This study aims to provide international evidence on the effectiveness of mentoring programs in increasing new teachers’ expected career length, focusing on the mediating roles of job satisfaction and organizational commitment.
This study utilizes multilevel structural equation modeling to analyze international data from the Teaching and Learning International Survey 2018.
The results indicate a direct effect of mentoring program participation on novice teachers’ expected career length. The study reveals that this relationship is mediated by organizational commitment, while job satisfaction does not serve as a mediating factor. These findings deepen our understanding of the mechanisms through which mentoring programs relate to novice teachers’ expected career length.
School leaders and policymakers are encouraged to provide and tailor mentoring programs in a way to promote organizational commitment, thereby enhancing the effectiveness of mentoring on novice teachers’ retention.
The insights will be informative for school leaders and administrators in retaining novice teachers, supporting more efficient school administration, management and improvement.
Teacher attrition, particularly among novice educators, stands as a critical global issue contributing to the worldwide shortage of teaching professionals (Amitai and Van Houtte, 2022; Nguyen and Springer, 2023). UNESCO reports that approximately 44 million new teachers need to be recruited by 2030 to maintain global educational quality and accessibility, largely due to the replacement of departing teachers (UNESCO and International Task Force on Teachers for Education, 2030, 2024). Statistics also show that nearly fifty percent of new teachers leave the profession within their initial five years (Ingersoll et al., 2018). This incurs substantial costs related to recruitment, replacement, and training (Nguyen et al., 2020). It also adversely affects student learning, leading to exposure to inexperienced educators, instructional instability, and disruption of school climate (Carver-Thomas and Darling-Hammond, 2019; Sorensen and Ladd, 2020).
To address this issue, various forms of institutional support have been developed and implemented by national governments, states, districts, and school leaders (Spooner-Lane, 2017). Notably, mentoring programs have shown promise in enhancing retention among novice teachers (Ronfeldt and McQueen, 2017), while teachers’ retention is determined by various factors rather than a single aspect (Casely-Hayford et al., 2022; Nguyen et al., 2020). However, the mechanisms linking mentoring to retention intention among new teachers are still unclear. As Long and colleagues (2012) note “the effect of induction (including mentoring) programs is unclear in the light of multiple factors that influence teachers’ staying or leaving” (p. 21). Ignoring these indirect pathways of mentoring on teacher retention, such as its role in developing other influential factors, may lead to an underestimation of its effectiveness. A comprehensive understanding of these indirect pathways could unveil nuanced strategies to enhance the overall effectiveness of mentoring programs.
We address this gap by examining job satisfaction and organizational commitment as potential mediators between mentoring and the expected career length among novice teachers. Research highlights job satisfaction and organizational commitment as key predictors of new teachers’ retention decisions (Ashiedu and Scott-Ladd, 2012; Da’as et al., 2020; Sass et al., 2011). The literature also documents the potential of mentoring programs to enhance job satisfaction and organizational commitment (Burger et al., 2021; Hong and Matsko, 2019). However, there exists a dearth of comprehensive international evidence regarding the extent to which job satisfaction and organizational commitment mediate the relationship between mentoring and the intention of new teachers to remain. Building upon existing knowledge, our study seeks to explore this dynamic, contributing to a holistic understanding of these interconnections.
We use data from the Teaching and Learning International Survey (TALIS) 2018, which includes lower secondary schools across 48 countries/economies. We seek to provide conceptual and practical insights for policymakers and educational leaders, offering guidance on refining mentoring programs to optimize their impact on novice teacher retention. Our research is guided by two questions:
Does participation in a mentoring program correlate with an increase in the expected career length of new teachers?
Do job satisfaction and organizational commitment mediate this relationship?
Literature review and hypotheses
Teacher retention and attrition
In the U.S., nearly half of novice teachers depart the profession within their initial five years (Ingersoll and Strong, 2011). This disconcerting attrition trend is observed in various countries, such as Australia, Chile, and the United Kingdom (Ávalos and Valenzuela, 2016; Department for Education, 2019; Mason and Matas, 2015). Extensive research underscores the adverse ramifications of such novice teacher attrition, influencing various aspects of schools, including organizational effectiveness, financial stability, and students’ academic performance (Hanushek et al., 2016; Ronfeldt et al., 2013; Watlington et al., 2010).
Scholars have investigated the determinants of teacher retention and attrition (Ingersoll, 2001). The findings emphasize that teacher retention is a complex interplay of individual emotional and psychological factors intertwined with institutional support and resources (Ashiedu and Scott-Ladd, 2012; Nguyen et al., 2024). At the individual level, research indicates that teachers with high levels of job satisfaction exhibit a greater inclination to remain within their roles (Nguyen et al., 2020). Waddell’s (2010) work elucidates that a sense of commitment to their schools correlates with a stronger intention to continue teaching. This study also suggests that fostering such commitment and satisfaction can be facilitated through institutional support, such as professional learning communities and mentoring/induction programs. Echoing this perspective, research underscores the role of school-based support, including mentoring and induction, in retaining novice teachers (Desimone et al., 2014; Ronfeldt and McQueen, 2017). Notably, a comprehensive review study on mentoring and induction conducted by Ingersoll and Strong (2011) concludes that novice teachers participating in induction programs demonstrate enhanced retention rates. This evidence pertains to both the actual decision and the self-reported intentions to persist or depart.
