Toward Sustainable Human Resource Development: The Influence of Workplace Friendship on Early Childhood Educators’ Retention Intention, with Workplace Well-Being and Job Embeddedness as Parallel Mediators

Abstract

Within the context of sustainable educational workforce development, enhancing the retention intention of early childhood educators is essential for ensuring educational quality and long-term talent sustainability. This study surveyed 200 early childhood educators in Taiwan and developed a parallel mediation model to examine how workplace friendship influences retention intention through workplace well-being and job embeddedness. Confirmatory factor analysis and structural equation modeling were conducted using AMOS 24.0. The results indicate that workplace friendship does not exert a direct effect on retention intention; however, it significantly enhances workplace well-being and job embeddedness, which in turn fully mediate the relationship. In line with the JD-R framework, workplace well-being is conceptualized as a core psychological resource, while job embeddedness reflects a structural resource shaping employees’ attachment to their organization. These findings suggest that workplace friendship must be transformed into a psychological and structural resource in order to promote retention. By identifying workplace friendship as an initial social resource that fosters well-being and embeddedness, this study contributes to sustainable human resource management and supports the stable development of the early childhood education system.

1. Introduction

Within the broader context of sustainable development, talent sustainability has become a critical issue for maintaining long-term organizational competitiveness, with employees’ retention intention serving as a core determinant. This challenge is particularly salient in the field of education, where educators are not only transmitters of knowledge but also fundamental contributors to long-term human capital development. In early childhood education, high turnover rates have become a global concern. A substantial body of research has shown that early childhood educators frequently face low wages, heavy workloads, and limited career advancement opportunities, all of which contribute to professional attrition and undermine educational quality and children’s learning opportunities [1,2,3]. Strengthening the retention intention of early childhood teachers has therefore become not only an organizational imperative but also a pressing issue for educational system stability and talent sustainability. Because retention reflects a motivational and resource-driven process, understanding the psychological and contextual factors that sustain educators becomes essential.

The antecedents of retention intention are multifaceted. Prior studies have emphasized individual-level determinants such as psychological empowerment [4], sense of calling [5], and professional commitment [6]. However, emerging evidence suggests that early childhood educators’ decisions to remain are shaped not only by personal characteristics but also by the availability of job resources within the work environment. Organizational climate, peer relationships, and interpersonal support [7,8] have been shown to influence teachers’ emotional attachment and their willingness to stay. Building on this perspective, the present study focuses on workplace friendship as a form of social job resource and examines how such social relationships are converted into psychological and organizational resources that can strengthen educators’ retention intention.

Despite increasing scholarly attention, several important gaps remain. First, most studies have examined workplace friendship in relation to turnover intention, while much less is known about its role in predicting retention intention. Although turnover and retention are often treated as conceptual opposites, recent scholarship warns against assuming equivalence because they involve distinct psychological processes. Retention reflects a resource-based motivational state, whereas turnover reflects withdrawal tendencies. Second, although existing studies suggest that workplace friendship may reduce turnover or enhance retention, limited research has examined how workplace friendship is transformed into employee resources that subsequently promote the intention to stay. This lack of theoretical clarity constrains our understanding of the mechanisms linking social relationships and retention outcomes.

To address these gaps, this study adopts the Job Demands–Resources (JD-R) framework and conceptualizes workplace friendship as an initial social job resource. Drawing from prior research, two constructs—workplace well-being [6,9,10,11] and job embeddedness [12,13,14]—emerge as key determinants of retention intention. Workplace well-being reflects educators’ positive psychological functioning in the work context and thus represents a core psychological resource within the JD-R model. Job embeddedness captures employees’ structural ties to the organization and constitutes a structural resource. Accordingly, this study proposes a parallel mediation model to examine how these psychological and structural resources jointly mediate the relationship between workplace friendship and retention intention.

Drawing from prior literature, two constructs job embeddedness and workplace well-being emerge as key determinants of retention intention, representing employees’ structural connections within the organization and their positive psychological states, respectively. Accordingly, this study proposes a parallel mediation model to investigate how workplace well-being and job embeddedness jointly mediate the relationship between workplace friendship and retention intention.

2. Literature Review and Hypotheses Development

2.1. Conceptual Distinction Between Turnover Intention and Retention Intention

Although turnover intention and retention intention are often treated as conceptual opposites, research increasingly recognizes that they are distinct constructs driven by different psychological mechanisms. Turnover intention typically reflects a withdrawal-oriented process, shaped by negative job experiences such as stress, burnout, dissatisfaction, or perceived injustice [15,16]. Retention intention, in contrast, represents a commitment- and attachment-oriented process, influenced by positive experiences such as organizational support, well-being, meaningful work, and job embeddedness [4,17]. Because the antecedents of staying are not merely the inverse of the antecedents of leaving, scholars argue that turnover and retention should not be assumed interchangeable [18,19]. The educational workforce literature further reinforces this distinction: educators may choose to stay not only because they lack alternatives, but because they experience belonging, professional fulfillment, or relational stability [20,21,22]. For this reason, retention intention reflects the motivational and resource-based aspects of persistence, making it especially relevant for understanding talent sustainability in early childhood education contexts. Focusing on retention therefore enriches the field by shifting attention from deficit-focused predictors of departure to resource-driven pathways that promote workforce stability, consistent with the JD-R gain cycle perspective.

