Volunteering is a planned behavior. In accordance with the Theory of Planned Behavior (Fishbein & Ajzen, 1975), many studies have shown that intention to remain is the best predictor of actual retention in both employees (Biron & Boon, 2013; Harhara et al., 2015; Rahman & Nas, 2013; Wells & Peachey, 2011) and volunteers (Chacón et al., 2007; Hyde & Knowles, 2013; MacGillivray & Lynd-Stevenson, 2013; Marta et al., 2014; Reuveni & Werner, 2015; Veludo-de-Oliveira et al., 2013; White et al., 2017).

Since the 1990s, various explanatory models on volunteers’ retention and intention to remain have been developed (Chacón et al., 2007; Finkelstein et al., 2005; Omoto & Snyder, 1995). Most explanatory variables included in these models are personal or individual factors. Only a few of the variables are organizational variables, such as integration into the organization (Omoto & Snyder, 1995), satisfaction with the management of the organization, and organizational commitment (Alfes et al., 2015; Chacón et al., 2007).

The lack of studies on organizational variables in the field of volunteering contrasts with the numerous studies that analyze the relationship between these types of variables and the tenure of employees in companies. As reviews of the literature have shown, organizational variables significantly influence employees’ real retention and intention to remain (Yahaya & Ebrahim, 2016).

Leadership is one of the most influential organizational factors in employees’ intention to remain, absenteeism, turnover intention, satisfaction and organizational commitment (Allen & Mueller, 2013; Elpers & Westhuis, 2008; Kelloway et al., 2012; Nesbit & Gazley, 2012). However, as with other organizational variables, there are few studies that address the relationship between volunteer coordinators’ leadership style and volunteers’ retention (Smith, 2017). Extant studies also focus on very specific leadership styles: ethical leadership (Benevene et al., 2018), autonomy-supportive leadership (Oostlander et al., 2014), and servant leadership (Hines, 2017; Schneider & George, 2011).

One of the most studied leadership styles in multiple fields is transformational leadership. Bass (1985) introduced a leadership theory based on Burns’ (1978) concept of transformational leadership. Transformational leaders are highly inspiring and motivating for followers, helping them to meet higher performance targets. Moreover, transformational leaders serve as change agents who articulate an organization’s vision, create employee awareness of problems within organizations, challenge the status quo, and inspire and motivate followers to be innovative in achieving their enormous potential (Lussier & Achua, 2012).

Much of the leadership research shows that the transformational leadership style is one of the most effective in increasing employees’ intention to remain in profit-making organizations (Tse & Lam, 2008; Wells & Peachey, 2011). If we generalize these results to non-profit organizations, we can infer that transformational leadership attributes will be stimulating and influence volunteers’ intention to remain.

Thus, the first hypothesis of this study is:

H1: The transformational leadership style predicts volunteers’ intention to remain.

The literature on the corporate sector has shown that job satisfaction and organizational commitment influence employees’ turnover intentions (Chan et al., 2010; Wells & Peachey, 2011; Yucel, 2012; Zimmerman & Darnold, 2009). Identity also seems to have a high predictive value on employees’ drop-out intention (Van Knippenberg et al., 2007).

Many studies have found that transformational leadership predicts satisfaction, organizational commitment, and identity of employees (Abouraia & Othman, 2017; Braun et al., 2013; Dwyer et al., 2013; Nguni et al., 2006). Moreover, in relation to transformational leadership and employees’ intention to remain, satisfaction, organizational commitment, and role identity have been found to have significant mediation effects (Cheng et al., 2016; Gyensare et al., 2016; Wells & Peachey, 2011).

In volunteering, Chacón et al. (2007) developed the three-stage model, suggesting that the best predictors of volunteers’ intention to remain are satisfaction, organizational commitment, and role identity. The relationships between variables proposed by this model have been empirically tested (Fuller, 2011; Hyde et al., 2016; Larkin, 2015; Malinen & Harju, 2017; Miller et al. 2014; Vecina & Chacón, 2017). However, the role of transformational leadership as an antecedent of satisfaction, organizational commitment, and role identity, and the possible indirect effect of transformational leadership on volunteers’ intention to remain through these three variables, have not been sufficiently investigated.

