Daily receiving and providing of social support at work: identifying support exchange patterns in hierarchical data

ABSTRACT

The norm of reciprocity proposes that one who receives support feels obliged to return obtained benefits. Yet how employees regulate the mutual flow of social support with one another does not always follow a reciprocal dynamic, which may have different implications for employee outcomes based on whether social exchanges gain, drain or conserve resources. To better understand support exchange patterns and their relationship with basic need satisfaction (for autonomy, competence and relatedness) and emotional exhaustion in employees’ daily lives, we performed a multilevel latent profile analysis (N = 124 employees, 894 days). At the day level, we identified four support exchange profiles differing in low vs high received and provided social support (i.e., low-reciprocating, under-reciprocating, over-reciprocating and high-reciprocating days). At the person level, we identified three support exchange classes of employees, differing in the degree to which employees showed certain support exchange profiles over time (i.e., over-reciprocators, low-reciprocators and high-reciprocators). Pairwise comparisons with multinomial logistic regression revealed that over-reciprocating days were associated with the highest levels of need satisfaction for autonomy and relatedness and the lowest levels of emotional exhaustion. Moreover, over-reciprocators were most likely to satisfy their needs for autonomy. We discuss theoretical and practical contributions of our findings.

KEYWORDS:

• Social support
• reciprocity
• COR theory
• diary study
• multilevel latent profile analysis

Introduction

Employees participate in and contribute to their workplace by turning to each other for resources. Social support is a widespread, everyday job resource that is associated with a range of positive outcomes for individuals and organizations alike (Jolly et al., 2020). Having supportive colleagues affects employees’ motivation and well-being in the workplace, as it contributes to the satisfaction of basic needs for autonomy, competence and relatedness (Van den Broeck et al., 2016), and to the prevention of burnout (Halbesleben, 2006; Mathieu et al., 2019). Through the availability of social support, employees seek to acquire the resources they are lacking and thereby increase their existing resource pool (Hobfoll, 2001; Jolly et al., 2020). For example, when targeting a work-related problem, employees may exchange task-relevant and emotional resources in the form of practical assistance and sympathetic words, which facilitate experiences of volition, effectiveness and connection in interacting with their work environment.

While there is ample evidence that speaks to the positive role of social support at work, its consequences may not always be beneficial (Hobfoll et al., 2018). Consistent with the basic tenets of conservation of resources theory (COR; Hobfoll, 1989), individuals are motivated to gain and protect resources for which they must, however, potentially invest resources. In other words, employees engage in a strategic resource investment process by providing support in order to receive support (Halbesleben & Wheeler, 2015). This dynamic of how employees exchange resources oftentimes follows the norm of reciprocity (Gouldner, 1960). Reciprocity, as outlined in social exchange theory (Blau, 1964), can be understood as a normative rule within exchange relations that mandates that, over time, all exchanges reach a fair balance between give and take (Cropanzano & Mitchell, 2005). Past research indicates that support exchanges at work can follow a pattern that is perceived as reciprocating (i.e., support received equals support provided), under-reciprocating (i.e., support provided exceeds support received) or over-reciprocating (i.e., support received exceeds support provided), with differing effects on employee outcomes depending on whether support exchanges evoke feelings of being under- or over-benefitted (Bowling et al., 2005; Buunk et al., 1993; Nahum-Shani et al., 2011; Tomprou et al., 2020). In particular, exchanges that are perceived as under-reciprocating have been shown to elicit unfavourable effects in employee functioning, such as increased physical symptoms, because employees’ resource reservoirs become depleted when they invest resources but do not experience any reciprocation in return (Tomprou et al., 2020).

