The Structure of Vocational Interests in Finland: A Study of the Fit of Holland’s RIASEC Model Based on a Representative Nationwide Sample of Career Counselling Clients

Abstract

This is the first study to examine the fit of Holland’s (1997) RIASEC model of vocational interests within the Finnish cultural context. We conducted secondary analyses of vocational interest data collected over a 20-year period by the Finnish Public Employment Services (PES). The data were extracted from the PES client registry and comprise a nationally representative sample of 20,108 career counselling clients (aged 10–62 years; 62% women), who had completed a RIASEC-based, interest inventory Kiinnostukset [Interests] as part of career counselling offered by psychologists in the Finnish PES. Results from the Randomization Test of Hypothesized Order Relations (RTHOR), Multidimensional Scaling (MDS) and Cosine Function Models (CFM) showed that Finnish vocational interests conform satisfactorily to Holland’s RIASEC circumplex structure in terms of ordinal relationships. Overall, the degree of fit (RTHOR Correspondence Index = .67) was comparable to levels obtained from U.S.-based samples. Differences in model fit by gender or by level of education were not statistically significant. Furthermore, MDS analyses replicated the expected RIASEC circular ordering, although both MDS and CFM results suggest an approximate rather than perfect circumplex fit. Overall, our findings offer support for the continued use of Holland’s RIASEC model in Finnish career counselling and guidance practice.

Keywords

Holland code vocational interests Finland RIASEC circumplex validity

Introduction

Career decisions made by individuals in terms of career choices have far-reaching consequences both for the individuals themselves and for the societies they live in. Indeed, person-vocation (P-V) fit, i.e., the level of congruence between individuals’ interests, abilities and the characteristics and the requirements of their chosen vocations (Holland, 1997; Vogel & Feldman, 2009) has been demonstrated to predict many important work-related outcomes such as work performance (Nye et al., 2012; Van Iddekinge et al., 2011b) and work satisfaction (Tsabari et al., 2005).

In career counselling and guidance practice, the assessment of individuals’ vocational interests by means of various interest inventories is a common way for career practitioners to help individuals choose careers that fit them well (Hanna & Rounds, 2020). While commonly used interest inventories have been demonstrated to both measure similar basic vocational interests (e.g., construction and music) (Chu et al., 2022) and to accurately predict occupational and educational choices (Hanna & Rounds, 2020), they are not interchangeable (Chu et al., 2022) nor infallible. Considering how important career decisions are to the individual’s life, career practitioners should always ensure that the interest inventories they use show evidence of their reliability and validity in their practice context, so that their use in helping clients can be ethically justified.

For career practitioners working outside the United States, where many popular models and measures of vocational interest have been developed, the question of cultural equivalency (i.e., whether these models accurately describe the structure of vocational interests in other cultural contexts) is also vitally important to consider (Tao et al., 2019). In general, the fit of U.S.-based interest models in other cultures varies highly (Glosenberg et al., 2019; Rounds & Tracey, 1996) and is often not supported (Tao et al., 2019). Thus, where no evidence of equivalency exists, no presumption of such should be made. Therefore, the purpose of this article is to examine the degree of fit of Holland’s (1997) prominent RIASEC model of vocational interests in the cultural context of Finland, where no previously published research exists to either support or discourage its use - and consequently the use of the Finnish RIASEC-based interest inventory Kiinnostukset [Interests]. In a broader sense, we also aim to add to the body of cross-cultural research on the structure of vocational interests, where few studies exist examining the fit of the RIASEC model in the Nordic countries (with the exception of Iceland, see Einarsdóttir et al., 2002).

Holland’s RIASEC Model of Vocational Interests

Vocational interests can be defined as “trait-like preferences to engage in activities, contexts in which activities occur, or outcomes associated with preferred activities that motivate goal-oriented behaviors” (Rounds & Su, 2014). From a developmental perspective, vocational interests become relatively stable during early adolescence (around the ages of 13 to 14), with their stability increasing considerably after the age of 18 and generally enduring for most people over several decades (Hoff et al., 2018; Low et al., 2005; Rottinghaus et al., 2007). Both birth year and age are minimally related to vocational interests, suggesting that differences in interests across age groups are negligible (Leuty & Hansen, 2014). The common practice of using an individual’s current vocational interest profile as a basis for long-term educational and career planning can therefore be generally supported (Rottinghaus et al., 2007).

While several theories of vocational interests have been presented, the most widely used model is Holland’s (1997) RIASEC model (Hanna et al., 2024), which has been used to organize interest types in vocational interest inventories since the 1970s (Chu et al., 2022). Holland’s (1997) RIASEC model sets forth several key assumptions about the nature and consequences of vocational interests. First, that there are six distinct vocational interest types (Realistic, Investigative, Artistic, Social, Enterprising and Conventional, hence “RIASEC”) that can be used to measure and categorize most people’s vocational interests as well as different work environments dominated by these types. The Realistic (R) type shows preference to work involving the hands-on use of tools and machines (e.g., mechanic). The Investigative (I) type prefers work related to research and thinking (e.g., biologist). The Artistic (A) type prefers work with creative freedom (e.g., writer). The Social (S) type shows preference for work involving helping other people or contributing to society (e.g., teacher). The Enterprising (E) type prefers work involving leadership and persuasion (e.g., manager). Last, the Conventional (C) type prefers work related to systematic manipulation of data (e.g., bookkeeper).

Furthermore, a key assumption of Holland’s (1997) model (known as the “calculus hypothesis”) is that these six vocational interest types can be arranged in a hexagonal structure based on their relative similarity, in which interest types adjacent to each other (such as Realistic and Investigative) share a higher degree of similarity than interest types alternate (e.g., Realistic and Artistic) or opposite (e.g., Realistic and Social) to each other. Indeed, the hexagonal structure of the six types is one of the best replicated findings in all of vocational psychology (Rounds, 1995). Hence Holland’s (1997) RIASEC model is sometimes also called the hexagonal, circular or circumplex model of vocational interests. A brief description of the interest types and the theoretical relationships between them is presented in Figure 1.

Figure 1.

RIASEC interest types and their relative hexagonal arrangement

According to Holland (1997), people seek out environments which match the patterns of their interests, allowing them to e.g., utilize their skills and abilities, embody their values and take on agreeable roles. Both positive and negatives outcomes can thus be predicted based on the degree of fit between the individual and the environment (see e.g., Nye et al., 2012; Tsabari et al., 2005; Van Iddekinge et al., 2011b). This is known as the “congruency hypothesis”. Indeed, vocational interests have been shown to have incremental validity over both cognitive ability and personality traits in predicting job performance (Van Iddekinge et al., 2011a) and career success (Rounds & Su, 2014; Stoll et al., 2016). This makes the assessment of vocational interests an appealing practice in career counselling, provided that the theoretical model used in conceptualizing interests captures the structure of vocational interests in the local population. Within the United States, past research has demonstrated that Holland’s hexagonal model adequately represents the interest structure of men and women in racially and ethnically diverse samples (Day et al., 1998; Fouad & Mohler, 2004; Morris, 2016), however, certain enduring gender differences are often observed, which we briefly describe next.

Gender and Cultural Differences in Vocational Interests

Several large-scale studies have investigated gender differences in vocational interests (Baerg MacDonald et al., 2023; Morris, 2016; Su et al., 2009; Tao et al., 2022). As a general trend, men tend to show higher levels of Realistic and Investigative interests, while women tend to have stronger Artistic and Social interests, with results being more mixed for the Enterprising and Conventional domains. Gender differences in vocational interests are at their apex in adolescence and decrease in adulthood (Hoff et al., 2018). Furthermore, Morris (2016) demonstrated that these gender differences have been stable over the past 80 years. However, while the overall pattern of gender differences in vocational interests is rather robust, results from cross-cultural research also suggest that cultural factors such as country-level gender equality (Du et al., 2025) and cultural dimensions such as power distance, individualism, motivation toward achievement and success and uncertainty avoidance (Hofstede & McCrae, 2004) may affect the magnitude gender differences in vocational interests. For example, Tao et al. (2022) demonstrated moderating effects, beyond the effect of country-level gender inequality, between the cultural dimensions of uncertainty avoidance, indulgence and power distance and the size of gender differences, with higher uncertainty avoidance, higher indulgence and lower power distance contributing to larger differences.

In the Nordic context, Icelandic studies (Einarsdóttir & Rounds, 2020; Hoff et al., 2020) focusing on gender differences in vocational interests followed the commonly observed pattern of gender differences: women exhibited higher Social and Artistic interests while men showed higher Realistic and Investigative interests. However, while Iceland and Finland are both Nordic countries, research has found that in terms of Hofstede’s (2001) cultural dimensions, Iceland is more culturally similar to countries such as the United States, Canada and United Kingdom than Denmark, Sweden and Norway (Guðmundsdóttir et al., 2015) particularly due to comparatively higher levels of cultural individualism. Finland, in turn, also differs from the other Nordic countries by sharing more cultural similarities with the Netherlands and Ireland in terms of lower levels individualism and higher levels of uncertainty avoidance that the other Nordic countries (Guðmundsdóttir et al., 2015). Thus, while the Nordic countries share some cultural similarities, e.g., in terms of low power distance (Guðmundsdóttir et al., 2015) and generally egalitarian societies, it would be important to not uncritically use results from Icelandic studies as a stand-in for all other Nordic countries. While we do not expect Finland to dramatically differ from the overall pattern of gender differences in vocational interests established in the current literature, these cultural differences may affect their magnitude (Tao et al., 2022) – provided, of course, that the RIASEC model is appropriate to use in Finland.

As a general rule, the local cultural context should be taken into account in career practice. In spite of this, it is all too common for measures developed and based in one culture (chiefly the United States) to be used in other cultures with a tacit assumption of equally valid administration (Tao et al., 2019). However, as previously mentioned, this equivalency should not be taken for granted. We next turn to the issue of cultural equivalency of the structure of vocational interests.