The reliability of retention intention (e.g. expected career length) as an indicator for actual decisions has sparked ongoing debate among scholars, presenting divergent viewpoints that require careful consideration. Some studies show that retention intention reliably predicts actual turnover (Cho and Lewis, 2012; Sun and Wang, 2017). Conversely, some scholarly inquiries cast doubt on its reliability, highlighting instances where the alignment between intent and actual retention is not unequivocal (Grant and Brantlinger, 2023). Despite these reservations, research often employs teachers’ expected career length as a proxy for retention due to the scarcity of accessible data (Richter et al., 2022). Research aiming to provide international evidence on teacher retention exclusively depends on intentions as a metric, given the lack of cross-national longitudinal data capturing teachers’ trajectories (e.g. Van den Borre et al., 2021).
Mentoring
Mentoring is an important aspect of the induction process, recognized as a tool across various countries for supporting new teachers as they begin their careers (Shanks et al., 2022). Teacher induction refers to structured professional development programs designed to support the transition of novice teachers into the profession, with the goal of enhancing teacher effectiveness and retention (Van den Borre et al., 2021). These programs typically involve orientation, workshops, collaborative activities, and mentoring (Ingersoll and Strong, 2011). Smith and Ingersoll (2004) demonstrated that induction support significantly reduced first-year attrition and migration.
According to a TALIS report from the Organisation for Economic Co-operation and Development (OECD), mentoring is described as a support system where experienced teachers help less experienced colleagues succeed in schools (OECD, 2019a). The report also notes that principals globally recognize the significance of mentoring in teachers’ professional development while 22% of new teachers have an assigned mentor across OECD countries. With the widespread recognition of mentoring, it has emerged as a strategic countermeasure to the challenge of attrition among novice educators, effectively nurturing the enduring commitment of novice teachers (Ingersoll and Strong, 2011; Pogodzinski, 2015).
Smith and Ingersoll (2004) demonstrated that new educators who received mentorship within their subject domains and participated in collaborative induction endeavors were less inclined to leave the teaching profession after their initial year. Likewise, Ronfeldt and McQueen (2017) found that new teachers who engaged in mentoring programs exhibited heightened retention rates during their inaugural year. Broadening the scope internationally, Van den Borre et al. (2021) employed cross-national data from TALIS 2018 to reveal that involvement in mentoring initiatives constituted a predictive factor for the retention intentions of early career educators at an individual level. While these studies focused on mentoring participation itself using a dichotomous variable, research also explored various features of mentoring, such as frequency and quality of interactions (Desimone et al., 2014). Overall, the literature underscores the integral role of mentoring in fostering the initial professional development of teachers and sustaining their dedication through the exchange of insights on crucial topics such as student dynamics, classroom pedagogy, teacher assessment, and alignment of curriculum standards for planning, application, and evaluation (Spooner-Lane, 2017).
Empirical investigations have consistently revealed the positive impact of mentoring programs on enhancing job satisfaction and organizational commitment among novice educators (Burger et al., 2021; Hong and Matsko, 2019; Smith and Ingersoll, 2004). Alhija and Fresko (2010) conducted a study in Israel, establishing a correlation between active participation in a school-based mentoring program and high levels of job satisfaction and commitment. Their study accentuated the multifaceted effects of mentoring, functioning as a mechanism to aid novice teachers in their adjustment to the unique cultural and communal nuances of their educational environment while fostering their growth in both pedagogical and personal dimensions. Moreover, through the analysis of survey and administrative data from Chicago Public Schools, Hong and Matsko (2019) revealed that mentoring programs characterized by frequent and substantive interactions with mentors directly contributed to the organizational commitment of novice educators. Importantly, their findings emphasized the need for additional empirical exploration into the intricate relationship between mentoring programs and various outcomes within diverse national contexts. They advocated for forthcoming research that delves deeper into the dynamic interplay between mentoring, organizational commitment, and teacher retention.
Building on the literature, our study postulates that participation in mentoring programs serves as a catalyst for nurturing retention intentions (Van den Borre et al., 2021). In addition, the present study conceptualizes participation in a mentoring program as a critical factor in fostering both job satisfaction and organizational commitment (Alhija and Fresko, 2010; Hong and Matsko, 2019). This assumption is reinforced by the notion that teachers who receive mentoring can integrate into their professional roles and respective organizations. Accordingly, the hypotheses formulated are as follows:
Participation in a mentoring program is positively associated with novice teachers’ expected number of years to continue teaching.
Participation in a mentoring program is positively associated with teachers’ job satisfaction.
Participation in a mentoring program is positively associated with teachers’ organizational commitment.
Job satisfaction and organizational commitment
Job satisfaction is characterized as a positive emotional state and a favorable attitude toward one’s own work experiences and responsibilities (Locke, 1976). On the other hand, organizational commitment entails a sense of identification with and active engagement in the organization (Mowday et al., 1979). While job satisfaction predominantly draws from an individual’s personal beliefs about their role, organizational commitment signifies an alignment with the organization’s values and objectives (Yoon and Thye, 2002). Recognizing these inherent distinctions, research has treated them as distinct constructs while incorporating them within a unified framework to predict employees’ outcomes, recognizing that individual behaviors, performance, and attitudes are influenced by both personal attributes and interactions with the organizational context in which they belong (Yiing and Ahmad, 2009).
