The relationship between working alliance and client outcomes in coaching: A meta-analysis

Abstract

A growing number of studies emphasize the working alliance between the client and the coach to be a key factor in coaching. Synthesizing 27 samples (N = 3563 coaching processes), this meta-analysis sheds light on the relationship between working alliance and a broad range of coaching outcomes for clients. The meta-analytic results indicate a moderate and consistent overall relationship between a high-quality working alliance and coaching outcomes for clients (r =.41, 95% CI [.34,.48], p <.001). Working alliance was positively related to all desirable coaching outcomes (range: r =.32 to.64), with the strongest relationship to affective and cognitive coaching outcomes. Moreover, working alliance was negatively related to unintended negative effects of coaching (r = –.29). Results revealed no differences regarding the type of clients, coaches’ expertise, number of coaching sessions, and clients’ or coaches’ perspectives. Similar to other helping relationships like psychotherapy or mentoring, the results support the importance of a high-quality working alliance in coaching.

Full text

https://doi.org/10.1177/0018726718819725
human relations
1 –24
© The Author(s) 2019
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0018726718819725
journals.sagepub.com/home/hum
human relations
The relationship between working
alliance and client outcomes in
coaching: A meta-analysis
Carolin Graßmann
SRH University Berlin, Germany
Franziska Schölmerich
Leuphana University Lüneburg, Germany
Carsten C Schermuly
SRH University Berlin, Germany
Abstract
A growing number of studies emphasize the working alliance between the client and
the coach to be a key factor in coaching. Synthesizing 27 samples (N = 3563 coaching
processes), this meta-analysis sheds light on the relationship between working alliance
and a broad range of coaching outcomes for clients. The meta-analytic results indicate
a moderate and consistent overall relationship between a high-quality working alliance
and coaching outcomes for clients (r = .41, 95% CI [.34, .48], p < .001). Working alliance
was positively related to all desirable coaching outcomes (range: r = .32 to .64), with the
strongest relationship to affective and cognitive coaching outcomes. Moreover, working
alliance was negatively related to unintended negative effects of coaching (r = –.29).
Results revealed no differences regarding the type of clients, coaches’ expertise, number
of coaching sessions, and clients’ or coaches’ perspectives. Similar to other helping
relationships like psychotherapy or mentoring, the results support the importance of a
high-quality working alliance in coaching.
Keywords
client outcomes, coaching, meta-analysis, relationship quality, working alliance
Corresponding author:
Carolin Graßmann, SRH University Berlin, Ernst-Reuter-Platz 10, Berlin 10587, Germany.
Email: carolin.grassmann@srh.de
819725HUM0010.1177/0018726718819725Human RelationsGraßmann et al.
research-article2019

2 Human Relations 00(0)
I propose that the working alliance between the person who seeks change and the one who
offers to be a change agent is one of the keys, if not the key, to the change process.
Bordin (1979: 252)
The use of coaching methodologies has become popular among clients and their organi-
zations to optimize clients’ learning and performance. Approximately 53,300 coaches
are in business worldwide, representing an increase of 11% in only 4 years (ICF, 2012,
2016). Recent meta-analyses contributed to the literature by showing that coaching is
effective in general, and supported that clients can successfully use coaching for achiev-
ing desirable outcomes, such as higher performance and well-being (Jones et al., 2015;
Sonesh et al., 2015a; Theeboom et al., 2014). Yet this effect was largely heterogeneous,
indicating that some coaching processes were highly successful and others were not.
This heterogeneity raises the question of which aspects of coaching drive the positive
impact on clients. Only internal coaches and the avoidance of multisource feedback have
been supported in these meta-analyses (Jones et al., 2015). However, less is known about
other drivers of coaching success. As Theeboom et al. (2014: 14) conclude, ‘it is the time
to shift attention from the question “does it work?” to “how does it work?”’.
The current meta-analysis moves on to answer this question and examines the work-
ing alliance between client and coaches, which has been recognized as a key factor in
coaching (e.g. Bluckert, 2005; O’Broin and Palmer, 2006). Although theoretical interest
was high early on (Kilburg, 1996; Wasylyshyn, 2003), quantitative investigations of the
working alliance in coaching have been started only in recent years (for the first study on
this topic, see Baron and Morin, 2009). Relationships between working alliance and
outcomes were supported in other helping relationships, like psychotherapy (Horvath
et al., 2011; Martin et al., 2000), mentoring (Eby et al., 2013), and supervision (Ramos-
Sánchez et al., 2002).2 Working alliance also plays a critical role in teaching (Rogers,
2015) and leadership (Graen and Uhl-Bien, 1995; Schermuly and Meyer, 2015), demon-
strating its key role in professional human interactions. In coaching, studies yielded con-
tradictory findings on the strength of the working alliance–outcome relationship, where
some studies showed no or small-sized relationships (e.g. Berry et al., 2011) and others
revealed large effect sizes (e.g. De Haan et al., 20161). Therefore, we test the general
assumption that working alliance is a key factor in coaching (Bluckert, 2005; O’Broin
and Palmer, 2006), and aim to better understand the magnitude of its relationship to
coaching outcomes for clients.
We also shed light on which outcomes have the strongest relationship with working alli-
ance. Differential effects are likely, as coaching outcomes can be evaluated on different
levels, which correlate only moderately with each other (Alliger et al., 1997). We use the
frameworks of Kraiger et al. (1993) and Kirkpatrick (1967) for this purpose. For instance,
working alliance may be most strongly related to affective coaching outcomes but less to
results outcomes, such as goal attainment, which is a more specific coaching outcome.
Moreover, research on unintended negative effects started recently (e.g. Graßmann and
Schermuly, 20161, 20181; Schermuly et al., 2014). Little knowledge is available on how to
prevent them (Graßmann and Schermuly, 20161), and we study whether working alliance
may be helpful in this regard. This advances our understanding of which outcomes are
actually related to the working alliance between clients and coaches.

Graßmann et al. 3
Furthermore, this meta-analysis strives to question the robustness of the relationship
between working alliance and client outcomes. In particular, this study investigates if
there is a difference between coaches’ and clients’ perspectives on the working alliance–
outcome relationship. The ‘true’ working alliance is hard to determine, because it is only
accessible by the clients’ and coaches’ perceptions. When these perceptions solely rely
on the client or the coach, common method variance is a strong cause for concern
(Podsakoff et al., 2012). We therefore explore clients’ and coaches’ perspectives on the
strength of the working alliance–outcome relationship.
The present meta-analysis makes three contributions to the coaching literature. First,
we synthesize studies on the working alliance in coaching and determine the strength of
its relationship to client outcomes. This tests our understanding of the presumed impor-
tance of the working alliance in coaching. Second, we shed light on which coaching
outcomes have the strongest relationship to working alliance. We explicitly test its
importance for coaching results, such as goal attainment, and preventing negative effects
to illuminate the coaching outcomes that working alliance is actually related to. Third,
we investigate the conditions under which this relationship is strongest, to test the robust-
ness of this relationship. Most notably, this study explores potential differences between
coaches’ and clients’ perspectives on the working alliance–outcome relationship. These
findings aim to deepen our understanding of the working alliance in coaching to help
clients and coaches to achieve the best possible outcomes in coaching.
Coaching
Coaching can be defined in many ways (see Passmore and Fillery-Travis, 2011, for an
overview). Following the definition of Grant and Stober (2006), we consider coaching as
a dyadic, egalitarian relationship between a client and a professional coach, which
involves a systematic process that focuses on collaborative goal setting, constructing solu-
tions, and fostering clients’ self-directed learning and personal growth. Coaches facilitate
these processes rather than giving advice (Bluckert, 2005). Coaching centers around cli-
ents’ goal attainment as the main coaching purpose: coaches help clients to define and set
individual goals and assist them in achieving those goals. This often comprises optimizing
clients’ satisfaction, well-being, and job performance (Kilburg, 1996), or, stated differ-
ently, maximizing clients’ personal and professional potential (ICF, 2012).
Coaching can be applied to a multitude of personal and organizational domains
(Theeboom et al., 2014). For instance, Grant (2005) distinguishes between executive, work-
place, and life coaching, whereof executive and workplace coaching take place in workplace
contexts (with executive and non-executive employees) and life coaching is typically about
more personal issues and can be conducted outside of the workplace. Similar to Theeboom
et al. (2014), we did not exclusively focus on coaching in an organizational setting. Besides
executives who use coaching for their professional development, coaching is also popular
for clients without managerial responsibility (ICF, 2016). We therefore also incorporated
studies that consider workplace coaching. Life coaching usually focuses on more holistic
issues (Grant, 2005) and often covers work-related issues. Key issues in life coaching are
typically work–life balance or developing new career opportunities (Grant, 2005). For
instance, this can be the case when coaching is not contracted by the clients’ organizations.
We therefore also included coaching outside of organizational settings, for example, career