The Fit of Holland’s (1997) RIASEC Model Outside the U.S.

Famously, in the meta-analysis by Rounds and Tracey (1996) the cross-cultural equivalence of Holland’s (1997) circular model was generally not supported, with only Iceland, Japan and Israel demonstrating an acceptable level of fit to the RIASEC model (notably with Iceland showing superior fit to the U.S. benchmark). More recently Glosenberg et al. (2019) studied the comparative fit of various vocational interest models (including the RIASEC model) and demonstrated varying levels of fit between countries, with higher country-level economic development being associated with better fit. Based on their results, they also cautioned against assuming generalizability of the model without supportive evidence and called for additional research on the cross-cultural validity of vocational interest models (Glosenberg et al., 2019).

This lack of supporting evidence is also an issue in the Nordic countries (i.e., Denmark, Finland, Iceland, Norway and Sweden), where the validity of the RIASEC model has only been formally studied (and supported) in Iceland (Einarsdóttir et al., 2002). In spite of this, Finland, Sweden and Norway have all offered publicly available RIASEC-based online interest assessments as part of their public employment services (Einarsdóttir et al., 2020). For example, in Finland the online vocational choice test AVO Ammatinvalintaohjelma [AVO Vocational Choice Program] (KEHA-keskus (n.d.)) and its associated interest inventory Kiinnostukset [Interests] served the general public for over 26 years, but somewhat shockingly no published reports exist regarding its reliability or validity (Harry Pulliainen, personal communication, March 13, 2025). Indeed, no formal and current published scientific evidence in support of or against the use of Holland’s RIASEC model appears to exist in Finland, Sweden, Norway or Denmark, except for the data provided by Glosenberg et al. (2019) showing some initial evidence of acceptable fit levels in these countries. However, while welcome, such data is not enough to draw any final conclusions. As Glosenberg et al. (2019) point out, they administered their interest inventory in English to participants recruited from an English-language website, thus their sample is likely to consist of English-speaking residents of countries where English is not necessarily a common first language. Consequently, they conclude their results are likely to be meaningfully different from the results that may be gained from testing locals in their first language. Furthermore, Einarsdóttir et al. (2020) note that despite its fit, Holland’s RIASEC model only partially describes the Icelandic vocational interest structure, the full dimensions of which are better captured by indigenously developed emic interest inventories and models (including basic vocational interests such as fishing, humanities and public administration, see Einarsdóttir & Rounds, 2013) and that this may also be the case in other Nordic countries. They conclude that it is necessary for the other Nordic countries to critically evaluate such theories and to empirically test instruments that accompany them. Based on the above considerations we believe that research on the fit of Holland’s RIASEC model in Finland’s cultural context is not only warranted, but long overdue.

Validating Holland’s (1997) RIASEC Model – Factorial and Spatial Approaches

The basic task of studying the (cross-cultural) validity of Holland’s (1997) RIASEC model is to establish if and how well a particular dataset conforms to what is predicted by the RIASEC model. Two forms of structural hypotheses have been proposed: the circular order model and the circumplex model (also known as the hexagonal model) (Rounds et al., 1992). According to the circular order model, which derives from Holland’s “calculus hypothesis”, it is expected that the magnitude of correlations between adjacent (e.g., R and I), alternate (e.g., R and A) and opposite (R and S) pairs of RIASEC types varies systematically based on the theoretical closeness of the types (with correlations being higher the more closely related the two types are, see Figure 1). As a consequence, any statistical test assessing the fit of the calculus hypothesis simultaneously tests whether or not the six types are correctly ordered in a circular fashion according to the RIASEC model (i.e., R-I-A-S-E-C) (Rounds et al., 1992). While spatial analysis techniques such Multidimension Scaling (Jaworska & Chupetlovska-Anastasova, 2009; Mugavin, 2008) can be used to create a spatial representation of the RIASEC circumplex for visual examination, Rounds and Tracey (1996) have argued that a method that explicitly evaluates Holland’s calculus hypothesis is to compare the predicted order relations with the actual magnitude of the correlations (i.e., correlation matrices) among the types using methods such as the Randomization Test of Hypothesized Order Relations (Hubert & Arabie, 1987). To this day, this remains one of the key methods of assessing the fit of the RIASEC model.

The circumplex (or hexagonal) model, in turn, is a more restrictive form of the circular order model, with the added assumption that the six interest types form a circle with equidistant points around the circumplex (thus also forming a hexagon within the circle). The circumplex model therefore adds the expectations that the interpoint distances between interest types are equal – and thus adjacent, alternate and opposite variables around the circle are all equally correlated, with the magnitude of correlations following Holland’s (1997) calculus hypothesis (Guttman, 1954; Rounds et al., 1992). This kind of strict circumplex fit can be studied using Cosine Function Models (a form of Confirmatory Factor Analysis), which test whether the correlations in the data follow the expected cosine function (i.e., that similarity between types depends on their angular distance in a perfect circumplex) (Nagy et al., 2009).

A great deal of RIASEC validation work has thus been based on the spatial model, which views the six interest types as occupying positions in a two-dimensional circumplex structure, with similarity represented as geometric proximity. Indeed, the more traditional factorial model, where vocational interests are treated as six correlated latent dimensions, appears to be less commonly used. Su et al. (2019) note that traditional confirmatory factor analysis (CFA) of vocational interests have generally yielded poor results (Warlick et al., 2018), which Su and colleagues (2019) suggest is due to the inherent complexity of broad-band interests dimensions combined with the restrictive assumptions of CFA that each latent factor must be highly unidimensional, while secondary loadings in interest domains are common. Consequently, Su et al. (2019) recommend the use of Exploratory Structural Equation Modeling (ESEM), which integrates both confirmatory and exploratory approaches, allowing for the specifying of underlying factor structures while simultaneously allowing secondary loadings to be freely estimated when investigating complex dimensional structures.

The Present Study

The overall goal of the present study was to empirically examine the structure of vocational interests in Finland and its correspondence with Holland’s (1997) RIASEC model of vocational interests. In practical terms, this meant our first goal was to evaluate the psychometric properties of the Finnish computer-based vocational interest inventory Kiinnostukset using confirmatory factor analysis and exploratory structural equation modelling to establish its reliability and validity as a RIASEC interest measure for both research and counselling purposes.

Our second goal was to study the fit of Holland’s RIASEC model of vocational interests in the Finnish cultural context based on response data from the aforementioned Kiinnostukset interest inventory. Specifically, we used a variety of sophisticated statistical methods (i.e., the randomization test of hypothesized order relations (RTHOR), multidimensional scaling (MDS) and cosine function models (CFM)) to analyze response data from a nationally representative sample (n = 20,108) of people (aged between 10 and 62, 64% women) who had sought career counselling at the Finnish Public Employment Services (PES) between August 2002 and May 2022. We hypothesized that the RIASEC model would fit the data, as was the case in Iceland, since these two Nordic countries share many cultural similarities and prior research including some Finnish-respondents has shown promising results in this regard (Glosenberg et al., 2019).

In sum, our study thus aims to establish the reliability and validity of the use of Holland’s (1997) RIASEC model to both measure and theoretically conceptualize vocational interests in Finland, both for the benefit of Finnish career counselling and guidance practice and to add a previously missing Finnish perspective to the body of literature assessing the cross-cultural fit of Holland’s (1997) RIASEC model in the Nordic countries.

Method

Participants

Our sample consisted of data from 20,108 Finnish-speaking career counselling clients who had completed a bespoke, computer-based Finnish vocational interest inventory Kiinnostukset [Interests] as part of career counselling provided by the Finnish Public Employment Services’ (PES) career counselling psychologists. The data in our sample covers a period of nearly 20 years, from fifth of August 2002 to 10th of May 2022. All testing was conducted on-site at various PES office locations nationwide using dedicated testing computers. The demographic data of clients were entered by the counselling psychologists responsible for setting up the career assessment batteries for each client.

The gender distribution of the clients in the data was 12,817 women (64%), 7290 men (36%) and one client with missing gender data. The reported age range of the clients was 10 to 62 years (M = 29.28, Mdn = 26.00, Mode = 19.00, SD = 10.90). In terms of age groups, 3.3% of the sample was aged 15 years or below, 18.4% was aged 16–19 years, 22.6% was aged 20–24 years, 25.1% was aged 25–34 years, 18.2% was aged 35–44 years, 11.1% was aged 45–54 years, and 1.2% was aged 55 years or above.

The clients’ levels of vocational or professional education (i.e., education excluding basic education and general upper secondary education) was reported as follows: 9600 clients (45%) had no vocational or professional education, 8332 (39%) had a vocational upper secondary qualification or equivalent (European Qualifications Framework (EQF) level 4 or 5), 2121 (10%) had a post-secondary education (a now defunct level of education previously offered in Finland) or university of applied sciences degree (EQF level 6) and 1141 (5%) had any level of university degree (EQF level 6, 7 or 8). Finally, the education data for 188 clients (1%) were missing.

The geographical distribution of the clients in the sample was obtained based on the locations of the PES offices where the interest inventory was taken. The geographical distribution of the sample is described in Table 1, which also includes an at-glance comparison to the distribution of the Finnish-speaking population provided by Statistics Finland Statistics Finland (n.d.). Location data were missing from 176 people, who were consequently omitted from Table 1.

Table 1.

