Abstract
Despite the prevalence of coaching in workforce development and the extant theory underpinning the effective execution of workplace coaching, the evaluation of the impact of workplace coaching on purported business outcomes is lacking. We draw from two prevailing theories guiding the coaching landscape: Self-Determination Theory (SDT) and Intentional Change Theory (ICT) to examine the impact of workplace coaching on individual (employee engagement) and organizational outcomes (performance, turnover, and promotion rates). We investigated the impact of workplace coaching participation over time among 823 coaching participants, as well as in comparison to a matched comparison sample composed of 824 employees in a large healthcare organization. Results suggest participation in coaching positively impacts employee retention and promotion rates. Workplace coaching was also significantly linked to employee engagement. Findings further suggest employee engagement partially mediates the relationship between workplace coaching participation and employee turnover. An agenda for future research and implications for practice are discussed.
Workplace coaching (hereafter referred to as coaching) has emerged as a common intervention in organizations (Boyatzis et al., 2024; Boyatzis et al., 2022; Theeboom et al., 2014). Jones et al. (2016) define the coaching relationship as a one-to-one intervention that uses reflection and goals to achieve work-based outcomes valued by the client (Bono et al., 2009; Smither, 2011). Supporters of coaching believe it is fundamental to developing leaders, playing a pivotal role in honing leadership capabilities (Boyatzis et al., 2024; Bozer & Delegach, 2019). To that end, coaching is widely adopted and financially supported by organizations, with substantial financial resources allocated to its implementation (Zhou, 2023). In fact, the estimated global total revenue from coaching in 2019 was $2.849 billion U.S. dollars (International Coaching Federation, 2020). However, the capacity for coaching to have a meaningful impact on key organizational outcomes—specifically performance, engagement, promotion, and turnover—remains less clear. By understanding the potential impact of coaching on these four outcomes, organizations can make informed decisions on resource allocation and program design, ultimately enhancing the overall effectiveness of their coaching initiatives.
A recent call for coaching research emphasizes the need to move beyond perceptions of effectiveness and investigate outcomes in organizations, with pre–post and comparison group designs (Boyatzis et al., 2022). Previous coaching outcomes research has been hindered by limitations, in part due to the scarcity of reliable data and small sample sizes (Bozer et al., 2013; Evers et al., 2006; MacKie, 2014; Moen & Skaalvik, 2009). Critical questions have also been raised about the use of self-report evaluations and the limited design/methodology employed in coaching research (Blackman et al., 2016; Pandolfi, 2020). This study aims to address that call for research in three ways: (a) we leverage the theories of Self-Determination Theory (SDT) and Intentional Change Theory (ICT) to support our arguments hypothesizing the impact of coaching on organizational outcomes; (b) the research design incorporates methodology involving both coaching participants and a matched comparison group within a large healthcare organization; (c) we investigate the impact of coaching on multiple organizational outcomes (i.e., engagement, performance, turnover, and promotion).
The Past Landscape
Coaching is a developmental intervention in which individuals, through intentional action, seek to bring about lasting positive change. Two prevailing theories have been instrumental in explaining the use of coaching as a developmental intervention: ICT and SDT, and we call upon both to support our argument for the impact of coaching participation on focal study outcomes. ICT, as proposed by Boyatzis (2006), provides a structured approach to intentional and sustained personal development, growth, and learning. A heightened focus on the self is germane to key tenets of ICT characterized as discoveries that take place as individuals intentionally undergo change (Boyatzis, 2006). Specifically, individuals must identify their ideal self, an awareness of their current self in relation to this ideal self (strengths and weaknesses), develop a plan based on these strengths and weaknesses to achieve the articulated ideal, space to engage in experimentation and practice of learned behaviors, and relationships that shape the change journey through connection, feedback, and support (Boyatzis, 2006). In coaching, practitioners utilize ICT to help clients identify their aspirations, craft a vision for their future selves, and implement deliberate strategies for achieving lasting positive change, thereby aligning personal and organizational goals (Boyatzis, 2019; Boyatzis et al., 2024).
SDT posits that individuals have three basic psychological needs: autonomy, competence, and relatedness; satisfying these needs fosters intrinsic motivation and sustained behavioral change (Deci & Ryan, 1985; Deci & Ryan, 2012). Within SDT, autonomy encompasses the core social need for control over one's actions. Competence considers one's psychological need to feel that they are capable of achieving desired aims and can effectively operate within their environment, and incorporates desires for learning to promote this belief in one's abilities. Relatedness captures the psychological need one has to feel a sense of belonging and connectedness to others.
