groups were combined into three categories: Non‐European (Somali and Iraqi), European (English and Russian) and Finnish (the reference group). Aggregating these groups was useful for capturing empirically meaningful trends that were evident at the descriptive level but not detectable in disaggregated regression models. This empirical strategy was also guided by sociocultural considerations. Somali and Iraqi applicants were combined because, in the Finnish context, they are frequently viewed as culturally distant minorities, often associated with refugee migration and subject to negative societal perceptions. The inclusion of English and Russian applicants in the European immigrant group follows a similar logic, as they are generally perceived as more culturally proximate, sharing a common religious background and exhibiting greater similarity in values and social norms with the Finnish mainstream population. To ensure the appropriateness of this grouping strategy, separate logistic regression analyses (not shown here) were conducted within the non‐European and European subgroups. These analyses confirmed that there were no significant gender differences in callback effects between Somali and Iraqi applicants, nor between English and Russian applicants. Thus, it was considered methodologically appropriate to operationalise ethnicity using a trichotomous variable for the subgroup comparisons in Table 4. TABLE 4Change in the odds of receiving a callback between 2016 and 2024, by gender and by ethnic group (non‐European, European and Finnish), logistic regression analysis.95% Confidence Interval Ethnic group Parameter B LowerUpper t‐statistic Sig. Exp(B) Non‐European(Intercept)−1.711−1.929−1.494−15.4400.0000.181Dataset 2016 (ref.)Dataset 20240.3370.0420.6322.2440.0251.401Female (ref.)Male−0.731−1.100−0.362−3.8850.0000.481Dataset 2024 * Male0.5050.0330.9772.0970.0361.657European(Intercept)−0.954−1.129−0.780−10.7160.0000.385Dataset 2016 (ref.)Dataset 20240.069−0.1730.3100.5580.5771.071Female (ref.)Male−0.317−0.570−0.065−2.4620.0140.728Dataset 2024 * Male0.006−0.3500.3620.0320.9751.006Finnish(Intercept)−0.233−0.410−0.056−2.5870.0100.792Dataset 2016 (ref.)Dataset 20240.065−0.1850.3140.5080.6111.067Female (ref.)Male−0.439−0.695−0.183−3.3630.0010.645Dataset 2024 * Male0.092−0.2670.4520.5040.6141.097 Note: Dependent Variable: Response to the application submitted (reference category = no callback).; Model: (Intercept), Dataset, Gender, Dataset * Gender; The logistic regression model accounts for the clustering of the data by employer. For this purpose, a complex sample module of the SPSS was used. Change in the odds of receiving a callback between 2016 and 2024, by gender and by ethnic group (non‐European, European and Finnish), logistic regression analysis. Ethnic group Parameter B t‐statistic Sig. Exp(B) Note: Dependent Variable: Response to the application submitted (reference category = no callback).; Model: (Intercept), Dataset, Gender, Dataset * Gender; The logistic regression model accounts for the clustering of the data by employer. For this purpose, a complex sample module of the SPSS was used. Table 4 presents the results of logistic regression analyses examining the effects of applicant gender, time period and their interaction on the odds of receiving a callback for a job candidate. The null hypothesis in Table 4 assumes that there is no change in gender differences in callback rates within each ethnic group across time. Again, the main effects refer to the situation in which the other variable in the interaction term is fixed at its reference value; in this case, gender differences refer to the year 2016. In 2016, male applicants had less than half the odds of receiving a callback compared to female applicants (p = 0.000, Exp(B) = 0.481). However, the interaction term between gender and time period is also statistically significant (p = 0.036, Exp(B) = 1.657), which means that male applicants’ callback odds improved significantly relative to those of female applicants between 2016 and 2024. These findings support Hypothesis 3 for non‐European applicants, suggesting that increased awareness of gender equality, perhaps driven by public discourse and institutional efforts, may have contributed to a reduction in the relative disadvantage faced by non‐European men. Importantly, no evidence supports the DBH framework in the context of this comparative study: Non‐European women were not doubly disadvantaged, but rather consistently outperformed their male counterparts in both time periods. The preference for female over male immigrants across various occupations has also been confirmed by a recent labour market study on employer discrimination in Germany, Sweden and Austria (Fossati et al. 2024). Among European applicants, women also receive more callbacks than men in both periods, as Table 3 shows. For English applicants, male callback rates increased slightly from 24.6% in 2016 to 26.4% in 2024, whereas female callback rates increased from 29.2% to 34.2%. In contrast, Russian male applicants showed only a marginal increase from 19.2% to 20.0%, while Russian female applicants showed a