Global Versus Domain-Specific Identity Processes

Salience of Identity Domains in EA

Relative salience of different identity domains can vary across developmental periods (Goossens, 2001; McAdams & Zapata-Gietl, 2015). Thus, it is important to consider domain salience in the context of a particular developmental stage, in this case, EA. Erikson (1968) emphasized occupational domain to be of particular importance for a young person’s identity, but he also talked about political, ethnic, and gender aspects of youth identity formation. In addition, he considered the relational domain to be very important because the formation of identity should prepare a young person for entering and sustaining intimacy (Erikson, 1968).

In line with this, more recently, EA was defined as a period of intensified identity exploration in the areas of love and work (Arnett, 2004). For Arnett, an intensified identity search in these two areas is related to the main features of EA, such as instability in major life domains often accompanied by anxiety, feeling “in-between” (not an adult, but not a teenager either), a tendency for self-focus, and perceptions of wide possibilities. Besides love and work, Arnett emphasized the importance of other life domains relevant to emerging adolescents, such as education, friendship, and relationships with parents. Thus, EA is conceptualized as a period of diverse experiences and varied developmental pathways across several life domains, all of which may contribute to the processes of identity development.

These different identity domains have traditionally been classified into two broad “umbrella” domains—interpersonal and ideological domains (Schwartz, 2001). In a recent qualitative study, McLean, Syed, Yoder, and Greenhoot (2016) found that in the interpersonal domain, the aspects of family, friends, and dating were the most important for emerging adults, while sex roles and recreation were of limited importance; in the ideological domain, occupation was found to be the most salient aspect, while religion and politics were the least salient. To sum up, the domains that acquired the strongest theoretical and empirical support as key areas for identity development in EA are occupation (work and education) and close relationships (with a romantic partner, family, friends).

Notably, convergence between the processes of identity development in occupational and relational domains was found to be low in EA (Fadjukoff et al., 2016; Luyckx et al., 2014). Thus, it is difficult to predict the extent to which identity formation in these two broad domains can be explained by some shared/global identity processes. Neither it is clear which of these domains represent global identity better in EA. Occupational domain was found to be the most salient among several ideological domains in terms of closest links with global identity processes (e.g., Skorikov & Vondracek, 1998), however, relational and occupational domains have not been directly compared in this respect.

Becoming an Adult in Lithuania

The cultural context may also be important in determining the salience of different identity domains in EA. Like many other Eastern European countries, Lithuania experienced various political and cultural shifts during the last century. The latest sociocultural shifts were marked by the increasing Westernization in many spheres of life, including the change in timing of adulthood transitions. During 1990–2015, the mean age of first marriage has increased by 6 years and is now about 29 years (Statistics Lithuania, 2006, 2016). On the other hand, the mean age of first marriage is still different from Western and Northern European countries (e.g., Scandinavian countries where it is about 4 years later). Lithuanians are similar to other post-Soviet European countries characterized by conservative views toward family formation. In terms of familism (an ideology that puts priority on the family instead of the individual), Lithuania stands together with Hungary, the Czech Republic, and Poland, and it is different from countries such as the Netherlands or Finland, which are considered much more individualistic (Stankūnienė & Maslauskaitė, 2008).

The delay of school-to-work transitions is evidently associated with the increased proportion of youth seeking higher education (Kraniauskienė, 2011b). Currently, the percentage of youth enrolled in tertiary education is one of the highest in Europe (41% of those aged 20–24 were enrolled in tertiary education in 2014; Eurostat, 2016a). At the same time, the level of unemployment among Lithuanian 15- to 24-year-olds is rather low (in 2015, it was about 19.3%, which is less than the European Union average; Eurostat, 2016). In recent decades, leaving the parental home has also been delayed, and during 2009–2013, the overall percentage of youth aged 20–29 years, who live in the parental home, was around 60% (Eurostat, 2016). The main reason for prolonged living in the parental home is financial difficulties in obtaining or renting an independent household (Kraniauskienė, 2011a), which suggests that even though youth unemployment rate is rather low, the actual income is too small for independent living.

