Abstract
This article reports on the results of the first meta-analysis of the association between working alliance and outcomes of individual career counseling. This random-effects meta-analysis included 18 published and unpublished studies that produced a weighted mean effect size of r = .42. This effect size was heterogeneous across studies. Separate meta-analyses were conducted for several types of outcomes: Career outcomes, mental health outcomes, and client-perceived quality of the intervention. Average effect sizes for the association between working alliance and types of outcomes were .28, .18 and .62, respectively. Moderator analyses indicated that the overall mean effect size (r =.42) varied in a large proportion as a function of the type of outcomes and the time of assessment of working alliance (first session, mid or at termination of the counseling service). Our results confirm that working alliance is associated to career counseling effectiveness and suggest that career counselors should emphasize on the working alliance during the career counseling process. In conclusion, this article provides suggestions for practice in individual career counseling and avenues of research on working alliance in this context.
Many individuals consult in career counseling seeking help in dealing with career difficulties (Gati & Levin, 2015). Meta-analyses conducted in the past decades indicate that career counseling interventions are generally effective (Brown & Ryan Krane, 2000; Oliver & Spokane, 1988; Whiston et al., 1998, 2017). Individual career counseling appears to be the most effective modality (Whiston & Rahardja, 2008) with medium to large effect sizes (ds) ranging from 0.41 (Brown & Ryan Krane, 2000) to 1.08 (Whiston et al., 1998). Large effect sizes were also observed in recent studies that were not included in these previous meta-analyses (Alchin et al., 2018; Masdonati et al., 2014; Milot-Lapointe et al., 2016, 2018, 2020).
Although previous research supported the effectiveness of individual career counseling, several studies demonstrated that there is variability in clients’ responses to individual career counseling and that some clients do not benefit significantly from this intervention (Covali et al., 2011; Milot-Lapointe et al., 2016, 2019; Multon et al., 2001, 2007). To ensure that what career counselors do is effective for every client and to help them know what works with their clients, it is important to identify the process variables that contribute to individual career counseling effectiveness (Milot-Lapointe et al., 2019; Whiston et al., 2016). In that regard, Heppner and Heppner (2003) provided 10 avenues of research of which the first one involves examining the contribution of working alliance to career counseling effectiveness. More recently, other researchers (Whiston & Rahardja, 2008; Whiston & Rose, 2015; Whiston et al., 2016) also underlined the importance of examining whether working alliance is related to career counseling outcomes.
In line with Bordin’s (1979) conceptualization, working alliance is defined as a mutual collaboration between the client and the counselor based on the development of an emotional bond, as well as a shared commitment to the tasks and goals of the intervention. In this pan-theoretical conceptualization, working alliance is considered as a common factor that is not specific to a theoretical approach (Horvath & Luborsky, 1993). The common factor model proposes that what make counseling effective is principally the set of factors that are common or shared between counseling approaches (e.g. client-counselor relationship, technical factors) rather than the differences between them (Leibert & Dunne-Bryant, 2015). Bordin (1989) theorized on how the working alliance contributes to the effectiveness of all types of counseling (and approaches). He stated that working alliance is not curative in itself but rather the ingredient that makes it possible for clients to accept and follow the intervention components in counseling. Thereby, the development of a “good enough” working alliance early in the process of career counseling should be seen as a facilitator or a condition to the effectiveness of career counseling interventions.
Meta-analyses in the field of personal counseling and psychotherapy found that working alliance is a significant predictor of the outcomes of various therapies with a mean effect size ranging from r = .21 (Horvath & Bedi, 2002) to r = .28 (Flückiger et al., 2018; Horvath et al., 2011). The magnitude of the association between working alliance and outcomes was found to be higher when working alliance was measured later in the psychotherapy (r = .39 later in therapy; r = .25 early in therapy) (Horvath et al., 2011). The size of the relationship between working alliance and outcomes of psychotherapy also varied as a function of the type of outcomes (rs ranged between .18 (dropout) and .42 (depressive symptoms).