Extant research has considered the satisfaction of basic needs within the coaching paradigm (Diller et al., 2021; Gabriel et al., 2014; Mühlberger et al., 2025). From a theoretical perspective, researchers assert that client engagement with a coach can support fulfillment of these basic psychological needs (Ryan & Deci, 2019). Distilling aims to achieve positive future states is a cornerstone of coaching (Ely et al., 2010) in concert with one's need for self-governance under the framing of SDT. Coaching further entails the identification of client strengths and encourages a growth mindset inherent to learning to accomplish established goals (Boyatzis, 2019), aligning with the psychological need for competence. The coaching relationship between coach and client is predicated on shared vision, understanding, and support (Boyatzis, 2019); cultivating connection and a sense of belongingness satisfies a client's basic psychological need for relatedness. Consequently, practitioners leverage SDT to design coaching interventions that empower individuals, enhance their sense of control, competence, and connection, promoting lasting positive transformations (Gabriel et al., 2014; Spence & Oades, 2011). Taken together, we find significant support for examining self-centered psychological concepts in professional coaching.
Hypotheses Development
The Impact on Engagement
Concordant with SDT, intrinsic motivation is cultivated through the satisfaction of psychological needs (Gagné & Deci, 2005; Ryan & Deci, 2000), which is essential to work engagement. Indeed, empirical research has found that the satisfaction of these needs directly relates to work engagement (Deci et al., 2001). Several scholars have hypothesized how coaching specifically has the potential to increase client engagement (Boyatzis et al., 2013; Crabb, 2011; Lyons & Bandura, 2022). Through the lens of SDT, coaching can increase positive job-related sentiments such as feelings of empowerment, competence, connection, and control (de Haan & Nilsson, 2023; Jones et al., 2016; Spence & Oades, 2011; Theeboom et al., 2014). Although these meta-analyses consistently find positive effects of coaching on improving work-related attitudes such as engagement, they have had to rely on a few studies with small sample sizes (e.g., de Haan & Nilsson, 2023; Jones et al., 2016; Theeboom et al., 2014). We expect to find a similar positive impact of coaching on employee engagement over time, as well as in comparison with employees who did not participate in coaching. Hypothesis 1: Coaching participation will positively predict employee engagement.
The Role in Performance
Because coaching should enhance clients’ feelings of competence, per SDT, and help clients align their personal and organizational goals, per ICT, it could also lead to increases in job performance. The specific connection of coaching to job performance has been studied mostly indirectly or on a small scale. In one study that purported to investigate job performance, 73 participants who had completed a 4-hr performance training, coupled with a 30-min coaching phone call, were compared to 42 control individuals who received neither the training nor coaching (Bright & Crockett, 2012). In the self-reported outcome measuring effectiveness in handling different aspects of work stress, participants rated themselves higher in half the areas assessed than the control group (Bright & Crockett, 2012). In another study, 31 managers completed a managerial training program that included 8 weeks of coaching (Olivero et al., 1997). These managers increased the timely completion of patient evaluation forms from 17.6% pre training to 75% post training (Olivero et al., 1997). Only one study examined the impact on manager-rated performance; Bozer and Sarros (2012) found increased pre–post supervisor-rated job performance ratings amongst executive leaders receiving approximately seven coaching sessions over 4 months, though the participant sample with supervisor ratings was limited in size (n = 25). We build upon this research to consider the potential impact of coaching on managerially rated employee performance. Hypothesis 2: Coaching participation will positively predict manager-rated performance.
The Interplay Between Coaching, Engagement, and Performance
There is abundant literature linking employee engagement to performance, measured in numerous ways, including in-role task performance, customer satisfaction, productivity, customer loyalty, and organizational citizenship behaviors (Carter et al., 2018; Dajani, 2015; Harter et al., 2002; Saks, 2006; Salanova et al., 2005). Put simply, more engaged individuals are more likely to expend greater effort toward performance-related work tasks. Despite the well-established link between engagement and performance, and research beginning to establish the linkage between coaching and work-related attitudes, there are no published explorations of engagement as a mediator through which coaching could influence performance.