The second objective of this study is to verify, using a sample of volunteers, whether the relationship between transformational leadership and intention to remain is mediated by satisfaction, organizational commitment, and role identity.

The following hypothesis was formulated to achieve this objective:

H2: Components of the three-stage model of volunteers’ duration (satisfaction, organizational commitment, and role identity) mediate the relationship between the transformational leadership style and volunteers’ intention to remain.

H2a: Satisfaction mediates the relationship between transformational leadership and intention to remain.

H2b: Organizational commitment mediates the relationship between transformational leadership and intention to remain.

H2c: Role identity mediates the relationship between transformational leadership and intention to remain.

Moreover, two hypotheses from the three-stage volunteering model will also be tested in this study:

H3: Satisfaction predicts organizational commitment.

H4: Organizational commitment predicts role identity.

Finally, data adjustment to the theoretical model that underlies these hypotheses as a whole will be tested (see Figure 1).

Data were collected from volunteers (N = 417) providing their services voluntarily in different non-profit organizations (N = 17) working mainly in health (36%), education (14%), environment (19%), social welfare (16%), psychological services (5%), sports (8%), and immigration (12%). Information regarding gender depicted a higher number of female participants (73%) than male participants (27%). Average age for the total sample was 44 (SD = 17.50), depicting an age range between 15 to 86 years old. Average time dedicated to volunteering was 3.5 hours per week (SD = 2.5, range 1-20), with an average experience of 5 years (SD = 4.5, range = 1-32). For further demographics details, see Table 1.

Response rate was 28%, a relatively low rate, but similar to that found in different studies for the method used in this study (Fernández et al., 2009).

With regard to sample representativeness, it is worth highlighting the heterogeneity of the different types of organization participating in this study, the fact that main areas of volunteering have been included (education, health, social, environmental, immigration, etc.), and similarity between socio-demographic age and gender data in the sample and those in the 2018 Report: portrait of volunteering in Spain by the Volunteering Platform in Spain (60% women and average age of 45 years).

The following instruments were used to gather information from participants.

General group characteristics. A checklist was included to examine socio-demographic characteristics of the sample. Respondents were asked to provide information regarding age, gender, education, hours dedicated to volunteering and experience in years working as a volunteer.

Multifactor Leadership Questionnaire (MLQ; Bass & Avolio, 1995). The Spanish version of MLQ, adapted by Molero et al. (2010), was used. This is a five-point Likert type scale with a total of 36 items. MLQ scales are scored as follows: 1 = not at all, 2 = once a while, 3 = sometimes, 4 = fairly often, 5 = frequently, if not always. In data analysis, 20 items were used for measuring transformational leadership. Examples of items include “My leader goes beyond self-interest for the good of the group.” Reliability, as measured by Cronbach’s alpha value, was .93.

Organizational Commitment Questionnaire (OCQ; Mowday et al., 1979). The OCQ version used for this study is a reduced version of 9 items adapted for samples of volunteers in the Spanish population by Dávila and Chacón (2003). It is a 7-point Likert scale with responses ranging from 1 for totally disagree to 7 for totally agree. Examples of items include “I worry about the future of this organization,” “My values and the values of the organization are very similar.” Reliability, as measured by Cronbach’s alpha, was .87.

Satisfaction. It was measured using three sub-scales containing 18 items measuring three types of satisfaction: task satisfaction, organizational satisfaction, and motivation satisfaction (Vecina et al., 2009) in a Likert-type response format, with choices ranging from 1 (not at all satisfied) to 7 (totally satisfied). An example of an item is “When I am working as a volunteer, the time just flies by.” Reliability, as measured by Cronbach’s alpha, was .84.

Role identity. In order to operationalize the role identity concept, a Spanish translated and adapted version of the scale by Grube and Piliavin (2000) was used. This instrument was adapted to Spanish samples by Dávila and Chacón (2004). It consists of five items, with examples including “Volunteerism is something I often think about” and “Volunteerism is an important part of my own identity” in a 7-point Likert-type scale with responses ranging from 1 (totally disagree) to 7 (totally agree). Reliability, as measured by Cronbach’s alpha, was .81.