Empirical investigations have demonstrated that both job satisfaction and organizational commitment significantly influence various employee outcomes, such as motivation, well-being, and retention, within diverse work settings, including educational institutions (Bolt et al., 2022; Madigan and Kim, 2021). In the context of teacher retention, research conducted across various countries, such as Australia, the United States, Nigeria, Oman, and Sweden, has found positive associations between job satisfaction, organizational commitment (referred to as school commitment in the context of teachers), and teachers’ decisions and intentions to remain within the profession (Casely-Hayford et al., 2022; Imran et al., 2017; Kelly et al., 2019).
Based on existing literature on the importance of job satisfaction and organizational commitment in teacher retention (Kelly et al., 2019; Madigan and Kim, 2021), this study incorporates these constructs into an analytical framework to better understand the effect of mentoring on the retention of novice teachers. It investigates how and to what extent job satisfaction and organizational commitment explain the relationship between participation in mentoring programs and retention intention by addressing the following questions:
The relationship between participation in the mentoring program and novice teachers’ expected number of years to continue teaching will be partially mediated by their levels of job satisfaction.
The relationship between participation in the mentoring program and novice teachers’ expected number of years to continue teaching will be partially mediated by their organizational commitment.
Methods
Data and sample
This study utilized cross-national data from TALIS 2018, designed by the OECD, involving 48 countries/economies participating in the core survey within lower secondary schools. TALIS 2018 provided publicly available data encompassing responses from individual teachers and school principals, mainly focusing on learning and working environments within schools (OECD, 2019b).
The sampling strategy of TALIS 2018 was a stratified two-stage probability sampling design, indicating that about 200 schools were selected with a probability proportional to size per country. Within each school, about 20 teachers were randomly selected. To align with prior research on novice teachers and acknowledge the significant attrition rates within the first five years of their careers (Van den Borre et al., 2021), the sample was limited to teachers with less than six years of total work experience. As a result, the analytic sample comprised 20,612 teachers in 6,227 schools within 41 countries [1].
The amount of missing data ranged from 0% (e.g. expected number of years to continue teaching and mentoring) to 1% (e.g. job satisfaction and organizational commitment). The results of missing completely at random test indicated that the missing data were not missing completely at random, (1,474) = 4655.708, p < 0.001. To deal with the missing data, this study employed the full information maximum likelihood (FIML) estimation (Graham, 2003). Using FIML generates less biased estimates than other methods like listwise deletion, pairwise deletion, and similar response pattern imputation (Enders and Bandalos, 2001).
Measures
The dependent variable was derived from TALIS 2018 Teacher Questionnaire Data, which asked teachers about their expected number of years to continue teaching using a continuous scale: “For how many more years do you want to continue to work as a teacher?” Responses that were clearly outliers and seemingly unrealistic were excluded to reduce biases in the estimates (i.e. where the expected career length exceeds 60 years).
The main independent variable was a dichotomous variable indicating participation in a mentoring program (i.e. I currently have an assigned mentor to support me): 0 = no; 1 = yes. As highlighted, the survey design allowed us to identify participants at the time they were surveyed, which may not include those who had mentoring prior to the survey. Therefore, the interpretation of the results should be understood as the difference between participants and non-participants in mentoring at the time of the survey.
We created two latent variables: job satisfaction and organizational commitment as mediating variables. Job satisfaction was created using three indicators (e.g. If I could decide again, I would still choose to work as a teacher) rated on a 4-point Likert scale (from 1 = strongly disagree to 4 = strongly agree) (Online Supplementary Appendix A for more details). While the TALIS report presented a latent job satisfaction construct encompassing four indicators (OECD, 2019b), we conducted exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) due to the specific focus on novice teachers in the sample. Convergent validity was assessed using statistically significant factor loadings that should exceed 0.55 (Tabachnick and Fidell, 2007), leading to the exclusion of one indicator. We calculated the average variance extracted (AVE) values, which should preferably be 0.5 or higher, and construct reliability (CR) values, which should be 0.7 or higher (Hair et al., 2006). All indicator factor loadings were statistically significant (p < 0.001, 0.730–0.750). The AVE value was 0.563 and the CR value was 0.795. The Cronbach’s alpha, and the Omega reliability coefficient (ω) were calculated and yielded a value of 0.799 and 0.800, respectively. All results collectively indicated the construct’s acceptability (see Online Supplementary Appendix A).
Organizational commitment was derived from TALIS 2018 Teacher Questionnaire Data, which consisted of three indicators (e.g. I enjoy working at this school) rated on a 4-point Likert scale ranging from 1 for “strongly disagree” to 4 for “strongly agree” (refer to Online Supplementary Appendix A for more details). The organizational commitment scale was designed to measure the extent of affective attitudes/commitment to their schools. We adopted three items that probed teachers’ affective attitudes and commitment towards their respective schools, guided by the findings from EFA (Liu and Watson, 2020). All factor loadings were found to be statistically significant (p < 0.001, 0.641–0.873). The Cronbach’s alpha coefficient calculated for this scale was 0.807, while the Omega reliability coefficient (ω) registered at 0.798. The AVE value was computed as 0.596, and the CR value was 0.814, both of which align with the recommended thresholds (Hair et al., 2006). Taken together, these outcomes validate the acceptability of the organizational commitment construct (see Online Supplementary Appendix A).