Promisingly, it has been found that training conducted under SDT as a guiding framework has evidenced positive influence on motivation and increases in employee engagement (Ryan & Deci, 2017). Through an SDT lens, performance is one outcome viewed as a byproduct of increased employee engagement, achieved through coaching by increasing an individual's empowerment, sense of control, competence, and connection. We consider the potential for engagement to serve as a mediating mechanism through which an intervention rooted in SDT (i.e., coaching) positively predicts employee performance. Hypothesis 3: Coaching participation will positively predict manager-rated performance directly, and partially through increasing employee engagement.
The Impact on Turnover
Many theories of employee turnover exist (Hom et al., 2017; Mitchell et al., 2001), as well as investigations into the antecedents of turnover (Heavey et al., 2013; Rubenstein et al., 2018). A recent meta-analysis found support for a reciprocal perspective wherein employees are less likely to turnover as a function of increased organizational attachment fostered through greater perceived organizational support related to employer-sponsored development programs (Ng et al., 2024). If, as theorized by ICT, coached individuals are more committed, motivated, and connected to their work and colleagues, then coaching participation may also be directly linked to lower turnover. Lyons and Bandura (2022) argue how managers’ use of coaching skills with employees could reduce employee turnover; specifically, that coaching interactions with an employee may positively impact factors associated with staying (e.g., commitment, perceived support) for that employee. We have found no primary studies connecting participation in coaching and actual turnover behavior, but this relationship could exist, given the potential for similar underlying mechanisms present in manager-as-coach research. Hypothesis 4: Coaching participation will predict lower turnover rates.
The Relationships Between Coaching, Engagement, and Turnover
In line with the increased commitment, motivation, and connection to work and colleagues espoused by the ICT framework of coaching, we consider the indirect linkage between coaching and turnover through employee engagement. Not only might coaching lead to increased employee engagement, but it could also, in turn, predict lower turnover. Employee engagement has been found to partially mediate the relationship between perceived support for participation in Human Resources Development activities (e.g., training) and intentions to turnover (Kumar et al., 2018; Shuck et al., 2014). Evidence linking employee engagement to lower turnover intentions has spanned decades (Fulmore et al., 2023; Harter et al., 2002; Memon et al., 2014; Saks, 2006; Wen et al., 2022). Although no existing research has explored this mediating relationship for coaching specifically, we expect to find similar linkages for coaching. Hypothesis 5: Coaching participation will predict lower turnover rates directly, as well as partially through increasing employee engagement.
The Role in Promotion
Promotion, an objective measure of career success (Judge et al., 1995), is another individual outcome that, along with its antecedents, has been long studied both theoretically and empirically (Ng et al., 2005; Seibert et al., 1999). A meta-analysis investigating the antecedents of promotion revealed training and skill development opportunities had the largest effect size (corrected r = 0.23; Ng et al., 2005). Receiving coaching could provide crucial support needed for coaching clients to progress in their careers per ICT because it provides an opportunity for clients to align personal with organizational goals and implement and sustain positive changes. In a large survey of U.S. veterans, self-reported participation in the category of employment-related activities “career planning exploration, mentor/coach” predicted self-disclosed promotion (Morgan et al., 2022). However, we could not find any primary studies connecting participation in coaching to promotion rates directly. Hypothesis 6: Coaching participation will predict higher promotion rates.
Methods
Our sample consisted of healthcare professionals within a large academic hospital within the South Central United States. The Office of Human Subjects Protection Institutional Review Board determined that the proposed activity is not research involving human subjects as defined by the U.S. Department of Health and Human Services (DHHS) and the U.S. Food and Drug Administration (FDA) regulations, falling under the protocol 2021–1071, as the present research does not entail direct contact with patients and consent from employees was obtained through agreement to participate in the program. We collected coaching participation data across 4 years, beginning the year that coaching was formally aligned with the organization's centralized learning and leadership development department (2018–2022), resulting in a sample of 823 coaching participants. A matched comparison sample was composed of 824 employees who were eligible for but had not yet participated in coaching. The comparison sample was created based on eligibility requirements to engage in coaching, and employees were randomly sampled within these parameters. See Table 1 for descriptors of all sample demographic and employee characteristics.
Demographic and Employee Characteristics for Coached and Comparison Sample.
Note. N sizes and percentages of total sample (Coached, Comparison, or Total) are reported for categorical variables (Gender, Race/Ethnicity, and Manager Level), and means and standard deviations are reported for continuous variables (age and organizational tenure). M = mean, SD = standard deviation, N = sample size.