Distribution of Study Sample by Administrative Region and Gender With Comparisons to the General Finnish-Speaking Population*

Administrative region of Finland	Number of clients from region in data	% of total sample data	% of entire Finnish-speaking population living in region	% of women in data from region	% of women in Finnish-speaking population in region
Uusimaa	4911	24.64	27.62	66.40	51.88
Southwest Finland	1104	5.54	8.70	67.20	51.52
Satakunta	1659	8.32	4.20	64.60	50.54
Kanta-Häme	236	1.18	3.35	67.80	50.62
Pirkanmaa	1187	5.96	10.53	68.50	50.92
Päijät-Häme	833	4.18	3.98	64.00	51.15
Kymenlaakso	2236	11.22	3.07	66.90	50.32
South Karelia	1487	7.46	2.41	60.10	49.80
South Savo	694	3.48	2.59	64.80	50.52
North Savo	369	1.85	4.98	61.50	50.52
North Karelia	263	1.32	3.21	70.70	50.00
Central Finland	1693	8.49	5.46	56.50	50.18
South Ostrobothnia	319	1.60	3.84	73.40	49.95
Ostrobothnia	234	1.17	1.51	61.50	50.17
Central Ostrobothnia	18	0.09	1.24	66.70	50.17
North Ostrobothnia	596	2.99	8.39	59.90	49.57
Kainuu	1657	8.31	1.40	56.80	49.25
Lapland	376	1.89	3.51	54.00	49.97
Åland	60	0.30	0.03	61.70	64.86

Note. *Based on population data current as of 2023.

Notably, the roughly 2:1 ratio of women to men in our sample is consistently higher than the ratio in the general population of each region, which may reflect, e.g., a difference in the rates of independently seeking or being offered (and subsequently agreeing to participate in) career counselling services offered by the Public Employment Services.

Furthermore, we note that although the geographic distribution of our sample is statistically different from the distribution of the Finnish-speaking population (χ²(18, N = 4 777 408) = 17,584.1, p < .001, Cramér’s V = 0.06), the effect size is below the conventional threshold for a small effect size (V = 0.10), suggesting minimal practical difference between the two distributions. We conclude that because our data sample is large, regionally diverse and collected as part of genuine career counselling processes, it is particularly well-suited to the task of assessing the structure of vocational interests in Finland.

Procedure

The data for this study were obtained from the Finnish Development and Administration Centre for ELY Centres and TE Offices (KEHA Centre), which is the competent authority for granting access to the Finnish PES registry data for research purposes. In Finland, neither legislation nor the guidelines of The Finnish National Board on Research Integrity (TENK) require ethical review by an ethics committee for research based purely on registry and documentary data or archive data, unless there are data security risks associated with data that is combined from several sources. Furthermore, according to Finnish legislation the requirement for informed consent is not required in case of registry-based research. In order to manage possible data security risks, we drafted and submitted a Data Management Plan to the KEHA Centre and minimized the amount of demographic data requested in our research permit application to include only basic de-identified data necessary to appropriately describe the data sample. The research permit was approved by the KEHA Centre on 28th of July 2022 and a copy of the data was securely transferred over to the principal researcher for analysis.

Measures

The premier Finnish vocational interest inventory, Kiinnostukset [Interests], is a computer-based interest inventory originally developed by the Finnish Ministry of Labor based on Holland’s (1997) RIASEC model. The test is freely available online as part of the public vocational choice program AVO Ammatinvalintaohjelma [AVO Vocational Choice Program] (KEHA-keskus (n.d.)). It consists of 72 items assessing levels of interest in various vocational activities or qualities (e.g., “Would you be interested in editing photos?”) on a 4-point Likert scale (1 = Not at all interested, 4 = Very much interested). The sum scores from six sets of 12 items each form the six main interest scales based on the RIASEC model (Realistic, Investigative, Artistic, Social, Enterprising and Conventional).

Data Analysis

We used various complementary statistical methods and software tools to study the fit of Holland’s (1997) RIASEC model to our data. Basic descriptive statistics were obtained using IBM SPSS Statistics (IBM SPSS Statistics for Windows Version 28, 2020). Possible gender differences in vocational interests were analyzed using a one-way multivariate analysis of variance (MANOVA) using gender as the independent variable and the six RIASEC scales as dependent variables.

In order to examine the structural validity of the Kiinnostukset interest inventory, we ran both Confirmatory Factor Analysis (CFA) and Exploratory Structural Equation Modeling (ESEM) using the Mplus statistical software (Muthén & Muthén, 1998-2017). Su et al. (2019) reviewed past literature and noted that in vocational interest research exploratory methods are more common than using CFA, the latter of which has generally yielded poor model fit results. However, similar poor results have also been demonstrated when running CFAs on several widely used personality inventories (Hopwood & Donnellan, 2010). Su et al. (2019) suggest that this may be partially explained due to the inherent complexity of interest and personality dimensions and by the restrictive assumptions of CFA requiring high levels of unidimensionality for each latent factor. Consequently, Su et al. (2019) recommend the use of ESEM, as it integrates both confirmatory and exploratory approaches, allowing for the specifying of underlying factor structures while simultaneously allowing secondary loadings to be freely estimated. We thus performed both CFA and ESEM, using the following fit indices and values for acceptable model fit: the root mean square error of approximation (RMSEA) (≤.08), the Comparative Fit Index (CFI) (≥.90), the Tucker-Lewis index (TLI) (≥.90) and the standardized root mean square residual (SRMR) (≤.08) (Bentler, 1990; Brown, 2006; Browne & Cudeck, 1992; Hu & Bentler, 1999).

For assessing the fit of Holland’s (1997) RIASEC model to our data, our first method of analysis was the Randomization Test of Hypothesized Order Relations (RTHOR) (Hubert & Arabie, 1987), computed in R Statistical Software (version 4.4.2) (R Core Team, 2025) using the RANDALL and RANDDIF programs developed by Tracey (1997). The RTHOR is a confirmatory analysis of a hypothesized pattern of order relations of any (dis)similarity matrix (e.g., a correlation matrix) (Tracey, 1997) and has long been used to study the fit of interest data to the RIASEC and other models of vocational interests. The RTHOR yields a correspondence index (CI) ranging from −1 to 1, with a CI value of 1 indicating perfect fit and a value of −1 indicating perfect misfit to the model defined in the analysis. Furthermore, the RTHOR produces a p-value as the probability of obtaining the CI value under the null hypothesis of random ordering – in other words, a p-value below .05 is seen as evidence that the correlation matrix is consistent with the circular model defined in the analysis (Hubert & Arabie, 1987). In our case, we are interested in our data’s fit to Holland’s circular order hypothesis (i.e., the correlations between adjacent interest types (e.g., R and I) are expected to be stronger than the correlations between alternate interest types (e.g., R and A), and the correlations of alternate types are expected to be stronger than the correlations between the opposite types (e.g., R and S), see also Figure 1) (Deng et al., 2007) - in other words verifying whether the data conforms to the circumplex structure in terms of ordinal relationships. A benchmark value of .70 with a lower bound 99% confidence interval of .66 demonstrates a degree of fit similar to those obtained in US-based samples (Rounds & Tracey, 1996).

While the RTHOR offers information about the general degree of circumplex fit, it does not help easily visualize the degree of potential (mis)fit. Therefore, the second statistical method we used to study was ordinal (i.e., non-metric) multidimensional scaling (MDS) using the R package smacof (Mair et al., 2022). MDS is an exploratory data analysis technique that can be used to condense large amounts of data into a relatively simple spatial map that relays important relationships in the data (Mugavin, 2008; for a more detailed review of the MDS procedure see Jaworska & Chupetlovska-Anastasova, 2009). Two diagnostic fit indexes can be used to assess the quality of the MDS model: Stress and Variance Accounted For (VAF), the latter of which is expressed as the R² coefficient. Stress values indicate the model’s goodness-of-fit and range from 0 to 1, with values equal or below .10 considered excellent and values from .15 and up unacceptable (Kruskal & Wish, 1978). In turn, the minimum acceptable R² coefficient is .6 (Meyer et al., 1992), with values closer to 1 indicating better fit. In the present study, we are particularly interested in using the visuospatial map for evaluating Holland’s hypothesis that the six interest types are ordered in a circular manner and in the expected order of R-I-A-S-E-C (see Figure 1 before).

While MDS helps visualize the ordering of types, importantly, it does not explicitly enforce or test for a circular structure in the data. Therefore, for our final analysis we utilized a cosine function model (CFM; Nagy et al., 2009), a special case of Browne’s circular stochastic process with a Fourier series (CSPF; Browne, 1992) using a combination of the R package holland (Hartmann et al., 2021) and Mplus version 8.6 (Muthén & Muthén, 1998-2018). The CFM is a form on Confirmatory Factor Analysis (CFA) (Nagy et al., 2009) that can be used to evaluate the fit of data to various circumplex models (such as Holland’s (1997) RIASEC hexagon) by testing whether the correlations in the data follow the expected cosine function (i.e., that similarity between types depends on their angular distance in a perfect circumplex). Model fit was evaluated using the χ² statistic, the root mean error of approximation (RMSEA; Browne & Cudeck, 1992), the Comparative Fit Index (CFI; Bentler, 1990) and the standardized root mean square residual (SRMR; Jöreskog & Sörbom, 1993). Acceptable fit for the aforementioned indicators are values ≤.08 for RMSEA and SRMR (Browne & Cudeck, 1992; Hu & Bentler, 1999) and values ≥.90 for CFI (Bentler, 1990; Brown, 2006).

Results

Descriptive Statistics and Reliabilities

The means and standard deviations of all RIASEC interest scale scores by gender are shown in Table 2, while scale reliabilities by gender are presented in Table 3. Furthermore, correlations between RIASEC scale scores at the total sample level and by gender are presented in Table 4.

Table 2.

Means and Standard Deviations of the RIASEC Scale Scores by Gender

	Realistic	Investigative	Artistic	Social	Enterprising	Conventional
Total	14.64 (6.62)	14.54 (7.81)	13.32 (8.71)	15.59 (8.09)	10.77 (7.15)	10.73 (6.36)
Women	13.73 (6.29)	13.82 (7.69)	14.64 (8.63)	17.27 (7.76)	11.14 (7.10)	11.22 (6.58)
Men	16.23 (6.89)	15.79 (8.12)	10.99 (8.37)	12.64 (7.81)	10.13 (7.18)	9.87 (5.85)

Note. Score range for each scale is 0 to 36.