Guided by the existing literature, we incorporated five teacher background variables as potential confounding factors: gender, age, years of teaching experience, full-time status, and possession of an advanced degree (Pas et al., 2012). Additionally, we considered four school background variables, encompassing school type, school socioeconomic status (ESES), school size, and school location (city or rural; Brault et al., 2014). Descriptive statistics for all teacher-level and school-level variables analyzed are presented in Table 1. Furthermore, Online Supplementary Appendix A outlines the correlations among the primary variables. The results of the inter-variable correlations demonstrated no indications of collinearity.
Descriptive statistics of the variables in the study
| Variables | n | Mean | SD | Min | Max | Description |
|---|---|---|---|---|---|---|
| Expected number of years to continue teaching | 20,612 | 20.54 | 12.82 | 0 | 60 | |
| Mentoring (%) | 20,612 | 0.28 | – | 0 | 1 | |
| Job satisfaction 1 | 20,410 | 3.04 | 0.82 | 1 | 4 | |
| Job satisfaction 2 | 20,410 | 3.39 | 0.72 | 1 | 4 | |
| Job satisfaction 3 | 20,397 | 2.89 | 0.91 | 1 | 4 | |
| Organizational commitment 1 | 20,406 | 3.01 | 0.88 | 1 | 4 | |
| Organizational commitment 2 | 20,410 | 3.23 | 0.70 | 1 | 4 | |
| Organizational commitment 3 | 20,397 | 3.10 | 0.77 | 1 | 4 | |
| Female (%) | 20,612 | 0.67 | – | 0 | 1 | Dummy variable, 0 = male, 1 = female |
| Age group | 20,612 | 1.71 | 0.95 | 1 | 5 | Categorical variable, 1 = under 30, 2 = 30 to 39, 3 = 40 to 49, 4 = 50 to 59, 6 = 60 and above |
| Years of teaching experience | 20,612 | 2.89 | 1.48 | 1 | 5 | Ranging from first-year teachers to those with five years of experience |
| Fulltime (%) | 20,612 | 0.75 | – | 0 | 1 | Dummy variable, 0 = part-time, 1 = full-time |
| Advanced degree | 20,612 | 0.37 | – | 0 | 1 | Dummy variable, 0 = less, 1 = master’s or doctoral |
| School type (%) | 6,227 | 0.82 | – | 0 | 1 | Dummy variable, 0 = private, 1 = public |
| ESES | 6,227 | 2.07 | 0.55 | 1 | 3 | Categorical variable, indicating the proportion of low-socioeconomic status students, 1 = None, 2 = 1%–30%, 3 = 31% to more than 60% |
| School size | 6,227 | 2.89 | 1.39 | 1 | 5 | Categorical variable, indicating the number of enrolled students, 1 = below 250, 2 = 250 to 499 3 = 500 to 749, 4 = 750–999, 5 = 1,000 and above |
| School rural (%) | 6,227 | 0.34 | – | 0 | 1 | Dummy variable, 0 = else, 1 = in a village/small town |
| School city (%) | 6,227 | 0.41 | – | 0 | 1 | Dummy variable, 0 = else, 1 = in a city/large city |
| Variables | n | Mean | SD | Min | Max | Description |
|---|---|---|---|---|---|---|
| Expected number of years to continue teaching | 20,612 | 20.54 | 12.82 | 0 | 60 | |
| Mentoring (%) | 20,612 | 0.28 | – | 0 | 1 | |
| Job satisfaction 1 | 20,410 | 3.04 | 0.82 | 1 | 4 | |
| Job satisfaction 2 | 20,410 | 3.39 | 0.72 | 1 | 4 | |
| Job satisfaction 3 | 20,397 | 2.89 | 0.91 | 1 | 4 | |
| Organizational commitment 1 | 20,406 | 3.01 | 0.88 | 1 | 4 | |
| Organizational commitment 2 | 20,410 | 3.23 | 0.70 | 1 | 4 | |
| Organizational commitment 3 | 20,397 | 3.10 | 0.77 | 1 | 4 | |
| Female (%) | 20,612 | 0.67 | – | 0 | 1 | Dummy variable, 0 = male, 1 = female |
| Age group | 20,612 | 1.71 | 0.95 | 1 | 5 | Categorical variable, 1 = under 30, 2 = 30 to 39, 3 = 40 to 49, 4 = 50 to 59, 6 = 60 and above |
| Years of teaching experience | 20,612 | 2.89 | 1.48 | 1 | 5 | Ranging from first-year teachers to those with five years of experience |
| Fulltime (%) | 20,612 | 0.75 | – | 0 | 1 | Dummy variable, 0 = part-time, 1 = full-time |
| Advanced degree | 20,612 | 0.37 | – | 0 | 1 | Dummy variable, 0 = less, 1 = master’s or doctoral |
| School type (%) | 6,227 | 0.82 | – | 0 | 1 | Dummy variable, 0 = private, 1 = public |
| ESES | 6,227 | 2.07 | 0.55 | 1 | 3 | Categorical variable, indicating the proportion of low-socioeconomic status students, 1 = None, 2 = 1%–30%, 3 = 31% to more than 60% |
| School size | 6,227 | 2.89 | 1.39 | 1 | 5 | Categorical variable, indicating the number of enrolled students, 1 = below 250, 2 = 250 to 499 3 = 500 to 749, 4 = 750–999, 5 = 1,000 and above |
| School rural (%) | 6,227 | 0.34 | – | 0 | 1 | Dummy variable, 0 = else, 1 = in a village/small town |
| School city (%) | 6,227 | 0.41 | – | 0 | 1 | Dummy variable, 0 = else, 1 = in a city/large city |
Analytic approach
The initial stage of analysis was the examination of the instrument’s construct validity. This assessment was conducted through a combination of single-level CFA and multilevel confirmatory factor analysis (MCFA). We evaluated intra-class variance to determine the necessity of adopting a multi-level modeling strategy. The intra-class correlations (ICCs) were computed across three levels: teacher-level, school-level, and country-level, as detailed in Online Supplementary Appendix A. The outcomes revealed significant variances in all individual indicators across schools and countries, with ICCs spanning from 0.03 to 0.171. The intra-class variance of the dependent variable surpassed 0.05 at both the school and country levels. It indicated the appropriateness of employing multilevel modeling to account for the nested data structure (Hox, 2013).