slight decline from 26.4% to 24.2%. Overall, these descriptive patterns suggest a persistent gender gap with little evidence of narrowing. Consistent with these descriptive patterns, Table 4 provides evidence for a statistically significant gender gap among European applicants in 2016: men are less likely than women to be invited to a job interview. They have approximately 31% lower odds of receiving a callback compared to female applicants (Exp(B) = 0.728, p = 0.014). There is still a persistent gender‐based disadvantage across the two periods. However, unlike the non‐European group, the interaction between gender and time is not significant (p = 0.975, Exp(B) = 1.006). This indicates that the magnitude of the male disadvantage did not change significantly across the two study periods. The findings are thus consistent with the SMTH framework, indicating that employer aversion continues to shape the prospects of equally qualified European immigrant men in the Finnish labour market, as they persistently lag behind their female counterparts in callback rates. Discussion The aim of the study reported in this article was to assess whether name‐based hiring penalties in Finland changed between 2016 and 2024 by replicating a large‐scale correspondence design across the same groups and job sectors using an identical sample size. The results point to an uneven trajectory of change. On the one hand, discrimination against the most disadvantaged non‐European‐origin groups declined, suggesting that employer responses are not entirely static. On the other hand, discrimination persists across all immigrant origin groups, and the broader stratification pattern privileging European‐origin candidates over non‐European‐origin applicants remains intact. These findings align with recent meta‐analyses, which show that hiring discrimination is both persistent and context dependent, and that change over time may be heterogeneous rather than uniform across groups and countries (Quillian et al. 2019; Quillian and Lee 2023). Overall, the findings provide mixed support for the study's hypotheses. The evidence is consistent with Hypothesis 1a for Somali and Iraqi applicants but not for English or Russian applicants, while the continued penalties observed across groups align with Hypothesis 1b. The patterning of outcomes is also consistent with Hypothesis 2, indicating a durable European‐non‐European hierarchy in discriminatory treatment across both waves. Finally, the results are in line with Hypothesis 3 in a group‐specific way: the gender gap narrows among non‐European‐origin applicants, whereas it remains largely unchanged among European‐origin applicants. The most notable change over the eight‐year period concerns Somali and Iraqi applicants, whose relative position improved compared to 2016. One plausible interpretation is that the combination of equality initiatives, organisational DEI efforts and sustained public discourse on non‐discrimination over the past eight years may have contributed to reducing the sharpest penalties for groups that faced the steepest disadvantages at baseline. At the same time, the persistence of substantial disadvantages even in 2024 underscores that any gains should not be read as evidence that discriminatory hiring has been eliminated. In other words, improved callback rates coexist with continued barriers at the earliest stage of recruitment, which remains a key gatekeeping point for second‐generation applicants whose qualifications and country‐specific human capital were held constant in the design. For English and Russian applicants, the absence of improvement—together with the descriptive indication of deterioration for Russians in some comparisons—suggests that change is not a general feature of the Finnish labour market over the period. However, because the interaction effect for Russian applicants is not statistically significant, any interpretation of this descriptive pattern must remain cautious. One plausible contextual factor is the geopolitical situation following the 2022 war in Ukraine, which may have shaped public sentiment and employer perceptions, but this cannot be confirmed within the limits of the experimental design. The results are broadly compatible with the pure discrimination model in the sense that unequal treatment remains observable even when applicants are matched on qualifications, including language skills, Finnish education and work experience. According to the pure discrimination model, a ‘taste’ for discrimination may persist among some recruiters or clients despite formal equality measures and public commitments to inclusion. While the evidence here cannot adjudicate motives directly, the combination of persistent penalties and limited improvement for some groups is less consistent with a view in which market forces or general policy efforts reliably erode discriminatory behaviour over time. The empirical observations of this study, therefore, reinforce the practical limits of relying on broad institutional reform alone when discriminatory decision making is embedded in routine recruitment practices and normative expectations. At the