Table 3.

Reliabilities of the RIASEC Interest Scales by Gender (Mcdonald’s ω Total)

	Realistic	Investigative	Artistic	Social	Enterprising	Conventional
Total sample	.83	.90	.90	.92	.90	.87
Women	.82	.90	.90	.90	.89	.88
Men	.86	.91	.91	.92	.91	.86

Table 4.

Correlations Between RIASEC Scale Scores

Total sample
	R	I	A	S	E	C
R
I	.32**
A	.17**	.46**
S	.12**	.34**	.37**
E	.19**	.30**	.26**	.62**
C	.22**	.38**	.10**	.24**	.40**

By gender (women above diagonal)
	R	I	A	S	E	C
R		.34**	.28**	.14**	.14**	.23**
I	.27**		.49**	.33**	.27**	.36**
A	.10**	.51**		.28**	.24**	.04**
S	.23**	.48**	.42**		.61**	.17**
E	.31**	.37**	.28**	.66**		.35**
C	.29**	.46**	.16**	.33**	.48**

Note. **p < .01. R = Realistic, I = Investigative, A = Artistic, S = Social, E = Enterprising, C = Conventional.

The multivariate effect of gender on the RIASEC scale scores was significant (Wilks’ L = .756, F (6, 20,100) = 1080.18, p < .001, η² = .244), indicating that approximately 25% of the overall variability in vocational interest scores could be attributed to gender. Univariate ANOVAs found significant gender differences across all RIASEC scales: In general, men scored higher on the Realistic (F (1, 20,105) = 684.35, p < .001, η² = .033) and Investigative scales (F (1, 20,105) = 292.99, p < .001, η² = .014) while women scored higher in the Artistic (F (1, 20,105) = 851.57, p < .001, η² = .041), Social (F (1, 20,105) = 1648.63, p < .001, η² = .076), Enterprising (F (1, 20,105) = 93.99, p < .001, η² = .005) and Conventional scales (F (1, 20,105) = 210.88, p < .001, η² = .010). To facilitate comparisons of effect sizes with Su et al. (2009), the partial η² values were converted to approximate Cohen’s d values. The effect sizes ranged from moderate in the Social scale (d = −.57) to small in the Artistic (d = −.41), Realistic (d = .37), Investigative (d = .24) and Conventional scales (d = −.21) and negligible in the Enterprising scale (d = −.14).

All correlations between the interest scales were positive and statistically significant at the .01 level, both when examined at the total sample level and when examined separately by gender. The highest correlation at the total sample level appears between the Social and Enterprising scales (r = .62). For the most part, the correlations in our sample appeared to be consistent with the expectations of Holland’s (1997) hexagonal RIASEC model in that correlations between adjacent interest scales (e.g., Realistic and Investigative, r = .32) appeared to be larger than correlations between alternate (e.g., Realistic and Artistic, r = .17) and opposite (e.g., Realistic and Social, r = .12) scales.

The Structural Validity of the Kiinnostukset Interest Inventory

Fit indices for the CFA and ESEM are reported in Table 5. To account for local dependencies among items with similar content, correlated residuals (which were generally small) were added between items within the same interest factor. These modifications were guided by theoretical considerations and modification indices, helping to improve model fit beyond what was explained by the primary latent factors alone. For transparency, we have provided the original ESEM fit indices in Table 5. The ESEM model with correlated residuals showed a superior model fit to the CFA, which in turn failed to achieve acceptable fit. The results from the ESEM with correlated residuals are presented in Table 6. Overall, the ESEM results showed that all six RIASEC dimensions clearly emerged with items generally showing highest loadings on the correct factors and few strong cross-loadings (except for item 55.”Do a hobby requiring craftsmanship” on the Realistic scale, which showed a strong cross-loading with the Artistic scale, likely reflecting craftmanship’s link to creativity and aesthetics). Although two items (34. Work to help developing countries; 37. Lift weights or do martial arts) showed atypical loading patterns, with the former loading stronger on the Enterprising scale than the intended Social scale (likely reflecting a potential reading of managing aid work rather than giving out the aid) and the latter item loading weakly on multiple scales (suggesting poor item functioning), these deviations do not compromise the theoretical coherence of the overall model. As shown earlier in Table 3, the interest scale reliabilities were satisfactory, ranging from .83 (Realistic) to .92 (Social). We conclude that the RIASEC factor structure for the Kiinnostukset interest inventory was therefore supported, although with some suggestions for future item improvements Having established adequate structural validity, we began to formally test the fit of Holland’s (1997) RIASEC model to our data by using the randomization test of hypothesized order relations (RTHOR).

Table 5.

Fit Indices for the Kiinnostukset Interest Inventory

Model	χ²	df	p	CFI	TLI	RMSEA	SRMR	Acceptable
CFA	154,982.82	2411	.00	.81	.79	.06	.08	No
ESEM (original)	127,446.43	2139	.00	.84	.81	.05	.03	No
ESEM with correlated residuals	66,847.04	2081	.00	.92	.90	.04	.02	Yes

Note. N = 20,108. CFA = confirmatory factor analysis; ESEM = exploratory structural equation modeling; CFI = comparative fit index; TLI = Tucker-Lewis index; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual. Acceptable model fit criteria: CFI ≥.90, TLI ≥.90, RMSEA ≤.08, SRMR ≤.08.

Table 6.

Results From Exploratory Structural Equation Modeling (ESEM) of the Kiinnostukset Interest Inventory (With Correlated Residuals)

Item scale	Item text*	Factor loading
Item scale	Item text*	R	I	A	S	E	C
R	1. Repair machines or devices?	.43	.21	−.24	−.14	.05	−.04
R	7. Work requiring physical effort?	.34	.00	−.12	.11	.06	−.11
R	13. Do handicrafts or work with wood or metal?	.71	−.07	.21	−.04	−.06	.02
R	19. Practical, not theoretical work?	.37	−.17	.05	.23	.00	−.05
R	25. Learn to use new appliances?	.40	.08	−.13	.01	.03	.19
R	31. Refurbish furniture or an apartment?	.67	−.09	.13	.07	.00	.01
R	37. Lift weights or do martial arts?	.21	.22	.02	.02	.19	−.19
R	43. Troubleshoot why a device isn’t working?	.45	.29	−.25	−.09	.03	.05
R	49. Manufacture products by hand or with machines?	.67	−.03	.16	−.04	−.07	.05
R	55. Do a hobby requiring craftsmanship?	.62	−.05	.35	.01	−.02	−.01
R	61. Get to know how a product is manufactured?	.44	.26	−.06	−.02	.04	.12
R	67. Transport people or cargo using a vehicle?	.32	−.03	−.19	.02	.07	.02
I	2. Work requiring broad theoretical studies?	.16	.46	.08	.04	−.12	.00
I	8. Read about the newest achievements in science?	.10	.62	−.07	.01	.02	.08
I	14. Research a topic that interests you?	−.15	.40	.02	.03	.13	.14
I	20. Study the meanings and origins of words?	.05	.77	.01	−.08	.01	−.04
I	26. Read up on the thoughts of a philosopher or a researcher?	.05	.47	.13	.08	.10	.05
I	32. Perform chemical, physical or biological experiments?	−.04	.46	.24	.06	−.10	.17
I	38. Visit science centers or exhibits?	−.13	.75	.19	.06	−.02	−.06
I	44. Solve problems requiring ingenuity and logical thinking?	.18	.54	−.13	−.02	−.01	−.01
I	50. Learn to know the laws of nature or human society?	.06	.65	.21	−.02	−.05	.00
I	56. Ponder things deeply and develop new ideas?	.20	.41	.01	.02	.11	.13
I	62. Watch science or nature documentaries?	.05	.64	.05	.11	.02	.03
I	68. Get to know what a research institution does?	−.06	.50	.22	.16	.14	.02
A	3. Edit photos or videos?	−.05	.14	.63	.10	−.03	.06
A	9. Act or dance?	.05	.07	.52	.08	.05	−.07
A	15. Be involved in creating a music video?	.00	.20	.62	.01	−.07	.04
A	21. Read the culture section of a newspaper?	−.13	.13	.63	.05	−.01	.03
A	27. Take part in a writing, drawing or singing competition?	.14	.07	.89	−.02	−.02	−.06
A	33. Write movie, music or theater reviews?	.23	.01	.76	.00	−.11	−.02
A	39. Visit art exhibits or concerts?	.19	.06	.60	.00	.05	.03
A	45. Practice playing a musical instrument or singing?	−.02	−.08	.58	.14	.18	−.04
A	51. Actively follow discussions in different media on topics related to art?	.03	.01	.54	−.01	.21	−.09
A	57. Write poems, stories or plays?	−.08	.24	.50	.05	.00	.09
A	63. Express yourself artistically?	.00	−.04	.65	.02	.06	.03
A	69. Draw or paint?	−.12	.14	.52	−.03	.12	.02
S	4. Work to improve people’s living conditions?	.03	−.09	−.02	.80	−.07	−.05
S	10. Get to know refugees from different countries?	.06	−.13	−.01	.51	.24	.01
S	16. Work as a support person for youth in trouble?	−.05	.03	.08	.59	.17	.00
S	22. Work a job, where you have to get along with all kinds of people?	.00	.15	.17	.39	.00	−.04
S	28. Work requiring a good eye for people?	−.03	−.05	−.01	.89	−.16	.05
S	34. Work to help developing countries?	.00	.07	.11	.24	.39	.00
S	40. Listen to the problems and sorrows of other people?	.22	.04	−.02	.43	.07	.02
S	46. Teach or act as a trainer?	−.05	.05	−.07	.67	.11	−.01
S	52. Develop your first aid skills?	.00	−.01	−.06	.95	−.13	.03
S	58. Mediate disagreements between people?	−.03	.03	−.04	.58	.18	.07
S	64. Help people, who need help with their problems?	.03	.20	.04	.56	.02	.01
S	70. Give instructions and advice to people?	−.02	.14	.13	.40	.00	−.01
E	5. Start your own business?	.13	.00	.13	−.10	.43	−.05
E	11. Act as a team leader?	−.01	.01	.07	.08	.74	−.07
E	17. Make decisions concerning other people?	−.09	.07	−.09	.29	.40	.04
E	23. Plan and supervise the work of others?	.05	.07	−.06	−.01	.56	.09
E	29. Do sales work?	.11	−.23	.01	.02	.36	.11
E	35. Act as a union steward in the workplace?	−.01	.01	−.03	.24	.45	.07
E	41. Be involved in organizing various events?	.08	−.08	.25	.10	.48	.05
E	47. Negotiate a sale or a contract?	.04	.04	−.11	−.04	.50	.11
E	53. Act as an exhibitor at an event or in an exhibit?	.05	−.10	.18	.05	.42	.06
E	59. Lead the realization of a common project?	.00	.06	.01	.00	.68	.04
E	65. Act as the president of a society or an association?	−.05	.06	.03	−.07	.54	.05
E	71. Work that requires persuasion or affecting the opinions of people?	−.08	.14	−.06	.21	.40	−.02
C	6. Work that involves a lot of text or number processing?	−.10	−.05	.15	.00	−.07	.65
C	12. Track income and spending?	.28	.04	.06	−.01	.00	.30
C	18. Act as a secretary for a work group or a meeting?	−.06	.07	.00	−.05	−.02	.63
C	24. Work requiring following instructions exactly?	.09	−.02	−.10	−.01	.10	.49
C	30. Compile, process or save data?	−.05	−.10	.06	.03	.19	.38
C	36. Act as the treasurer of an association or society?	.15	.09	−.14	.08	.02	.33
C	42. Check for errors in texts or calculations?	.03	.12	.01	−.04	−.03	.59
C	48. Keep papers and items in good order?	.02	−.03	−.06	−.03	.20	.33
C	54. Do office work?	−.03	.02	.00	.01	−.09	.61
C	60. Work on the same task for a long time?	.17	−.09	−.03	.14	−.02	.43
C	66. Draft memos or documents?	−.10	−.05	−.01	−.01	−.01	.64
C	72. Work that requires attention to detail and being meticulous?	.02	.14	−.03	.07	−.03	.22