4 Human Relations 00(0)
coaching with students as clients (e.g. Graßmann and Schermuly, 20181). Although these
coaching types differ in their field of application, they typically incorporate workplace
issues and adhere to the idea of a dyadic egalitarian relationship with collaborative goal set-
ting, constructing solutions, and fostering clients’ self-directed learning and personal growth.
We therefore include them in this meta-analysis.
According to the understanding of coaching presented above, we do not cover manage-
rial coaching, group coaching, sports coaching, and clinical populations. Managerial
coaching is hierarchical in nature, as it takes place between subordinates and their for-
mally appointed direct supervisors (Gregory and Levy, 2010). We therefore see it as a
component of leadership that differs from coaching with external or internal coaches –
mainly in unequal power, organization-driven goals, and lack of confidentiality (McCarthy
and Milner, 2013). Group coaching does not match the dyad character of our definition.
We do not cover sports coaching, because sports coaches also provide technical advice
(Nicholls, 2017). We also do not cover clinical interventions, as our coaching definition
focuses on well-functioning clients (Grant and Stober, 2006).
Working alliance
The concept of working alliance can be traced back to psychotherapy research, begin-
ning with its earliest idea in Freud’s writings (for a description of the history of the work-
ing alliance in psychotherapy, see Horvath et al., 2011). Later on, working alliance was
extended as a pan-theoretical concept beyond psychoanalytic approaches with an empha-
sis on collaboration and consensus (see Horvath et al., 2011). It was positioned as a com-
mon success factor across all therapeutic approaches (Rosenzweig, 1936). Bordin (1979)
termed the concept as ‘working alliance’, and went on by proposing that the positive
function of working alliance may be generalizable across all helping relationships and is
not limited to psychotherapy alone. Coaching researchers became interested in the role
of the working alliance from the very beginning of this research discipline (e.g. Kilburg,
1996; Wasylyshyn, 2003). The working alliance in coaching differs, though, from other
helping relationships as coaching is more egalitarian (Grant and Cavanagh, 2004).
O’Broin and Palmer (2006, 2007) nevertheless proposed that working alliance is readily
transferable into the coaching context and that it very likely proves to be effective in
promoting coaching outcomes. They also highlighted the paradox of an almost universal
acknowledgment of the importance of working alliance for coaching outcomes, and the
serious lack of empirical studies on this relationship at that time (O’Broin and Palmer,
2006, 2007). This lack of empirical studies diminished continuously after that, which the
current meta-analysis seeks to review empirically.
However, there is some ambiguity regarding the concept of working alliance (Horvath
et al., 2011) and its labeling, where different labels describe the same or a related con-
cept, such as coaching relationship (e.g. De Haan et al., 20161), relationship quality (e.g.
Graßmann and Schermuly, 20161), or working alliance (e.g. Baron et al., 20111). Despite
this ambiguity, some characteristics of a high-quality working alliance emerged in all of
these conceptualizations. In a high-quality working alliance, clients and coaches mutu-
ally agree on the goals they want to achieve in coaching, they agree on the tasks that will
help to reach those goals, and finally create a bond that entails trust, respect, and liking

Graßmann et al. 5
for each other (Baron and Morin, 2009; Bordin, 1979; Horvath and Greenberg, 1989).
This involves a sense of partnership between coach and client, in which they share the
feeling that they care for each other, are committed to their responsibilities, and are
actively engaged in the process (Horvath and Bedi, 2002; Kokotovic and Tracey, 1990).
In synthesizing these characteristics, O’Broin and Palmer (2007: 305) stated that the
working alliance in coaching ‘reflects the quality of the client and coach’s engagement
in collaborative, purposive work within the coaching relationship, and is jointly negoti-
ated, and renegotiated throughout the coaching process over time’.
Working alliance has been assessed with different measures in the past. The most
established measure adapted the Working Alliance Inventory, which has been developed
in psychotherapy research (Horvath and Greenberg, 1989). This measure is based on
Bordin’s (1979) conceptualization of working alliance, and has been transferred to
coaching (e.g. Baron et al., 20111; Berry et al., 2011, De Haan et al., 20161). Although not
originally developed for coaching, this measure has been validated as a useful measure
of working alliance. Another measure developed in German (Jansen et al., 20041) has
been developed specifically for coaching, and measures the working alliance between
clients and coaches holistically, focusing on the sense of a collaborative partnership and
mutual respect. This scale was used in German-speaking countries (e.g. Graßmann and
Schermuly, 20161; Maurer, 20091). Other measures are less common or have been devel-
oped specifically for the purpose of each study. For instance, some researchers devel-
oped their own scales to assess working alliance and ensured their qualities with validity
assessments or pilot studies (e.g. Dingman, 20041; Ghods, 20091). Unfortunately, little is
known about how well these measures relate to each other empirically. We use them
conjointly in our analyses as they are theoretically based on our understanding of work-
ing alliance as the quality of the collaborative partnership between a professional coach
and a client for the purpose of working jointly towards the client’s goals for which coach-
ing has been contracted.
Working alliance and coaching outcomes for clients
The working alliance between the client and the coach may be a key factor for attaining
coaching outcomes, and social exchange theory helps to explain its importance. It assumes
that the interactions between individuals are interdependent and contingent on the actions
of the interaction partner (Blau, 1964; Cropanzano and Mitchell, 2005), which aligns well
with the collaborative character of coaching. The idea of beneficial social exchanges in
helping relationships can be traced back to Bordin (1979), who proposed that working
alliance involves an agreed-upon contract with concrete exchanges between both partners.
The core of the explanation is the access to psychological benefits (Graen and Uhl-Bien,
1995) that pave the way for achieving desirable coaching outcomes. When clients and
coaches are engaged in high-quality working alliances, clients share trust and openness
(Alvey and Barclay, 2007). Clients exchange sensitive information that they would other-
wise prefer to keep private. Moreover, clients disclose their uncertainty, helplessness, or
their current inability to cope with work-related challenges. They may be more open to the
idea of change when feeling safe (Kretzschmar, 2010), more receptive of the coaches’
actions, and willing to adopt different perspectives. Clients are also less likely to drop out

6 Human Relations 00(0)
from coaching (Schermuly, 20181). When clients exchange these resources, coaches can
exchange other psychological benefits in return. Although this does not mean giving
advice on how to achieve the clients’ goals, coaches may be better able to facilitate the
change process. For instance, they can share new perspectives on the situation, shed light
on clients’ resources to handle their challenges, and help them find workable solutions.
Receiving more feedback from their clients, they can be better attuned to their clients’
needs and use more appropriate and effective techniques. In sum, coaches can deliver
opportunities to learn and grow. These exchanges in high-quality working alliances are
likely to foster desirable coaching outcomes for clients.
Coaching outcomes for clients can be manifold, because they differ depending on the
client and the subject that he or she wants to discuss in coaching (Greif, 2016). To com-
pare different levels of outcomes and following the approach of Jones et al. (2015), we
use the outcome evaluation framework by Kraiger et al. (1993) and Kirkpatrick (1967).
One level may not be fully captured by a single criterion, such that different criteria on
the same level could show different results (Alliger and Janak, 1989). Therefore, we
consider more than one outcome category when possible and classify them along our
framework (see Table 1). Kraiger et al. (1993) distinguish between affective, cognitive,
and skill-based learning outcomes, originally introduced for evaluating training out-
comes. Affective outcomes include attitudes and motivation (Kraiger et al., 1993). For
example, research showed that working alliance is related to higher satisfaction with
coaching (Boyce et al., 20101), perceived effectiveness (Ghods, 20091), and self-efficacy
(Baron et al., 20111). Cognitive outcomes capture knowledge acquisition and organiza-
tion (Kraiger et al., 1993). For instance, working alliance is related to clients’ self-reflec-
tion and insight into their own thoughts, feelings, and behavior (Grant, 20141; Sonesh
et al., 2015b1). Skill-based outcomes refer to the development of technical and motor
skills. In this meta-analysis, we were not able to test the relationship between working
alliance and skill-based outcomes because of missing studies. Additionally, and follow-
ing the approach of Jones et al. (2015), we incorporated the results level of Kirkpatrick’s
(1967) evaluation framework. For the purpose of this meta-analysis, we only focus on
results for clients. The results level primarily captures goal attainment, which can be seen
as the key outcome in coaching (Spence, 2007). Regarding the beneficial exchanges in
Table 1. Coaching outcome categories used in this meta-analysis.
Outcome level Description
Affective outcomes Attitudinal and motivational outcomes
Coaching satisfaction Clients’ overall satisfaction with coaching process
Perceived effectiveness Perceived overall effectiveness of coaching for attaining
coaching outcomes
Self-efficacy Clients’ perceived performance capabilities
Cognitive outcomes Knowledge acquisition and organization, such as clients’
self-reflection and insight
Results outcomes Individual client’s accomplishments, such as goal attainment
Unintended negative effects Harmful and unwanted results for clients that are directly
caused by coaching