same time, persistence of discrimination is not uniform across groups, and the consistent patterning of penalties points to a durable hierarchy in perceived proximity to the majority population. Across both study waves, non‐European‐origin groups were more heavily penalised than their European‐origin counterparts, providing support for the ethnic hierarchies and social distance thesis. The narrowing gap between Russians and Iraqis by 2024 suggests that hierarchies can shift at the margins, yet the overall ordering remains: applicants perceived as more culturally, racially or religiously distant continue to face systematically higher penalties. This persistence is consequential because it implies that observed improvement for some groups does not necessarily represent a dismantling of stratification. Rather, it may reflect modest movement within a durable hierarchy in which the majority in‐group continues to allocate opportunities unevenly across minority out‐groups. This interpretation is consistent with the broader correspondence study literature, which shows that discrimination varies by group and context and that visible minority groups often occupy the most disadvantaged positions. More broadly, the magnitude of the callback disadvantages observed in Finland places the country at the higher‐discrimination end of the comparative correspondence‐study literature. This is particularly evident for Somali and Iraqi applicants, whose relative callback burdens remain substantial even in 2024. In this respect, the Finnish case is not exceptional in showing the presence of discrimination, but it does stand out in the severity and persistence of penalties for some groups relative to many other Western settings (Quillian et al. 2019; Zschirnt and Ruedin 2016). The Finnish case thus contributes longitudinal evidence that hierarchies can remain stable even when some of the steepest penalties soften. The findings also show a hiring penalty along gender lines. Across all groups and periods, women receive more callbacks than men, offering no support for the double burden hypothesis at least in the context of the Finnish labour market. Instead, the pattern is consistent with the subordinate male target hypothesis, in which immigrant men face greater disadvantage relative to women of the same origin. Crucially, the interaction results indicate that temporal change in the gender gap is group‐specific. The narrowing of the male penalty among non‐European‐origin immigrant applicants suggests that gendered ethnic disadvantage is, under some conditions, mutable over time. One plausible explanation in this regard is that non‐European‐origin men faced the steepest baseline penalties in 2016, likely due to their racialised visibility, cultural and religious distance and associations with refugee migration. Therefore, this severe initial penalty may have created greater potential for improvement over time. In the context of heightened public discourse on equality issues, institutional initiatives and targeted labour‐market integration measures in recent years, employers may have become more cautious or reflective when evaluating male applicants from these groups in 2024 compared to 2016. A second possibility is a ‘familiarity effect’: with increased exposure to Somali and Iraqi immigrants in both public life and the labour market, employers’ tendency to associate these groups with cultural distance or social risk may diminish over time. A third important factor may be labour demand in a rapidly ageing Finnish society, which could increase employers’ willingness to consider candidates from groups that had previously been strongly disadvantaged. Although none of these mechanisms can be tested directly with correspondence data, they provide coherent interpretations of why the gender gap narrowed for non‐European‐origin immigrant applicants while remaining stable for European‐origin applicants. In contrast, no such temporal improvement is observed for European‐origin male candidates. The interaction between gender and time for this group is not statistically significant, indicating that their relative disadvantage compared to female applicants has remained unchanged over time. The gender gap among them was less extreme to begin with, leaving less room for meaningful improvement to be detected through interaction effects. Also, applicants of English and Russian backgrounds were already perceived as culturally proximate and employable in 2016 and were less exposed to exclusionary mechanisms. Moreover, since European men are not viewed as particularly vulnerable or disadvantaged, employers may not have revised their perceptions or behaviours over time, leading to a consistent pattern of lower callbacks that does not shift across the two study periods. The contrast between these groups may underscore the role of both visibility and targeted intervention in shaping how employer discrimination evolves. This study also has some limitations. While it captures discrimination at the initial stage of recruitment, which is substantively important, it cannot speak to later stages such as interview treatment, wage offers or