In the second stage, we constructed a 1-1-1 multilevel structural equation model (MSEM) based on the MCFA models from the first stage. To address the hierarchical nature of the data and avoid estimation bias, we used the “TWOLEVEL COMPLEX” and “CLUSTER” commands in Mplus. The “TWOLEVEL” component allowed us to model teacher and school parameters at Levels 1 and 2, respectively, while the “COMPLEX” component enabled the use of robust standard errors due to clustering within countries. Although we did not include country-level variables for a three-level model, the “COMPLEX” command adjusted standard errors for nonindependence within countries, accounting for the sampling process (Martin et al., 2021; Muthén and Muthén, 1988–2017).
To address unequal probabilities of selection and nonresponse, and to ensure the sample’s representativeness, teacher-level and school-level weights were incorporated into the MSEM separately. For data preparation, STATA 18.0 was utilized, while all subsequent data analysis procedures were conducted using Mplus 8.10. We employed the maximum likelihood estimation with robust standard errors and adjusted chi-square statistics (MRL) method to account for data non-normality and observation non-independence (Muthén and Muthén, 1988–2017). Furthermore, to facilitate intercept interpretability and potential convergence concerns, grand mean centering was applied to all control variables at both teacher-level and school-level, except for the dummy variables (Enders, 2013).
Results
Descriptive statistics
Table 1 shows the descriptive statistics. The expected number of years to continue teaching was 20.54. Among participants, 28% of novice teachers had an assigned mentor. Specifically, 40%, 35%, 24%, 20%, and 18% in years 1, 2, 3, 4, and 5, respectively, had mentors. Female novice teachers made up 67% of the sample, while males were 33%. Most teachers were under 30 or in their 30s, with an average age group of 1.71. The average teaching experience was 2.89 years (SD = 1.48). Full-time novice teachers comprised 75%, and 37% had an advanced degree.
Eighty-two percent of novice teachers worked in public schools. Regarding school socioeconomic composition, 11.7%, 70.1%, and 18.2% of novice teachers worked in a school where students from low socioeconomic backgrounds accounted for none, 1%–30%, and more than 31%, respectively. Schools averaged mid-sized (250–749 students). About 34% of teachers worked in rural areas (population <15,000), and 41% worked in cities with populations over 100,000.
Measurement model
First, the results of CFA indicated that the measurement model achieved a satisfactory fit, (df = 7) = 556.08, p < 0.001, standardized root mean square residual (SRMR) = 0.027, comparative fit index (CFI) = 0.980, Tucker-Lewis Index (TLI) = 0.958, root mean squared error of approximation (RMSEA) = 0.060 (Hu and Bentler, 1999). Chi-square test of model fit is sensitive to the sample size and is not used to evaluate the model due to the large sample included in this study (Barrett, 2007). The results of MCFA also achieved satisfactory fit, (df = 7) = 48.974, p < 0.001, = 0.024, = 0.000, CFI = 0.985, TLI = 0.968, RMSEA = 0.017.
Structural model
The MSEM model indicated a satisfactory fit, (df = 44) = 219.815, p < 0.001, = 0.040, = 0.020, CFI = 0.956, TLI = 0.932, RMSEA = 0.014. To address the research questions simultaneously, a 1-1-1 MSEM model was estimated, as presented in Table 2. The results revealed a significant direct relationship between participation in the mentoring program and the expected number of years to continue teaching (B = 0.889, p < 0.01, 95% CI [0.395, 1.383], H1). It indicated that teachers who engaged in the mentoring program had an increase of 0.889 years in their intention to stay compared to those who did not participate at the time of the survey. These findings support the first hypothesis, indicating a positive association between participating in the mentoring program and the expected career length.