Graßmann et al. 7
high-quality working alliances as outlined above, we expect working alliance to be posi-
tively related to all of our proposed client outcomes:
Hypothesis 1: Working alliance is positively related to (a) affective outcomes, (b)
cognitive outcomes, and (c) results outcomes.
Besides a broad range of positive coaching outcomes, coaching can also lead to unin-
tended negative effects for clients. The frameworks by Kirkpatrick (1967) and Kraiger
et al. (1993) do not consider them, but working alliance may prevent or reduce unin-
tended negative effects for clients. Therefore, we additionally include them in our meta-
analysis. Unintended negative effects, or sometimes termed negative side effects, are all
harmful or unwanted results for clients that are directly caused by coaching and occur
parallel to or after coaching (Schermuly et al., 2014). Although clients perceive these
unintended effects currently as negative, it is not excluded that unintended negative
effects could exhibit positive consequences in the long term (Schermuly et al., 2014). For
instance, clients reported that they were less satisfied with their job, that they experi-
enced their job as less meaningful, or that in-depth problems were triggered that could
not be dealt with in coaching (Graßmann and Schermuly, 20161). Prior research showed
that half of clients experience at least one unintended negative effect from coaching, usu-
ally several at once (Graßmann and Schermuly, 20161, 20181; Graßmann et al., 2018b1;
Schermuly et al., 2014). Although the literature on the consequences of unintended nega-
tive effects for clients is scarce, first empirical evidence suggests that they are related to
client dropout from coaching, which is associated with high costs for clients and coaches
(Schermuly, 20181). However, little is available to help coaches deal with these issues
(Kilburg, 2002).
Working alliance may be beneficial to prevent unintended negative effects for clients.
Building upon social exchange theory, coaches and clients exchange more psychological
benefits in high-quality working alliances. These exchanges, as outlined above, include
the disclosure of sensitive inner feelings and thoughts that make clients vulnerable in front
of their coach. When clients do not share them with their coaches, coaches may not be
able to obtain sufficient information about the clients’ situations. This lack of information
may impede dealing properly with the clients’ issues and make unintended negative
effects more likely (Graßmann and Schermuly, 20161). In line with this assumption, work-
ing alliance was associated with negative experiences in mentoring (Eby and McManus,
2004) and supervision (Ramos-Sánchez et al., 2002). We therefore assume that:
Hypothesis 2: Working alliance is negatively related to unintended negative effects
for clients.
The differential perspectives of clients and coaches
Clients and coaches may differ in their evaluation of working alliance and coaching out-
comes (O’Broin and Palmer, 2006). Based on social exchange theory, both interaction
partners exchange resources and experience positive and negative effects (Blau, 1964;
Cropanzano and Mitchell, 2005). However, clients and coaches do not necessarily agree

8 Human Relations 00(0)
in their evaluations of these exchanges and experiences. For instance, clients may not
always share the necessary information with their coaches, and coaches can only take into
account what clients have communicated. Research has already demonstrated that coaches
and clients can differ in their evaluation of the working alliance (Baron et al., 20111) and
coaches’ empathy (Will et al., 2016). It may be the case that clients’ evaluations of work-
ing alliance may be a better predictor of outcomes because their evaluation, and not the
evaluation of their coaches, shapes how they act during the coaching process, what they
will disclose, and what they are going to change. In support of this assumption, clients’
evaluations of session quality were more strongly related to their own evaluation of work-
ing alliance than when evaluated by their therapists (Kivlighan et al., 2016, 2017).
Beyond this rationale to explore the differences between clients’ and coaches’ eval-
uations of the working alliance–outcome relationship, it is hard to evaluate the ‘true’
working alliance objectively as it is only accessible by the clients’ and coaches’ per-
ceptions. In dealing with perceptions, there might be a strong concern for common
method variance when using data from a single source (Podsakoff et al., 2012). We
therefore question the stability of the potential key role of the working alliance in
coaching, and explore if the working alliance–outcome relationship differs between
clients’ and coaches’ perspectives:
Research Question 1: Is there a difference in the strength of the working alliance–
outcome relationship from the clients’ and coaches’ perspectives?
Method
Selection of studies
Literature search. For finding relevant studies to include in our meta-analysis, we searched
several databases (EBSCO, including PsycINFO and PsychArticles; Web of Science;
ProQuest Dissertations; Mendeley; ICF Research Portal as a coaching-specific database)
and coaching-specific journals (Coaching: An International Journal of Theory, Research
and Practice, International Journal of Evidence Based Coaching and Mentoring, Interna-
tional Coaching Psychology Review, Coaching: Theorie & Praxis). We used the keyword
coaching in combination with each of the following additional keywords: working alli-
ance, relationship quality, bond, coaching alliance, or their German equivalents. To find
studies that are not retrievable via these channels (e.g. unpublished data), we additionally
searched conference proceedings (of the Academy of Management, Society of Industrial
and Organizational Psychology, European Association of Work and Organizational Psy-
chology, German Psychological Society, as well as the conference of their section of
Industrial and Organizational Psychology), contacted scholars known to be active in the
field of coaching research, and sent out a request for published and unpublished studies
via the Organizational Behavior mailing-list service of the Academy of Management.
Inclusion criteria. We used the following guidelines as inclusion criteria: (a) the working
alliance had to be referred to as working alliance, relationship, relationship quality, bond,
or simply alliance; (b) the coaching intervention matched our coaching definition as

Graßmann et al. 9
described above (a dyadic, egalitarian relationship between a client and a professional
coach, which involves a systematic process that focuses on collaborative goal setting,
constructing solutions, and fostering clients’ self-directed learning and personal growth,
focusing on well-functioning clients); (c) the study included a quantifiable measure of
the relationship between working alliance and a coaching outcome for clients (r statistic
or other statistic that could be converted to r); and (d) the study was presented in English
or German. We excluded all studies that did not fit our inclusion criteria (see Figure 1).
The most common reason for the elimination of a study was that there was no quantifi-
able measure of the relationship between working alliance and coaching outcome.
Our literature search strategy presented above also resulted in so-called grey literature,
for example, unpublished data or dissertations, which we incorporated, for two reasons.
First, empirical research in the field of coaching, especially regarding the role of working
alliance, is rather young. The inclusion of grey literature allowed us to maximize the num-
ber of studies considered in this meta-analysis. Second, their inclusion helps to overcome
some of the problems of publication bias (Hopewell et al., 2006). Only samples that have
been reported in dissertations, books (theses published as books), or not yet published data
were included. Following the recommendations of Hopewell et al. (2006), we used the
same criteria for inclusion for the assessment of grey literature, and calculated the effect
sizes with publication status as a moderator to additionally ensure quality of the data.
Figure 1. Search strategy for the inclusion of studies in the present meta-analysis