2.2. Workplace Friendship and Retention Intention

Workplace friendship refers to non-romantic, voluntary, and informal relationships that develop among employees within organizational settings [23]. Unlike formal work interactions, such relationships emphasize autonomy and emotional exchange between colleagues. Prior research indicates that workplace friendship fosters trust, mutual assistance, and emotional support among peers, cultivating a sense of belonging and connectedness [24]. In addition, such friendships promote close and open communication, enabling employees to access greater social support and emotional comfort in the workplace, thereby contributing to a positive interpersonal climate [25]. Overall, workplace friendship can be regarded as a critical informal social resource within organizations, playing an important role in interpersonal interactions and psychological need fulfillment.

A substantial body of literature suggests that workplace friendship positively influences both individuals and organizations. Specifically, higher levels of workplace friendship have been found to enhance employees’ work engagement, well-being, organizational identification, and organizational commitment, thus strengthening their positive attitudes toward work [24,26]. At the same time, workplace friendship can reduce burnout and stress, alleviate feelings of job insecurity, and effectively decrease turnover tendencies [26]. Moreover, workplace friendship plays a meaningful role in employees’ turnover behaviors and organizational attachment. Several studies report that friendship among coworkers significantly reduces turnover intention [27,28], suggesting that it functions as a supportive resource that mitigates negative psychological responses arising from work-related stress or role conflict. However, empirical studies directly examining the impact of workplace friendship on retention intention remain relatively scarce. Only a limited number of studies have shown that such friendships enhance employees’ intention to stay through emotional support and a strengthened sense of belonging [24,29]. These findings indicate that although prior research has predominantly focused on reducing turnover intention, workplace friendship may also serve as a positive psychological resource that actively promotes employees’ desire to remain.

From the perspective of the JD-R model, job resources not only help buffer the adverse effects of job demands but also stimulate employees’ motivation, thereby promoting positive attitudes and behaviors [30,31]. As a social resource, workplace friendship provides emotional support, trust, and reciprocal assistance, which can enhance employees’ psychological capital and well-being. Building on this theoretical foundation, the present study argues that workplace friendship not only reduces turnover tendencies but also strengthens employees’ likelihood of staying through mechanisms related to resource conservation and motivational gains. Accordingly, this study proposes the following hypothesis:

H1: 

Workplace friendship has a significant positive effect on retention intention.

2.3. The Mediating Role of Workplace Well-Being

Well-being is a multidimensional and broadly defined construct that has increasingly been recognized as a critical factor influencing employees’ attitudes and behaviors in organizational research. Employee well-being is commonly categorized into three major forms: workplace well-being, subjective well-being, and psychological well-being. Subjective well-being (SWB) reflects an individual’s overall evaluation of life quality based on personal standards and comprises emotional experiences and cognitive assessments of life satisfaction. In contrast, psychological well-being (PWB) emphasizes positive psychological functioning and the realization of one’s potential, including self-acceptance, purpose in life, environmental mastery, personal growth, and positive interpersonal relationships [32]. Building on these perspectives, workplace well-being directly reflects employees’ sense of satisfaction with their work and their positive emotional experiences within the work environment, highlighting the contextualized nature of well-being in organizational settings [33].

Bartels et al. [34] conceptualized workplace well-being through two core dimensions: interpersonal workplace well-being and individual workplace well-being. The interpersonal dimension focuses on social aspects such as support and belonging, whereas the individual dimension centers on personal growth and goal fulfillment. Thus, workplace well-being represents employees’ subjective evaluation of their development and optimal functioning in the workplace, encompassing both personal and social aspects. It not only reflects psychological health but also embeds well-being within the specific context of work.

Recent studies suggest that workplace friendship enhances well-being, which subsequently reinforces employees’ positive organizational attitudes and behaviors. Çınar and Basım [29] demonstrated that relational job crafting increases employees’ intention to stay primarily through enhanced workplace friendship, suggesting that social connections formed at work can transform proactive relational behaviors into stronger retention intentions. For instance, Chen et al. [24] demonstrated that workplace friendship promotes organizational commitment through increased well-being, thereby strengthening employees’ willingness to maintain long-term relationships with the organization.

From the perspective of the JD-R model, workplace friendship constitutes a social job resource that fulfills employees’ emotional and social support needs, consequently enhancing their well-being. As a positive psychological resource, well-being fosters motivation, strengthens organizational attachment, and reduces the likelihood of negative behaviors [35,36]. Accordingly, workplace well-being is conceptualized as a key mediating mechanism through which workplace friendship shapes retention intention, transforming social resources into positive psychological and behavioral outcomes. Based on this reasoning, the following hypotheses are proposed:

H2: 

Workplace friendship has a significant positive effect on workplace well-being.

H3: 

Workplace well-being mediates the relationship between workplace friendship and retention intention.