Multilevel structural equation model results between mentoring and expected number of years to continue teaching
| Independent variable | Mediating variable | Dependent variable | B (standard error) | (standard error) | 95% CI | |
|---|---|---|---|---|---|---|
| Direct effects | Mentoring | Expected number of years to continue teaching | 0.889** (0.300) | 0.034** (0.011) | [0.016, 0.052] | |
| Mentoring | Job satisfaction | 0.049 (0.042) | 0.038 (0.032) | [−0.014, 0.091] | ||
| Mentoring | Organizational commitment | 0.077* (0.033) | 0.065* (0.026) | [0.023, 0.108] | ||
| Indirect effects | Mentoring | Job satisfaction | Expected number of years to continue teaching | 0.308 (0.252) | 0.012 (0.010) | [−0.004, 0.027] |
| Indirect effects | Mentoring | Organizational commitment | Expected number of years to continue teaching | 0.373** (0.120) | 0.014** (0.005) | [0.006, 0.022] |
| Total indirect effects | 0.681* (0.312) | 0.026* (0.012) | [0.006, 0.045] | |||
| Total effects | 1.571*** (0.299) | 0.060*** (0.010) | [0.042, 0.077] | |||
| Independent variable | Mediating variable | Dependent variable | B (standard error) | 95% CI | ||
|---|---|---|---|---|---|---|
| Direct effects | Mentoring | Expected number of years to continue teaching | 0.889** (0.300) | 0.034** (0.011) | [0.016, 0.052] | |
| Mentoring | Job satisfaction | 0.049 (0.042) | 0.038 (0.032) | [−0.014, 0.091] | ||
| Mentoring | Organizational commitment | 0.077* (0.033) | 0.065* (0.026) | [0.023, 0.108] | ||
| Indirect effects | Mentoring | Job satisfaction | Expected number of years to continue teaching | 0.308 (0.252) | 0.012 (0.010) | [−0.004, 0.027] |
| Indirect effects | Mentoring | Organizational commitment | Expected number of years to continue teaching | 0.373** (0.120) | 0.014** (0.005) | [0.006, 0.022] |
| Total indirect effects | 0.681* (0.312) | 0.026* (0.012) | [0.006, 0.045] | |||
| Total effects | 1.571*** (0.299) | 0.060*** (0.010) | [0.042, 0.077] | |||
Note(s): *p < 0.05, **p < 0.01, ***p < 0.001. B: Unstandardized estimate; : Standardized estimate; CI: confidence interval (standardized estimate)
However, the results showed that the relationship between participation in the mentoring program and teachers’ job satisfaction was not statistically significant (B = 0.049, p > 0.05, 95% CI [−0.019, 0.118], H2). On the other hand, the results showed a positive association between participation in the mentoring program and teachers’ organizational commitment (B = 0.077, p < 0.05, 95% CI [0.023, 0.132], H3).
The analysis showed that the pathway from participation in the mentoring program to the expected number of years to continue teaching through job satisfaction was not statistically significant (B = 0.308, p > 0.05, 95% CI [−0.106, 0.722], H4). It indicates job satisfaction did not mediate the relationship between mentoring program participation and teachers’ expected career length. However, the pathway from participating in the mentoring program to the expected number of years to continue teaching via organizational commitment was statistically significant (B = 0.373, p < 0.01, 95% CI [0.177, 0.570], H5). Participation in the mentoring program was associated with a 0.373-year increase in intention to stay in the profession through organizational commitment, supporting hypothesis five.
The total effects of the mentoring program on the expected number of years to continue teaching were statistically significant (B = 1.571, p < 0.001, 95% CI [1.079, 2.062]). Participation in the mentoring program was associated with an increase in years of intention to stay by approximately 1.6 years through job satisfaction and organizational commitment. This suggests that teachers who participated in mentoring were willing to work around 1.6 years–approximately 19 months–longer than those who did not participate in the program. Given the intercept of 19.804 years and the range of average years of intention to stay in the profession across the 41 countries in this study (12.65 years–27.33 years, see Online Supplementary Appendix B), an extension of 1.6 years for novice teachers may not be interpreted as trivial. Additionally, the total indirect effects from the mentoring program to the expected number of years to continue teaching were also statistically significant (B = 0.681, p < 0.05, 95% CI [0.169, 1.194]), indicating that 43% of the total effect of participation in the mentoring program was explained by the indirect effects. In terms of control variables, female teachers, age, and city (compared to town) were negatively associated with the outcome, while years of teaching experience were positively related to the outcome (see Figure 1).
The flow chart displays a legend at the bottom. It indicates the following: The latent variables are represented by ovals; manifest variables are represented by rectangles; control variables are represented by double-bordered rectangles; solid arrows represent significant paths; and dashed arrows represent non-significant paths. The flow chart depicts two concentric rectangles with rounded edges. The inner rectangle is labeled “Within School,” and the outer one is labeled “Between School.” The region outside the outer rectangle is labeled “Between Country.” The rectangle labeled “Within School” displays a rectangle labeled “Mentoring” positioned on the center-left. A solid rightward arrow labeled “H 1, 0.889 double asterisk” points from “Mentoring” to another rectangle labeled “Expected number of years to continue teaching,” positioned on the center right. Two ovals are positioned at the center, arranged in a vertical series. The oval at the top is labeled “Job satisfaction,” and the one at the bottom is labeled “Organizational commitment.” A dashed arrow labeled “H 2, 0.049” points upward from “Mentoring” to “Job satisfaction.” A solid arrow labeled “H 3, 0.077 asterisk” points downward and right from “Mentoring” to “Organizational commitment.” Additionally, a concave-down, rightward, dashed arrow and a concave-up, rightward, solid arrow, labeled “H 4” and “H 5,” respectively, point from “Mentoring” to “Expected number of years to continue teaching.” Individual solid arrows labeled “6.250 triple asterisk” and “4.820 triple asterisk” point from “Job satisfaction” and “Organizational commitment” to “Expected number of years to continue teaching.” At the bottom right, below “Expected number of years to continue teaching,” five control variable rectangles are present. From left to right, they are labeled as follows: “Female,” “Age,” “Years of teaching experience,” “Fulltime,” and “Advanced degree.” Individual upward arrows from these rectangles