Studies on adulthood criteria suggest that Lithuanian emerging adults perceive becoming an adult primarily as the acquisition of certain character traits rather than going through role changes. As in most studies conducted in other cultural settings (e.g., Arnett, 2003), the most highly endorsed criterion for a person to be considered an adult was accepting responsibility for the consequences of one’s actions (Vosylis, 2014; Vosylis, Kaniušonytė, & Raižienė, 2016). However, contrary to the findings in the United States, the criteria reflecting family capacities, on average, were more important than the achievement of independence and self-sufficiency. Among the top 10 criteria were “being capable of caring for children” for both men and women, “providing financial support” and “keeping the family away from physical harm” for men, and “becoming capable of running a household” for women (Vosylis, 2014). Overall, these data suggest that in the Eastern European context, the romantic relationship domain might be more salient in relation to features of EA as compared to other domains.

The Approach to Content Domains in Process-Oriented Models of Identity

The identity status model has been one of the most important operationalizations of Erikson’s concept of identity that inspired a significant amount of research (Kroger & Marcia, 2011). The status model conceptualizes identity formation as based on identity exploration and commitment making. Different combinations of these two processes referred to as identity statuses describe identity situation of a person in a given moment in life (Kroger & Marcia, 2011). However, identity statuses were understood more as outcomes of identity formation process rather than developed for studying the process itself (Bosma & Kunnen, 2001). This consideration led to the development of new process-oriented models of identity development, among which the five-dimensional (Luyckx et al., 2008) and the three-factor (Crocetti, Rubini, & Meeus, 2008) models have been proposed as advancements of Marcia’s status approach.

These models share a view on identity development as a dual-cycle process, that is, whereas in the status approach commitments are seen as the outcome of exploration, the process models assume that commitments are developed in an iterative process of formation, revision, and maintenance (Meeus, 2011). Particularly, in the three-factor model (Crocetti et al., 2008), the existing identity commitments (i.e., enduring choices made in specific domains) can be reflected and evaluated through in-depth exploration (a process of thinking about the enacted commitments) or revised through reconsideration of commitments (comparison of current commitments with possible alternatives).

Although being very similar in terms of theoretical basis, the process-oriented models somewhat differently operationalize identity formation in terms of domains. In the instrument developed to assess identity formation in the five-dimensional model (Dimensions of Identity Development Scale [DIDS]; Luyckx et al., 2008), usually only the domain of general future plans is assessed. In comparison, the instrument developed to evaluate identity formation in the three-factor model (Utrecht-Management of Identity Commitments Scale [U-MICS]; Crocetti et al., 2008) assesses specific domains, which are then averaged to measure global identity. One advantage of the U-MICS is its flexibility with regard to identity content domains—it is easy for a researcher to adapt the scale to measure different domains. However, the original and most frequently applied version includes only two content areas, one for the ideological (e.g., education) and one for the interpersonal (e.g., friendship) domain.

In the current study, we applied the three-factor model to address three questions: (1) To which extent do domain-specific identity processes reflect identity processes on a more global level? (2) Which identity domains have the closest links with global identity? (3) What are the links between domain-specific and global identity to features of EA (i.e., instability, feeling in-between, self-focus, and perceived possibilities)? In addressing these questions, we selected domains that received the strongest theoretical and empirical support as key areas for identity development in EA. To cover the occupational domain, we included educational and work identity, and to cover the relational domain, we included best friend and romantic partnership identity.