As compared to psychotherapy, few studies investigated the association between working alliance and outcomes in the field of career counseling (Whiston et al., 2016). In an article published in this journal, Whiston et al. (2016) conducted a systematic review of the articles and theses on the working alliance in career counseling published in English and French between 1990 and 2014. They also included one article in vocational rehabilitation counseling, an intervention they considered similar to career counseling. Based on the results of eight studies, Whiston et al. (2016) found a moderate mean association (r = .36) between clients’ perception of the working alliance and various outcomes of individual career counseling (e.g. career indecision, psychological distress, satisfaction with the intervention). They underlined that the true effect size of working alliance is probably around .30 (as in psychotherapy) because some of the studies with small sample size (ns were between 26 and 2,732) had higher correlation coefficient.
Although Whiston et al.’s (2016) results suggest that working alliance is an important process variable in predicting the effectiveness of career counseling, the association between working alliance and outcomes is not clearly supported with all types of outcomes. Indeed, while Whiston et al.’s review clearly supported the association between working alliance and client-perceived usefulness of career counseling sessions (rs ranged from .59 to .70;), non significant, small or moderate associations were observed with career outcomes (e.g. career indecision, career maturity) and mental health outcomes (psychological distress, life satisfaction). Indeed, Masdonati et al. (2009, 2014) found a moderate association between working alliance and career decision difficulties (r = −.25 and −.28 respectively), whereas in Heppner et al. (2004) study, working alliance was not significantly correlated with career decidedness (r = .11) and clarity of vocational identity (r = .06). Similarly, Multon et al. (2001) and Heppner et al. (2004) reported that working alliance was not significantly correlated with psychological distress (r = −.07 and −.15 respectively) while a small association (r = .20) was observed between working alliance and life satisfaction in Masdonati et al.’s (2014) study and a moderate association (r = .32) in Masdonati et al.’s (2009). Finally, Covali et al. (2011) found a small association between working alliance and career maturity (r = .13) and vocational self-efficacy (r = .11), while Elad-Strenger and Littman-Ovadia (2012) found working alliance be moderately (r = .28) correlated with client career exploration.
Because of the high variability in the correlation coefficients observed in these studies (rs ranged from .06 to .70), it appears important to replicate and extend Whiston et al.’s (2016) results to obtain a clear estimate of the association between working alliance and different types of career counseling outcomes. Indeed, although Whiston et al.’s (2016) review of research made a substantial contribution to the field, the mean r of .36 (or .30) they reported might not be a precise estimate of the association between working alliance and all types of outcomes of career counseling since this association appears to be overinflated by the very high correlation coefficients (between .59 and .70) observed in outcomes related to client-perceived usefulness of sessions. In addition, some relevant studies that reported correlation coefficients on the association between working alliance and various outcomes of career counseling were not included in their literature review (Kim et al., 2002; Li & Kim, 2004; Massoudi et al., 2008; Michaud et al., 2011; Milot-Lapointe, 2013; Rochlen et al., 2004), which might also have affected their results. Finally, the sampling error and the non independency of outcomes from a same study were not taken into account in the calculation of the overall effect size.
To obtain a more precise and appropriate weighted indicator of the effect size of working alliance, Whiston et al. (2016) suggested that a meta-analysis of the relationship between working alliance and career counseling outcomes should be conducted when additional studies would be available. Indeed, a meta-analysis would be helpful to obtain a weighted estimate (with confidence intervals) of the effect of working alliance and to explain the possible variability between effect sizes across studies with moderator variables (e.g. type of outcome; time of assessment of the working alliance). While the number of studies was possibly not sufficient at the time of Whiston et al.’s (2016) literature review, there is now enough studies to perform a meta-analysis considering recently published articles (Alchin et al., 2018; Iwanaga et al., 2019; Marcilio et al., 2019; Milot-Lapointe et al., 2020) and studies that were not included by Whiston et al. (2016). Indeed, using Monte Carlo simulations, Field (2001) found that a minimum of 15 samples (with a mean sample size over 80) in meta-analyses of correlation coefficients was enough to control for Type I error and to reach good statistical power when expecting a medium effect size.