2.4. The Mediating Role of Job Embeddedness

Job embeddedness (JE) refers to the combined forces that connect an individual to their organization and surrounding environment, making it difficult for them to detach or voluntarily leave their job or workplace [37,38]. Employees become “embedded” within an organization through three primary forces: links, fit, and sacrifice [37,39]. Links denote the quantity and quality of connections employees have with their job environment, organization, and interpersonal networks, such as interactions with colleagues and supervisors or roles and responsibilities held within the organization. Fit represents the degree to which employees’ values, career goals, and personal characteristics align with their job role, organizational culture, and work environment. Sacrifice refers to the perceived losses associated with leaving the organization, including compensation, benefits, social relationships, and developmental opportunities. Collectively, these forces strengthen employees’ attachment to both the organization and the broader community, thereby reducing the likelihood of turnover and promoting retention [40].

In the context of early childhood education, these three components of job embeddedness manifest in distinct and highly salient ways. First, links often take the form of collegial relationships, mentoring ties, and team-based collaboration common in early childhood education and care (ECEC) settings [41]. Early childhood educators typically work closely with co-teachers, assistants, administrative staff, and families, creating dense interpersonal networks that can strongly anchor them to the workplace. Second, fit is reflected in the alignment between educators’ beliefs about child development, their pedagogical philosophy, and the center’s educational approach or culture [42]. A strong match between an educator’s values—such as play-based learning, developmental appropriateness, or family engagement—and the organizational culture enhances their sense of comfort and belonging in the center. Finally, sacrifice in ECEC settings extends beyond material or financial considerations. Leaving a kindergarten or childcare center often requires educators to relinquish established bonds with children, trust built with parents, familiarity with routines and teaching teams, and accumulated knowledge of classroom dynamics [43]. These relational and emotional losses represent substantial sacrifices, making turnover more costly and retention more desirable.

Although job embeddedness is commonly conceptualized through three sub-dimensions—links, fit, and sacrifice—the present study employs the Global Job Embeddedness Scale (GJES) developed by Crossley et al. [40]. This scale was intentionally designed as a unidimensional measure to capture employees’ overall psychological attachment to their job and organization. Prior validation studies have demonstrated that this global operationalization provides strong predictive validity for turnover and retention outcomes and helps avoid potential construct misspecification that may arise when the three elements are modeled as separate factors [44,45]. Therefore, the present study follows the original design of the GJES and adopts a global, unidimensional score of job embeddedness as the basis for subsequent analyses.

Extant research widely acknowledges that employees with higher levels of job embeddedness tend to form deeper emotional connections with their work and organization, which fosters more positive work attitudes and behaviors. Job embeddedness effectively enhances organizational attachment and increases employees’ willingness to remain when faced with external job opportunities [13]. Empirical studies have consistently demonstrated stable positive relationships between job embeddedness, organizational loyalty, and retention intention [12,13,14]. Although the effects of job embeddedness have been extensively supported, limited research has explored whether workplace environmental factors particularly a friendly and supportive social climate can further strengthen employees’ embeddedness.

Empirical studies indicate that organizational, supervisory, and coworker support can strengthen job embeddedness, which in turn reduces turnover intention [46]. Workplace friendship provides a form of social and emotional support that operates similarly to coworker support but further deepens employees’ sense of belonging. Friendship enhances embeddedness across the dimensions of links, fit, and sacrifice by fostering strong interpersonal ties, value alignment, and perceived losses associated with leaving. Research also shows that friendships—more than family relationships—reinforce community attachment, thereby influencing organizational fit and perceived sacrifice and indirectly lowering turnover intention [47]. Qualitative evidence from the hospitality sector likewise demonstrates that friendship cultivates interpersonal networks and contextual fit, ultimately reducing employees’ tendencies to leave [48]. Overall, workplace friendship emerges as a salient social resource that contributes to job embeddedness through emotional and relational mechanisms.

From the perspective of the JD-R model, job resources help buffer the adverse effects of job demands and cultivate motivational processes that foster positive attitudes and behaviors [35,36]. As a social job resource, workplace friendship provides emotional support and mutual trust, enhances employees’ sense of fit and connection, and subsequently strengthens their job embeddedness. Through this mechanism, workplace friendship may indirectly increase employees’ retention intention, illustrating its role in resource conservation and motivational enhancement. Based on these arguments, the following hypotheses are proposed:

H4: 

Workplace friendship has a significant positive effect on job embeddedness.

H5: 

Job embeddedness mediates the relationship between workplace friendship and retention intention.

Grounded in the JD-R theoretical framework and supported by empirical evidence, this study integrates workplace friendship, workplace well-being, job embeddedness, and retention intention into a unified research model. The proposed research framework is presented in Figure 1.

3. Results

3.1. Participants and Procedure

The participants of this study were ECEC personnel currently employed in early childhood education and childcare institutions in Taiwan, including directors, teachers, educare staff, and assistant educare staff. Student interns were excluded from participation. To ensure that respondents possessed sufficient familiarity with their work environment, an inclusion criterion of at least three months of tenure in their current institution was applied. In Taiwan, a three-month probation period is the standard timeframe for evaluating newly hired educators, during which they undergo workplace adjustment, demonstrate essential job competencies, and become acquainted with institutional routines. Accordingly, this requirement ensured that participants were sufficiently familiar with workplace relationships, organizational culture, and job responsibilities to provide meaningful assessments of the study variables.