Participants and Procedure

Participants were 1,217 emerging adults (49.9% women; M _age = 25.89, SD _age = 3.51, range 20–31 years). The data were gathered in 2014 and 2015, with the aid of university students. Each student was asked to find several participants to fill in a paper-and-pencil questionnaire. There were no restrictions on the participant recruitment procedures, and students were encouraged to use their social networks. However, in order to acquire a sample with similar characteristics of the general population in terms of gender, employment, education, residential, and family status, detailed guidelines regarding participant characteristics (e.g., an equal number of men and women) were provided. The recruited sample was diverse in terms of occupation, residential, and family status, and it was comparable to the general population of this age, with underrepresentation of emerging adults living with their parents and slight overrepresentation of employed and married emerging adults (Table 1).

OPEN IN VIEWER

Table 1.

Employment, Education, Partner, and Residential Status Distribution Across Study Participants and Comparative National Data.

Sample Characteristics				Population Characteristics
Status		n	Percentage
Employment	Full-time employed	793	65.2	47% full-time employed
	Part-time employed	136	11.2	4% part-time employed (Statistics Lithuania, 2014)
	Unemployed	288	23.7
Education	In education	366	31.1	30% students (Statistics Lithuania, 2014)
	Not studying	851	69.9
Partner	Single (6.2% have children)	356	29.3	77% percent never married
	In a relationship (3.4% have children)	234	19.2	20% married (Statistics Lithuania, 2014)
	Cohabiting (6.2% have children)	257	21.1
	Married (79.5% have children)	370	30.4
Residential	Living with parents	215	17.7	58% living with parents (Eurostat, 2013)
	Living in temporary household	587	48.2
	Living in personal household	415	34.1

Note. Statistics Lithuania 2014 data are calculated using public data files from annual representative income and living conditions survey 2014 conducted by Statistics Lithuania. Statistics are calculated for participants aged between 20 and 31. Access to the data set: http://osp.stat.gov.lt/viesos-duomenu-rinkmenos. Eurostat 2013 statistics are based on Eurostat information provided in share of young people living with their parents by sex and age table, for participants aged between 20 and 29, for 2013. Access to the data: http://ec.europa.eu/eurostat/data/database.

Measures

Identity processes of commitment, in-depth exploration, and reconsideration of commitment in multiple domains were assessed by the U-MICS questionnaire (Crocetti et al., 2008). In total, we presented 52 items, which targeted these identity processes with regard to friendship, partnership, education, and work. For each domain, participants filled 13 items: 5 for commitment (e.g., “My best friend [or partner/work/education] makes me feel sure of myself”), 5 for in-depth exploration (e.g., “I often reflect on my best friend [or partner/work/education]”), and 3 for reconsideration of commitment (e.g., “I often think it would be better to try to find a different best friend [or partner/work/education]”). The response options ranged from 1 (completely untrue) to 5 (completely true). Cronbach’s α coefficients are provided in Table 2.

OPEN IN VIEWER

Table 2.

Reliability of U-MICS Scales and Confirmatory Factor Analysis Results for Testing U-MICS Items for Identity in Different Domains.

Model Tested for	Cronbach’s α				Model Fit Statistics
	n	Commitment	In-Depth Exploration	Reconsideration	χ2	df	CFI	RMSEA (90% CI)
Best friend	1,198	.88	.76	.88	304.80***	59	.957	.059 [.053, .066]
Partner	836	.92	.63	.86	177.09***	59	.967	.049 [.041, .057]
Educational	444	.93	.76	.84	194.46***	59	.949	.072 [.061, .083]
Work	953	.91	.77	.89	283.20***	59	.959	.063 [.056, .071]

Note. CFI = comparative fit index; RMSEA = root mean square error of approximation; CI = confidence interval.

***p < .001.

Instability was assessed by measuring anxiety/depression symptoms using a scale from the Adult Self-Report Form (ASR; Achenbach & Rescorla, 2003). ASR scales measure “syndromes” that are constructed through statistical analysis of symptoms of behavioral and emotional problems. The scale, which combines both anxiety/depression symptoms, consisted of 16 items (e.g., “I worry about my future”; α = .85). Response options ranged from 0 (not true) to 2 (very true or often true).