10 Human Relations 00(0)
Twenty-three articles met the inclusion criteria, representing 27 separate samples. Of
these samples, 18 were published in peer-reviewed journals. Included samples that did
not pass peer review were unpublished (n = 4), or were published as dissertations (n = 2)
or books (n = 3). All samples included in the final analysis are indicated with an * in our
list of references and notes.
Coaching outcome categories
The studies included in this meta-analysis used a broad range of coaching outcomes. We
categorized the coaching outcomes into affective, cognitive, and results outcomes, as
well as unintended negative effects (see again Table 1). On the affective outcomes level,
we further distinguished between clients’ satisfaction with coaching, perceived effective-
ness, and self-efficacy as different indicators on that level. The first and the second
authors independently categorized the outcomes into the categories. The common kappa
statistic was used to determine interrater agreement, where zero equals chance agree-
ment and one represents perfect agreement between the judges (Cohen, 1960). The
agreement between the coders was high: κ = .82. In the case of discrepancies, the coders
discussed cases of disagreement to reach a consensus for the final coding.
Statistical analyses
Meta-analytic approach. We employed the Hedges and Olkin (1985) approach to meta-
analysis to calculate the effect sizes. Each effect size was weighted by its precision, so
that studies with larger sample sizes contributed more to the estimate of the population
effect size. This approach does not allow for artificial sources of variance, that tend to
result in an inflation of effect size estimates (Hunter and Schmidt, 1990). Furthermore,
we adopted the random effects model as the statistical approach for this meta-analysis.
The random effects model allows the true effect size to differ from study to study (Boren-
stein et al., 2010), which seems to be justified because of the variability in coaching
processes. We used comprehensive meta-analysis software (Borenstein et al., 2014) for
our calculations.
Estimation of effect sizes. The included studies reported correlations or beta coefficients
as effects sizes. We therefore used the product–moment correlation (r) as the effect size
estimate. For those few studies in which multiple time points were used, we considered
the correlation at the first measurement time for being comparable to the other studies.
We used linear composites of correlations if the same focal construct was measured by
multiple indicators. When studies reported more than one outcome, we examined the
effect sizes from that study individually and as a combined effect size for that study. We
combined the outcomes for each study into an overall effect size per study and entered it
into an overall working alliance–coaching outcome analysis. In addition, the overall
working alliance–coaching outcome relationship was disaggregated by type of outcome
and reanalyzed individually. As clients and coaches can differ to a great extent in their
assessment of working alliance (e.g. Graßmann and Schermuly, 20181), we chose not to
average their evaluations, in order to avoid the risk of meaningless aggregations. When

Graßmann et al. 11
studies included clients’ and coaches’ perspectives, we chose the effect sizes from the
clients’ perspective for estimating the overall effect. However, we analyzed the data
regarding the differential perspectives of clients and coaches. For calculating the respec-
tive effect sizes, we used the individual effect sizes per perspective and entered them into
analyses.
Results
Demographic description of included samples
Table 2 summarizes the characteristics of the selected studies. Total sample size was
based on N = 3563 coaching processes. Sample sizes per study ranged from 14 to 1741
coaching processes (median = 51). In the majority of the included samples, clients evalu-
ated working alliance and outcomes (48%), followed by both clients and coaches (30%),
or coaches only (22%). In most cases, a field sample of clients (67%) was used, working
in diverse industries (56%), universities (33%), or in military, manufacturing, or tele-
communication (4% each). Clients were students in 33% of the studies. We defined pro-
fessional coaches as coaches who were not newly trained as coaches, for example, for the
purpose of the study or student coaches. Professional coaches (56%) were used in most
samples. For those samples that used professional coaches and also provided data on
their coaching experience (n = 6), the coaches had on average 10.7 (SD = 1.7) years of
experience. The mean number of coaching sessions reported for 16 samples was 8.6 (SD
= 6.7) sessions. The mean age of coaches was 40.6 (SD = 10.7) years, reported for 14
samples. The mean age of clients was 35.6 (SD = 8.5) years, reported for 15 samples.
Seventy-four percent of the manuscripts were written in English.
Relationship between working alliance and coaching outcomes
The overall aggregated correlation between working alliance and coaching outcomes for
clients was r = .41 (k = 27, 95% CI [.34, .48], p < .001). This suggests that the working
alliance between clients and coaches, in general, has a significant positive relationship
with coaching outcomes for clients. We analyzed heterogeneity between studies with
Cochran Q, I2, and T2 statistics. The Q statistic provides a test of significance if the true
effect size varies from study to study, I2 represents the proportion of the observed vari-
ance that can be attributed to differences in true effect sizes rather than sampling error,
and T2 is the variance of true effect sizes (Borenstein et al., 2009). The heterogeneity in
effect sizes was significant and large in magnitude, Q = 94.97, p < .001, I2 = 72.62, T2 =
.03, leaving room for the analyses of possible moderating variables.
Another main goal of this meta-analysis is to shed light on the differential relation-
ships between working alliance and different levels of coaching outcomes for clients (see
Table 3). In support of Hypotheses 1 and 2, working alliance was significantly associated
with all considered coaching outcomes. The relationship was strongest between working
alliance and affective coaching outcomes (r = .53, 95% CI [.44, .60], p < .001), as well
as cognitive outcomes (r = .49, 95% CI [.19, .71], p < .001). On a deeper level, we were
able to test for different indicators of affective coaching outcomes: working alliance was

12 Human Relations 00(0)
Table 2. Study characteristics and outcome of studies included in the meta-analysis.
Study nOutcome categories Perspective Client
sample
Coach
expertise
Coaching durationaPeer review Language
Baron etal. (2011) 30 Self-efficacy Both Field Novice 5.8 sessions,
75 min,
8 months
Yes English
Beinicke etal. (2017) 27 Satisfaction, perceived
effectiveness, cognitive
outcomes
Both Student Novice – No German
Berry etal. (2011); Berry (2005)
– Distance group
– Face-to-face group 51
51
Results outcomes Coach Field Professional
21 sessions
26 sessions
Yes English
Boyce etal. (2010) 74 Satisfaction, perceived
effectiveness
Both Field – 8 sessions,
10–90 min
Yes English
De Haan etal. (2013) 156 Perceived effectiveness,
self-efficacy
Both Field Professional
(10.3 years)
8.6 sessions,
75–120 min
Yes English
De Haan etal. (2016) 1741 Perceived effectiveness,
self-efficacy
Both Field Professional
(13.3 years)
8 sessions,
modal duration: 4–6
months
(n = 554)
Yes English
Dingman (2004) 92 Self-efficacy Client Field – – No English
Gan and Chong (2015) 172 Perceived effectiveness Client Field Professional – Yes English
Gessnitzer and Kauffeld (2015) 31 Results outcomes Both Student Novice 5 sessions,
81.8 min,
3 months
Yes English
Ghods (2009) 152 Satisfaction, perceived
effectiveness
Client Field Professional 6 sessions,
4–6 months
No English
Grant (2014) 49 Results outcomes,
cognitive outcomes
Client Student Novice 4 sessions,
45–60 min,
10–12 weeks
Yes English
(Continued)

Graßmann et al. 13
Study nOutcome categories Perspective Client
sample
Coach
expertise
Coaching durationaPeer review Language
Graßmann and Schermuly (2016) 111 Results outcomes,
negative effects
Client Field Professional Modal duration: less
than 5 months (43%)
Yes English
Graßmann and Schermuly (2018) 28 Results outcomes,
negative effects
Both Student Novice 3.3 sessions Yes English
Graßmann etal. (2018a) 53 Results outcomes Coach Field Professional
(10.9 years)
8.1 sessions,
7.7 months
No German
Graßmann etal. (2018b) 275 Results outcomes,
negative effects
Coach Field Professional
(9.1 years)
6.5 months Yes English
Ianiro etal. (2013) 33 Results outcomes Client Student Novice 5 sessions,
60–120 min,
3–4 months
Yes English
Ianiro etal. (2015) 30 Results outcomes Client Student Novice 5 sessions,
60–120 min,
3–4 months
Yes English
Jansen etal. (2004) 71 Satisfaction, results
outcomes
Both Field Professional – No German
Maurer (2009)
– Questions group
– Reflection group
14
14
Satisfaction Client Field Professional 90 min No German
Putz and Berost (2017) 40 Self-efficacy, results
outcomes
Client Student Novice 1 No German
Schermuly (2018)
– Dropout group
– Without dropout
66
49
Negative effects Coach Field Professional
(12.0 years,
8.7 years)
7.5 sessions
15.3 sessions
Yes English
Sonesh etal. (2015b)
– Study I
– Study II
44
89
cognitive outcomes,
results outcomes
Client Student
Field
Novice
Professional
– Yes English
Webers (2008) 19 cognitive outcomes,
self-efficacy
Client Student Novice 4–7 sessions,
60–90 min,
2–3 months
No German
aWhen being reported, descriptive characteristics of the included studies are provided for the mean number of sessions, duration per coaching session, and the duration of the coaching process.
Table 2. (Continued)