Coaching Intervention
Participation in coaching was voluntary. Coaching exists in a variety of standalone offerings and embedded in leadership development programs (see Table 2 for descriptions of our coaching engagement types). Regardless of the method in which it was delivered, as coaching was structured in different ways to match various institutional and client needs, it consistently aligned with the overarching definition of coaching articulated by Jones et al. (2016). Specifically, they distill four features of coaching, consisting of: (1) formation and maintenance of a helping relationship between the coach and client; (2) a formally defined coaching agreement or contract, setting personal development objectives; (3) the fulfilment of this agreement (i.e., achievement of the objectives) through a development process focusing on interpersonal and intrapersonal issues; (4) striving for growth of the client by providing the tools, skills, and opportunities they need to develop themselves and become more effective. (Jones et al., 2016, p. 250)
Descriptions of Coaching Engagement Types.
Triad meetings are check-in meetings between the coach, client, and client's 1-up supervisor.
Coaches were internal to the institution and either held an external certification or were internally certified through a formal internal leader-coach education program delivered in partnership with a university-based coaching institution designated as a Level One accredited coach education provider with the International Coaching Federation (ICF). The coach training program focuses on coaching competencies, such as maintaining presence, listening actively, evoking awareness, and facilitating client growth (International Coaching Federation, 2019). During the time period in which the study took place, there were 114 active internal coaches. During the course of the study, as of fiscal year 2022 (September 2021–August 2022), coaching was also made available to all employees. Internal coaches commit to coaching at least three clients per year, making the scalability, bandwidth, and accessibility for offering coaching to all employees feasible. Coupling this with explicit, frequent support from leadership to engage in coaching as a valuable form of leadership development, we believe has directly attributed to the large voluntary engagement in coaching we were able to collect over 4 fiscal years (September 2018–August 2022).
Prior to COVID-19, all leadership development activities, including coaching, took place in-person. At the onset of COVID-19, all development efforts transitioned to virtual (i.e., Zoom). Given the timing of data collection (September 2018–August 2022), the vast majority of coaching engagements were virtual.
Comparison Sample
We employed a hybrid approach to identify a matched comparison sample (see Breaugh & Arnold, 2007), combining group matching with the implementation of statistical control. Little et al. (2000) delineate three factors to guide the decision-making for controlling for nuisance variables: sample size, number of control variables, and number of groups being compared. These factors predominantly concern sample attrition. Matching can significantly decrease sample size depending on the matching strategy (one-to-one vs. group), so if the treatment condition is already regarded as small in sample size, statistical control is recommended (Little et al., 2000). Due to our longer duration of data collection, we have retained a large sample, supporting our identification of a matched comparison. However, we acknowledge the high degree of covariates in our study and align with Little et al.'s (2000) recommendation for the use of statistical control in instances of several control variables to inhibit attrition. Little et al. (2000) recommend the use of statistical control when more than two groups are compared to reduce attrition. As our study examines two groups (coached vs. comparison), we find additional support for our usage of a matched comparison group. Due to the inherent benefits of both strategies and acknowledging the potential overlapping nature of each factor, researchers have suggested that a hybrid approach be considered (Breagh & Arnold, 2007; Little et al., 2000). Keeping with these propositions, we utilized a hybrid approach to guard against sample attrition by identifying two parameters and nested randomization within those parameters to guide sample selection, and employed statistical control for the remaining three covariates (gender, race/ethnicity, and tenure).
The comparison sample was matched based on eligibility requirements to engage in coaching, and employees were randomly sampled within these parameters. Subject Matter Experts (SMEs) identified employee level as a parameter, as employee level serves as a core guiding eligibility criterion for engaging in leadership development within the institution. Employee levels, as they relate to leadership development (inclusive of coaching engagement), are subdivided into three categories: individual contributor; manager (supervisors, managers, faculty, and section chiefs); and executive (chairs, department administrators, directors, executive directors, and clinical medical directors). Coaching accessibility changed over the course of the study as a function of employee level. For the first 2 fiscal years (September 2018–August 2020), coaching was only available to manager- and executive-level employees. Starting in fiscal year 2021 (September 2020–August 2021), coaching became available for certain individual contributor-level employees, dependent on performance parameters (performance minimum rating of 2.5), through participation in a cohort leadership development training program in which coaching was embedded. As of fiscal year 2022 (September 2021–August 2022), coaching was accessible to employees at all levels with no performance requirements. Employee level also reflects employee scope of responsibility, and given the usage of coaching predominantly being delivered to employees based on level (i.e., executive level employees), SMEs found ensuring the comparison sample was matched based on scope of responsibility, as reflected in employee level, to be an important matching parameter. The second parameter employed to identify the comparison sample was the number of coaching participants per fiscal year. Combining these parameters, the comparison group was randomly sampled based on the number of employees per employee level by fiscal year. In sum, the matching criteria to identify the original pool were based on whether individuals were eligible to participate in coaching during the given fiscal year. Beyond matching criteria based on access to coaching, we controlled for other demographic and employee characteristics, as covariates, during the analysis phase to further account for potential confounds. See Table 3 for sample sizes by fiscal year for coaching participants and the comparison group.