The Present Study
The current study aimed to replicate and extend Whiston et al.’s (2016) results using a meta-analytical statistical approach. Following the suggestion of Whiston et al. (2016), this study was carried out in order to provide a more robust weighted estimate of the association between working alliance and individual career counseling outcomes, as well as to test for potential moderators that could explain the heterogeneity in the strength of this association across studies.
Based on empirical results in psychotherapy described earlier (Flückiger et al., 2018; Horvath et al., 2011), the type of outcomes and the time of assessment of working alliance will be included as potential moderators of the association between working alliance and outcomes of individual career counseling. It may be hypothesized that client-perceived usefulness of career counseling sessions is more likely to be associated with working alliance as it depends largely on the satisfaction of client expectations about tasks and goals of counseling (Chiesa et al., 2020), while positive career and mental health outcomes require both the use of specific intervention components (e.g. written exercises in session) by counselors and a high working alliance between counselor and clients (Milot-Lapointe et al., 2018, 2020). Furthermore, as initials tasks and goals of career counseling can be modified during the course of counseling (Gysbers et al., 2014) and working alliance tends to fluctuate during the process (Heppner et al., 2004), it may be hypothesized that earlier assessments of working alliance are less likely to be associated with final outcomes.
In view of the above, the aims of the current study were threefold: a) to examine the significance and size of the association between working alliance and outcomes of individual career counseling using a meta-analysis, b) to examine the homogeneity of the association between working alliance and outcomes of individual career counseling across studies, c) to test for a potential moderator effect of the type of outcome and the time of assessment of working alliance on the association between working alliance and outcomes of individual career counseling.
Method
Literature Search
The search for relevant studies for this meta-analysis was conducted in four phases. First, we searched in APA PsycINFO, ERIC, Academic Search Complete and Behavioral Science Collection databases for peer reviewed articles published in English, French, Portuguese, Russian, or Spanish over the last 30 years (1990 through 2020). Second, we searched in ProQuest Dissertations and Thesis and Google Scholar to access unpublished dissertations or theses. The keywords employed in these databases were career counseling, employment counseling or vocational rehabilitation counseling and either working alliance or therapeutic alliance.
Third, we searched by hands journals that typically publish career counseling research in English (British Journal of Guidance & Counselling, Career Development Quarterly, Journal of Career Assessment, Journal of Career Development, Journal of Counseling Psychology, Journal of Employment Counseling, Journal of Occupational Rehabilitation, Journal of Vocational Behavior) or in French (L’orientation scolaire et professionnelle, Revue Canadienne de counseling et de psychothérapie, Revue canadienne de développement de carrière). Fourth, we examined reference lists of the selected articles and of reviews of research on working alliance in career counseling (Whiston & Rahardja, 2008; Whiston & Rose, 2015; Whiston et al., 2016) in search of further articles.
For a study to be selected, it had to provide statistics describing the association between working alliance and individual career counseling or individual vocational rehabilitation counseling outcomes (in line with Whiston et al., 2016), such as a r zero order correlation coefficient or statistics that could be converted to r. Both published and unpublished studies were included. In addition, one ongoing study (that has not yet been submitted to a journal) conducted by the authors of this meta-analysis was included. Studies on the working alliance between career counselors in training and their supervisors were not included.
These four phases resulted in the identification of 18 correlational studies (14 articles, two research report, one dissertation and one ongoing study) on the association between working alliance and outcomes of individual career counseling. The mean sample size of the 18 studies was 241.28 for a total of 4,331 participants. Studies included in this meta-analysis were conducted in United-States (k = 7; n = 3,399 participants), Canada (k = 4; n = 289), Switzerland (k = 2; n = 289), France (k = 2; n = 153), Israel (k = 1; n = 94), Australia (k = 1; n = 87), and Brazil (k = 1; n = 32).