Investigating empirically-occurring combinations of receiving support and providing support helps one to understand which patterns of support exchanges exist in employees’ daily work lives and how these joint experiences of receiving and providing support are associated with employee motivation and well-being. We draw on the theoretical idea that support exchange patterns can be conceptualized as resource gain, resource drain or resource conservation (Hobfoll, 1989; Tomprou et al., 2020). Depending on whether one’s resource pool increases, decreases or is conserved as a result of the conjunction of receiving and providing support, different combinations – which naturally occur in the form of day-level profiles among the variables of interest that give rise to person-level classes – should have different effects on work-related outcomes. Basing on central tenets of COR (Hobfoll, 1989) and self-determination theory (Deci et al., 2001), we propose that receiving co-worker support can promote and transform into another resource, such as higher levels of needs satisfaction, and can also help employees cope with stressful situations by reinforcing positive aspects of their self-image, thus leading to lower levels of emotional exhaustion (Halbesleben, 2006). We further contend that reciprocity is crucial for better understanding the daily dynamics of support exchange – outcome relationships (Gabriel et al., 2018; Halbesleben & Wheeler, 2011; Schaufeli, 2006). More precisely, we propose that different support exchange patterns exist, consisting of varying levels of daily receiving and providing social support. An exchange can be conceptualized as a beneficial experience if it follows reciprocity norms (i.e., resource conservation) and/or the employee is mainly the recipient of social support (i.e., resource gain); it can be a depleting experience if it violates reciprocity norms and the employee is mainly the provider of social support (i.e., resource drain).

To date, much research has been devoted to social support at the workplace from the recipient perspective (for a review, see Jolly et al., 2020), while little is known about the effects of providing social support (for an exception, see N. P. Podsakoff et al., 2009, who discuss organizational citizenship behaviours that indicate supportive actions towards colleagues) or the patterns of social support that is received and provided in employees’ work lives. Past studies that did look at support exchanges at work yielded important findings from a qualitative perspective (e.g., Nahum-Shani et al., 2011), applied cross-sectional methods (e.g., Bowling et al., 2004; Buunk et al., 1993) or focused on how support behaviour unfolds over longer time periods (e.g., Yang et al., 2018). Recently, Halbesleben and Wheeler (2015) investigated the reciprocal nature of perceived support, citizenship behaviour and trust across work days and recommended taking into account how helping behaviour fluctuates from one day to another and is thus best examined in these short cycled time frames for a better understanding of the daily receiving and providing of support among employees (Shrout et al., 2010; Xanthopoulou et al., 2008).

The present study contributes to research on workplace social support in the following ways. First, this study takes a new analytic look at social support and support exchange behaviour by combining the advantages of diary studies and so-called “person-oriented” (vs “variable-oriented”) methods to improve our conceptual understanding of the daily dynamics of receiving and providing support. Specifically, we investigate which day-specific combinations of different levels of receiving and providing support exist that generate meaningful typologies of employee behaviour. In other words, we answer the question of how specific days differ from others when it comes to receiving and providing support. Some days may be characterized by high- or low-reciprocating support exchanges, while others may be characterized by under- or over-reciprocating support exchanges (i.e., day-level profiles). This approach takes into account that receiving and providing support in the workplace do not occur in isolation, but may co-occur in employees’ everyday lives. To elaborate further, employees’ general propensity to engage in high-reciprocating, low-reciprocating, under-reciprocating, or over-reciprocating day-level profiles may vary over time (giving rise to person-level classes). Applying multilevel latent profile analysis (MLPA) to daily diary study data enables us to identify profiles of daily support exchanges (based on daily patterns of received and provided support) and to identify classes of employees who differ in their propensity to engage in certain daily support exchange profiles. We hereby follow directions put forward by other researchers, who encourage the use of “person-oriented” approaches that account for nested data, in order to address novel research questions in the study of human behaviour in organizational settings (Howard & Hoffman, 2018; Mäkikangas et al., 2018).

Second, the use of MLPA not only allows for the identification of patterns and subgroups (i.e., profiles and classes) based on the give and take of co-worker support, but it also allows for the linking of outcomes to profile and class membership (Mäkikangas et al., 2018). Thus, our approach enables us to broaden our understanding of the importance of everyday co-worker support by examining how day-level profiles and person-level classes of support exchange behaviours are associated with employees’ motivation (i.e., need satisfaction for autonomy, competence and relatedness) and well-being (i.e., emotional exhaustion) in the workplace. We can thus gain a better understanding of when the beneficial association of support exchanges with employee outcomes may turn detrimental by taking into account whether reciprocity norms are adhered to or are violated (Hobfoll et al., 2018; Van Veldhoven et al., 2017).