14 Human Relations 00(0)
significantly related to coaching satisfaction (r = .64, 95% CI [.49, .75], p < .001), per-
ceived effectiveness (r = .58, 95% CI [.50, .65], p < .001), and self-efficacy (r = .32, 95%
CI [.24, .39], p < .001). Furthermore, we found that working alliance was positively
related to coaching outcomes on the results level (r = .32, 95% CI [.25, .39], p < .001)
and negatively related to unintended negative effects for clients (r = –.29, 95% CI [–.42,
–.14], p < .001).
The influence of perspectives
This meta-analysis also aims to explore if the working alliance–outcome relationship dif-
fers depending on whether either clients or coaches evaluate their coaching processes.
Thus, we examined the differences in effect sizes between studies that used either the
clients’ or coaches’ perspective (see Table 4). The perspective can refer to either working
alliance or coaching outcomes. We therefore analyzed the overall relationship between
Table 4. Effect sizes sorted by perspective on working alliance and outcome (aggregated over
outcome).
Perspective on working alliance Perspective on outcome
Client Coach
r95% CI k r 95% CI k
Client .47*** .40, .54 21 – – –
Coach .26** .08, .42 6 .34*** .16, .50 11
r = correlation coefficient, CI = 95% confidence interval, k = number of studies included in the analysis.
***p < .001, **p < .01.
Table 3. Overall effect size of working alliance on outcome and disaggregated by outcome
category.
Outcome k N r CI (95%)
Lower Upper
Affective outcomes 13 3544 .53*** .44 .60
Coaching satisfaction 6 353 .64*** .49 .75
Perceived effectiveness 6 2322 .58*** .50 .65
Self-efficacy 6 2059 .32*** .24 .39
Cognitive outcomes 5 228 .49*** .19 .71
Results outcomes 14 957 .32*** .25 .39
Unintended negative effects 5 529 −.29*** −.42 −.14
Total 27 3563 .41*** .34 .48
k = number of studies included in the analysis, N = total number of coaching processes in k studies,
r = correlation coefficient, CI = 95% confidence interval. Studies could include multiple outcomes.
***p < .001.

Graßmann et al. 15
working alliance and coaching outcomes for four possibilities: the perspective on working
alliance (client or coach) and the perspective on coaching outcomes (client or coach).
There was a significant working alliance–outcome relationship from all considered per-
spectives: when clients evaluated both working alliance and outcomes (r = .47, 95% CI
[.40, .54], p < .001), when coaches evaluated both working alliance and outcomes (r = .34,
95% CI [.16, .50], p < .001), and when coaches evaluated working alliance and their cli-
ents evaluated coaching outcomes (r = .26, 95% CI [.08, .42], p = .004). Their respective
confidence intervals overlapped, which indicates that the working alliance–outcome rela-
tionship did not differ in their strength between the clients’ and coaches’ perspectives. As
the working alliance–outcome relationship can be supported from different perspectives,
common method variance seems to be unlikely as the only cause for this relationship. We
could not compute the effect size for the case when clients evaluated working alliance and
coaches evaluated coaching outcomes because of the lack of studies.
Publication bias and sensitivity analysis
To investigate if publication bias was present in the data, we visually inspected the funnel
plot. The funnel plot appeared to be distributed symmetrically. We also used the trim-and-
fill-method (Duval and Tweedie, 2000) that estimates and adjusts a funnel plot for missing
studies. No studies seemed to be missing, and the effect size remained unchanged. We then
performed subgroup analyses with language and peer review as moderators (see Table 5).
There was no difference between studies that were written in English (r = .42, 95% CI [.34,
.50]) or German (r = .37, 95% CI [.20, .53], p = .61). There was no difference between
studies that passed peer review (r = .42, 95% CI [.33, .49]) and studies that did not pass
peer review (dissertations, books, and unpublished data; r = .40, 95% CI [.24, .54], p = .87).
Based on these calculations, there is no indication for publication bias in the data.
We then performed subgroup analyses for client sample type, coach expertise, and
number of coaching sessions (see again Table 5). There was no difference between studies
that used field samples of clients (r = .42, 95% CI [.33, .50]) or students as clients (r = .39,
95% CI [.28, .49], p = .71). There was also no difference between the studies that used
novice (r = .40, 95% CI [.26, .53]) and professional coaches (r = .40, 95% CI [.30, .49],
p = .57). To examine the number of sessions as a moderator and following the procedures
of Theeboom et al. (2014), we used five or fewer and more than five sessions for compari-
son. Analyses revealed no difference between coaching processes that lasted five sessions
or fewer (r = .34, 95% CI [.15, .51]) and more than five coaching sessions (r = .39, 95%
CI [.27, .50], p = .49). Because of the large sample size of the study by De Haan et al.
(2016)1, we also examined the overall effect without this study. There was no difference
in the overall effect (r = .41, 95% CI [.32, 49]), suggesting a robust relationship between
working alliance and coaching outcomes for clients.
Discussion
The current meta-analysis aimed to investigate the strength of the relationship between
working alliance and coaching outcomes for clients, analyze how strong working alliance
is related to different coaching outcomes, and explore the conditions under which this

16 Human Relations 00(0)
relationship is strongest. The results indicate a moderate and robust working alliance–
outcome relationship, across all investigated coaching outcomes. The relationship was
strongest for affective and cognitive outcomes. We also found a positive relationship
between working alliance and results outcomes. Moreover, a high-quality working alli-
ance was negatively related to unintended negative effects for clients.
Theoretical implications
These findings speak for the importance of the working alliance in coaching. We based
our rationale for this relationship on social exchange theory (Blau, 1964; Cropanzano
and Mitchell, 2005). We assumed that clients and coaches are more likely to exchange
psychological benefits (e.g. openness and trust) in high-quality working alliances that
enhance the possibility to achieve desirable coaching outcomes. The moderate magni-
tude of the relationship that we found in this meta-analysis matches the moderate effect
sizes reported in meta-analyses in other helping relationships, such as in psychotherapy
(Horvath et al., 2011; Martin et al., 2000) or mentoring (Eby et al., 2013). This supports
Bordin’s (1979: 252) early hypotheses that working alliance is ‘one of the keys, if not the
key, to the change process’ and that this may be true for every helping relationship.
Nevertheless, the relationship between working alliance and client outcomes in coaching
seems to be nearly twice as strong as in psychotherapy (Horvath et al., 2011, r = .28;
Table 5. Effect sizes of working alliance on outcome for different moderators (aggregated
over outcome).
Outcome k r CI (95%)
Lower Upper
Client sample
Field sample 18 .42*** .34 .50
Student sample 9 .39*** .24 .52
Coach expertise
Novices 10 .40*** .26 .53
Professionals 15 .40*** .30 .49
Number of sessions
≤ 5 sessions 6 .34*** .15 .51
> 5 sessions 10 .39*** .27 .50
Language
English 20 .42*** .34 .49
German 7 .37*** .20 .53
Peer reviewed
Yes 18 .42*** .33 .49
No 9 .40*** .26 .53
k = number of studies included in the analysis, r = correlation coefficient, CI = 95% confidence interval.
Unpublished studies conducted in Germany were coded as written in German.
***p < .001, **p < .01.