point to “Expected number of years to continue teaching.” A solid arrow labeled “negative 2.522 triple asterisk” points from “Female.” A solid arrow labeled “negative 4.743triple asterisk” points from “Age.” A solid arrow labeled “0.290 asterisk” points from “Years of teaching experience.” Individual dashed arrows point from “Fulltime” and “Advanced degree.” Above “Expected number of years to continue teaching,” there are six control variable rectangles arranged in a horizontal series. These are positioned above the “Within School” rectangle but within the “Between School” rectangle. From left to right, they are labeled as follows: “School type,” “E S E S,” “School size,” “School rural,” and “School city.” Individual dashed downward arrows point from “School type,” “E S E S,” “School size,” and “School rural” to “Expected number of years to continue teaching.” A solid downward arrow from “School city” labeled “negative 1.723 double asterisk” points to “Expected number of years to continue teaching.”MSEM Diagram with Unstandardized Results (RQ1, RQ2). Source. Created by authors
The flow chart displays a legend at the bottom. It indicates the following: The latent variables are represented by ovals; manifest variables are represented by rectangles; control variables are represented by double-bordered rectangles; solid arrows represent significant paths; and dashed arrows represent non-significant paths. The flow chart depicts two concentric rectangles with rounded edges. The inner rectangle is labeled “Within School,” and the outer one is labeled “Between School.” The region outside the outer rectangle is labeled “Between Country.” The rectangle labeled “Within School” displays a rectangle labeled “Mentoring” positioned on the center-left. A solid rightward arrow labeled “H 1, 0.889 double asterisk” points from “Mentoring” to another rectangle labeled “Expected number of years to continue teaching,” positioned on the center right. Two ovals are positioned at the center, arranged in a vertical series. The oval at the top is labeled “Job satisfaction,” and the one at the bottom is labeled “Organizational commitment.” A dashed arrow labeled “H 2, 0.049” points upward from “Mentoring” to “Job satisfaction.” A solid arrow labeled “H 3, 0.077 asterisk” points downward and right from “Mentoring” to “Organizational commitment.” Additionally, a concave-down, rightward, dashed arrow and a concave-up, rightward, solid arrow, labeled “H 4” and “H 5,” respectively, point from “Mentoring” to “Expected number of years to continue teaching.” Individual solid arrows labeled “6.250 triple asterisk” and “4.820 triple asterisk” point from “Job satisfaction” and “Organizational commitment” to “Expected number of years to continue teaching.” At the bottom right, below “Expected number of years to continue teaching,” five control variable rectangles are present. From left to right, they are labeled as follows: “Female,” “Age,” “Years of teaching experience,” “Fulltime,” and “Advanced degree.” Individual upward arrows from these rectangles point to “Expected number of years to continue teaching.” A solid arrow labeled “negative 2.522 triple asterisk” points from “Female.” A solid arrow labeled “negative 4.743triple asterisk” points from “Age.” A solid arrow labeled “0.290 asterisk” points from “Years of teaching experience.” Individual dashed arrows point from “Fulltime” and “Advanced degree.” Above “Expected number of years to continue teaching,” there are six control variable rectangles arranged in a horizontal series. These are positioned above the “Within School” rectangle but within the “Between School” rectangle. From left to right, they are labeled as follows: “School type,” “E S E S,” “School size,” “School rural,” and “School city.” Individual dashed downward arrows point from “School type,” “E S E S,” “School size,” and “School rural” to “Expected number of years to continue teaching.” A solid downward arrow from “School city” labeled “negative 1.723 double asterisk” points to “Expected number of years to continue teaching.”MSEM Diagram with Unstandardized Results (RQ1, RQ2). Source. Created by authors
In addition to the primary analysis, supplementary analyses were conducted to determine whether the results varied by countries/economies and years of teaching experience. Additional analysis among participant groups compared estimates of participants with 5 years of experience as a reference to assess how the effectiveness of mentoring differed based on the timing of participation. The results showed variations in magnitude and direction across countries, years, and timing of participation. These mixed results may be due to the decrease in sample sizes and/or measurement noise related to participation in mentoring. Due to space constraints and limited contextual knowledge, detailed discussions on the results for each specific group/region are not provided. However, the supplementary analysis offers valuable insights for researchers and policymakers in different countries, encouraging further studies to explore the reasons behind these discrepancies. Detailed information can be found in the Online Supplementary Materials (see Online Supplementary Appendices C and D).
Discussion
This study provides global evidence that new teachers who participated in mentoring programs were more likely to report a longer expected career length than their counterparts. Specifically, it finds that participation in mentoring programs is associated with an approximately 0.89 years–11 months–increase in the number of years to continue teaching across countries. The effect sizes are similar to those of previous cross-national research, underscoring the potential importance of mentoring in supporting new teacher retention (Van den Borre et al., 2021). The increase of approximately five percent in the average expected career length should not be underestimated, especially considering that the average retention intention across countries was around 20 years. The retention of novice teachers is a critical issue globally, prompting governments, states, districts, and school leaders to implement various strategies to address this concern. These results offer policymakers and school leaders comprehensive information to effectively design and develop strategies to address the global issue of teacher attrition among novice teachers (Räsänen et al., 2020). We suggest that mentoring programs may serve as a useful approach to support efforts to alleviate the global teacher shortage and address associated challenges such as financial costs, staff morale, and school management stability.