Finally, through a daily diary study design, we capture the nuanced interplay of social support received and provided and measure these behaviours closely to their occurrences, which minimizes recall bias and improves response accuracy (Ohly et al., 2010). Moreover, daily assessments avoid the shortcomings associated with global trait self-report measures of employees’ behaviours and experiences, thereby increasing ecological validity (Grommisch et al., 2020).

Theoretical background: Social support and support exchange behaviour

Considerable empirical evidence substantiates the notion that social support is an important element in creating a resourceful work environment that activates processes impacting critical outcomes such as employee well-being, job attitudes and work relationships (Schaufeli & Bakker, 2004; Xanthopoulou et al., 2009). Accordingly, being able to rely on colleagues for help with a task or venting one’s emotions in front of them, is considered functional and stimulating, which is why employees may be especially motivated to obtain social resources by returning supportive actions (Mathieu et al., 2019).

This idea is in line with COR (Hobfoll, 1989) and its theoretical advances on social support (Hobfoll et al., 1990), which posit that people actively strive to gain personal and social resources and to avoid their losses. Having a strong resource pool is desirable because individuals who are rich in valued resources are in a better position to respond to demanding situations and even become richer in resources, as possessing resources serves to protect other resources (Hobfoll et al., 1990). Following this theorization, social support is not only valuable from an instrumental and emotional point of view, but it is considered necessary to assure people’s sense of identity based on the social attachments related to it (Hobfoll et al., 1990). Hence, in the realm of the workplace, social support is a vital resource that is based within interpersonal relationships at work, and while it can replace other resources, its availability alone may already be important in indicating the ability of employees to draw upon colleagues in order to gain further resources (Jolly et al., 2020).

More recently, however, the truism of workplace social support as being uniformly beneficial has been challenged, especially when considering the role of providing support. While providers of social support may “feel good”, as indicated by increased positive affect, they may not “do good” when it comes to accomplishing work tasks, which further influences their daily well-being at work (Koopman et al., 2016). Joining in questioning the prevailing positive view of helping behaviour at work, Gabriel et al. (2018), as well as Lanaj et al. (2016), have identified the provision of help as a behaviour that depletes resources and thus bears potential costs to both employees’ well-being and work performance. From the standpoint of COR (Hobfoll, 1989), providing social support can be understood as an act that includes an investment of resources by one person in another; hence helping co-workers somewhat depicts a resource-draining experience, as the trade-off that comes with helping potentially consumes resources such as time and effort (Koopman et al., 2016).

Buunk et al. (1993) have already pointed out that the general pattern of support exchanges needs to be taken into account in order to fully grasp the quite complex nature of helping behaviour. That is, receiving social support occurs within the broader context of workplace relationships, where employees also “give back” to colleagues in need by providing social support (Nahum-Shani & Bamberger, 2011). According to social exchange theory and, more specifically, the norm of reciprocity (Gouldner, 1960), social support is argued to be especially beneficial for employee outcomes when support exchanges are perceived as reciprocating because work relationships are considered to be primarily exchange relationships that are expected to follow an equitable support pattern (Buunk et al., 1993). In contrast, support exchanges that are perceived as under-reciprocating or over-reciprocating may induce feelings of resentment or guilt, respectively (Buunk et al., 1993; Nahum-Shani et al., 2011). From the perspective of COR (Hobfoll, 1989), resource-gaining, resource-draining and resource-conserving considerations are inherent to the performance of helping behaviour and associated work-related outcomes. We propose that there are benefits for employees’ motivation (Liu et al., 2020) and well-being (Schaufeli, 2006) when support exchanges are perceived as low-reciprocating, high-reciprocating and over-reciprocating, meaning that resources are either conserved or gained (resulting in higher levels of need satisfaction for autonomy, competence and relatedness and lower levels of emotional exhaustion). However, there also may be certain costs for employees’ motivation and well-being when support exchanges are perceived as under-reciprocating, meaning that resources are drained (resulting in lower levels of need satisfaction for autonomy, competence and relatedness and higher levels of emotional exhaustion).