Graßmann et al. 17
Martin et al., 2000, r = .22). This seems paradoxical. Coaching issues are usually less
severe than in psychotherapy, which should reduce the threshold for self-disclosure and
therefore the dependency on a high-quality working alliance. This may not be entirely
attributable to the lower number of sessions in coaching, which may heighten the neces-
sity to open up more quickly, as we did not find differences for the number of coaching
sessions. The stronger relationship in coaching may be owing to differing expectations in
coaching than in psychotherapy. Coaching clients may expect less depth and necessity to
open up, holding back sensitive but valuable information. When working alliance is
strong, clients may still share their feelings and thoughts with their coaches, which may
contribute to a stronger working alliance–outcome relationship in coaching. Future
research should follow this path and investigate which expectations clients bring into
coaching and how these may alter coaching outcomes.
Although we found that working alliance was related to all coaching outcomes con-
sidered, the effect sizes varied in size. This aligns well with both the coaching and train-
ing literature, which have found varying effect sizes for different levels (Alliger and
Janak, 1989; Jones et al., 2015). We found the strongest relationships between working
alliance and affective and cognitive outcomes. Working alliance may exhibit the strong-
est relationships to these coaching outcomes because they are assessed more holistically,
especially in the case of overall satisfaction with coaching and perceived overall effec-
tiveness. These measures may depend the least on other variables. For instance, using
new skills in the organization may also depend on the clients’ tasks or their supervisory
support. We were also able to test different indicators of affective outcomes, namely
clients’ satisfaction with coaching, perceived effectiveness, and self-efficacy. In line with
Alliger and Janak’s assumption in training that a single criterion may not fully capture
one evaluation level and may show different results, working alliance was most strongly
related to overall satisfaction with coaching and perceived effectiveness, and less with
clients’ self-efficacy. This supports the idea that holistic measures may exhibit stronger
effects than more specific ones and researchers should use different indicators when pos-
sible. Coaching outcomes on the results level were related to working alliance with a
lower, but still moderate, strength. These outcomes are less holistic in nature, because
they depend more on the specific client and his or her individual goal setting. To find that
working alliance is also related to the results level (with the highest number of studies)
suggests a robust ingredient that coaches can use to promote these key coaching out-
comes. This meta-analysis could not test for skill-based coaching outcomes, and future
research may wish to investigate the strength of their relationship to working alliance.
Besides the positive coaching outcomes for clients, working alliance was also nega-
tively associated with unintended negative effects for clients. This demonstrates its broad
spectrum of effects. A high-quality working alliance may let clients be more open and
coaches more focused on how they can help their clients, lowering the possibility for the
emergence of unintended negative effects. We cannot answer whether working alliance
may be more important for their reduction once they emerged. High-quality working
alliances may deliver the safe environment where such a taboo topic (Kilburg, 2002) can
be openly discussed, which allows the joint work on their reduction. Future research
should investigate the differential effects of working alliance for the emergence and
reduction of unintended negative effects for clients. This may include longitudinal

18 Human Relations 00(0)
designs, where clients and coaches evaluate their working alliance and outcomes after
each session, and qualitative designs to explore how clients and coaches react once unin-
tended negative effects emerge (see Schermuly and Graßmann, 2018, for a research
model on unintended negative effects).
The strength of the working alliance–outcome relationship did not differ when clients’
or coaches’ perspectives were used. However, there was still a trend for a stronger rela-
tionship from the clients’ perspective. The clients’ perspective on working alliance might
be more meaningful for coaching outcomes, as the clients themselves need to transfer the
skills and knowledge that they acquired in coaching into practice. Clients might therefore
be best suited to evaluate their coaching processes. Research on psychotherapy revealed
similar findings: clients’ evaluations of working alliance predicted outcomes better than
therapists’ evaluations (e.g. Zilcha-Mano et al., 2015). We suggest that future research
continues to assess working alliance from clients’ and coaches’ perspectives, because
both are related to coaching outcomes. Future studies should, though, examine the role
of the perspective in more detail, as clients and coaches may differ in their evaluation of
working alliance and coaching outcomes. It would be interesting to study if a higher
agreement between clients and coaches on their high-quality working alliance would
positively impact coaching outcomes because coaches better attune to their clients, as it
was shown in psychotherapy (e.g. Marmarosh and Kivlighan, 2012). Critically, common
method variance may be a potential concern if only one perspective is used, as it may
inflate relationships (Podsakoff et al., 2012). This meta-analysis found a medium-sized
relationship not only from clients’ or coaches’ perspectives alone, but also when clients
evaluated working alliance and coaches evaluated coaching outcomes. This indicates
that common method variance cannot be the sole explanation for this relationship, and
lends further support for the working alliance–outcome relationship.
Practical implications
Considering the robust relationships between working alliance and all investigated
coaching outcomes, coaches should actively monitor their working alliance with their
clients throughout coaching. For clients, we suggest opting for coaches who they feel
they can best relate to. For this purpose, coaches usually offer a first session to get to
know each other. Although working alliance may develop over time as it positively
relates to the number of coaching sessions (Baron and Morin, 2009), clients should use
this first contact to evaluate whether they will establish a high-quality working alliance
with their coaches. Working alliance seems to be a good measure of how well coaches
and clients work together, so coaching supervisors may want to use working alliance as
a starting point in supervision. This may be best indicated when clients seem to be dis-
satisfied with coaching or its effectiveness, as we found the strongest relationship for
these outcomes.
The possible difference in perspectives indicates that coaches should reevaluate the
working alliance and outcomes constantly. Coaches should realize that their clients’
perspective might differ from their own perspective. Coaches need to develop such cali-
bration skills so they can be more responsive to their clients’ needs (Lawley and Linder-
Pelz, 2016). Feedback may help coaches to do so. Psychotherapy research showed that

Graßmann et al. 19
feedback strengthened the association between therapist-evaluated working alliance
and outcome, suggesting that feedback helped therapists to identify clients who were
not progressing as expected and guided them in modifying the treatment as needed
(Zilcha-Mano and Errázuriz, 2015).
Coaches should use the necessary time to establish and improve the working alliance
to their clients. There are some suggestions on which techniques lead to a high-quality
working alliance in coaching. For instance, working alliance was related to coaches’
dominant-friendly behavior and coaches’ pleasant mood (Ianiro and Kauffeld, 2014),
reciprocal friendliness (Ianiro et al., 20151), behavioral similarity regarding dominance
and affiliation (Ianiro et al., 20131), and clients’ perceived range of coaches’ techniques
(De Haan et al., 20161). This meta-analysis demonstrated how important working alli-
ance is for coaching outcomes, so we call for more research on its predictors to help
coaches establish high-quality working alliances.
Limitations
This study is not without limitations. Because research on the working alliance in coach-
ing is not a decade old, this meta-analysis included a small number of studies. Whilst it
appears sufficiently large to detect meaningful relationships, more studies would allow
for the investigation of additional moderators. This seems necessary owing to the high
heterogeneity between the studies. The vocational access to coaching is not secured by
entry regulations and leads to diverse backgrounds, education, and approaches of
coaches. The large diversity of clients’ issues and personal needs contributes to individ-
ual and heterogeneous coaching processes. We cannot dissolve the differential effects of
working alliance across this heterogeneity and encourage researchers to investigate addi-
tional moderators, such as coaching experience or client–coach matching.
Meta-analyses can only be as good as the studies that they include. The included stud-
ies relied on correlational data that prohibit causal assumptions. Although our theoretical
reasoning suggests working alliance as a predictor for coaching outcomes, the opposite
direction might be possible as well. When clients feel that coaching has strong positive
outcomes for them, this may change the coaching process and strengthen their working
alliance with their coaches in turn. Because experimental manipulation of working alli-
ance will be hard to apply because of ethical considerations, cross-lagged study designs
should help to determine the reciprocity of this relationship.
Conclusion
The current meta-analysis indicates that the working alliance between clients and coaches
can be effectively used to promote client outcomes in coaching, as we found a robust
medium-sized overall relationship. Furthermore, this study has pointed out that working
alliance is related to a broad range of outcomes with varying effects sizes. Although
working alliance seems to be most strongly related to affective and cognitive coaching
outcomes, there were still medium-sized relationships to results level outcomes, such as
goal attainment, and unintended negative effects for clients. Our examination of clients’
and coaches’ perspectives contributes to a more refined understanding of the working