Note. *These items have not gone through a full adaptation process into English and are presented here only as initial translations to allow for examination of factor content. Strongest significant scale loadings are displayed in bold.

Randomization Test of Hypothesized Order Relations (RTHOR)

Table 7 presents the results from RTHOR for the correlations between the RIASEC scores in our sample by gender and level of education. For the full sample, we obtained a Correspondence Index (CI) of .67 (p = .02), which is within the 99% confidence interval of the .70 US benchmark set by Rounds and Tracey (1996) and above the meta-analytic average CI of .63 (Tracey & Rounds, 1993), although our CI is considerably lower than the .78 value obtained by (Glosenberg et al., 2019) in their study. In other words, our data appears to conform to the RIASEC circumplex structure in terms of ordinal relationships to a similar degree as US-based samples do.

Table 7.

Results From the Randomization Test of Hypothesized Order Relations (RTHOR) by Gender and Level of Education

	Predictions
	Group size	Met	Tied	Correspondence index (CI)	p
Total	20,108	60	0	.67	.02
Gender
Men	7290	52	0	.44	.03
Women	12,817	60	0	.67	.02

	Group size	Met	Tied	Correspondence index (CI)	p
Vocational/Professional education
None	9082	61	0	.69	.02
Men	3606	54	0	.50	.02
Women	5476	61	0	.69	.02
Vocational	7743	62	0	.72	.02
Men	2900	53	0	.47	.02
Women	4843	59	0	.64	.02
Postsecondary or UAS	2006	60	0	.67	.02
Men	473	54	0	.50	.03
Women	1533	59	0	.64	.02
University	1088	56	0	.56	.02
Men	237	44	0	.22	.15
Women	851	56	0	.56	.02

Examined by gender, we note a rather large difference in the fit value between women (CI .67, p = .02) and men (CI .44, p = .03), although the test of difference in fit indicated a similar number of met predictions between genders (p = 0.10). Notably, the trend of higher CI in women remains consistent when examined by level of education, with university-educated men showing a particularly poor fit (CI .22) (see Table 7).

An examination of CI values by level of education produced values of .69 (no vocational or professional education), .72 (vocational upper secondary qualification or equivalent), .67 (post-secondary or university of applied sciences degree) and .56 (university degree), indicating an acceptable fit value for all but university-educated clients, although again the test of difference in fit indicates a similar number of predictions met across differently educated groups, with p-values ranging from .91 to .95.

Multidimensional Scaling

We performed ordinal (i.e., non-metric) multidimensional scaling (MDS) to further examine and visualize the degree of fit of Holland’s (1997) RIASEC model to our data. MDS was performed to the total sample and separately be gender. Stress values were .01 for the total sample, .02 for women and .06 for men, indicating an excellent goodness-of-fit (i.e., values <.10) with few discrepancies between the original data and the two-dimensional MDS solution. Variance Accounted For (i.e., R²) coefficients were .79 for the total sample, .65 for men and .81 for women, in turn indicating acceptable levels of fit (≥.60).

Examined visually, the interest domains in the overall sample and when separated by gender appear to follow the RIASEC circular ordering (i.e., the interest domains are positioned in the expected order relative to each other, see Figure 1), although they were spatially arranged in a relatively similar misshapen polygon with the Investigative type closer to the other types than expected instead of the theoretically ideal equilateral hexagon (see Figures 2 –4). This kind of finding is not unexpected (see e.g., Rounds et al., 2021) in fact, Holland himself stated that “at best, the hegaxons resulting from real-word data are misshapen polygons – the hexagon is an ideal” (Holland, 1979, p. 43). In sum, based on the MDS results we conclude that the overall data appears to reflect Holland’s (1997) circumplex to an adequate degree, but more so for women than men (as was to be expected based on the RTHOR results presented earlier).

Figure 2.

MDS solution for RIASEC scales of the Kiinnostukset interest inventory (whole sample)

Figure 3.

MDS solution for RIASEC scales of the Kiinnostukset interest inventory (men)

Figure 4.

MDS solution for RIASEC scales of the Kiinnostukset interest inventory (women)

Cosine Function Modelling

Finally, we used cosine function modelling to explicitly test whether the correlations between the six RIASEC types fit the cosine function that would be expected if the data were to fit a perfect circumplex model as opposed to being simply approximately circular. Specifically, we used the CFM procedure as described by Nagy et al. (2009) and implemented by the R package holland (Hartmann et al., 2021). The results of these analyses are presented in Table 8.

Table 8.

Fit Indices for Alternative Cosine Function Models

	Model description	χ²	df	RMSEA	CFI	SRMR	Δ χ²	Δ df	Δ CFI
M1	Unconstrained	133.26	3	.05	.99	.01	—	—
M2	Equality of scaling parameters	2745.69	8	.13	.90	.12	2612.43	5	−.09
M3	Equality of communalities	27,637.17	13	.13	.83	.08	24,891.49	5	−.07
M4	Perfect circumplex	3657.38	13	.12	.87	.11	−23979.79	0	.03

The fit indices of the CFMs indicate that the unconstrained model (M1) showed a reasonable fit, suggesting a circumplex pattern in the data. However, imposing constraints of equality in scaling parameters (M2) and communalities (M3) significantly worsened fit, indicating that the six RIASEC types vary in their contributions and explained variance within the circumplex. While the perfect circumplex model (M4) demonstrated a somewhat better fit than M3, it failed to reach the cutoffs of good fit. The hypotheses of equal scaling parameters (M2), equal scale communalities (M3) and of a perfect circumplex structure (M4) were therefore not supported. Taken together, the results support the idea of a flexible, rather than perfect circumplex structure (cf. Holland, 1979, p. 43; Rounds et al., 2021).

Discussion

This study is, to our best knowledge, the first one to explore the structure of vocational interests in Finland and its fit to Holland’s (1997) RIASEC model of vocational interest with the goal of establishing its validity for the benefit of Finnish career counselling and guidance practice. Furthermore, the present study is one of the few studies investigating the fit of Holland’s (1997) RIASEC model within the cultural context of the Nordic countries (Denmark, Finland, Iceland, Norway and Sweden), where a notable gap in the literature has existed until now for all countries but Iceland (Einarsdóttir et al., 2002).

The Reliability and Validity of the Finnish Kiinnostukset Interest Inventory

To assess whether Finnish vocational interest data conforms to the expectations of Holland’s (1997) RIASEC model, we first performed a psychometric evaluation of the Finnish computer-based vocational interest test Kiinnostukset to determine whether it is a reliable and valid measure of the six RIASEC vocational interest types. As we hypothesized, this was indeed the case:

Exploratory structural equation modeling (ESEM) showed acceptable model fit, with all six RIASEC dimensions clearly emerging and items generally showing highest loadings on the correct factors with few aberrant or cross-loadings. In the overall sample, the RIASEC scale reliabilities (McDonald’s ω total) ranged from .83 (Realistic) to .92 (Social), with an average value of .89, suggesting that the scales capture their underlying interest constructs well. Taken together, we conclude that the Kiinnostukset interest inventory can be used by Finnish career practitioners and researchers as a reliable measure of the RIASEC interests for both research and counselling purposes, although we suggest reviewing the use of items 37 (poor item functioning), 34 (loads primarily on E instead of S) and 55 (correct primary loading on R, but also strong secondary loading on A) and to use scale means instead of sum scores for results where strict item adherence to the RIASEC scales with no strong cross-loadings is desired.