Feeling in-between was assessed by using a Sense of Adulthood in Peer and Adult Contexts questionnaire (Vosylis, Klimstra, & Raižienė, 2016) that assesses feelings of being an adult in peer and adult contexts (e.g., “How often do you feel like an adult when you are with your parents?”). Seven items in total were provided (α = .86). Response options ranged from 1 (never) to 5 (always).

Self-focus and possibilities, the last two dimensions of EA, were assessed using scales from the short version (Crocetti, et al., 2015b) of the Inventory of Dimensions of Emerging Adulthood (Reifman, Arnett, & Colwell, 2007): self-focus (e.g., “Is this period of your life a time of independence?” 3 items, α = .72) and perceived possibilities (e.g., “…a time of many possibilities?” 3 items, α = .75). The response scale was from 1 (strongly disagree) to 5 (strongly agree).

Testing a Factor Structure of U-MICS and Modeling Method Effects

Structural equation modeling (SEM) analyses using Mplus 7.4 with maximum likelihood robust estimator were conducted to address our research questions. The comparative fit index (CFI) and root mean square error of approximation (RMSEA) were used to assess model fit. CFI higher than .90 and RMSEA lower than .08 indicated acceptable fit, and CFI higher than .95 and RMSEA lower than .05 indicated good fit (Little, 2013). The difference between nested models was tested using the scaled χ² difference test (Satorra & Bentler, 2001) and the difference in model fit statistics (ΔCFI ≥ −.01 indicates substantial decrease in model–data fit; Little, 2013).

First, a three-factor structure of identity processes was tested, running separate confirmatory factor analyses (CFAs) for each domain. In line with prior studies, all models contained 3 latent variables for each identity process, 13 indicators (items), and 3 error correlations (for more details, see Crocetti, et al., 2015a). All four models showed acceptable or good model fit (Table 2). After establishing the structural validity of the U-MICS in each domain separately, we turned to models where identity processes for different domains were analyzed in concert.

Some parallel forms of the U-MICS questionnaire were not applicable to some participants (e.g., partner-related items were not applicable to those who did not have a partner). Therefore, the analyzed data set contained missing data that were not missing at random (the response rate to domain-specific scales is presented in Table 2). To avoid loss of power and other drawbacks associated with listwise deletion, full information maximum likelihood estimation was used. To avoid parameter bias due to missing values, three binary auxiliary variables were included: partner status: single (0) or with partner/married (1), education status: not studying (0) or studying (1), and employment status: not employed (0) or fully/partially employed (1).

First, we focused on modeling the method effects. The U-MICS questionnaire is designed to have parallel forms that address identity processes in different domains. Since the similarity between the methods used (items in this case) can amplify the correlation between the traits assessed (Marsh & Grayson, 1995), we modeled these effects before proceeding with further analyses. To model the method effects, we relied on multitrait–multimethod analytic practices. First, we tested a baseline correlated traits (CT) model, which assumes no presence of method effects (Marsh & Grayson, 1995). Our CT model was a combination of four previous CFA models consisting of 12 latent variables with 52 indicators. CT model had an acceptable fit. Then we chose CT-constrained uncorrelated methods (CTCoUM) model, which was the same as CT model but had 13 additional orthogonal latent variables reflecting method effects, that is, one latent method variable for each parallel item of the U-MICS questionnaire; for details on CTCoUM specification, see Lance and Fan (2015). The CTCoUM model converged and had a good fit; however, the method factor loadings for Item 11 were equal to 0; therefore, this factor was removed. This model resulted in an admissible solution and had a better fit compared to the CT model (Table 3), indicating the presence of significant method effects. Standardized loadings are provided in the Appendix.

OPEN IN VIEWER

Table 3.

Results of Structural Equation Modeling Analysis.