Data were collected through anonymous self-administered questionnaires using convenience sampling. Recruitment was conducted with the assistance of ECEC personnel known to the researcher, who helped distribute the survey. Data collection was carried out from 1 April 2025 to 3 June 2025. All participants received a study description and provided informed consent before completing the survey. They were informed that participation was voluntary and could be discontinued at any time without penalty. The study adhered to the principle of minimal risk, and all data were used exclusively for academic purposes.

To minimize common method variance (CMV), several procedural remedies were implemented during questionnaire design and administration. First, reverse-coded items from the original scales were retained, and different scoring formats were used across scales to reduce response patterning. Items were also randomly ordered to mitigate potential context effects between adjacent questions. Second, anonymity and strict confidentiality were ensured throughout data collection. Participants were informed, in neutral language at the beginning of the questionnaire, that the survey was solely for academic research and unrelated to employment decisions or performance evaluations. This approach aimed to reduce social desirability bias and evaluation apprehension. Together, these procedures helped reduce potential CMV and enhance the validity of the data.

A total of 250 questionnaires were distributed. After excluding invalid responses—such as those containing uniform answer patterns or substantial missing data—200 valid questionnaires were retained, yielding an effective response rate of 80%. To ensure that the sample size was adequate for Structural Equation Modeling (SEM), this study followed Kline [49] recommendation that at least 150 cases are sufficient when the model is not overly complex. Hair et al. [50] further emphasized that a sample size of approximately 200 is desirable when the ratio of observed to latent variables is appropriate.

To further verify sample adequacy, the ratio of observed variables to latent variables was calculated. The measurement model contained 26 observed indicators and 4 latent constructs, resulting in a ratio of approximately 6.5:1, which meets commonly recommended guidelines for SEM identification. Taken together, the final sample of 200 valid cases not only exceeded the minimum threshold but also met the ideal sample size suggested in methodological guidelines, thereby supporting the reliability and stability of the SEM parameter estimates.

The demographic characteristics of the ECEC personnel are presented in

Table 1. Demographic Characteristics of Participants.

Variable	Category	n	%
Age	30 years and below	62	31.0
	31–50 years	105	52.5
	51 years and above	33	16.5
Tenure in Current Institution	Less than 3 years	131	65.5
	3–10 years	43	21.5
	More than 11 years	26	13.0
Education Level	High school/junior college or below	54	27.0
	University degree or above	146	73.0
Job Position	Director	30	15.0
	Teacher	123	61.5
	Educare staff	47	23.5
Monthly Salary (NT$)	≤35,000	51	25.5
	35,001–40,000	110	55.0
	≥40,001	39	19.5

. The participants were primarily aged 31–50 years (52.5%), followed by those under 30 years (31.0%) and those aged 51 years or above (16.5%). Most participants had fewer than three years of work experience (65.5%), while 21.5% had three to ten years of tenure, and 13.0% had more than eleven years. In terms of educational level, the majority held a university degree or higher (73.0%), whereas 27.0% had a high school diploma or junior college degree. Regarding job positions, teachers constituted the largest group (61.5%), followed by educare staff (23.5%) and center directors (15.0%). Monthly income levels were concentrated in the NT$35,001–40,000 range (55.0%), with 25.5% earning NT$35,000 or below and 19.5% earning above NT$40,001. Overall, the sample mainly comprised teachers aged 31–50 with university-level education, relatively short job tenure (less than three years), and monthly earnings between NT$35,001 and NT$40,000, reflecting the typical profile of ECEC personnel in Taiwan.

3.2. Questionnaire Design

Workplace friendship was measured using the scale developed by Nielsen et al. [51], which assesses the extent to which early childhood educators perceive friendship among colleagues in their work environment. The scale consists of six items, such as “I have formed a strong friendship with my colleagues” and “Being able to see my colleagues is one reason why I look forward to my job.” All items were rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree), with higher scores indicating stronger perceptions of workplace friendship. Reliability analysis showed excellent internal consistency, with a Cronbach’s α of 0.948.

Workplace well-being was assessed using the scale developed by Pradhan and Hati [52], which captures the degree of well-being experienced by early childhood educators in their work context. The scale includes nine items, with sample statements such as “I am quite satisfied with my job” and “I enjoy meaningful work.” Participants responded using a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree), where higher scores reflect greater workplace well-being. Cronbach’s α for this study was 0.947, indicating excellent internal consistency.

Job embeddedness was measured using the scale developed by Crossley et al. [40], which evaluates the extent to which early childhood educators feel connected to and embedded within their job and work environment. The scale comprises seven items, such as “I feel a strong sense of belonging to my job as an early childhood educator” and “I feel closely connected to my job as an early childhood educator.” One item—“It would be easy for me to leave my job as an early childhood educator (R)”—was reverse coded. Responses were recorded on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree), with higher scores indicating greater levels of job embeddedness. The Cronbach’s α for this sample was 0.925, demonstrating strong internal consistency.

Retention intention was measured using a four-item scale adapted from Cheng et al. [53], assessing the extent to which early childhood educators intend to remain in their current position over the next three years. Sample items include “I will continue working in my current kindergarten” and “I intend to remain in this kindergarten for at least the next two to three years.” One item—“I will look for a new job in the near future (R)”—was reverse coded. All items were rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Higher scores indicate stronger retention intention. The scale demonstrated good internal consistency, with a Cronbach’s α of 0.812.