20 Human Relations 00(0)
alliance in coaching. Given its profound relation to client outcomes, the time has come
to identify what improves the working alliance between clients and coaches.
Funding
This research received no specific grant from any funding agency in the public, commercial, or
not-for-profit sectors.
Notes
1 This is a study included in the meta-analysis.
2 We distinguish coaching from other helping relationships by working with non-clinical cli-
ents on work-related topics (contrary to psychotherapy), focusing on the clients’ rather than
the organizations’ goals (contrary to mentoring), and excluding the purpose of quality control
(contrary to supervision).
References
Alliger GM and Janak EA (1989) Kirkpatrick’s levels of training criteria: Thirty years later.
Personnel Psychology 42(2): 331–342.
Alliger GM, Tannenbaum SI, Bennett W Jr, et al. (1997) A meta-analysis of the relations among
training criteria. Personnel Psychology 50(2): 341–358.
Alvey S and Barclay K (2007) The characteristica of dyadic trust in executive coaching. Journal
of Leadership Studies 1(1): 18–27.
Baron L and Morin L (2009) The coach–coachee relationship in executive coaching: A field study.
Human Resource Development Quarterly 20(1): 85–106.
Baron L, Morin L and Morin D (2011) Executive coaching: The effect of working alliance discrep-
ancy on the development of coachees’ self-efficacy. Journal of Management Development
30(9): 847–864.
Beinicke A, Schubert L and Putz D (2017) Spotlight on coaching: A perspective of clients and
coaches. Unpublished raw data, Department of Psychology, University of Würzburg,Germany.
Berry RM (2005) A comparison of face-to-face and distance coaching practices: The role of the
working alliance in problem resolution. Doctoral thesis, Georgia State University, Atlanta.
Berry RM, Ashby JS, Gnilka PB and Matheny KB (2011) A comparison of face-to-face and dis-
tance coaching practices: Coaches’ perceptions of the role of the working alliance in problem
resolution. Consulting Psychology Journal: Practice and Research 63(4): 243–253.
Blau PM (1964) Exchange and Power in Social Life. New York: John Wiley.
Bluckert P (2005) Critical factors in executive coaching – the coaching relationship. Industrial and
Commercial Training 37(7): 336–340.
Bordin ES (1979) The generalizability of the psychoanalytic concept of the working alliance.
Psychotherapy: Theory, Research and Practice 16(3): 252–260.
Borenstein M, Hedges LV, Higgins JPT and Rothstein HR (2009) Introduction to Meta-analysis.
Chichester: Wiley.
Borenstein M, Hedges LV, Higgins JPT and Rothstein HR (2010) A basic introduction to fixed-
effect and random-effects models for meta-analysis. Research Synthesis Methods 1(2): 97–111.
Borenstein M, Hedges LV, Higgins JPT and Rothstein HR (2014) Comprehensive meta-analysis
(version 3) [Computer software]. Englewood, NJ: Biostat.
Boyce LA, Jackson RJ and Neal LJ (2010) Building successful leadership coaching relation-
ships: Examining impact of matching criteria in a leadership coaching program. Journal of
Management Development 29(10): 914–931.

Graßmann et al. 21
Cohen J (1960) A coefficient of agreement for nominal scales. Educational and Psychological
Measurement 20(1): 37–46.
Cropanzano R and Mitchell MS (2005) Social exchange theory: An interdisciplinary review.
Journal of Management 31(6): 874–900.
De Haan E, Duckworth A, Birch D and Jones C (2013) Executive coaching outcome research:
The contribution of common factors such as relationship, personality match, and self-efficacy.
Consulting Psychology Journal: Practice and Research 65(1): 40–57.
De Haan E, Grant AM, Burger Y and Eriksson P-O (2016) A large-scale study of executive and
workplace coaching: The relative contributions of relationship, personality match, and self-
efficacy. Consulting Psychology Journal: Practice and Research 68(3): 189–207.
Dingman ME (2004) The effects of executive coaching on job-related attitudes. Doctoral thesis,
Regent University, Virginia Beach, VA.
Duval S and Tweedie R (2000) Trim and fill: A simple funnel-plot-based method of testing and
adjusting for publication bias in meta-analysis. Biometrics 56(2): 455–463.
Eby LT and McManus SE (2004) The protégé’s role in negative mentoring experiences. Journal
of Vocational Behavior 65(2): 255–275.
Eby LT, Allen TD, Hoffman BJ, et al. (2013) An interdisciplinary meta-analysis of the potential
antecedents, correlates, and consequences of protégé perceptions of mentoring. Psychological
Bulletin 139(2): 441–476.
Gan GC and Chong CW (2015) Coaching relationship in executive coaching: A Malaysian study.
Journal of Management Development 34(4): 476–493.
Gessnitzer S and Kauffeld S (2015) The working alliance in coaching: Why behavior is the key to
success. The Journal of Applied Behavioral Science 51(2): 177–197.
Ghods N (2009) Distance coaching: The relationship between the coach–client relationship,client
satisfaction, and coaching outcomes. Unpublished doctoral thesis, Alliant International
University, San Diego, CA.
Graen GB and Uhl-Bien M (1995) Relationship based approach to leadership: Development of
leader-member exchange (LMX) theory of leadership over 25 years: Applying a multi-level
multi-domain perspective. Leadership Quarterly 6(2): 219–247.
Grant AM (2005) What is evidence-based executive, workplace, and life coaching? In: Cavanagh
M, Grant AM and Kemp T (eds) Evidence-based Coaching, Vol. 1: Theory, Research and
Practice from the Behavioural Sciences. Bowen Hills, Queensland: Australian Academic
Press, 1–12.
Grant AM (2014) Autonomy support, relationship satisfaction and goal focus in the coach–coachee
relationship: Which best predicts coaching success? Coaching: An International Journal of
Theory, Research and Practice 7(1): 18–38.
Grant AM and Cavanagh MJ (2004) Toward a profession of coaching: Sixty-five years of pro-
gress and challenges for the future. International Journal of Evidence Based Coaching and
Mentoring 2(1): 1–16.
Grant AM and Stober DR (2006) Introduction. In: Stober DR and Grant AM (eds) Evidence Based
Coaching Handbook: Putting Best Practices to Work for Your Clients. Hoboken, NJ: John
Wiley and Sons, 1–14.
Graßmann C and Schermuly CC (2016) Side effects of business coaching and their predictors from
the coachees’ perspective. Journal of Personnel Psychology 15(4): 152–163.
Graßmann C and Schermuly CC (2018) The role of neuroticism and supervision in the relationship
between negative effects for clients and novice coaches. Coaching: An International Journal
of Theory, Research and Practice 11(1): 74–88.
Graßmann C, Schermuly CC and Rogenhofer A (2018a) Unpublished raw data. SRH University
Berlin, Germany.

22 Human Relations 00(0)
Graßmann C, Schermuly CC and Wach D (2018b) Potential antecedents and consequences of nega-
tive effect for coaches. Coaching: An International Journal of Theory, Research and Practice.
Epub ahead of print 19 June 2018, https://doi.org/10.1080/17521882.2018.1489870.
Gregory JB and Levy PE (2010) Employee coaching relationships: Enhancing construct clarity
and measurement. Coaching: An International Journal of Theory, Research and Practice
3(2): 109–123.
Greif S (2016) Researching outcomes of coaching. In: Bachkirova T, Spence G and Drake D (eds)
The SAGE Handbook of Coaching. London: SAGE, 569–588.
Hedges L and Olkin I (1985) Statistical Methods for Meta-analysis. San Diego, CA: Academic
Press.
Hopewell S, Clarke M and Mallett S (2006) Grey literature and systematic reviews. In: Rothstein H,
Sutton AJ and Borenstein M (eds) Publication Bias in Meta-analysis: Prevention, Assessment
and Adjustments. Hoboken, NJ: John Wiley and Sons, 49–72.
Horvath AO and Bedi RP (2002) The alliance. In: Norcross JC (ed.) Psychotherapy Relationships
that Work: Therapist Contributions and Responsiveness to Patients. New York, NY: Oxford
University Press, 37–69.
Horvath AO and Greenberg LS (1989) Development and validation of the Working Alliance
Inventory. Journal of Counseling Psychology 36(2): 223–233.
Horvath AO, Del Re AC, Flückiger C and Symonds D (2011) Alliance in individual psycho-
therapy. Psychotherapy 48(1): 9–16.
Hunter JE and Schmidt FL (1990) Methods of Meta-analysis: Correcting Error and Bias in
Research Findings. Newbury Park, CA: SAGE.
Ianiro PM and Kauffeld S (2014) Take care what you bring with you: How coaches’ mood and
interpersonal behavior affect coaching success. Consulting Psychology Journal: Practice and
Research 66(3): 231–257.
Ianiro PM, Schermuly CC and Kauffeld S (2013) Why interpersonal dominance and affiliation
matter: An interaction analysis of the coach-client relationship. Coaching: An International
Journal of Theory, Research and Practice 6(1): 25–46.
Ianiro PM, Lehmann-Willenbrock N and Kauffeld S (2015) Coaches and clients in action: A
sequential analysis of interpersonal coach and client behavior. Journal of Business and
Psychology 30(3): 435–456.
ICF (2012) Global coaching study. Available at: http://coachfederation.org/files/includes/media
/docs/2012ICFGlobalCoachingStudy-ExecutiveSummary.pdf (accessed 10 February 2018).
ICF (2016) Global coaching study. Available at: https://coachfederation.org/files/FileDownloads
/2016ICFGlobalCoachingStudy_ExecutiveSummary.pdf (accessed 10 February 2018).
Jansen A, Mäthner E and Bachmann T (2004) Erfolgreiches Coaching: Wirkfaktoren im Einzel-
Coaching [Successful coaching: Effect research in coaching]. Kröning, Germany: Asanger.
Jones RJ, Woods SA and Guillaume YRF (2015) The effectiveness of workplace coaching: A
meta-analysis of learning and performance outcomes from coaching. Journal of Occupational
and Organizational Psychology 89(2): 249–277.
Kilburg RR (1996) Toward a conceptual understanding and definition of executive coaching.
Consulting Psychology Journal: Practice and Research 48(2): 134–144.
Kilburg RR (2002) Failure and negative outcomes: The taboo topic in executive coaching. In:
Fitzgerald C and Berger JG (eds) Executive Coaching: Practices and Perspectives. Palo Alto,
CA: Davis-Black, 283–301.
Kirkpatrick DL (1967) Evaluation of training. In: Craig RL and Bittel LR (eds) Training and
Development Handbook. New York, NY: McGraw-Hill, 87–112.
Kivlighan DM, Hill CE, Gelso CJ and Baumann E (2016) Working alliance, real relationship, ses-
sion quality, and client improvement in psychodynamic psychotherapy: A longitudinal actor
partner interdependence model. Journal of Counseling Psychology 63(2): 149–161.