A new analytic approach to support exchange behaviour

Given the lack of studies based on a “person-oriented” approach to support exchange behaviour in employees’ daily lives, our first goal was to address which support exchange behaviours are typically shown by employees on a day-to-day basis. Based on the theoretical and empirical accounts delineated above, we propose that different support exchange patterns emerge when accounting for different levels of receiving and providing social support. Building on the idea that receiving and providing social support are two constructs that are exhibited independently at the day level, for simplicity and illustrative purposes we have assumed a two-by-two matrix of low vs high levels of receiving social support and low vs high levels of providing social support (see Figure 1). Days on which employees not only receive considerable amounts of assistance but also provide help to colleagues would result in a profile high in both constructs (i.e., high day-level receiving support and high day-level providing support results in a high-reciprocating day). At the other extreme, some days could be characterized by negligible amounts of receiving and providing social support, which would reflect a profile with low levels of both indicators (i.e., low day-level receiving support and low day-level providing support results in a low-reciprocating day). Furthermore, days could reveal unequal amounts of support received and provided (i.e., low day-level receiving support and high day-level providing support results in an under-reciprocating day, while high day-level receiving support and low day-level providing support results in an over-reciprocating day). Thus, we asked:

Figure 1. Possible support exchange patterns emerging from different combinations of low vs high receiving and providing support.

Research question 1: Do distinct profiles of days (day-level profiles) based on varying levels of receiving and providing social support exist?

Furthermore, we sought to explore whether employees differ as regards to their propensity to engage in certain day-level profiles over time. We expect that subgroups of employees (termed classes) can be identified based on their overall tendency to engage in certain day-level profiles. For example, a class of employees may predominantly show high-reciprocating days over time, while another class of employees may be characterized by predominantly showing another day-level profile or a diversity of (i.e., mix of different) day-level profiles over time. Accordingly, our second research question asked:

Research question 2: Do distinct classes of employees (person-level classes) based on the proportion of different day-level support exchange profiles exist?

Support exchange behaviour, need satisfaction and emotional exhaustion

Based on the idea that social support both enables psychological and social functioning and mitigates stress reactions (Jolly et al., 2020), in the following, we propose that day-level support exchange profiles are and membership in person-level classes is related to basic need satisfaction and emotional exhaustion. From the perspective of self-determination theory (Deci et al., 2001), an optimal workplace is an environment that can promote the satisfaction of employees’ psychological needs for autonomy, competence and relatedness (Williams et al., 2014). Some have argued that co-worker social support is important for meeting each need directly (Knight et al., 2017). Through co-workers helping each other, new and appropriate ways of working can be collaboratively established, thus creating a sense of control over one’s behaviour and job (need satisfaction for autonomy). Further, a supportive work environment that includes constructive feedback from others about work-related issues may promote learning and development and, thus, a sense of being able to act effectively (need satisfaction for competence). Finally, exchanging social support increases opportunities to learn from and about each other, which helps build work relationships and enhances the sense of belonging among colleagues (need satisfaction for relatedness).

Given that the norm of reciprocity acts as a regulating rule in support exchange relationships, whether an employee experiences equal or unequal support exchanges should affect need satisfaction differently. Supporting this assumption, Liu et al. (2020) have found that resource exchanges characterized by equal transactions and immediate returns (i.e., resource gain) enhance need satisfaction, whereas exchanges reflecting self-interest at one’s expense (i.e., resource drain) attenuate need satisfaction. While they relied on a need satisfaction measure that merged all three needs into one (i.e., basic psychological need satisfaction), we investigate whether nuances in the level of need satisfaction for autonomy, competence and relatedness, respectively, exist.