Most studies assessed more than one type of outcome. Following Whiston et al. (2016) and other meta-analyses on the working alliance (Flückiger et al., 2018; Grabmann et al., 2020; Horvath et al., 2011), we aggregated several types of outcomes to produce an overall effect size of working alliance. Additionally, we grouped the outcomes in three categories: Career outcomes, mental health outcomes, and client-perceived quality of the intervention. Career outcomes include the subcategories Career indecision and Career readiness, which are two correlated problems in decision-making (Brown & Rector, 2008). In the subcategory Career indecision, we included career decision difficulties and career decision needs, two correlated indicators of career indecision (Brown & Rector, 2008; Brown et al., 2012). Career Readiness includes career maturity, vocational self-efficacy, vocational outcome expectancies, career exploration, career transition readiness, clarity of vocational identity, goal instability, career thoughts and client perception of employment prospect (which is a career thought). These variables are indicators of client confidence and readiness to deal with career development tasks (e.g. made a career decision, find a job) (Brown & Rector, 2008). These variables had strong loadings on a same factor (Lack of readiness) in Brown and Rector’s (2008) meta-analysis of problems in career decision-making. This factor has also been validated in a subsequent study (Brown et al., 2012).
Mental health outcomes include psychological distress (negative state) and life satisfaction (positive state), two distinct but, at least moderately correlated indicators of mental health in several studies (rs = ranged from −.42 to −.59 in Fotiadou et al., 2008; Grevenstein et al., 2018; Headey et al., 1993; Milot-Lapointe & Le Corff, ongoing study).
Client-perceived quality of the intervention includes variables related to client’s subjective perception that the intervention was helpful (satisfaction with the intervention, perception of the usefulness and the effectiveness of the intervention, perception of the quality of sessions). Finally, some variables (job satisfaction, stages of change employment, vocational rehabilitation engagement, and vocational autonomous motivation) were not included in this meta-analysis because they were not conceptually and statistically related to the other outcome variables.
All studies measured working alliance in accordance with Bordin’s theoretical model. The Working Alliance Inventory (WAI; Horvath & Greenberg, 1989) or its short form (WAI-S; Tracey & Kokotovic, 1989) were used in 15 studies. The Working Alliance Survey (WAS; Lustig et al., 2002) was used in one study to measure individuals with disabilities’ perception of the working alliance. A homemade working alliance questionnaire (Covali et al., 2011) that only includes agreement on goals and tasks was used in one study. The Relationship Scale (RS; Hill & Kellems, 2002) which includes items on bond and agreement on tasks between client and counselor was used in one study. Concerning the time of assessment, five studies measured working alliance at the end of the first career counseling session while five studies measured working alliance at mid of the process, seven studies at termination, and one study did not provide this information.
Data Analytic Strategy
When conducting a meta-analysis, researchers must decide between using a fixed-effects model or a random-effects model (Martincin & Stead, 2015). The fixed-effects model assumes that studies in the meta-analysis are sampled from a population in which the true effect size is the same in all studies (Field & Gillett, 2010). In the random-effects model, it is assumed that studies in the meta-analysis are drawn from a population of studies in which the average effect size varies randomly from study to study (Field & Gillett, 2010). Based on results of previous meta-analysis in psychotherapy (Horvath et al., 2011) that showed heterogeneity in the associations between working alliance and outcomes, this meta-analysis used a random-effects model computed with the program developed by Field and Gillett (2010) (Whiston et al., 2017). The Hedges and Vevea’s (1998) random-effects method was chosen over the Hunter and Schmidt’s (2000) method because of previous findings (Field, 2005) suggesting it offers a better control for Type 1 error when the number of studies included in the meta-analysis is around 20.
In many studies, there were multiple outcome measures. To avoid dependencies among effect sizes from a same study, in accordance with similar meta-analyses on working alliance in psychotherapy (Flückiger et al., 2020; Horvath et al., 2011), we used the Hunter and Schmidt’s (2004) aggregation procedure to obtain one effect size per study. This procedure considers the correlation among within-study outcomes measures, and thus provides a more precise estimate of the parameter in the population (Horvath et al., 2011).The rs effect size of each study were then transformed to a Fisher’s z (Fisher, 1924) (Flückiger et al., 2020; Horvath et al., 2011; Whiston et al., 2017). The transformed scores were used to calculate an average effect size (that account for sampling error) in which each effect size was weighted by a variance component that incorporates both between- and within-study variance (Field & Gillett, 2010). The average effect size was converted back to the r metric for interpretive purposes, and a 95% confidence interval was calculated to estimate the precision of the r average effect size.