Model Tested	Model Fit Statistics
	χ2	df	Δχ2	Δdf	CFI	ΔCFI	RMSEA (90% CI)
Modeling method effects
CT	3,100.91***	1,196	—	—	.914	—	.036 [.035, .038]
CTCoUM (M0)	2,421.11***	1,184	568.55***	13	.944	.030	.029 [.028, .031]
Testing alternative structures of U-MICS questionnaire
M1	13,907.76***	1,253	8,080.26***	68	.438	−.506	.091 [.090, .092]
M2	9,636.677***	1,238	4,814.15***	54	.627	−.317	.075 [.073, .076]
M3	2,478.60***	1,223	63.67**	39	.944	.000	.029 [.027, .031]
Measurement model for other features of emerging adulthood
M4	2,27.41***	71	—	—	.942	—	.043 [.036, .049]
Latent multivariate multiple regression models: Identity processes are used as predictors of other features of emerging adulthood
M5	3,827.37***	2,010	—	—	.937	—	.027 [.026, .028]
M6	3,601.93***	1,935	230.78***	75	.942	.005	.027 [.025, .028]

Note. Δχ², Δdf, and ΔCFI for models M¹, M², and M³ show difference from model M⁰; Δχ², Δdf, and ΔCFI for model M⁶ show difference from model M⁵. CT = correlated traits model; CTCoUM = correlated traits-constrained uncorrelated-methods model (M⁰); M¹ = three general processes/no context-specific processes; M² = three processes in ideological and three in relational domains; M³ = global identity processes modeled as second-order variables; M⁵ = global identity processes as predictors of other features of emerging adulthood; M⁶ = domain-specific identity processes as predictors of other features of emerging adulthood (global identity processes are not present in the model); CFI = comparative fit index; RMSEA = root mean square error of approximation; CI = confidence interval.

***p < .001. **p < .01.

Global and Domain-Specific Identity Processes

To assess if the U-MICS questionnaire differentiates processes in different identity domains, several alternative structures of these identity processes were tested and compared in terms of fit against the CTCoUM model, which served as a baseline model (M⁰) in this step of the analysis (Table 3). The first model (M¹) was a CFA model, in which all items measuring the same identity process but in different domains were grouped to represent one latent variable (e.g., latent variable for identity commitments was constructed by using all commitment items in four domains). This model had a very poor fit. Next, we tested a model (M²) where items measuring different identity processes were grouped based on two more general domains (relational and occupational) they represent. The fit of this model was better when compared to the previous one but still poor. Taken together, these analyses supported the significant differentiation of identity processes in different domains that we assessed.

Next, we tested a second-order model (M³) in which 12 domain-specific processes were modeled as first-order latent variables, three global processes were modeled as second-order ones, including residual correlations for domain-specific processes in a specific domain (Figure 1). With such model, we theorized that some correlations between identity processes in different domains can be attributed to global identity processes; however, each domain-specific process has a residual variance that covariates with residual variances of other processes in the same domain. This model fit the data well and was not substantially different from the M⁰ model. An inspection of factors loadings highlighted that for each identity global factor (commitment, in-depth exploration, and reconsideration of commitment), the relational domains (partner and best friend) were better indicators than the ideological domains (education and work). Taken together, these analyses not only indicated the presence of global processes (as a shared variance between domain-specific processes) but also supported the presence of the domain-specific variance. It also suggests that the relational domain was the most salient in our sample.

OPEN IN VIEWER

Figure 1.

Model depicting global and domain-specific identity processes measured by U-MICS questionnaire. Standardized model parameters and determination coefficients (in parenthesis) are displayed. Although not displayed for reasons of clarity, this model includes latent method factor variables. All parameters are statistically significant at p < .01.

Associations of Identity Processes With Features of EA

In the final step of our analysis, we focused on relationships between global versus domain-specific identity processes and features of EA. First, new latent variables were created for features of EA. Three parcels were formed for a sense of adulthood and another three for instability by using item-to-construct balance approach based on corrected item–total score correlations (Little, 2013). Scale items were used as indicators for being self-focused and for perceived possibilities, as they had only 3 items per scale. Model–data fit of this measurement model was good (M⁴ in Table 3). To assess relationships between identity processes and main features of EA, we tested two models. In the first model (M⁵), global identity processes (as measured in M³) were used as predictors of features of EA, while in the second model (M⁶), domain-specific processes (as measured in the CTCoUM) were considered. Both models fit the data well (Table 3).