3.3. Common Method Bias Test

To assess the potential influence of CMV, this study first conducted Harman’s single-factor test. The unrotated factor solution indicated that the first factor accounted for 55.111% of the variance, slightly above the conventional 50% threshold [54,55]. Given that Harman’s test is widely regarded as an insufficient and overly sensitive diagnostic, particularly in complex models, additional and more rigorous statistical procedures were implemented as recommended in the methodological literature.

First, confirmatory factor analysis (CFA)-based competing model comparisons were performed. As shown in

Table 2. Comparison of Competing Measurement Models.

Model	χ2	χ2/df	GFI	AGFI	RMSEA	CFI	IFI	NFI	RFI	NNFI (TLI)
Model 1: Four-Factor Model	1080.319	3.687	0.691	0.651	0.116	0.844	0.845	0.799	0.776	0.827
Model 2: Three-Factor Model	1273.832	4.303	0.642	0.602	0.129	0.806	0.807	0.762	0.739	0.787
Model 3: Two-Factor Model	1386.013	4.651	0.627	0.587	0.135	0.784	0.785	0.741	0.718	0.764
Model 4: Single-Factor Model	18,777.929	62.802	0.503	0.453	0.162	0.686	0.688	0.650	0.619	0.659

Note: Model 1 (Four-Factor Model) represents the hypothesized measurement structure of this study, in which work place friendship, workplace well-being, job embeddedness, and retention intention are specified as four distinct latent constructs; Model 2 (Three-Factor Model) combines workplace friendship, workplace well-being, and job embeddedness into a single factor while retaining retention intention as a separate factor; Model 3 (Two-Factor Model) combines workplace friendship, workplace well-being, and job embeddedness into one factor, with retention intention specified as the second factor; Model 4 (Single-Factor Model) specifies all measurement items from the four constructs as indicators of a single latent factor.

, the hypothesized four-factor measurement model demonstrated substantially superior fit compared with alternative three-factor, two-factor, and especially the single-factor model. These results indicate that a single latent factor cannot account for the covariance among the items, providing strong evidence that CMV does not threaten the validity of the findings [56].

Second, variance inflation factors (VIFs) were examined as an additional diagnostic tool. All VIF values ranged from 2.230 to 3.666, well below the conservative threshold of 5 [50], suggesting no substantial collinearity or inflation in shared variance attributable to common method bias.

Taken together, the CFA model comparisons and VIF diagnostics demonstrate that although the initial Harman’s test suggests a moderate level of shared variance, CMV is unlikely to pose a serious threat to the integrity of the results. Therefore, the observed relationships among the study variables can be interpreted with confidence.

3.4. Data Analysis

This study employed AMOS 24.0 to conduct confirmatory factor analysis (CFA) and to test the structural paths and mediation effects within the structural equation model (SEM). First, for the measurement model, maximum likelihood estimation was used to perform CFA in order to assess convergent validity, discriminant validity, and overall model fit for the latent constructs.

For the structural model, a theory-driven SEM was developed to examine the hypothesized path relationships and mediation effects. Both path coefficients and indirect effects were evaluated using the bootstrap procedure with 5000 resamples. Bias-corrected 95% confidence intervals (CIs) were used as the criterion for significance; when the CI did not include zero, the corresponding path or indirect effect was considered statistically significant. In addition to reporting indirect effects, this study also presents the direct effects and total effects to determine whether the observed mediation reflects full or partial mediation. This analytic strategy avoids reliance solely on p-values and provides a more comprehensive evaluation of the precision and stability of the estimated effects.

4. Research Results

4.1. Confirmatory Factor Analysis

Results of the confirmatory factor analysis indicated that all standardized factor loadings for the four latent constructs were statistically significant (see

Table 3. Measurement model.

Factors	Items	M	SD	Skewness	Kurtosis	Factor Loading
Workplace friendship (CR = 0.949, AVE = 0.756)
	Workplace friendship 1	3.760	1.004	−0.406	−0.513	0.889
	Workplace friendship 2	3.530	1.169	−0.445	−0.562	0.840
	Workplace friendship 3	3.755	1.044	−0.430	−0.557	0.889
	Workplace friendship 4	3.665	1.015	−0.312	−0.519	0.889
	Workplace friendship 5	3.610	1.069	−0.365	−0.366	0.848
	Workplace friendship 6	3.390	1.151	−0.287	−0.521	0.860
Workplace well-being (CR = 0.947, AVE = 0.667)
	Workplace well-being 1	4.920	1.481	−0.358	−0.620	0.805
	Workplace well-being 2	5.605	1.424	−0.931	0.057	0.742
	Workplace well-being 3	5.495	1.425	−0.700	−0.330	0.764
	Workplace well-being 4	5.865	1.340	−1.004	0.084	0.733
	Workplace well-being 5	5.270	1.506	−0.674	−0.267	0.889
	Workplace well-being 6	5.230	1.499	−0.544	−0.415	0.908
	Workplace well-being 7	5.265	1.409	−0.404	−0.572	0.796
	Workplace well-being 8	4.745	1.681	−0.355	−0.674	0.816
	Workplace well-being 9	5.235	1.425	−0.496	−0.565	0.878
Job embeddedness (CR = 0.925, AVE = 0.640)
	Job embeddedness 1	3.345	1.054	−0.209	−0.302	0.783
	Job embeddedness 2	3.060	1.050	0.064	−0.249	0.692
	Job embeddedness 3	3.825	1.034	−0.661	−0.137	0.858
	Job embeddedness 4	3.660	1.086	−0.572	−0.221	0.904
	Job embeddedness 5	3.115	1.081	−0.014	−0.348	0.700
	Job embeddedness 6	3.750	1.045	−0.443	−0.341	0.876
	Job embeddedness 7	3.930	1.039	−0.645	−0.430	0.758
Retention intention (CR = 0.815, AVE = 0.530)
	Retention intention 1	3.110	0.986	−0.096	0.095	0.773
	Retention intention 2	3.405	1.028	−0.232	−0.246	0.855
	Retention intention 3	3.750	1.065	−0.369	−0.703	0.596
	Retention intention 4	3.870	1.175	−0.609	−0.683	0.659