The heterogeneity of effect sizes was estimated using the Q statistic and the extent of heterogeneity in the average effect size was calculated using the I2 index (which estimates the proportion of heterogeneity that is not caused by sampling error). I2 scores of 25%, 50%, and 75% are considered to reflect low, moderate, and high heterogeneity, respectively (Huedo-Medina et al., 2006).
Results
First Objective: Association Between Working Alliance and Outcomes
Data used in this meta-analysis are displayed in Table 1. The mean effect size for the association between working alliance and individual career counseling outcomes across the 18 studies was r = .42, and its 95% confidence interval ranged from .33 to .51. This medium-large effect size (Cohen, 1988) was statistically significant (z = 8.22; p < .001), and indicated that the higher the working alliance, the better the various career counseling outcomes were.
Correlations Between Working Alliance and Outcomes of Individual Career Counseling.
1 Zero order correlation between working alliance and pre–post change in outcomes. This explains why some correlation coefficient for the same outcome are not in the same direction (e.g. alliance-psychological distress relation is negative in Milot-Lapointe, 2013 and Heppner et al., 2004 and positive in Milot-Lapointe et al. 2020).
As can be seen in Table 2, results also indicated a significant association between working alliance and all categories and subcategories of outcomes: Career (z = 4.80; p < .001; k = 12), career indecision (z = 6.06; p < .001; k = 5), career readiness (z = 3.51; p < .001; k = 9), mental health (z = 4.39; p < .001; k = 6), and client-perceived quality of the intervention (z = 9.27; p < .001; k = 8). The mean effect size for the association between working alliance and client-perceived quality of the intervention was r = .62 [.58; .70] indicating a large association with this type of outcome. A moderate association was observed between working alliance and career (r = .28 [.18; .39]), career indecision (r = −.24[−.16; −.31]), career readiness (r = .29[.13; .42]) and mental health (r = .18 [.10; .26]) outcomes. As the confidence intervals indicate, the upper r value of career (.39), career indecision (-.31), career readiness (.42) and mental health (.26) outcomes did not reach the lower r value of client-perceived quality of the intervention (.58), which suggests that working alliance was more strongly associated to this type of outcome. The magnitude of the association between working alliance and career, career indecision, career readiness and mental health outcomes did not significantly differ from one another given the overlap between their confidence intervals.
Effect Size of Working Alliance by Outcome Category.
The estimate of an effect size in a meta-analysis may be vulnerable to publication bias (Horvath et al., 2011). To determine whether there was a publication bias in the estimation of the overall effect size of the association between working alliance and outcomes in our data, we performed Kendall’s t rank correlation between the adjusted global effect size and the adjusted variance of estimate (Martincin & Stead, 2015). This approach was preferred to the inspection of plot asymmetry, because it overcomes the subjectivity of visual inspections (Martincin & Stead, 2015). The correlation coefficient was not statistically significant (t = .111, p = .52), which suggests that the significant association between working alliance and outcomes of individual career counseling in the current study did not result from publication bias.
Second Objective: Estimating the Variability of Effect Sizes
The chi-scare test of homogeneity of the overall effect size of the association between working alliance and all outcomes was statistically significant (Q = 27.69, p < .05) and the I2 score was 39%, providing evidence of a moderate heterogeneity between the effect sizes of the 18 studies. The chi-square test of homogeneity of effect size for career outcomes (Q = 5.77, p > .05), mental health outcomes (Q = 0.97, p > .05), and client perceived quality of the intervention (Q = 8.25, p > .05) were not statistically significant in all three cases, indicating similar effect size across studies.