The results of M⁵, where global identity processes were used as predictors, indicated that identity commitment was negatively related to instability and positively related to the sense of adulthood; in-depth exploration and reconsideration of commitment were positively associated with instability, self-focus, and perceived possibilities and negatively linked to the sense of adulthood (Table 4). Percentages of the explained variance ranged from small to large (from 5% to 32%).

OPEN IN VIEWER

Table 4.

Parameter Estimates (Standardized Slope Coefficients) in Structural Equation Modeling Linking Global and Domain-Specific Identity Processes to EA Features.

		EA Features
		Instability	Sense of Adulthood	Self-Focus	Perceived Possibilities
Domains	Identity Process	Standardized Slope Coefficients
Global	Commitment	−.30***	.33***	.13	.04
	In-depth exploration	.20**	−.19**	.14*	.21**
	Reconsideration of commitment	.38***	−.15*	.15*	.26***
	R 2	.32***	.15***	.05*	.09**
Best friend	Commitment	.09	−.04	.16**	.27***
	In-depth exploration	.03	−.05	−.04	−.01
	Reconsideration of commitment	.04	−.06	.05	.00
Partner	Commitment	−.16*	.04	−.13*	−.20**
	In-depth exploration	.08	−.06	.08	.09
	Reconsideration of commitment	.15*	−.10	−.02	.04
Education	Commitment	−.20**	.26**	.11	−.03
	In-depth exploration	.16**	−.05	.07	.14
	Reconsideration of commitment	.13*	−.04	.04	.06
Work	Commitment	−.11	.22**	.08	.03
	In-depth exploration	−.01	−.05	.10	.07
	Reconsideration of commitment	.17**	.03	.08	.18**
	R 2	.27***	.15***	.07**	.14***

Note. EA = emerging adulthood.

*p < .05. **p < .01. ***p < .001.

The findings of M⁶, where domain-specific identity processes were used as predictors, indicated that the relationships between domain-specific identity processes and the features of EA varied by domain. Instability was significantly negatively related to commitments in partner and educational domains and positively related to reconsiderations in partner, educational, and work domains as well as in-depth exploration in the educational domain. The sense of adulthood was positively related only to commitments in educational and work domains. Interestingly, self-focus and perceived possibilities were positively related to commitments in the best friend domain and negatively to commitments in the partner domain. Perceived possibilities were also positively related to reconsideration of commitment in the work domain. Percentages of explained variance ranged from small to large (from 7% to 27%).

To Which Extent Do Domain-Specific Identity Processes Reflect Identity Processes on a Global Level?

Study results partly support the validity of hierarchical account of domain-specific versus global identity processes. Thus, it is meaningful to a certain extent to infer global identity situation based on domain-specific assessments. Yet the variance associated with global identity processes varies across domains significantly, thus identity domains differ in their potential to represent global identity processes. In our study, global processes were most strongly represented by relational domain processes as opposed to two occupational domains. Generally, the results also suggest that researchers ought to consider which domains they select to measure global identity and what the potential of those domains to represent global identity processes is.

Which Identity Domains Have the Closest Links With Global Identity?