Gender Differences in Vocational Interests in Finland

We found significant gender differences in vocational interests based on our data. The results were largely in line with those previously found in the meta-analysis by Su et al. (2009): namely, in our data men had higher scores in the Realistic (d = .37) and Investigative (d = .24) scales while women had higher scores in Artistic (d = −.41), Social (d = −.57) and Conventional (d = −.21) scales. However, in contrast to the results obtained by Su et al. (2009), there was a statistically significant gender difference in the Enterprising scale favoring women in our data, although with a negligible effect size (d = −.14).

The most notable difference between our results and those by Su et al. (2009) is the considerably lower magnitude of the effect size of gender difference in our Realistic scale (d = .37 vs. d = .84). In their study Su et al. (2009) demonstrate that this kind of finding could be partially explained by the item development strategies chosen while creating the interest inventory, although unfortunately we have no access to information on the item development process of Kiinnostukset and can therefore make no accurate determination. In the same vein, past research has shown that results from interest inventories measuring vocational interest via rating vocational activities (like Kiinnostukset does) tend to show smaller gender differences than results from inventories where occupations are rated directly (e.g., “How interesting would you find the job of a nurse?”) (Ludwikowski et al., 2020). Regardless of the reason, although the magnitude of the effect size was lower than in some previous studies, the gender difference in the Realistic scale was nevertheless observed.

This is not surprising, as similar gender difference patterns have also been robustly replicated in more recent studies: In their meta-analysis focusing on gender differences in adolescents’ vocational interests in an international context, Baerg MacDonald et al. (2023) demonstrated that adolescent men showed higher interests in Realistic and Investigative careers, while adolescent women showed higher interest in Social and Artistic careers regardless of the year of the study, national gender equality rating or interest scale type. Adolescent interest patterns therefore reflect those found in adults (Baerg MacDonald et al., 2023). In a more culturally similar context, Einarsdóttir and Rounds (2020) similarly showed that Icelandic men were higher on Realistic, Investigative and Enterprising interests, while Icelandic women were higher in Social and Artistic interests. They also note that women’s preference for working with people and men’s preference for working with things in general reflects the enduring and extensive gender segregation in the Nordic labor market (Melkas & Anker, 2001).

To summarize the general trend in literature, gender differences in vocational interests appear to be more robust for the Realistic, Investigative, Social and Artistic scales and more mixed for the Enterprising and Conventional scales. In line with this, vocational interests in Finland are also substantial, a rare exception to the findings demonstrating negligible gender differences in most psychological variables (Hyde, 2005). This is not to be taken as support for gender essentialist notions of “men’s and women’s jobs or interests”, rather these results are likely to be a testament to the powerful force of gendered socialization (Charles & Bradley, 2009). For example (Gottfredson, 2005), posits that children begin orienting to gender roles already at 6 to 8 years old, with the local culture pushing genetically diverse individuals to adhere towards common average gender types and therefore narrowing possible self-concepts and later vocational experiences in adolescence and beyond. Negative reactions from the environment may therefore discourage and suppress interest in activities seen as “inappropriate” for their gender, thus resulting in a more limited horizon of occupational choices.

When discussing gender differences, a somewhat controversial idea called “gender equality paradox” (GEP) is often brought up. According to the GEP hypothesis, gender differences in various psychological variables (such as vocational interests) become more pronounced in countries with higher levels of gender equality (Marsh et al., 2021); see Herlitz et al. (2025), for a review). Accordingly, since Finland ranks very high in country-level gender equality (World Economic Forum, 2023), gender differences in vocational differences should therefore be higher in Finland than in most other countries. However, our results indicate lower effect sizes of gender on vocational interests than those obtained by (Su et al., 2009) in both the Realistic and Social interests, which are the archetypal “things” and “people” interests where the most pronounced differences would be expected. Thus, the results from this study do not appear to support the GEP hypothesis. This is not and cannot be conclusive evidence, of course, and the debate around GEP still continues. We consequently refer interested readers to a recent methodological critique of GEP research by Ilmarinen and Lönnqvist (2024).

We end this section by noting that despite the evidence of gender differences in vocational interests, career practitioners work with unique individuals, not average statistics. While being aware of the general effect of gender on vocational interests is important, even more important is helping guide individuals with diverse interest profiles towards careers that best fit their interests regardless of their gender. After all, a good person-vocation fit is linked to job satisfaction and staying in the chosen occupation (Van Iddekinge et al., 2011b), which not only benefits the client but society at large.

Evidence Supporting the Use of Holland’s (1997) RIASEC Model in Finland

The Randomization Test of Hypothesized Order Relations (RTHOR)

Overall, the Correspondence Index (CI) of our sample (.67) was comparable to the US benchmark of .70 set by Rounds and Tracey (1996). Furthermore, comparing our results to those obtained in Iceland (Einarsdóttir & Rounds, 2020) - another Nordic country - we note that the Correspondence Index obtained from Kiinnostukset is similar in level to that obtained from the indigenous Icelandic interest inventory Bendill (Einarsdóttir & Rounds, 2013) in a sample of Icelandic working adults aged 18-61 (CI = .65). Both these CIs are, however, considerably lower than what was obtained by Glosenberg et al. (2019), who received CIs of .89 for Iceland and .78 for Finland. The magnitude of difference affecting both the Finnish and Icelandic CIs suggests to us that this may be due to the differences in sampling methods and scale used between the studies, since Glosenberg et al. (2019) recruited English-speaking participants online to respond to an interest inventory in English, while Einarsdóttir and Rounds (2013) and ourselves targeted local populations using local instruments offered in local languages.

When examined by gender, the CI values for men (.44) appeared to be considerably lower than those of women (.67), but this difference was not statistically significant (p = .10). Similar to our results, Einarsdóttir and Rounds (2020) note that “the hexagonal structure does not seem to hold up as well when tested for men and women separately using the randomization test.” Indeed, the aforementioned Icelandic working adult sample had CIs of .49 for women and .60 for men. The CI values did not vary significantly by level of education, although university-educated responders showed an absolute value (.56) below the desirable cutoff, in contrast to the other educational level groups. Given that the RTHOR evaluates whether the data conforms to the RIASEC circumplex structure in terms of ordinal relationships, the lower correspondence index for university-educated respondents may reflect broader and, consequently, less differentiated endorsement of interests across multiple domains, thereby weakening the level of circumplex correspondence. The particularly poor fit among university-educated men, in turn, may further reflect the fact that this was the smallest subgroup in the data (making up roughly 1%) and due to selection effects it may represent a relatively narrow range of academic or occupational backgrounds.

Nevertheless, we conclude that the sum of current evidence suggests that the structure of Finnish vocational interests overall conforms to the RIASEC circumplex structure in terms of ordinal relations to similar degree as U.S.-based samples do and to similar extent as the indigenous Icelandic interest inventory Bendill (Einarsdóttir & Rounds, 2013) does in Iceland with the local working adult population.

Multidimensional Scaling (MDS) and Cosine Function Models (CFM)

Multidimensional scaling showed that model solutions for the overall sample and by gender were appropriate, with stress values <.10 and variance accounted for (VAF, i.e. R²) values ≥.60. For all solutions, visual examination showed a circular ordering structure resembling Holland’s (1997) hexagon (i.e., the interest domains were positioned in the expected order of R-I-A-S-E-C), although the Investigative scale was “dented” in towards the other scales rather than “pointing” outwards, meaning the Investigative scale showed relatively weaker distinctiveness in comparison to the other interest domains. This may be due to, e.g., the Investigative scale items sharing features with items from the other interest scales or a general tendency in the sample to endorse Investigative items. However, as noted earlier, Holland stated that “hegaxons resulting from real-word data are misshapen polygons – the hexagon is an ideal” (Holland, 1979, p. 43). In fact, a rather similar dented shape can be seen in the MDS solution of the O*NET Interest Profiler Short Form (Rounds et al., 2021, p. 100), an interest inventory widely used for career exploration and planning purposes in the U.S. Taken together, the MDS solutions also offer support for the fit of Holland’s (1997) RIASEC model to Finnish vocational interest data.

Finally, we used cosine function models to test the fit of our data against a theoretically perfect circumplex. Given the previous results from Multidimensional Scaling, it is not surprising that cosine function models with more stringent constraints did not show acceptable levels of fit, while the unconstrained model in turn did. In sum, we conclude that Holland’s (1997) RIASEC model shows a satisfactory, flexible fit to Finnish vocational interest data.

Limitations and Future Directions

Some limitations should be noted regarding our study. First, ideally, we would have liked to establish the criterion validity of the Kiinnostukset interest inventory, either in terms of predictive or concurrent validity. Unfortunately, as our dataset only contains minimal demographic data in addition to the interest inventory responses, we are unable to do so. Thus, establishing the criterion validity of Kiinnostukset remains a task for future research. In a related vein, we believe it would be worthwhile for future studies to examine possible differences in RIASEC fit between Kiinnostukset and an adapted Finnish version of one of the “highly regarded” interest inventories listed in Su et al. (2009). This might help shed light as to why exactly our results somewhat differ from those obtained by Glosenberg et al. (2019) in terms of lower CI values.

Second, while our sample can be considered representative of the Finnish-speaking population, we note that Finland also has a sizeable Swedish-speaking minority (about 5% of the population) and an indigenous Sámi people population. As our data contains no information on the clients’ first language or indigenous identity, we are unable to determine how well these populations are represented in our data. Regardless, future research should adapt vocational interest assessment measures for and study these populations separately in consideration of their linguistic rights based on Finnish legislation.