Third Objective: Testing the Role of Moderators
Based on previous meta-analysis on the working alliance in psychotherapy (Horvath et al., 2011) we investigated the impact of two categorical moderators that could possibly moderate the relation between working alliance and outcomes: Type of outcome (career-mental health outcome or client-perceived quality of the intervention) and time of assessment of working alliance (first session, mid, or at termination of the counseling service). As mentioned earlier, most studies in this meta-analysis assessed more than one type of outcome. All moderation analyses were carried out in order to ensure independency of the data. For the type of outcome variable, we created a first category including studies whose mean r effect size was based solely on career, mental health or both outcomes (as previous analyses showed that the effect size of these outcomes did not differ from one other) (k = 10) and a second category of studies whose mean effect size was based on client perceived quality of the intervention solely or with career and mental health outcomes (k = 8) (the effect size of five studies was based solely on client perceived quality of the intervention while the effect size of three was based on client perceived quality of the intervention and career and mental health outcomes).
Considering that two studies measured working alliance at all times of assessment (first session, mid and at termination), we included the earliest alliance measure (as did Horvath et al., 2011). To ensure that the two potential moderator variables explain unique variance in effect sizes, we ran a multiple regression with the two variables entered in one block. The analysis showed that both the type of outcome (β = .40, p < .05) and the time of assessment of working alliance (β = .61, p < .01) significantly explained the variability in the alliance-outcome associations across studies. Tukey’s contrasts indicated that the effect size of the association between working alliance and career-mental health outcomes (r = .28) was significantly lower (p < .05) than the effect size (r = .60) of the association between working alliance and the second category of outcomes (including client-perceived quality of the intervention). Moreover, the overall association between working alliance and outcomes was significantly higher (p < .05) when alliance was measured at termination (r = .63; k = 7), than when it was measure at the end of the first session (r = .21; k = 5) or at mid (r = .32; k = 5) of the career counseling service.
Discussion
A few years ago, Whiston et al. (2016) made a substantial contribution to the field by conducting the first systematic review of research on the association between working alliance and individual career counseling outcomes. The current study aimed to replicate and extend Whiston et al.’s (2016) results using a meta-analytical statistical approach. One of the significant findings of this meta-analysis is that with more studies and using a more precise statistical approach, the overall effect size (r = .42) for the association between working alliance and individual career counseling outcomes was higher than the effect size (r = .30) estimated by Whiston et al. (2016). This result shows that the strength of the working alliance explains 18% of the variability in the effectiveness of career counseling, while Whiston et al.’s study suggested it explains 9% of this variability.
Although our results suggest that working alliance could play a more important role than suggested in previous studies, a significant variability was observed in the effect sizes of working alliance across studies. Indeed, our results provided evidence of a significant heterogeneity between the effect sizes of the 18 studies and indicated that the average effect size of the working alliance varied in a large proportion as a function of the type of individual career counseling outcomes and the time of assessment of working alliance.
Concerning the types of outcomes, results showed that working alliance was moderately associated with career and mental health outcomes, and largely correlated with client-perceived quality of the intervention. Thus, although working alliance was associated, at least moderately, to all types of outcomes of individual career counseling, our results showed that working alliance was more strongly associated to clients’ subjective perception that career counseling was helpful (e.g. client satisfaction about the intervention) than to tasks and goals-oriented outcomes (e.g. change in career decision difficulties or in satisfaction with life). This means that a strong working alliance between counselors and clients is a very good indicator that clients will find the intervention useful and, but to a lesser extent, that they will solve their career problems during career counseling.
Interestingly, a recent meta-analysis in coaching (Grabmann et al., 2020) also found a strong association between working alliance and client-perceived quality of the intervention (rs ranged between .58 and .64) while goals outcomes were moderately associated with working alliance (r = .32). According to Masdonati et al. (2009), the fact that client satisfaction is the outcome that is the most strongly associated to working alliance suggests that relational aspects are an important factor that clients take into account when judging the quality of an intervention. According to expectation confirmation theory (Oliver, 1980) clients’ satisfaction in a service results from the extent to which it meets its expectations (Chiesa et al., 2020). In that regard, Chiesa et al. (2020) demonstrated that clients’ satisfaction and perceived effectiveness about career counseling is highly associated with the capacity of career counselors to meet client expectations about goals and tasks of counseling (rs = .64 and .51). One may thus hypothesize that the strong association between working alliance and client satisfaction and perceived effectiveness about career counseling might be explained or mediated by the response to client expectations about counseling. Indeed, a good working alliance might be seen as a prerequisite to meet client expectation which in turn will explain client satisfaction about career counseling. However, this hypothesis needs empirical investigation.