In this study, we documented that relational domains are more strongly related to global identity than ideological domains. This result is consistent with prior research (e.g., Crocetti et al., 2012) showing that emerging adults in an identity configuration characterized by a stable relational identity but an unstable ideological identity report higher need satisfaction than their counterparts who have achieved a condition of identity stability only in the ideological domain. Similarly, McLean et al. (2016) found that the interpersonal domain was the most frequently present and the most dominant in the identity narratives of emerging adults compared to ideological and existential domains. Thus, while for adolescents the occupational identity was found to be the most salient and pioneering domain of identity formation (e.g., Skorikov & Vondracek, 1998), the findings of the present study suggest that in EA, relational domains may be taking over in relative salience and importance, at least when assessed among emerging adults with heterogeneous educational, work, and relationship statuses. In addition, the high salience of relational domains found in this study can be further understood, considering the specific context of our sample. In fact, in the Lithuanian and Eastern European cultural context, familistic-relational values are considered to be very important (Stankuniene & Maslauskaite, 2008).

How Are Global Versus Domain-Specific Identity Processes Related to Features of EA?

In this study, we found that while most associations between global identity processes and features of EA were replicated at the domain level, some other were domain-specific and provided new insights on the role played by specific relational and occupational domains in EA. In terms of similarities, global and domain-specific identity processes explained similar portions of variance on different features of EA, such as instability, sense of adulthood, self-focus, and possibilities. Furthermore, the overall pattern of associations was generally consistent, with commitment being related to more positive outcomes than in-depth exploration and reconsideration of commitment (for a discussion of the bright and dark sides of identity formation, see Crocetti, Beyers, & Cok, 2016).

In terms of results that are domain-specific, we first note that within the relational domain, the partner domain was more strongly related to features of EA than the best friend domain, confirming that in EA partners become more important than friends for EA psychosocial development (e.g., Crocetti & Meeus, 2014). Further, commitment to the partner and the best friend had opposite associations with self-focus and possibilities. Thus, these results highlighted that the lack of a significant association between global commitment and self-focus and perceived possibilities was hiding differential associations that domain-level analyses could unveil. This supports our previous speculations about the changing nature of personal relationships during the developmental period under study—higher commitments to a relationship with the best friend indicate holding in EA, while higher romantic partner commitments show moving toward adulthood (Lanz & Tagliabue, 2007).

Within the ideological domain, we found that identity in the educational domain was more strongly related to features of EA than identity in the work domain. This result is consistent with evidence from other Eastern emerging adults (Polish and Romanian; Karaś, Cieciuch, Negru, & Crocetti, 2015) and might point to the value of education as a means of social mobility for Eastern youth. Interestingly, reconsideration of commitment in the work domain was also related to perceived possibilities. Thus, although reconsideration of commitments represents a temporary crisis, it also represents a necessary step for improving the current situation by searching for more satisfying long-term commitments (Crocetti et al., 2008).

In synthesis, our findings show that if domain-specific identity processes correlate with external constructs in the same direction, then scores for global identity processes provide a summative result and show a stronger relationship with these external constructs. However, when domain-specific identity processes have opposite links with external constructs, corresponding links at the global level might be close to zero. Thus, looking at global identity processes may suggest that an effect is absent, while it may be present on a domain-specific level, as we found for relational commitments.

Limitations, Future Research Directions, and Conclusions

Although the present study reveals several important findings, some limitations need to be acknowledged. First, the study was cross-sectional, hence it is not possible to assess the true direction of relationships between variables. Second, our sample was nonrandomized, and it overrepresented employed emerging adults and those with independent residential status. Third, reliance on variable-oriented analyses in our study did not allow to reveal individual differences in the salience of specific identity domains and their links to features of EA (e.g., Meca et al., 2015). Finally, we only assessed four identity domains. Future studies should use longitudinal designs, representative samples, person-oriented methods, and more identity domains to provide further evidence on the relationships that we assessed. The salience of identity domains among younger emerging adults, or those still in education, as compared to those that have finished their studies, could also be assessed.

Concluding, this study shows that global identity processes only partly explain domain-specific identity processes. After accounting for global ones, domain-specific identity processes still have unique links with features of EA. Thus, consideration of different life domains to measure identity processes is highly relevant, since identity development in different domains can have different associations with outcomes of interest. These findings underline the importance of including multiple domains when assessing identity processes in EA.