Sample Sizes Per Condition (Coached and Comparison Groups).
Data Structure
Data was collected across 4 fiscal years (2019–2022). We undertook a two-pronged strategy to enhance our ability to further isolate the impact of coaching, acknowledging that pure intervention isolation is not fully achievable in the context of field studies. First, data were restructured from their fiscal year timepoints; that is, data were structured such that participation and outcomes were anchored around the timepoint at which coaching began. This reoriented our data from being structured around linear year time-point, which limited our capacity to examine the impact of coaching over time and allowed us to collapse across years to examine all coaching participants as one sample. For example, coaching participants in fiscal year 2020 (FY20) were treated so their FY19 performance was considered “Pre,” their performance in FY21 as “Post” and their performance in FY20 as “During.”
Measures
See Table 4 for correlations between all study variables.
Descriptives and Correlations Between Study Variables.
Note. *p ≤ .05, ** p ≤ .01. M = mean, SD = standard deviation, N = sample size. N sizes are calculated based on individuals with available data. The numbers do not add up to the total numbers in the participant and comparison samples due to missing data.
Engagement
Engagement was captured as part of the institution's biennial employee engagement survey (2019, 2021, and 2023). The measure consisted of two items rated on a 1 (strongly disagree) to 5 (strongly agree) scale: “My job provides me with a sense of accomplishment” and “I would recommend MD Anderson to others as a great place to work.” Although the scale is unpublished, these engagement items have been used year-over-year for benchmarking, and internal consistency for this measure pre (α = .78), during (α = .79), and post (α = .82) coaching was acceptable.
Performance
Performance, as evaluated by participants’ direct supervisors, was rated on a scale of 1.00–3.00 as part of employee's annual performance review. This is a weighted score comprised of the institution's leadership competencies, job-related competencies, and goal achievement.
Turnover
Turnover was a dichotomous variable with 0 indicating Stayed or 1 indicating Turned Over. The organization in which the study took place has historically low turnover rates, even in comparison to other organizations within the same industry, resulting in lower base rates of turnover. To account for lower turnover base rates, we collected turnover instances across 3 fiscal years including the time at which coaching began (or the comparison sample was matched); this captured if an employee left the institution at the timepoints through 3 years post engagement.
Promotion
Promotion was also a dichotomous variable with 0 indicated Not Promoted or 1 indicating Promoted. Given the longer period over which coaching takes place, we collected and collapsed promotion rates over 3 fiscal years including the time at which coaching began (or the comparison sample was matched); this captured if an employee received a promotion at the timepoints through 3 years post engagement.
Covariates
We conducted chi-square analyses to examine differences between the coaching group and the matched comparison sample to identify and control for influences of demographics (race/ethnicity and gender) and conducted an independent t-test for employee characteristics (organizational tenure). Due to uneven sample size sub-groupings, race/ethnicity was investigated as both dichotomized and all categories intact. Organizational tenure was treated as a continuous variable. Chi-square analyses detected a significant difference between coached and comparison samples for gender (χ2[1, N = 1,647] = 8.21, p = .004), but not for race/ethnicity when dichotomized (χ2[1, N = 1,647] = 0.51, p = .48) or when analyzing all categories (χ2[6, N = 1,647] = 7.32, p = .29). Finally, we conducted an independent samples t-test on tenure, which found a significant difference between coached (n = 823, M = 15.48) and comparison (n = 824, M = 17.00) groups; however Levene's test for equality of variance detected a violation for the assumption of homogeneity of variance, F(1, 1645) = 5.44, p = .02. Hence we report the Welch's corrected t-test statistic, Welch's t(1623.88) = –3.67, p < .001. However, given the various outcomes examined as well as the theoretical and organizational context, we elected to conservatively control for all three covariates across analyses (race/ethnicity, gender, and organizational tenure).