The moderate association between working alliance and career and mental health outcomes might be explained by recent results (Milot-Lapointe et al., 2018, 2020) indicating that positive changes on these variables depends on both the quality of the working alliance and the choice of the intervention components used by career counselors during the process. Indeed, Milot-Lapointe et al.’s (2018) study showed that, despite a high working alliance between clients and counselors, change in client career decision difficulties tends to be lower when the career counseling processes did not include empirically supported intervention components (e.g. written exercises in-session on occupational analyses). In other words, while a strong working alliance appears to lead to a high client satisfaction, our results suggest that it may not be sufficient to lead to optimal career and mental health outcomes such as lower career indecision or psychological distress.
Turning to the time of assessment of working alliance, our results showed that working alliance was largely associated with career counseling outcomes when measured at the end of the counseling process, and moderately correlated with outcomes when measured at the end of the first session or at mid of the process. This means that client perception of the working alliance during career counseling is a modest indicator of the outcomes that will occur at the end of the process, whereas client perception of the working alliance at the end of the process reflects in a large part the outcomes of counseling. Although this result has also been observed in previous meta-analyses in psychotherapy (Flückiger et al., 2018; Horvath et al., 2011), no theoretical hypotheses have been proposed to explain it. According to Gysbers et al. (2004), regardless of the number of sessions, working alliance evolves during career counseling, “moving from forming the working alliance, to strengthening it, to fulfilling it, to finally closing it upon completion of career counseling” (pp. 7–8). As the closing phase of the working alliance involves evaluating results of career counseling (i.e. if clients goals were achieved or problems are resolved) (Gysbers et al., 2014), it is possible that client perception of the alliance at the end of the process is influenced by the final results of the intervention. Hence, working alliance measured at the end of career counseling should probably be considered more an outcome variable than a process variable.
Interestingly, the overall results of this meta-analysis are in large part consistent with results of recent meta-analyses on working alliance in psychotherapy (Flückiger et al., 2018; Horvath et al., 2011) which included a high number of studies (k = 290 and 190 respectively). Although these meta-analyses found a moderate association (r = .28 for both studies) between working alliance and psychotherapy outcomes (while we found a medium-large association), this slight discrepancy might be explained by the fact that the overall effect size of the current study (r = .42) was overinflated by the results on client perceived quality of the intervention (r = .62) (this type of outcome was not present in the meta-analyses in psychotherapy). Indeed, our separated meta-analyses for career (r = .28) and mental health outcomes (r = .18) provided results that are consistent with those of previous meta-analyses (Flückiger et al., 2018; Horvath et al., 2011) on depressive symptoms (rs between .25 and .42) and psychological distress (rs = .24 and .27), indicating that working alliance has a similar degree of association with task- and goal-oriented outcomes in career counseling and psychotherapy. Moreover, the finding according to which the association between working alliance and outcomes of career counseling is moderated by the type of outcomes and the time of assessment of working alliance is also consistent with the results of Flückiger et al. (2018) and Horvath et al. (2011), who found higher effect sizes for specific type of outcomes (depressive symptoms) and when alliance was measured later in psychotherapy. These similarities strengthen our confidence in the validity and generalizability of our results.