), ranging from 0.596 to 0.908, which exceed the recommended threshold of 0.50. This demonstrates that each item adequately represents its corresponding latent construct [50]. The composite reliability (CR) values ranged from 0.815 to 0.949, all above the recommended cutoff of 0.70, indicating strong internal consistency among the measurement indicators.

To further assess convergent validity, the average variance extracted (AVE) values were examined. The AVE values for all constructs ranged from 0.530 to 0.756, exceeding the 0.50 benchmark [57], thereby supporting adequate convergent validity. The normality of the measurement items was also assessed. Skewness values ranged from −1.004 to 0.064, and kurtosis values ranged from −0.703 to 0.095, all within the acceptable ranges (skewness < 2, kurtosis < 7) [50], suggesting no severe violations of the assumption of normal distribution.

In addition, discriminant validity was assessed using the Heterotrait–Monotrait ratio (HTMT) (see

Table 4. Discriminant Validity.

Variable	Workplace Friendship	Workplace Well-Being	Job Embeddedness
1. Workplace friendship	-
2. Workplace well-being	0.664	-
3. Job embeddedness	0.789	0.842	-
4. Retention intention	0.574	0.676	0.733

Note: The lower triangular matrix presents Heterotrait–Monotrait Ratio of Correlations (HTMT).

), following the criterion proposed by Henseler et al. [58], which recommends a threshold of 0.85. The HTMT values for all constructs in this study ranged from 0.574 to 0.842, all below the suggested cutoff. These results indicate that the constructs demonstrate satisfactory discriminant validity.

4.2. Model Fit

Because the chi-square statistic in SEM is sensitive to violations of multivariate normality and tends to inflate model misfit when sample sizes are moderate or when item distributions deviate slightly from normality, the Bollen–Stine bootstrap procedure was applied to obtain a corrected chi-square value. This procedure generates an empirical distribution of the chi-square statistic through resampling, thereby improving the accuracy and conservativeness of model fit assessment under mild non-normality conditions [50,59].

As shown in

Table 5. Bollen–Stine Bootstrap Model Fit Indices.

Fit Index	Criterion	Bollen–Stine Bootstrap Model
χ2	Smaller is better	421.812
χ2/df	<3	1.434
GFI	>0.90	0.924
AGFI	>0.90	0.905
RMSEA	<0.08	0.052
CFI	>0.90	0.973
IFI	>0.90	0.973
NFI	>0.90	0.924
RFI	>0.90	0.914
NNFI(TLI)	>0.90	0.973

, the model demonstrated acceptable to excellent fit: χ²/df = 1.434 (<3), GFI = 0.924, AGFI = 0.905, RMSEA = 0.052 (<0.08), CFI = 0.973, IFI = 0.973, NFI = 0.924, RFI = 0.914, and TLI = 0.973. All indices met their commonly recommended thresholds [50]. These results indicate that the overall model fit is satisfactory, providing a robust foundation for subsequent path and mediation analyses.

4.3. Path Analysis

The structural model was further examined through path analysis, and the results are presented in

Table 6. Path Analysis Table.

Path Relationship	β	SE	z	95% CI
				Lower	Upper
Workplace Friendship → Workplace Well-Being	0.716	0.086	10.577 ***	0.735	1.113
Workplace Friendship → Job Embeddedness	0.798	0.063	10.841 ***	0.546	0.816
Workplace Friendship → Retention Intention	−0.022	0.102	−0.165	0.017	0.362
Workplace Well-Being → Retention Intention	0.305	0.057	3.187 **	−0.226	0.178
Job Embeddedness → Retention Intention	0.491	0.107	4.067 ***	0.150	0.759

*** p < 0.001, ** p < 0.01.

. First, regarding the relationships between the independent variable and the mediators, workplace friendship had a significant positive effect on workplace well-being (β = 0.716, z = 10.577, p < 0.001) and on job embeddedness (β = 0.798, z = 10.841, p < 0.001). These findings indicate that stronger workplace friendship enhances both well-being and embeddedness among early childhood educators. Thus, Hypotheses H2 and H4 were supported.