Analyses
Due to the nature of outcome variables of interest, separate analyses were warranted to test the overall conceptual model depicted in Figure 1. Specifically, engagement and performance were continuous variables and turnover and promotion were categorical, dichotomous variables. Mixed ANOVAs were conducted in IBM SPSS version 24 to test Hypotheses: H1 and H2, for impacts over time (pre–post coaching) as well as between groups (coached vs. comparison). To test Hypotheses: H3, and H5, SPSS macros developed by Hayes (2017) were used to conduct partial mediation models using regression, specifically, we employed model 4 to examine the data (see Figure 1). This macro detects indirect effects while testing conceptual models in a single step. Indirect effects are evident when the confidence interval excludes zero. Finally, logistic regression conducted in IBM SPSS version 24 was used to test the following Hypotheses: H4 and H6. For H6, to control for coaching sample bias specific to the context of promotion rates, one form of coaching (onboarding coaching) was excluded from analyses, as this form of coaching was inherently linked to promotion reception. All analyses control for race/ethnicity, gender, and tenure.
Conceptual model of hypotheses.
Exploratory Analyses
We also acknowledge the hierarchical organizational structure inherent to the healthcare industry, in the given context of leadership development initiatives, and coaching specifically. Therefore, we attempted to examine the impact of coaching on outcomes of interest by employee level. As identified in our Methods, employee level was categorized into three subgroupings of individual contributor, manager, and executive level and dummy coded with the manager level as the referent group. We conducted a chi-square analysis to examine differences between the coaching group and the matched comparison sample by employee level, which detected significant differences between groups on the distribution of employees by level between coached and comparison (χ2[2, N = 1,647] = 118.20, p < .001). Additionally, there were uneven groups at all levels and an insufficient sample size at the individual and executive levels, which inhibited us from exploring the relationship between coaching and outcomes of interest by level (see Table 1).
Results
See Table 4 for descriptive statistics and intercorrelations for study variables. According to Hypothesis 1, we posited that coaching participation would positively predict employee engagement. Participation in coaching was not found to be directly predictive of engagement over time or in relation to the comparison group. However, a significant interaction between time (pre–post coaching engagement) and condition (coached vs. comparison) was detected, F(1, 945) = 4.25, p = .04, η2 = .004. Post hoc tests of Simple Effects revealed a significant linkage between coaching and engagement scores, with engagement not significantly differing prior to coaching (Coached Mpre = 4.38; Comparison Mpre = 4.41) and significantly increasing post-coaching in relation to the comparison group (Coached Mpost = 4.52; Comparison Mpost = 4.46). Thus, H1 was partially supported, such that although coaching did not significantly predict employee engagement, we found support for the positive relationship between coaching and employee engagement.
For Hypothesis 2, which predicted coaching participation would positively impact manager-rated performance, Mauchly's test of sphericity indicated a violation in the assumption of sphericity χ2(2) = 33.59, p < .001, thus, we employed and report the Greenhouse-Geisser corrected degrees of freedom. With this correction, analyses revealed a significant change in performance over time pre–post coaching F(1.87, 873.28) = 3.68, p = .028, η2 = .008, and a main effect of participation was detected between groups, F(1, 467) = 18.59, p < .001, η2 = .038, such that coached individuals performed significantly higher than comparison, and further a significant interaction between time and condition was detected F(1.87, 873.28) = 4.19, p = .018, η2 = .009. Post hoc tests of Simple Effects were conducted supporting the difference between groups across timepoints. To further explore the interaction, post hoc Repeated Measures ANOVAs were conducted to explore performance over time for the coached group and the comparison group separately. Repeated Measures ANOVAs did not detect changes over time in performance for coaching participants, but did detect significant changes in performance over time for the comparison group, F(1.82, 451.38) = 4.07, p = .018; Mauchly's test of sphericity indicated a violation in the assumption of sphericity χ2(2) = 25.71, p < .001, thus, we employed and report the Greenhouse-Geisser corrected degrees of freedom. Therefore, post hoc tests revealed an effect of coaching on performance such that individuals who did not engage in coaching significantly increased in performance whereas coaching participant performance did not significantly change over time (see Table 5). Thus, H2 was partially supported, such that although performance was not found to increase after the coaching engagement began compared to pre-performance scores, employees who engaged with a coach did have significantly higher performance scores than the comparison sample.
Hypothesis 2 Estimated Marginal Means.
Note. N = 472 total cases (Coached = 220; Comparison Group = 252). M = mean. N sizes are calculated based on individuals with available data. For performance analyses, we examined employees with available data across three timepoints (pre, during, and post). The numbers do not add up to the total numbers in the participant and comparison samples due to missing data.