on‐the‐job outcomes. Correspondence testing can identify patterns that are consistent with particular theoretical accounts, but it cannot directly distinguish whether unequal treatment reflects taste‐based prejudice, statistical inference, customer preferences or perceived norms, nor can it measure how employers interpret specific name signals. A further consideration concerns the scope of occupational coverage. The sectors included in the experiment comprise a heterogeneous set of positions that range from entry‐level roles to more demanding customer‐facing and administrative jobs, and wage and skill levels may vary substantially within the same broad occupation and sector. Nevertheless, the study does not capture hiring processes in higher‐paid or higher‐status professional fields such as IT, engineering or other specialised occupations. To the extent that discriminatory screening differs across segments of the labour market, the level of discrimination documented here should be interpreted as specific to the sectors and job types covered by the design. Further correspondence tests in additional occupational domains would be needed to assess whether penalties are higher or lower in other parts of the labour market in Finland. Future research could combine correspondence designs with complementary approaches, including employer surveys, audit studies incorporating interviews or qualitative work on organisational recruitment routines. Additional longitudinal replications would also be valuable for assessing whether the observed improvements for non‐European‐origin groups represent an emerging trend or a period‐specific shift. Conclusions While previous correspondence experiments have documented persistent ethnic penalties in recruitment, few have examined how these patterns change over time. This comparative study addresses that gap by providing rare longitudinal evidence on hiring discrimination against second‐generation immigrant applicants in Finland by replicating a correspondence study in 2016 and 2024 using the same groups, occupational sectors and sample size. The findings show that name‐based hiring penalties remain substantial, but change is heterogeneous across groups. More specifically, three main patterns emerge. First, discrimination declined for Somali and Iraqi applicants, indicating that the steepest penalties can soften over time. However, these groups continue to face marked disadvantage, showing that progress does not equate to equal access. Second, European‐origin candidates continue to be favoured over non‐European‐origin applicants, supporting the persistence of an ethnic hierarchy shaped by social distance and symbolic boundaries. Third, women receive more callbacks than men across all groups, contradicting a double burden account in this setting. The gender gap narrows only among non‐European‐origin applicants, suggesting that gendered ethnic disadvantage can shift for the most penalised groups while remaining stable elsewhere. The broader implications of the findings of this study are that institutional initiatives and equality discourse may contribute to incremental change, but they are unlikely to eliminate discrimination without interventions that also address routine recruitment practices and the social expectations that sustain unequal treatment. For policy, the results underline the need for measures that embed equality principles in everyday hiring decisions and create accountability mechanisms that make discriminatory gatekeeping visible and consequential to employers. For research, the study underscores the value of longitudinal replications for distinguishing stable hierarchies from genuine behavioural change and for clarifying which groups benefit from shifting contexts and which remain persistently disadvantaged. In brief, the findings underscore both the scope for change and the persistence of bias, highlighting the importance of sustained efforts rather than one‐off policy fixes. Funding This work was supported by grants from the Kone Foundation (202407040) and Otto A. Malm Foundation (13852). Conflicts of Interest The author declares no financial or non‐financial competing interests nor any potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Acknowledgments Open access publishing facilitated by Helsingin yliopisto, as part of the Wiley ‐ FinELib agreement. Endnotes However, this pattern is not universal, and evidence from Canada points to more heterogeneous second‐generation trajectories, with average outcomes comparable to the third‐and‐higher generations alongside group‐specific penalties once education and location are taken into account (Chen and Hou 2019; Hou and Picot 2011; see also Aydemir and Sweetman 2007). Longitudinal correspondence studies involve repeated observations of the same groups, often over an extended period, to track changes over time. https://www.tek.fi/en/news‐blogs/finland‐drives‐out‐international‐experts https://www.internations.org/expat‐insider/2024/finland‐40460. https://satakieli.haaga‐helia.fi/article/we‐are‐not‐a‐burden‐it‐is‐hard‐for‐an‐immigrant‐to‐get‐a‐fair‐chance‐in‐the‐finnish‐job‐market. References