Second, in predicting the dependent variable—retention intention—the direct effect of workplace friendship was not significant (β = −0.022, z = −0.165, p > 0.05), suggesting that workplace friendship alone does not directly increase educators’ intention to stay. Therefore, Hypothesis H1 was not supported. In contrast, workplace well-being had a significant positive effect on retention intention (β = 0.305, z = 3.187, p < 0.01), and job embeddedness also demonstrated a significant positive effect (β = 0.491, z = 4.067, p < 0.001). These results indicate that retention intention is primarily influenced by workplace well-being and job embeddedness, which serve as mediating mechanisms through which workplace friendship exerts its indirect effects, rather than being explained by the direct effect of workplace friendship.

The structural model demonstrated satisfactory explanatory power for the endogenous variables. Workplace friendship accounted for approximately 54.5% of the variance in job embeddedness (R² = 0.545), representing a moderate to high level of explanatory strength. It explained 40.8% of the variance in workplace well-being (R² = 0.408), indicating moderate explanatory power. Finally, the model explained 43.0% of the variance in retention intention (R² = 0.430), suggesting a meaningful level of practical explanatory strength. These results support the theoretical proposition that workplace friendship—conceptualized as a key job resource—affects retention intention through both psychological (well-being) and structural (job embeddedness) pathways.

4.4. Mediation Analysis

After confirming the significance of the primary structural paths, this study further examined the mediating effects of workplace well-being and job embeddedness on the relationship between workplace friendship and retention intention. The results are presented in

Table 7. Indirect Effects (Bootstrap Results).

Path Relationship	β	SE	z	95% CI
				Lower	Upper
WF → WWB → RI	0.165	0.081	2.037 *	0.012	0.339
WF → JE → RI	0.295	0.113	2.610 **	0.106	0.551

Note: WF = Workplace Friendship, WWB = Workplace Well-Being, JE = Job Embeddedness, RI = Retention Intention. * p < 0.05, ** p < 0.01.

. First, the direct effect of workplace friendship on retention intention was not significant (β = −0.022, p > 0.05), suggesting that workplace friendship alone does not directly enhance educators’ intention to remain in their current positions. However, the indirect effects were significant. Workplace friendship exerted a significant positive indirect effect on retention intention through workplace well-being (β = 0.165, z = 2.037, p < 0.05). Similarly, workplace friendship demonstrated a significant positive indirect effect through job embeddedness (β = 0.295, z = 2.610, p < 0.01). These findings indicate that workplace friendship influences retention intention indirectly by increasing employees’ well-being and embeddedness.

When considering both direct and indirect effects, the total effect of workplace friendship on retention intention was significant (β = 0.588, z = 9.333, p < 0.001). Importantly, the significance of the total effect derived entirely from the indirect paths rather than from a direct effect. This pattern supports a full mediation model, indicating that workplace friendship affects retention intention only when translated into psychological (well-being) and structural (embeddedness) resources. Overall, these results highlight that workplace friendship is not merely an emotional or relational resource; rather, it influences retention by fostering positive psychological states and deeper connections to one’s work, which in turn shape educators’ decisions to remain in their roles.

5. Discussion

5.1. Theoretical Implications

The findings of this study offer several theoretical insights into the mechanisms through which workplace friendship influences early childhood educators’ retention intention. First, the results show that workplace friendship does not exert a significant direct effect on retention intention. This contrasts with prior studies suggesting that friendship reduces turnover tendencies [27,28], and indicates that friendly interpersonal climates alone are insufficient to anchor employees’ long-term commitment. From the JD-R perspective, workplace friendship constitutes a social job resource; however, social resources must be transformed into motivational or stress-buffering psychological resources before they meaningfully influence retention decisions.

Second, the study demonstrates that workplace friendship significantly enhances both workplace well-being and job embeddedness. This aligns with Chen, Wang and Chu [16], who showed that friendship improves commitment through increased well-being, and with findings by Asgharian et al. [48] and Dechawatanapaisal [46] highlighting that workplace support strengthens fit, attachment, and retention. These results reinforce the JD-R proposition that social resources stimulate both psychological resources (well-being) and structural resources (embeddedness), thereby reducing turnover tendencies.

It is important to note that the mechanisms identified in this study differ conceptually and empirically from those proposed by Çınar and Basım [29]. Their research positions workplace friendship as a mediator linking relational job crafting to employees’ intention to stay. By contrast, the present study conceptualizes workplace friendship as an antecedent social job resource, consistent with the JD-R model, and examines its downstream effects on retention intention through two distinct mediators—workplace well-being and job embeddedness. Rather than focusing on how friendship is formed, our interest lies in how existing friendship networks are translated into psychological (well-being) and structural (embeddedness) resources that ultimately shape retention decisions. Moreover, our findings reveal a full mediation pattern, demonstrating that friendship alone is insufficient to directly increase retention intention in early childhood education settings. Instead, its influence operates entirely through the accumulation of well-being and embeddedness—two mechanisms not simultaneously examined in Çınar and Basım’s work.