Hypothesis 3 posited that participation in coaching would positively predict manager-rated performance directly, and partially through increasing employee engagement. Partial mediation of Employee Engagement during coaching on the relationship between coaching participation and performance over time and compared to the comparison group was not detected, failing to support H3.
In support of Hypothesis 4, in which coaching was hypothesized to predict lower turnover, coaching participation was found to predict turnover rates directly such that individuals who engaged in coaching were significantly less likely to turnover in relation to the comparison group (see Table 6). The comparison group was 70% more likely to turnover in relation to coaching participants, with an odds ratio of 2.29, p < .001, and baseline frequencies revealed that employees who participate in coaching had a 12% turnover rate in relation to the comparison group (23%). Hypothesis 5, positing employee engagement as a mediator in the relationship between coaching participation and turnover, was also supported, b = .10, CI [.03, .18], with engagement during coaching partially mediating this effect. Specifically, individuals who engaged in coaching were less likely to turn over as a result of the positive impact of coaching on employee engagement, which, in turn, negatively influenced their likelihood to turn over.
Logistic Regression Analysis of Coaching Participation on Turnover Rates.
Note. N = 1,647 total cases. 282 cases of turnover (Coached = 95; Comparison Group = 187). B = beta weight, SE = standard error, χ2 = chi-square value, df = degrees of freedom, p = p-value, significant at p < .05 level, Exp(B) = odds ratio.
Hypothesis 6 stated that coaching would predict promotion rates. For our final hypothesis, we found that coaching participation increased promotion rates (see Table 7). The comparison group was 40% less likely to receive a promotion in comparison to coaching participants, with an odds ratio of 0.66, p = .001, and baseline frequencies revealed that employees who participate in coaching had a 31% promotion rate compared to the comparison group (22%), supporting H6.
Logistic Regression Analysis of Coaching Participation on Promotion Rates.
Note. N = 1,352 total cases (Coached n = 528; Comparison Group n = 824). 351 cases of promotion (Coached = 166; Comparison Group = 185). B = beta weight, SE = standard error, χ2 = chi-square value, df = degrees of freedom, p = p-value, significant at p < .05 level, Exp(B) = odds ratio. Analyses exclude Onboarding Coaching, a form of coaching linked to promotion reception.
Discussion
The present study leveraged the two theories of SDT and ICT to develop a holistic understanding of coaching and its expected impact on the outcomes of engagement, performance, turnover, and promotion. We investigated coaching across multiple timepoints, orienting our investigation around the initial timepoint at which an individual engaged with a coach, with a large sample representative of populations found in other large healthcare organizations (see Table 1 for sample demographic and employee characteristics). Our identification of a matched comparison sample lends confidence to our findings insofar as we strived to isolate the effects of coaching, although we acknowledge no field study can take place within a vacuum, especially within the intervention-saturated healthcare context.
Summary of Findings
We found that coaching is positively linked to employee engagement (Hypothesis 1, partially supported). Findings further support that coaching reduces turnover directly and indirectly through engagement (Hypothesis 4, Hypothesis 5) and predicted higher promotion rates (Hypothesis 6). Coaching participants performed significantly better than comparison employees overall, although performance over time did not improve significantly among coached employees (Hypothesis 2, partially supported). Mediation of engagement on performance was not detected (Hypothesis 3, not supported). Taken together, results highlight the benefits of coaching on turnover and promotion, as well as the linkage between coaching and engagement, while leaving performance change less clear.
In examining the marginal means and taking into account performance scaling and range restriction, we find the likely interpretation of these findings to be that coaching participants already have extremely high performance (ranging from 2.80 to 2.83) with less room to improve, whereas comparison group employees have more room to increase their performance and still remain significantly lower compared to coaching participants (ranging from 2.68 to 2.75). Importantly, within the present organizational context, these performance differences are practically significant insofar as performance is divided into three categories: Developing (1.00–1.74), Successful (1.75–2.74), and Exceptional (2.75–3.00). So, from a practical standpoint, employees who do not participate in coaching are predominantly performing in a lower performance bracket compared to coaching participants. This aligns with tenets of SDT, which suggest efforts to meet individual needs for autonomy, competence, and relatedness should, in turn, impact performance (Gagné & Deci, 2005).