Furthermore, support relationships at work are considered important for employee well-being (Halbesleben & Wheeler, 2011) and are associated with reduced vulnerability to burnout (Russell et al., 1987). Being able to rely on task-related and emotional assistance from colleagues when needed is considered a protective determinant of emotional exhaustion (Baeriswyl et al., 2017). Emotional exhaustion is a state of feeling emotionally drained and over-extended at work (Fernet et al., 2013). Accordingly, when considering the dynamics of helping behaviour that may represent resource-gaining, resource-draining or resource-conserving experiences, these combinations of providing and receiving support may have differing relations with employee outcomes. This is because, from a theoretical point of view, COR predicts that a greater resource pool leads to more resources and less vulnerability to stressors (Hobfoll, 1989; Salanova et al., 2010). In the context of the current study, resources should be more likely to be maintained or gained on resource-conserving and resource-gaining days; in contrast, resources should be more likely to be lost on resource-draining days. We thus propose that days characterized by resource-gaining and resource-conserving experiences are associated with higher levels of need satisfaction for autonomy, competence and relatedness and lower levels of emotional exhaustion. In contrast, days characterized by resource-draining experiences should be associated with lower levels of need satisfaction for autonomy, competence and relatedness and higher levels of emotional exhaustion. Thus, we considered the following third research question:

Research question 3: Do day-level profiles of support exchange behaviour differentially relate to day-level need satisfaction (for autonomy, competence, relatedness) as well as emotional exhaustion?

While our third research question asks for outcomes that differentiate day-level profiles, we also wanted to explore whether these outcomes would differentiate classes of people characterized by their propensity to engage in certain support exchange profiles that we assumed to exist (i.e., what consequences are associated with engaging more often in certain day-level profiles). Mirroring the reasoning above, we propose that employees with higher relative frequencies of resource-gaining and resource-conserving days should experience higher levels of need satisfaction for autonomy, competence and relatedness, as well as lower levels of emotional exhaustion. And vice versa, employees with higher relative frequencies of resource-draining days should experience lower levels of need satisfaction for autonomy, competence and relatedness, as well as higher levels of emotional exhaustion. This leads to our final research question:

Research question 4: Do person-level classes of support exchange behaviour differentially relate to person-level need satisfaction (for autonomy, competence, relatedness) as well as emotional exhaustion?

Method

Sample and procedure

This study was part of a larger project with the objective to assess the social context of employees both inside and outside the workplace (Patterer et al., 2021). To investigate our research questions proposed above, we conducted a daily diary study over a period of ten consecutive working days, following the recommendations offered by Gabriel et al. (2019). Participants were recruited through a convenience sampling approach and comprised family members, friends and acquaintances of the authors and their students, who were mainly based in Austria and Germany. To be eligible for this study, participants had to be employed and had to work a minimum of 20 hours per week. As an incentive, participants could take part in a lottery of gift vouchers for an online retailer.

A brief on the study purpose and a general questionnaire asking for sociodemographic data were sent out to participants. Upon completion, respondents could register for the daily surveys and indicate their preference to receive the questionnaire links via email or text message. The daily measurements consisted of two online questionnaires: after work and in the evening. In the after-work questionnaire (5 p.m.), participants were asked to rate their social support received from co-workers as well as experienced need satisfaction for autonomy, competence and relatedness while at work. In the evening questionnaire (9 p.m.), participants had to rate their social support provided to co-workers and emotional exhaustion with reference to the working day. We wanted to keep the daily questionnaires as short and balanced as possible across all measurement occasions in order to reduce participants’ burden. Each day, online questionnaires were accessible for a few hours for that day only, so that daily entries could not be filled in at wrong times or on wrong days (e.g., evening questionnaire on the next day). To match participants’ questionnaires, participants created a personal code based on answers to four questions that were not expected to change (e.g., the last two digits of their birth years). Participants had to enter this code to begin each entry. The personal code allowed the matching of participants’ responses without revealing participants’ identities.

The final sample of our study comprises 124 employees, each of which provided data on at least two working days (Mdn = 8 days). The final data set included 894 of 1,240 possible daily records (72% completion rate). Among the participants, 51% were male, with a mean age of 35.45 years (SD = 11.31, ranging from 22 to 62 years). Working hours per week was M = 39.67 hours (SD = 9.25), and average duration of employment was M = 7.56 years (SD = 8.33). Most of the participants completed university (68%) or had a higher education degree (26%). The largest occupational sectors in which participants worked were service (64%), followed by information and communication (20%) and industry (16%). The majority of participants had no managerial responsibilities (74%).