By identifying well-being and embeddedness as dual mediating mechanisms, this study advances the theoretical understanding of how workplace friendship functions as a job resource and clarifies the pathways through which social relationships influence educators’ intention to stay. These findings also contribute to broader discussions on talent sustainability by illustrating that the key challenge is not merely fostering positive interpersonal interactions but transforming such interactions into enduring psychological and organizational capital that strengthens educators’ attachment and stability within the system.

In sum, this study, grounded in the JD-R framework, examines the relationships among workplace friendship, workplace well-being, job embeddedness, and retention intention, and validates the mediating roles of well-being and embeddedness. The findings further reveal that although workplace friendship does not directly enhance retention intention, it indirectly contributes through the accumulation of psychological and structural resources. This study not only enriches the literature on workplace friendship but also offers practical implications for early childhood institutions, emphasizing the importance of cultivating a supportive and friendly work environment, promoting well-being, and strengthening job embeddedness as strategies for fostering workforce sustainability.

5.2. Managerial Implications

First, the findings highlight that workplace friendship serves as an essential foundation for fostering workplace well-being and job embeddedness. ECEC institutions should actively cultivate a friendly and trusting organizational climate rather than relying on spontaneous interpersonal bonding. In addition to regular professional learning communities and inter-role workshops, directors may implement structured peer-mentoring programs, pairing experienced educators with newer staff to promote professional support and interpersonal connection. Furthermore, informal social gatherings, such as team breakfasts, birthday celebrations, or short community-building activities during staff meetings, can help strengthen interpersonal ties in a natural and low-pressure manner. Encouraging supervisors to adopt open, empathetic, and supportive communication styles further reinforces a climate of trust and belonging, thereby reducing isolation and turnover risks [25].

Second, workplace well-being plays a central role in predicting retention intention. Institutions can enhance employees’ psychological resources by implementing flexible scheduling practices, providing professional development subsidies, and cultivating a culture of positive feedback. Regular well-being check-ins or staff well-being surveys may also help directors monitor emotional strain and identify areas requiring organizational support. These strategies can elevate satisfaction and motivation among staff [35].

Third, job embeddedness is another key mediator through which workplace friendship influences retention intention. Management interventions may focus on the three core dimensions of embeddedness—links, fit, and sacrifice. Specifically: (1) establishing mentoring or peer-support programs to increase interpersonal links; (2) ensuring alignment between employees’ values and the center’s culture through targeted training, cultural onboarding, and reflective dialogue sessions; (3) offering attractive benefits such as educational allowances, family-support policies, or tenure-based incentives to increase the perceived sacrifice of leaving. Such measures reinforce employees’ embeddedness and strengthen long-term retention [40,46].

Finally, these initiatives should be integrated into a broader Sustainable Human Resource Management (Sustainable HRM) framework. To move beyond a conceptual emphasis on “viewing” well-being and embeddedness as long-term human capital, directors may incorporate specific, measurable indicators of these constructs into institutional sustainability planning [60]. For example, well-being indicators—such as job satisfaction levels, emotional exhaustion scores, or perceived organizational support—can be included in annual quality assurance reports. Similarly, embeddedness-related indicators—such as retention rates, participation in center-level activities, or engagement in mentoring programs—can be tracked as part of long-term workforce planning. Embedding these metrics into regular HR audits and institutional dashboards enables organizations to monitor progress, allocate resources strategically, and align HR practices with the long-term objectives of SDG 4 (Quality Education) and SDG 8 (Decent Work and Economic Growth) [61,62]. By institutionalizing these indicators, ECEC organizations can ensure a stable workforce and strengthen the sustainability of the early childhood education system.

5.3. Research Limitations and Future Research Developments

First, this study relied on self-reported questionnaires. Although anonymity, reverse-coded items, and procedural controls were implemented, common method variance (CMV) cannot be fully eliminated. Future studies may use multi-source data, longitudinal designs, or objective indicators (e.g., actual retention records) to enhance validity.

Second, this study adopted the JD-R framework and focused on workplace friendship as a social job resource. Other job resources—such as organizational support, leadership styles, or career development opportunities—may also shape retention intention. Future research could incorporate additional resources to build a more comprehensive model.

Third, the use of convenience sampling limits the generalizability of the findings. The sample may not fully represent all ECEC personnel in Taiwan. Future research should consider probability or stratified sampling to improve representativeness.

Fourth, the tenure variable captured years within the current institution, and most participants (65.5%) had less than three years of institutional tenure. Although many educators may possess longer overall professional experience, shorter institution-specific tenure may weaken job embeddedness, as links, fit, and perceived sacrifice typically grow over time. Future studies should distinguish total professional tenure from institution-specific tenure.

A related limitation concerns the inclusion criterion of at least three months of tenure. While a three-month probation period is the standard practice in Taiwan and ensures basic workplace familiarity, probationary employees may not yet have fully formed embeddedness-related ties. Future research may adopt alternative thresholds or examine embeddedness development across tenure stages.

Finally, workplace friendship did not directly predict retention intention, suggesting possible moderating mechanisms. Future research could examine personal resources (e.g., self-efficacy, sense of calling, emotional intelligence) and organizational resources (e.g., fairness climate, perceived organizational support) as moderators. Sustainable HRM intervention studies—such as mentoring programs or community-based benefits—would also help assess long-term impacts on well-being, embeddedness, and retention.