Third, our sample only consisted of people who had participated in career counselling services offered by the Finnish Public Employment Services. While this is certainly a population of key significance in terms of assessing vocational interests, it also does not fully capture all contexts where the assessment of vocational interests could be important. For example, in Finland, all students must make their first major career decision at the end of ninth grade at around age 15 as they face the transition to upper secondary education (either general or vocational track). However, our data consisted only of 704 adolescents aged 15 or below (roughly 3% of our data). This can be linked to perhaps one of the key limitations of our sample, as it is by its very nature selective. While access to vocational guidance is a right under Article 9 of the European Social Charter (1996), it is not necessarily sought by nor offered to all those who could benefit from it, as the roughly 2:1 ratio of women to men in our sample suggests a gender-based selection effect in access to Public Employment Services provided career counselling psychologist services. In order to get the full picture of vocational interests, future studies should aim to study vocational interests in the Finnish general population as well as in educational settings at the primary, secondary and tertiary levels.

Conclusion

Ethical and evidence-based career practice requires sufficient research evidence in support of the theoretical models and tools used by career practitioners to help clients make important career decisions. In this study, we carefully analyzed interest data from a large, nationally representative sample of Finnish career counselling clients and consequently provided, for the first time, evidence of support to the continued use of Holland’s (1997) RIASEC model in Finnish career counselling and guidance practice.

In general, the foremost Finnish vocational interest inventory Kiinnostukset showed satisfactory psychometric properties and the statistical fit of the RIASEC circumplex structure in Finland was similar in level to US and Icelandic samples. Although the fit was less ideal for men and people with university-level education, these differences were not statistically significant. Having set this baseline, we now call for further research in service of both refining old and developing new interest assessment tools to further support ethical and evidence-based career practice in the Finnish cultural context.

Footnotes

Acknowledgements

We wish to thank Terence Tracey for providing us with the RANDALL and RANDDIF programs. Furthermore, we wish to thank Dragoş Iliescu for providing us with example syntaxes for the RANDALL program and Jörg Heine for providing us with technical insight to the R package holland. Finally, we wish to thank Hannu Auer for being our data liaison and Harry Pulliainen for providing us insight on the development of the Kiinnostukset interest inventory.

ORCID iD

Ville Puonti

Ethical Considerations

Our study is a secondary analysis of de-identified registry data. The study was approved by the Finnish Development and Administration Centre for ELY Centres and TE Offices (KEHA Centre), which is the competent authority for granting permits to access Finnish Public Employment Services registry data for research purposes. In Finland, neither legislation nor the guidelines of The Finnish National Board on Research Integrity (TENK) require ethical review by an ethics committee for research based purely on registry and documentary data or archive data.

Consent to Participate

According to Finnish legislation, informed consent is not required for registry-based studies. Data for historical or scientific research can be obtained from official personal data files by the competent authority’s permission (Data Protection Act 1050/2018).

Author’s Contributions

Ville Puonti: Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Validation, Writing - Original draft.

Asko Tolvanen: Methodology, Formal analysis, Writing – Review & Editing.

Johanna Rantanen: Conceptualization, Writing - Review & Editing.

Noona Kiuru: Conceptualization, Supervision, Writing - Review & Editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Finnish Cultural Foundation (grant number 00230949).

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The datasets analyzed in the current study are not available due to the terms of our research permit forbidding the sharing of this data to third parties.*

References

Baerg MacDonald

Benson

Sakaluk

J. K.

Schermer

J. A.

(2023). Pre-occupation: A meta-analysis and meta-regression of gender differences in adolescent vocational interests. Journal of Career Assessment, 31(4), 715–738. https://doi.org/10.1177/10690727221148717

Bentler

P. M.

(1990). Comparative fit indexes in structural models. Psychological Bulletin, 107(2), 238–246. https://doi.org/10.1037/0033-2909.107.2.238

Brown

T. A.

(2006). Confirmatory factor analysis for applied research (pp. xiii–475). The Guilford Press.

Browne

M. W.

(1992). Circumplex models for correlation matrices. Psychometrika, 57(4), 469–497. https://doi.org/10.1007/BF02294416

Browne

M. W.

Cudeck

(1992). Alternative ways of assessing model fit. Sociological Methods & Research, 21(2), 230–258. https://doi.org/10.1177/0049124192021002005

Charles

Bradley

(2009). Indulging our gendered selves? Sex segregation by field of study in 44 countries. AJS; American journal of sociology, 114(4), 924–976. https://doi.org/10.1086/595942

Chu

Russell

M. T.

Hoff

K. A.

Jonathan Phan

W. M.

Rounds

(2022). What do interest inventories measure? The convergence and content validity of four RIASEC inventories. Journal of Career Assessment, 30(4), 776–801. https://doi.org/10.1177/10690727221081554

Day

S. X.

Rounds

Swaney

(1998). The structure of vocational interests for diverse racial-ethnic groups. Psychological Science, 9(1), 40–44. https://doi.org/10.1111/1467-9280.00007

Deng

C.-P.

Armstrong

P. I.

Rounds

(2007). The fit of Holland’s RIASEC model to US occupations. Journal of Vocational Behavior, 71(1), 1–22. https://doi.org/10.1016/j.jvb.2007.04.002

10.

Y. Y. L.

Wee

Rounds

(2025). Gender differences in vocational interests across 57 countries: A test of the Interest–gender-equality paradox. Social Psychological and Personality Science, 16(1), 27–44. https://doi.org/10.1177/19485506241280321

11.

Einarsdóttir

Björnsdóttir

M. D.

Lerkkanen

(2020). Cross-cultural validation of assessment instruments used in career counselling and guidance in the Nordic countries: Etic and emic approaches. In Career and career guidance in the Nordic countries (9, pp. 177–189). Brill. https://doi.org/10.1163/9789004428096_012

12.

Einarsdóttir

Rounds

(2013). Bendill, rafræn áhugakönnun: Þróun og notkun [the development and use of an Icelandic interest inventory]. Testing Institute.

13.

Einarsdóttir

Rounds

(2020). A quantitative review of gender differences in vocational interests in Iceland: Pervasive and persistent. Nordic Journal of Transitions, Careers and Guidance, 1(1), 10–26. https://doi.org/10.16993/njtcg.29

14.

Einarsdóttir

Rounds

Ægisdóttir

Gerstein

L. H.

(2002). The structure of vocational interests in Iceland: Examining Holland’s and Gati’s RIASEC models. European Journal of Psychological Assessment, 18(1), 85–95. https://doi.org/10.1027/1015-5759.18.1.85

15.

European Social Charter . (1996). https://rm.coe.int/168007cf93.

16.

Fouad

N. A.

Mohler

C. J.

(2004). Cultural validity of Holland’s theory and the strong interest inventory for five Racial/Ethnic groups. Journal of Career Assessment, 12(4), 423–439. https://doi.org/10.1177/1069072704267736

17.

Glosenberg

Tracey

T. J. G.

Behrend

T. S.

Blustein

D. L.

Foster

L. L.

(2019). Person-vocation fit across the world of work: Evaluating the generalizability of the circular model of vocational interests and social cognitive career theory across 74 countries. Journal of Vocational Behavior, 112(June 2019), 92–108. https://doi.org/10.1016/j.jvb.2019.01.002

18.

Gottfredson

L. S.

(2005). Applying Gottfredson’s theory of circumscription and compromise in career guidance and counseling. In Career Development and Counseling: Putting Theory and Research to Work (pp. 71–100). John Wiley & Sons, Inc.

19.

Guttman

(1954). A new approach to factor analysis: The Radex. In Mathematical thinking in the social sciences (pp. 258–348). Free Press.

20.

Guðmundsdóttir

Gudlaugsson

Adalsteinsson

(2015). Icelandic National Culture compared to National Cultures of 25 OECD member states using VSM94. Veftímaritið Stjórnmál Og Stjórnsýsla, 11(1), 19. https://doi.org/10.13177/irpa.a.2015.11.1.2

21.

Hanna

Nye

C. D.

Samo

Chu

Hoff

K. A.

Rounds

Oswald

F. L.

(2024). Interests of the future: An integrative review and research agenda for an automated world of work. Journal of Vocational Behavior, 152, Article 104012. https://doi.org/10.1016/j.jvb.2024.104012

22.

Hanna

Rounds

(2020). Psychological bulletin how accurate are interest inventories? A quantitative review of career choice hit rates. Psychological Bulletin, 146(9), 765–796. https://doi.org/10.1037/bul0000269

23.

Hartmann

F. G.

Heine

J.-H.

Ertl

(2021). Concepts and coefficients based on John L. Holland’s theory of vocational choice—examining the R package holland. Psych, 3(4), 728–750. https://doi.org/10.3390/psych3040047

24.

Herlitz

Hönig

Hedebrant

Asperholm

(2025). A systematic review and new analyses of the gender-equality paradox. Perspectives on Psychological Science: A Journal of the Association for Psychological Science, 20(3), 503–539. https://doi.org/10.1177/17456916231202685

25.

Hoff

K. A.

Briley

D. A.

Wee

C. J. M.

Rounds

(2018). Normative changes in interests from adolescence to adulthood: A meta-analysis of longitudinal studies. Psychological Bulletin, 144(4), 426–451. https://doi.org/10.1037/bul0000140

26.

Hoff

K. A.

Song

Q. C.

Einarsdóttir

Briley

D. A.

Rounds

(2020). Developmental structure of personality and interests: A four-wave, 8-year longitudinal study. Journal of Personality and Social Psychology, 118(5), 1044–1064. https://doi.org/10.1037/pspp0000228

27.

Hofstede

(2001). Culture’s consequences: Comparing values, behaviors, institutions and organizations across nations. Sage Publications. https://books.google.fi/books?id=w6z18LJ_1VsC

28.

Hofstede

McCrae

R. R.

(2004). Personality and culture revisited: Linking traits and dimensions of culture. Cross-Cultural Research, 38(1), 52–88. https://doi.org/10.1177/1069397103259443

29.

Holland

J. L.

(1979). Professional manual, 1979 edition, the self-directed search. Consulting Psychologists Press.

30.

Holland

J. L.

(1997). Making vocational choices: A theory of vocational personalities and work environments, 3rd ed., xi–303. Psychological Assessment Resource.