Intention to remain. It was measured by a single item based on the theory of planned behavior from Ajzen (1985). The item asked volunteers specifically about the probability of continuing with their volunteer work within the organization after two years, in the scale of 1 to 7, where 1 meant zero probability and 7 meant maximum probability. Following Hayduk (1996) and Hayduk and Glaser (2000) it is assumed that a latent construct can be measured with a single observable indicator if that item is sufficiently representative of the underlying latent construct. Two-year intention was been selected because the effect of organizational commitment and role identity on long-term intention is greater than on short-term intention (Chacón et al., 2007).

Twenty non-profit organizations working mainly in the education, environment, social service, and health sectors were contacted to ask them to participate in the study. Consent to collect data was received from 17 of the organizations. The objectives of the study were explained to authorities and volunteers in question. Specific online web pages were developed for 13 non-profit organizations included in this study. Volunteers completed an anonymous online questionnaire hosted on specific web pages (approximate time for completing the questionnaire was 10-15 minutes). Rating forms were delivered by hand and by mail to the remaining four organizations for whom it was not appropriate to develop web links.

Data analyses included correlation analysis, multiple regression analysis and analysis of structural equations (path analysis). Data analyses were performed with the SPSS-AMOS program version 25.

Results of correlations between all the study variables are shown in Table 2. Most of the variables have reliabilities close to or greater than .80.

As can be seen from Table 2, all the relationships between the study variables are correlated significantly, with moderate to high levels of correlation. This allows us to continue to the subsequent multiple regression analysis. In addition, the magnitude and signs of the correlations coincide with what was expected given our hypotheses. Main socio-demographic variables (age, gender, and type of organization) were also included in a first analysis of correlations and in a first multiple regression analysis, to check and control their possible significant effect on relationships and predictions of main variables in the study. However, non-significant results were found between socio-demographic variables and main variables in all cases.

A two-stage multiple regression analysis was performed (see Table 3). This followed the hypotheses in which we expect to find a significant predictive effect for transformational leadership in relation to intention to remain (model 1) and a mediating effect for the variables in the three-stage model (satisfaction, organizational commitment, and role identity), between transformational leadership and intention to remain (model 2).

As observed in the results shown in Table 3, in addition to the already discussed result of the significant predictive effect of transformational leadership on the intention to remain (β = .195, p < .001), there are indications of the possible mediating effect of role identity (partial reduction in model 2 of the standardized beta between transformational leadership and intention to remain). Predictive effects of satisfaction and organizational commitment on intention to remain are not significant.

An analysis of structural equations, specifically through a path-analysis (Duncan, 1975) was carried out using the maximum likelihood method. The model was designed to jointly test the role of transformational leadership, satisfaction, organizational commitment, role identity, and volunteers’ intention to remain (see Figure 2).

The results of path analysis (Figure 2) shows that transformational leadership predicts intention to remain directly (β = .13, p < .01) but also indirectly through the mediating impact of satisfaction (β = .41, p < .001). The results also indicate that satisfaction predicts organizational commitment, and organizational commitment together with the influence of satisfaction predicts role identity. However, among the three components of the three-stage model, only role identity directly predicts intention to remain and proves to be a relatively more reliable predictor (β = .31, p < .001). This model explains 15% of the variance of intention to remain. The goodness of fit indices for the model measuring the best acceptable range for CFI, NFI, GFI is > .90, and for RMSEA is <.06 (Steiger & Lind, 1980). All indices for the model are adequate (p < .05), affirming that the model fits the data (Table 4). Table 5 shows all the total, direct, and indirect effects of the statistical model being tested.