Furthermore, this study uncovers the mediating role of organizational commitment in the relationship between mentoring and new teachers’ expected career length. While no statistically significant indirect effect via job satisfaction was detected, the findings suggest that participation in mentoring programs is associated with higher organizational commitment, which in turn is linked to a longer expected career length for new teachers. When considering the total effect, participating in a mentoring program is associated with an increase of 1.6 years, or approximately 19 months, in retention intention. This study builds upon the findings of Van den Borre et al. (2021) by uncovering the indirect pathways through which mentoring relates to retention intention—insights that would remain obscured without examining the underlying mechanisms. By integrating organizational commitment into our analysis and framework, we can more comprehensively evaluate the effectiveness of mentoring programs. This indirect pathway highlights that mentoring may support a stronger commitment to the current workplace and is associated with teachers’ expected career length. Given the multifaceted nature of teachers’ retention intentions and the diverse factors influencing them, evaluations of mentoring programs should adopt a comprehensive framework that goes beyond examining only direct effects. School leaders should ensure that mentoring programs foster a sense of attachment among novice teachers to their schools, thereby enhancing their effectiveness in promoting retention. This approach is particularly valuable for novice teachers, as their initial commitment is still developing and can be significantly influenced by organizational support. During this critical transitional period, mentors and school leaders can help new teachers adjust to their working environment, fostering growth, integration, and retention (Hong and Matsko, 2019).
The increase in retention and organizational commitment among novice educators may contribute to school improvement and effectiveness. This includes more effective teaching and learning, stable school management, and heightened morale among teaching staff (Bryk et al., 2015; Ingersoll, 2001). Specifically, enhancing teacher retention can be linked to promoting greater stability in school management, improving student learning outcomes, and benefiting teachers and administrators in various ways, such as workload distribution, school climate, and professional development (Ingersoll and Strong, 2011; Spooner-Lane, 2017). The potential positive outcomes associated with a mentoring program underscore its importance for consideration by school leaders and administrators, as it represents a school-based approach that leverages internal school resources. Based on our findings, we recommend that principals ensure that novice teachers have access to mentoring support, as such support may contribute to teacher retention and overall school improvement and effectiveness.
Limitations and future directions
This study provides valuable insights but also has certain limitations. The results are correlational rather than causal, as they are based on cross-sectional data. Consequently, it is not possible to assert any claims on the causal effect of mentoring programs on novice teachers’ retention intention. Nevertheless, we have minimized bias and enhanced rigor by employing comprehensive data analysis techniques, along with model fit tests. Future research could benefit from utilizing longitudinal or experimental designs to investigate the causal effects of mentoring programs on new teachers’ retention intentions.
Second, the indicators used in this study have limitations, requiring careful interpretation of the results. For example, a dichotomous indicator of mentoring programs could not capture other aspects, such as frequency or quality. Additionally, using retention intention as an indicator may not perfectly align with new teachers’ actual decisions to stay or leave the teaching profession in the future. The literature indicates ongoing debate regarding the reliability of this indicator for teacher retention, with mixed findings. While there are known limitations in using retention intention as a proxy measure, it is currently the best available measure to examine new teachers’ retention across different countries and to investigate the pathways between mentoring programs and retention intention, particularly concerning the emotional and psychological aspects of new teachers. Despite these limitations, utilizing cross-national data can provide international evidence to understand the determinants of new teachers’ retention intention in a more holistic way (Van den Borre et al., 2021).
Third, the findings are not able to offer a detailed explanation for why a specific pathway, such as the one between mentoring and job satisfaction, may not have shown a significant effect while others did. It is possible that mentoring, typically provided by more experienced teachers within the same school, tends to focus more on fostering positive school commitment rather than directly influencing job satisfaction. However, identifying the specific reasons for this discrepancy is beyond the scope of our study. Qualitative research that delves into more nuanced explanations of the pathways could shed light on areas for improvement in mentoring programs to enhance job satisfaction and increase new teachers’ retention. Furthermore, an important avenue for future research might be to explore the role of school-level contextual factors, such as leadership and school climate (Choi, 2023; Nguyen et al., 2024), in the relationship between mentoring and teacher retention. This suggestion is grounded in findings that school-based management including school leadership and a positive school climate are strong predictors of teachers’ organizational commitment (Devos et al., 2014; Dou et al., 2017). Investigating how these school characteristics interact with individual mentoring experiences would provide a more nuanced understanding of the conditions under which mentoring most effectively promotes novice teacher retention.
Conclusions and implications
This study provides an in-depth analysis of how mentoring programs relate to the expected career length of new teachers. The results indicate that the relationship between participation in a mentoring program and expected career length cannot be fully understood by examining only the direct effects. The potential effectiveness of mentoring on anticipated career longevity may be mediated by teachers’ identification with and engagement in their schools. Therefore, mentoring programs that cultivate a sense of commitment to the school environment may be more successful in enhancing new teachers’ intentions to remain in the profession.
Fostering commitment to schools plays a critical role in supporting new teachers’ retention intentions. Their belief in the values and objectives of their schools during the early stages of their careers significantly impacts their perception and appreciation of the teaching profession. This study recommends that policymakers and school leaders develop and implement mentoring programs that focus on enhancing organizational commitment, as such efforts may be associated with longer anticipated career durations among new teachers. By taking these steps, school leaders can better support novice teachers, improving their likelihood of remaining in the profession and mitigating the negative consequences of attrition for students and the school community (Carver-Thomas and Darling-Hammond, 2019).
Note
Among 48 countries/economies, 7 (Iceland, Israel, Italy, Netherlands, Romania, Singapore, Spain) were excluded from the final sample due to missing data on teacher and school background variables (e.g. school type).
The supplementary material for this article can be found online.