31.

Hopwood

C. J.

Donnellan

M. B.

(2010). How should the internal structure of personality inventories be evaluated? Personality and Social Psychology Review: An Official Journal of the Society for Personality and Social Psychology, Inc, 14(3), 332–346. https://doi.org/10.1177/1088868310361240

32.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/10705519909540118

33.

Hubert

Arabie

(1987). Evaluating order hypotheses within proximity matrices. Psychological Bulletin, 102(1), 172–178. https://doi.org/10.1037/0033-2909.102.1.172

34.

Hyde

J. S.

(2005). The gender similarities hypothesis. American Psychologist, 60(6), 581–592. https://doi.org/10.1037/0003-066X.60.6.581

35.

IBM SPSS Statistics for Windows (Version 28) . (2020) [Computer software]. IBM Corp.

36.

Ilmarinen

V.-J.

Lönnqvist

J.-E.

(2024). Deconstructing the gender-equality paradox. Journal of Personality and Social Psychology, 127(1), 217–237. https://doi.org/10.1037/pspp0000508

37.

Jaworska

Chupetlovska-Anastasova

(2009). A review of multidimensional scaling (MDS) and its utility in various psychological domains. Tutorials in Quantitative Methods for Psychology, 5(1), 1–10. https://doi.org/10.20982/tqmp.05.1.p001

38.

Jöreskog

K. G.

Sörbom

(1993). LISREL 8: Structural equation modeling with the SIMPLIS command language. In LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language (p. xvi-202). Scientific Software International; Lawrence Erlbaum Associates, Inc.

39.

KEHA-keskus. (n.d.). AVO Ammatinvalintaohjelma

. Retrieved May 25, 2025, from https://asiointi.mol.fi/avo/

40.

Kruskal

Wish

(1978). Multidimensional scaling. Sage Publications, Inc. https://doi.org/10.4135/9781412985130

41.

Leuty

Hansen

J.-I.

(2014). Teasing apart the relations between age, birth cohort, and vocational interests. Journal of Counseling Psychology, 61(2), 289–298. https://doi.org/10.1037/a0035341

42.

Low

K. S. D.

Yoon

Roberts

B. W.

Rounds

(2005). The stability of vocational interests from early adolescence to middle adulthood: A quantitative review of longitudinal studies. Psychological Bulletin, 131(5), 713–737. https://doi.org/10.1037/0033-2909.131.5.713

43.

Ludwikowski

W. M. A.

Schechinger

H. A.

Armstrong

P. I.

(2020). Are interest assessments propagating gender differences in occupations? Journal of Career Assessment, 28(1), 14–27. https://doi.org/10.1177/1069072718821600

44.

Mair

Groenen

P. J. F.

de Leeuw

(2022). More on multidimensional scaling and unfolding in R: smacof version 2. Journal of Statistical Software, 102(10), 1–47. https://doi.org/10.18637/jss.v102.i10

45.

Marsh

H. W.

Parker

P. D.

Guo

Basarkod

Niepel

Van Zanden

(2021). Illusory gender-equality paradox, math self-concept, and frame-of-reference effects: New integrative explanations for multiple paradoxes. Journal of Personality and Social Psychology, 121(1), 168–183. https://doi.org/10.1037/pspp0000306

46.

Melkas

Anker

(2001). Occupational segregation by sex in Nordic countries: An empirical investigation. In Fetherolf

(Ed.), Women, gender and work (p. 25). International Labour Office.

47.

Meyer

J. M.

Heath

A. C.

Eaves

L. J.

(1992). Using multidimensional scaling on data from pairs of relatives to explore the dimensionality of categorical multifactorial traits. Genetic Epidemiology, 9(2), 87–107. https://doi.org/10.1002/gepi.1370090203

48.

Morris

M. L.

(2016). Vocational interests in the United States: Sex, age, ethnicity, and year effects. Journal of Counseling Psychology, 63(5), 604–615. https://doi.org/10.1037/cou0000164

49.

Mugavin

M. E.

(2008). Multidimensional scaling: A brief overview. Nursing Research, 57(1), 64–68. https://doi.org/10.1097/01.NNR.0000280659.88760.7c

50.

Muthén

L. K.

Muthén

B. O.

(1998-2017). Mplus User’s guide. Eighth edition. Muthén & Muthén.

51.

Nagy

Marsh

H. W.

Lüdtke

Trautwein

(2009). Representing the circles in our minds: Confirmatory factor analysis of circumplex structures and profiles. In Teo

T. K.

(Ed.), Structural equation modeling in educational research: Concepts and applications (pp. 287–315). Sense Publishers. https://hdl.handle.net/11858/00-001M-0000-0024-F7B1-6

52.

Nye

C. D.

Rounds

Drasgow

(2012). Vocational interests and performance: A quantitative summary of over 60 years of research. Perspectives on Psychological Science: A Journal of the Association for Psychological Science, 7(4), 384–403. https://doi.org/10.1177/1745691612449021

53.

Official Statistics of Finland (OSF): Population structure [online publication]. ISSN=1797-5395 . (n.d.). Statistics Finland.

54.

R Core Team . (2025). R: A language and environment for statistical computing. R Foundation for Statistical Computing. (Version 4.4.2) [Computer software]. https://www.R-project.org/

55.

Rottinghaus

P. J.

Coon

K. L.

Gaffey

A. R.

Zytowski

D. G.

(2007). Thirty-year stability and predictive validity of vocational interests. Journal of Career Assessment, 15(1), 5–22. https://doi.org/10.1177/1069072706294517

56.

Rounds

(1995). Vocational interests: Evaluating structural hypotheses. In Assessing individual differences in human behavior: New concepts, methods, and findings (pp. 177–232). Davies-Black Publishing.

57.

Rounds

Hoff

Lewis

(Eds.), (2021). O*NET® interest profiler manual. Department of Labor National O*NET Resource Center. https://www.onetcenter.org/reports/IP_Manual.html

58.

Rounds

(2014). The nature and power of interests. Current Directions in Psychological Science, 23(2), 98–103. https://doi.org/10.1177/0963721414522812

59.

Rounds

Tracey

T. J.

(1996). Cross-cultural structural equivalence of RIASEC models and measures. Journal of Counseling Psychology, 43(3), 310–329. https://doi.org/10.1037/0022-0167.43.3.310

60.

Rounds

Tracey

T. J.

Hubert

(1992). Methods for evaluating vocational interest structural hypotheses. Journal of Vocational Behavior, 40(2), 239–259. https://doi.org/10.1016/0001-8791(92)90073-9

61.

Stoll

Rieger

Lüdtke

Nagengast

Trautwein

Roberts

(2016). Vocational interests assessed at the end of high school predict life outcomes assessed 10 years later over and above IQ and big five personality traits. Journal of Personality and Social Psychology, 113(1), 167–184. https://doi.org/10.1037/pspp0000117

62.

Rounds

Armstrong

P. I.

(2009). Men and things, women and people: A meta-analysis of sex differences in interests. Psychological Bulletin, 135(6), 859–884. https://doi.org/10.1037/a0017364

63.

Tay

Liao

H.-Y.

Zhang

Rounds

(2019). Toward a dimensional model of vocational interests. Journal of Applied Psychology, 104(5), 690–714. https://doi.org/10.1037/apl0000373

64.

Tao

Glosenberg

Tracey

T. J. G.

Blustein

D. L.

Foster

L. L.

(2022). Are gender differences in vocational interests universal? Moderating effects of cultural dimensions. Sex Roles, 87(5), 327–349. https://doi.org/10.1007/s11199-022-01318-w

65.

Tao

Gupta

Tracey

T. J. G.

(2019). Interest assessment in a cross-cultural context. In Athanasou

J. A.

Perera

H. N.

(Eds.), International handbook of career guidance (pp. 655–671). Springer International Publishing. https://doi.org/10.1007/978-3-030-25153-6_30

66.

Tracey

T. J.

Rounds

J. B.

(1993). Evaluating Holland’s and Gati’s vocational-interest models: A structural meta-analysis. Psychological Bulletin, 113(2), 229–246. https://doi.org/10.1037/0033-2909.113.2.229

67.

Tracey

T. J. G.

(1997). RANDALL: A microsoft FORTRAN program for a randomization test of hypothesized order relations. Educational and Psychological Measurement, 57(1), 164–168. https://doi.org/10.1177/0013164497057001012

68.

Tsabari

Tziner

Meir

E. I.

(2005). Updated meta-analysis on the relationship between congruence and satisfaction. Journal of Career Assessment, 13(2), 216–232. https://doi.org/10.1177/1069072704273165

69.

Van Iddekinge

C. H.

Putka

D. J.

Campbell

J. P.

(2011a). Reconsidering vocational interests for personnel selection: The validity of an interest-based selection test in relation to job knowledge, job performance, and continuance intentions. Journal of Applied Psychology, 96(1), 13–33. https://doi.org/10.1037/a0021193

70.

Van Iddekinge

C. H.

Roth

P. L.

Putka

D. J.

Lanivich

S. E.

(2011b). Are you interested? A meta-analysis of relations between vocational interests and employee performance and turnover. Journal of Applied Psychology, 96(6), 1167–1194. https://doi.org/10.1037/a0024343

71.

Vogel

R. M.

Feldman

D. C.

(2009). Integrating the levels of person-environment fit: The roles of vocational fit and group fit. Journal of Vocational Behavior, 75(1), 68–81. https://doi.org/10.1016/j.jvb.2009.03.007

72.

Warlick

C. A.

IngramIV

P. B.

Ternes

M. S.

Krieshok

T. S.

(2018). An investigation into the structural form of the O*NET–interest profiler–short form. Journal of Career Assessment, 26(3), 503–514. https://doi.org/10.1177/1069072717714544

73.

World Economic Forum . (2023). The global gender gap report 2023. https://www3.weforum.org/docs/WEF_GGGR_2023.pdf