Implications for Practice
Our findings suggest that working alliance plays a significant role in career counseling effectiveness and that career counselors should emphasis on the working alliance from the beginning of counseling and until the end of the process. Since working alliance is one of the most robust predictors of career counseling effectiveness, we recommend that career counselors invest significant time and efforts for building, maintaining and restoring the alliance during each career counseling session. Nonetheless, considering that working alliance was moderately associated with career and mental health outcomes, it is important that career counselors keep in mind that working alliance is a necessary but not a sufficient condition to solve client difficulties. Indeed, counselors should use empirically supported interventions in addition to paying attention to the quality of the working alliance with their clients (Milot-Lapointe et al., 2018).
Limitations
Results of this meta-analysis should be considered in the light of the limitations or characteristics of the studies that were included. A first limitation is related to the fact that all studies used correlational design (which was also the case in previous meta-analyses in related field). Although correlational studies have been recommended for the investigation of working alliance, there are limitations regarding the conclusions that can be drawn, as we cannot conclude from correlation studies that a strong working alliance in career counseling causes better outcomes (Whiston et al., 2016).
A second limitation of this meta-analysis is that some studies had nested data, which occur when one or many counselors delivered career counseling to more than one client (Whiston et al., 2016). Since these studies did not account for counselor effects in the outcomes, it is possible that a portion of the variability in career counseling effectiveness explained by working alliance is attributable to differences between counselors in their average impact on client outcomes.
A third limitation of this meta-analysis is the relatively low number of studies (in comparison with the field of psychotherapy) on the association between working alliance and individual career counseling outcomes. Although the number of studies was adequate to conduct our analyses (from a statistical viewpoint and regarding the similarities with the n of other meta-analyses on r coefficients in the field; e.g. Grabmann et al., 2020; Horvath & Symonds, 1991; Martincin & Stead, 2015), we had to merge career and mental health outcomes in a same category in the analysis of the moderating effect of the type of outcomes to reach sufficient statistical power. It is important to recall that the merging of these two types of outcomes in a same category was adequate because the size of their association with working alliance overlapped in this study.
Conclusion and Future Directions for Research
Despite the above-mentioned limitations, this first meta-analysis on the association between working alliance and individual career counseling outcomes contributed to meeting the need, highlighted by Whiston et al. (2016), to obtain a more precise estimate of the magnitude of this relationship. Additionally, this study identified two significant moderators (time of assessment of the working alliance and type of outcomes) of the association between working alliance and outcomes of individual career counseling. Since the conclusions of this meta-analysis rely on results of 18 studies (rather than one or few studies), career counselors can be confident that working alliance is an important variable to consider in their interventions with their clients. However, we concur with previous claims (Heppner & Heppner, 2003; Whiston & Rose, 2015; Whiston et al., 2016) that more research is needed on the working alliance in career counseling.
A first avenue of research would be to investigate the association between working alliance and long-term effects of career counseling. Although previous research suggested that career counseling has positive long-term effects for many clients (Perdrix et al., 2012), we do not know if process variables such as working alliance contribute to this effectiveness.
Since working alliance explained an important but modest proportion of the variance in career counseling effectiveness, it appears important to study whether the combination of working alliance with other variables explains more variance in outcomes. Indeed, although studies have found that other variables such as client personality (Stauffer et al., 2013) and career intervention components (Milot-Lapointe et al., 2018, 2020) are significantly associated with career counseling outcomes, no study included all these important variables in a same model aiming to explain final outcomes. Thus, a third avenue of research would be to investigate, using repeated measures designs and complex statistical models (e.g. structural equation modeling), whether the intervention components used by counselors, client personality and working alliance have a unique contribution and interact together in predicting the effectiveness of career counseling.
Finally, given the importance of working alliance to career counseling effectiveness, it seems particularly important that researchers identify which client variables affect the development of working alliance and what career counselors can do to develop a strong alliance with each client. Research (Masdonati et al., 2014; Stauffer et al., 2013) conducted on client variables showed that clients with high levels of career decision difficulties and neuroticism, and low levels of conscientiousness, evaluate more negatively the quality of the working alliance with their counselor. In this sense, a possible avenue of research would be to identify which specific type of client career decision difficulties (e.g. chronic career indecision) lead to a lower working alliance, and whether career decision difficulties interact with client personality in predicting working alliance.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