Measures

The questionnaires were administered in German, and all items were rated on a 7-point scale ranging from 1 “not at all” to 7 “very much”. All items were adapted to refer to the current working day.

Receiving social support

Social support that was received was measured using six items from the support appraisal for work stressors inventory (SAWS) by Lawrence et al. (2007). The items we selected represent emotional and instrumental support received from colleagues, which, following a recent meta-analytic work by Mathieu et al. (2019), were found to best reflect common support behaviours in the workplace. The following is an example item for emotional support received: “Today, how much did your colleagues help you to feel better when you experienced work-related problems?”. The following, meanwhile, is an example item for instrumental support received: “Today, how much did your colleagues give you practical assistance when you experienced work-related problems?”. The average within-level Cronbach’s alpha was .92 (SD = 0.04, Min = .84, Max = .96), and the between-level Cronbach’s alpha was .97.

Providing social support

Social support that was provided was measured using six items from the 2-way social support scale (2-way SSS) by Shakespeare-Finch and Obst (2011). The items were adapted to match the working context and represent emotional and instrumental support given to colleagues. The following is an example item for emotional support provided: “Today, at work, I was there to listen to my colleagues” problems’. The following is an example item for instrumental support provided: “Today, at work, I helped colleagues with their tasks when they were unable to complete them”. The average within-level Cronbach’s alpha was .87 (SD = 0.04, Min = .79, Max = .92), and the between-level Cronbach’s alpha was .93.

Emotional exhaustion

Emotional exhaustion was measured using three items from the Maslach Burnout Inventory (MBI) by Maslach and Jackson (1981). The following is an item example of this: “Today, I felt emotionally drained from my work”. The average within-level Cronbach’s alpha was .93 (SD = 0.03, Min = .89, Max = .98), and the between-level Cronbach’s alpha was .96.

Basic need satisfaction at work

Basic need satisfaction at work was measured using three items each for autonomy, competence and relatedness from the basic need satisfaction at work scale by Deci et al. (2001). The following are example items of need satisfaction for autonomy, competence and relatedness, respectively: “Today, I felt like I could pretty much be myself at work”, “Today, I felt a sense of accomplishment from working” and “Today, I got along with people at work”, respectively. The average within-level Cronbach’s alpha for need satisfaction for autonomy, competence and relatedness was .79 (SD = 0.04, Min = .71, Max = .89), .65 (SD = 0.05, Min = .56, Max = .75), and .79 (SD = 0.03, Min = .75, Max = .82), respectively. The between-level Cronbach’s alpha for need satisfaction for autonomy, competence and relatedness were .78, .64 and .80, respectively.

Analytic strategy

Nonparametric multilevel latent profile analysis (MLPA; Asparouhov & Muthén, 2008; Mäkikangas et al., 2018; Vermunt, 2008) was conducted to identify latent subpopulations on the day level (i.e., level 1: profiles) and latent subpopulations on the person level (i.e., level 2: classes). MLPA extends single-level latent profile analysis (LPA; Masyn, 2013) for independent observations to observations stemming from nested data structures (e.g., days nested in employees) by allowing latent class intercepts to vary across employees (i.e., random means).

The MLPA with latent classes at level 2 is based on the relative frequencies of level 1 profiles. Hence, it is recommended that one conduct single-level LPA prior to running the MLPA model (Mäkikangas et al., 2018). Accordingly, we first estimated single-level LPAs with k = 2 to k = 6 classes in order to identify the optimal number of latent profiles at the day level. LPA was conducted for six different within-profile variance-covariance structures (see Table 1), resulting in 5 (k = 2 to k = 6 classes) × 6 (within-profile variance-covariance structures) = 30 latent profile models. The within-profile variance-covariance structures represent different assumptions regarding the variance and covariance of the indicators both within and between latent profiles (see Table 1). As the best within-profile variance-covariance structure is not known a priori, all of the different structures must be tested to identify the best fitting model (Masyn, 2013).

Table 1. Within-profile variance-covariance structures.