Graßmann et al. 23
Kivlighan DM, Kline K, Gelso CJ and Hill CE (2017) Congruence and discrepancy between work-
ing alliance and real relationship: Variance decomposition and response surface analyses.
Journal of Counseling Psychology 64(4): 394–409.
Kokotovic AM and Tracey TJ (1990) Working alliance in the early phase of counseling. Journal
of Counseling Psychology 37(1): 16–21.
Kraiger K, Ford JK and Salas E (1993) Application of cognitive, skill-based, and affective theories
of learning outcomes to new methods of training evaluation. Journal of Applied Psychology
78(2): 311–328.
Kretzschmar I (2010) Exploring clients’ readiness for coaching. International Journal of Evidence
Based Coaching and Mentoring (Special Issue 4): 1–20.
Lawley J and Linder-Pelz S (2016) Evidence of competency: Exploring coach, coachee and
expert evaluations of coaching. Coaching: An International Journal of Theory, Research and
Practice 9(2): 110–128.
McCarthy G and Milner J (2013) Managerial coaching: Challenges, opportunities and training.
Journal of Management Development 32(7): 768–779.
Marmarosh CL and Kivlighan Jr DM (2012) Relationships among client and counselor agree-
ment about the working alliance, session evaluations, and change in client symptoms using
response surface analysis. Journal of Counseling Psychology 59(3): 352–367.
Martin DJ, Garske JP and Davis MK (2000) Relationship of the therapeutic alliance with outcome
and other variables: A meta-analytic review. Journal of Consulting and Clinical Psychology
68(3): 438–450.
Maurer I (2009) Führungskräftecoaching: Eine Studie zur Wirksamkeit von prozessorientierten
Interventionstechniken bei der Problemklärung [Executive coaching: A study on the efficacy
of process-oriented intervention techniques for problem resolution]. Baden-Baden, Germany:
Tectum.
Nicholls AR (2017) Psychology in Sports Coaching: Theory and Practice. London, UK:
Routledge.
O’Broin AO and Palmer S (2006) The coach-client relationship and contributions made by the
coach in improving coaching outcome. The Coaching Psychologist 2(2): 16–20.
O’Broin AO and Palmer S (2007) Reappraising the coach-client relationship: The unassum-
ing change agent in coaching. In: Palmer S and Whybrow A (eds) Handbook of Coaching
Psychology: A Guide for Practitioners. Hove: Routledge, 295–324.
Passmore J and Fillery-Travis A (2011) A critical review of executive coaching research: A decade
of progress and what’s to come. Coaching: An International Journal of Theory, Research and
Practice 4(2): 70–88.
Podsakoff PM, MacKenzie SB and Podsakoff NP (2012) Sources of method bias in social sci-
ence research and recommendations on how to control it. Annual Review of Psychology 63:
539–569.
Putz D and Berost L (2017) Unpublished raw data. Rheinische Fachhochschule Köln, Germany.
Ramos-Sánchez L, Esnil E, Goodwin A, et al. (2002) Negative supervisory events: Effects on
supervision and supervisory alliance. Professional Psychology: Research and Practice 33(2):
197–202.
Rogers DT (2015) Further validation of the learning alliance inventory: The roles of working
alliance, rapport, and immediacy in student learning. Teaching of Psychology 42(1): 19–25.
Rosenzweig S (1936) Some implicit common factors in diverse methods of psychotherapy.
American Journal of Orthopsychiatry 6(3): 412–415.
Schermuly CC (2018) Client dropout from business coaching. Consulting Psychology Journal:
Practice and Research 70(3): 250–267.
Schermuly CC and Graßmann C (2018) A literature review on negative effects of coaching:
What we know and what we need to know. Coaching: An International Journal of Theory,

24 Human Relations 00(0)
Research and Practice. Epub ahead of print 3 October 2018, https://doi:10.1080/17521882
.2018.1528621.
Schermuly CC and Meyer B (2015) Good relationships at work: The effects of Leader–Member
Exchange and Team–Member Exchange on psychological empowerment, emotional exhaus-
tion, and depression. Journal of Organizational Behavior 37(5): 673–691.
Schermuly CC, Schermuly-Haupt M-L, Schölmerich F and Rauterberg H (2014) Zu Risiken und
Nebenwirkungen lesen Sie … Negative Effekte von Coaching [For risks and side effects
read … Negative effects of coaching]. Zeitschrift für Arbeits- und Organisationspsychologie
58(1): 17–33.
Sonesh SC, Coultas CW, Lacerenza CN, et al. (2015a) The power of coaching: A meta-analytic
investigation. Coaching: An International Journal of Theory, Research and Practice 8(2):
73–95.
Sonesh SC, Coultas CW, Marlow SL, et al. (2015b) Coaching in the wild: Identifying factors that
lead to success. Consulting Psychology Journal: Practice and Research 67(3): 189–217.
Spence GB (2007) GAS powered coaching: Goal attainment scaling and its use in coaching
research and practice. International Coaching Psychology Review 2(2): 155–167.
Theeboom T, Beersma B and Van Vianen AEM (2014) Does coaching work? A meta-analysis on
the effects of coaching on individual level outcomes in an organizational context. The Journal
of Positive Psychology 9(1): 1–18.
Wasylyshyn KM (2003) Executive coaching: An outcome study. Consulting Psychology Journal:
Practice and Research 55(2): 94–106.
Webers G (2008) Die Untersuchung der Selbstwirksamkeitserwartungen, Selbststeuerungs-
kompetenzen und Selbstreflexion im Kontext eines Coachings zum Thema Procrastination.
Unpublished diploma thesis, Universität Osnabrück, Germany.
Will T, Gessnitzer S and Kauffeld S (2016) You think you are an empathic coach? Maybe you
should think again. The difference between perceptions of empathy vs. empathic behaviour
after a person-centred coaching training. Coaching: An International Journal of Theory,
Research and Practice 9(1): 53–68.
Zilcha-Mano S and Errázuriz P (2015) One size does not fit all: Examining heterogeneity and iden-
tifying moderators of the alliance–outcome association. Journal of Counseling Psychology
62(4): 579–591.
Zilcha-Mano S, Solomonov N, Chui H, et al. (2015) Therapist-reported alliance: Is it really a pre-
dictor of outcome? Journal of Counseling Psychology 62(4): 568–578.
Carolin Graßmann is postdoctoral researcher at SRH University of Applied Sciences Berlin,
Germany. Her main research focuses on the factors that determine coaching outcomes and nega-
tive effects of coaching. Her work has appeared in international journals such as Journal of
Personnel Psychology and Personnel Review. [Email: carolin.grassmann@srh.de]
Franziska Schölmerich works as postdoc researcher at the Digital Transformation Research Center at
Leuphana University Lüneburg, Germany. Her current research interests include diversity and agility
in teams, leadership, and coaching. Her work has been published in international journals such as
Human Performance and Small Group Research [Email: franziska.schoelmerich@leuphana.de]
Carsten C Schermuly is a Professor of Business Psychology at SRH University of Applied
Sciences Berlin, Germany. His main research focuses on psychological empowerment, interaction
processes and diversity in teams, and the effects of coaching. His work has been published in
Journal of Organizational Behavior, Human Performance, Cognitive Science, and European
Journal of Work and Organizational Psychology. [Email: carsten.schermuly@srh.de]