Our research findings have considerable practical implications for non-profit organizations. At a time when non-profit organizations are finding it a challenge to attract and retain volunteers (Benevene et al., 2018; Hidalgo & Moreno, 2009; Taylor & McGraw, 2006), the current study highlights measures that leaders or those in managerial positions in non-profit organizations can employ to ensure that their volunteers stay. For instance, in-time facilitation of volunteers after recruitment into their role is essential. Non-profit organizations should invest in training volunteers by carrying out induction programs, as suggested by Saksida et al. (2017). As part of the training, group meetings are helpful where leaders can play their role while introducing volunteers and explaining organization’s mission to them. Moreover, introductory and skill enhancement workshops are ideal for newly recruited and experienced volunteers in order to develop communication and teamwork. The idea is that such meetings and training events should not be limited only to the start of the project. In contrast, brief volunteer training sessions are helpful up to the end of the project to deal with issues that may arise and to clarify any ambiguities during volunteer work. Newton et al. (2014) also support the idea that ongoing training and professional development could influence volunteers’ decisions to stay with the volunteer organization.

Furthermore, the leadership role carried out properly is essential in energizing, motivating, and improving volunteers. The fact that an organization’s management takes into account volunteers’ concerns results in greater satisfaction among volunteers and a strengthening of their commitment to the organization. This leads to role identity, which ensures their retention in the organization. In addition, retaining volunteers is the goal of successful volunteer management.

Our findings show that the role of leadership is a significant component in managing and retaining volunteers. While taking advantage of this finding, non-profit organizations may wish to reconfigure their human resources management (HRM) departments to include leadership positions that will be responsible entirely for managing volunteers within an organization. This is significant because volunteers are typically not managed strategically within the HRM function (Saksida et al., 2017). Instead, their management is distributed across such functions as marketing, fundraising and campaigning, administration, and service delivery (Brewster & Lee, 2006). As a result, leadership may assume a more central role in volunteer management, treating volunteers as valuable and essential members of the organization. This may help develop their role identity through gleaning satisfaction and enhance their commitment, which in turn will strengthen their intention to remain as well as their actual behavior.

Despite its contribution in the context of volunteering, this study suffers from a few limitations. Firstly, the cross-sectional nature of the study and the self-reporting nature of data collection process may lead to a risk of bias due to common method variance. However, following recommendations by Podsakoff et al. (2003), and similarly to the study by Saksida et al. (2017), the following actions were carried out to minimize this slant: the procedure was clearly explained to participants, anonymity and confidentiality were ensured, and widely validated scales were used. Furthermore, confirmatory factor analysis could be performed in future cross-validation studies (Conway & Lance, 2010).

Secondly, a relatively low response rate is another limitation of the current study. Various factors could contribute to a web-based survey’s low response rate, such as less information regarding respondents’ computers or the feasibility of using webpages relating mainly to adults (Fan & Yan, 2010). Unlike the method used in this study, when instruments are sent in a personalized way, with incentives and weekly follow-ups, higher retention and response rates can be achieved (Fernández et al., 2009). For the purposes of this study, the rate achieved was considered adequate, although in subsequent cross-validations of the results of this research it will be necessary to achieve a higher rate.

Our sample also consisted mostly of adults of various ages volunteering in different non-profit organizations (e.g., health, education, environment, sports, social services). It was not feasible to identify how comfortable participants were by using web browsers and navigating between or within websites. However, we did not find evidence of non-response bias, and the study sample was representative of the volunteer population in this particular domain. Nevertheless, more studies in the future may address this issue by employing more demographically diverse samples to examine phenomena under study here.

This study represents an advance in the current understanding of volunteers’ sustained prosocial behavior. Findings may further assist the establishment of effective human resource management strategies in volunteering programs. The underpinning contribution of transformational leadership in the retention of volunteers seems to be significant. In general, our results indicate that leaders of non-profit organizations, particularly those with a transformational leadership style, predict the intention of volunteers to stay while influencing their level of satisfaction. Moreover, to achieve sustained volunteerism, volunteers’ socialization into their roles within a non-profit organization is essential. Unless and until volunteer role becomes part of a volunteer’s identity over the years through deriving satisfaction and commitment to the organization in which they work, there will be fewer chances for volunteer behavior to continue for a long time into the future.