Implications
Many organizations in the contemporary landscape are realizing the criticality and challenge of retention. Literature has already identified investment in employee development as a significant predictor of retention (Kyndt et al., 2009). We found coaching participation negatively predicted turnover likelihood, such that employees in the comparison group were 70% more likely to turn over compared to employees engaged in coaching after controlling for race/ethnicity, gender, and organizational tenure. This aligns with the application of ICT in the coaching context; sustained positive change leading to alignment between individual and organizational goals, as accomplished through coaching, has been found to decrease the odds employees will leave the organization. The field would benefit from future research exploring the nuance of this relationship to identify potential boundary conditions, as well as mediating mechanisms such as goal alignment and organizational commitment to expound the influence of coaching on turnover rates.
Establishing the empirical linkage between coaching and promotion rates lends support to ICT in the context of coaching (see Boyatzis, 2019). We interpret this finding to be demonstrative of the positive sustained change engendered through engaging with a coach. The ultimate overarching goal of coaching is positive change; however, research investigating outcomes of positive change in the workplace, such as promotion rates, is surprisingly limited. We address this gap, evidencing the positive impact of coaching on internal promotion rates in a larger organizational context.
Results indicate to practitioners and researchers that coaching may be an effective practice to increase engagement and career success and reduce turnover within organizations. Whether these effects are sustainable beyond 1 year post coaching in the context of engagement, and beyond 3 years for turnover and promotion, has yet to be determined. Therefore, practitioners should continue developing and enhancing coaching education and leadership development initiatives, as well as instituting sustainable methods for capturing and evaluating coaching participation and key organizational outcomes over time.
Future Research
Future studies should examine the long-term impact of coaching beyond 1 year for engagement and beyond 3 years for turnover and promotion. Further research should also explore mediating mechanisms, such as goal alignment, organizational commitment, and self-efficacy, that may connect coaching to promotions and turnover outcomes. Research could investigate what drives differences in performance outcomes between coached and non-coached employees, particularly when performance ratings are already high. Finally, research could incorporate diverse organizational contexts and larger longitudinal samples to strengthen external validity.
The observed relationship between coaching and promotion should be interpreted with caution, given the potential for selection bias. Employees who are already motivated toward advancement may be more likely to opt into coaching. Although we sought to mitigate this concern by including a comparison group, the possibility of bias cannot be fully ruled out. Future research using randomized assignment or stronger controls for pre-existing motivation would help clarify the extent to which coaching drives this outcome.
Limitations
Key limitations include restricted performance measurement. Specifically, the 3-point scale and already high baseline scores of coached employees limited the ability to detect significant improvements. Another limitation was in the sample size distribution, such that longitudinal analyses were limited by a substantial decrease in sample size over time. We were additionally limited by uneven sample sizes across employee levels, with most employees who participated in coaching residing at the manager or executive level and most in the comparison group residing at the managerial level. Third, because most coaching was conducted virtually due to the pandemic, differences in delivery format could not be examined. Additionally, both coached and comparison employees may have engaged in other development programs, making it challenging to fully isolate coaching effects. Finally, we acknowledge that our use of SDT and ICT was limited to framing our argument about the impact of coaching on engagement, turnover, promotion, and performance; these outcomes are not explicitly operationalized within either theory, and we did not measure core constructs central to these theories. Therefore, we advise caution in interpreting the present findings within these theoretical paradigms and encourage future coaching research to empirically assess these constructs.
Conclusion
Our investigation extends beyond the executive leader population to show the influence of coaching as a tool accessible at all employee levels. The findings from this study connect coaching to a key individual outcome—employee engagement—as well as improved organizational outcomes, as evidenced in higher promotion rates and lower turnover rates.
Footnotes
Acknowledgments
The authors thank the broader Leadership Institute team, who were instrumental in the design, coordination, management, and facilitation of our coaching efforts throughout the institution. We also thank our broader support network of the Leadership Institute for their guidance and continued contribution to bolster our efforts as an institution.
Ethical Considerations
The Office of Human Subjects Protection Institutional Review Board determined that the proposed activity is not research involving human subjects as defined by DHHS and FDA regulations, falling under the protocol 2021–1071, as the present research does not entail direct contact with patients and consent from employees was obtained through agreement to participate in the program.
Consent to Participate
The requirement for informed consent to participate has been waived by the Institutional Review Board.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The data supporting the findings of this study are not publicly available due to privacy and ethical restrictions. However, deidentified data may be made available by the corresponding author upon reasonable request, provided that such requests comply with applicable institutional and ethical guidelines.