Model	${\mathrm{\Sigma }}_k$
A: Profile Invariant ${\theta }_{mm}$, diagonal ${\mathrm{\Sigma }}_k$	$\left[\begin{array}{cccc}{\theta }_{11}& \hspace{0.17em}& \hspace{0.17em}& \hspace{0.17em}\\ 0& {\theta }_{22}& \hspace{0.17em}& \hspace{0.17em}\\ 0& 0& \ddots & \hspace{0.17em}\\ 0& 0& 0& {\theta }_{MM}\end{array}\right]$
B: Profile Varying ${\theta }_{mm}$, diagonal ${\mathrm{\Sigma }}_k$	$\left[\begin{array}{cccc}{\theta }_{11k}& \hspace{0.17em}& \hspace{0.17em}& \hspace{0.17em}\\ 0& {\theta }_{22k}& \hspace{0.17em}& \hspace{0.17em}\\ ⋮& ⋮& \ddots & \hspace{0.17em}\\ 0& 0& \cdots & {\theta }_{MMk}\end{array}\right],\mathrm{\forall }k\in \left(1,\dots ,K\right)$
C: Profile Invariant ${\theta }_{mm}$, Class Invariant Unrestricted ${\mathrm{\Sigma }}_k$	$\left[\begin{array}{cccc}{\theta }_{11}& \hspace{0.17em}& \hspace{0.17em}& \hspace{0.17em}\\ {\theta }_{21}& {\theta }_{22}& \hspace{0.17em}& \hspace{0.17em}\\ ⋮& ⋮& \ddots & \hspace{0.17em}\\ {\theta }_{M1}& {\theta }_{M2}& \cdots & {\theta }_{MM}\end{array}\right]$
D: Profile Invariant ${\theta }_{mm}$, Class Varying Unrestricted ${\mathrm{\Sigma }}_k$	$\left[\begin{array}{cccc}{\theta }_{11}& \hspace{0.17em}& \hspace{0.17em}& \hspace{0.17em}\\ {\theta }_{21k}& {\theta }_{22}& \hspace{0.17em}& \hspace{0.17em}\\ ⋮& ⋮& \ddots & \hspace{0.17em}\\ {\theta }_{M1k}& {\theta }_{M2k}& \cdots & {\theta }_{MM}\end{array}\right],\mathrm{\forall }k\in \left(1,\dots ,K\right)$
E: Profile Varying ${\theta }_{mm}$, Class Invariant Unrestricted ${\mathrm{\Sigma }}_k$	$\left[\begin{array}{cccc}{\theta }_{11k}& \hspace{0.17em}& \hspace{0.17em}& \hspace{0.17em}\\ {\theta }_{21}& {\theta }_{22k}& \hspace{0.17em}& \hspace{0.17em}\\ ⋮& ⋮& \ddots & \hspace{0.17em}\\ {\theta }_{M1}& {\theta }_{M2}& \cdots & {\theta }_{MMk}\end{array}\right],\mathrm{\forall }k\in \left(1,\dots ,K\right)$
F: Profile Varying ${\theta }_{mm}$, Class Varying Unrestricted ${\mathrm{\Sigma }}_k$	$\left[\begin{array}{cccc}{\theta }_{11k}& \hspace{0.17em}& \hspace{0.17em}& \hspace{0.17em}\\ {\theta }_{21k}& {\theta }_{22k}& \hspace{0.17em}& \hspace{0.17em}\\ ⋮& ⋮& \ddots & \hspace{0.17em}\\ {\theta }_{M1k}& {\theta }_{M2k}& \cdots & {\theta }_{MMk}\end{array}\right],\mathrm{\forall }k\in \left(1,\dots ,K\right)$

Note. ${\theta }_{mmk}$ is the variance of the latent profile indicator variable $m$ in Profile $k$ and ${\theta }_{mjk}$ is the covariance between the latent profile indicator variable $m$ and $j$ in Profile $k$; In Model A, the within-profile variance is constrained to be profile-invariant and covariances are constrained to be 0 in all profiles (i.e., equal variances across profiles and no covariances among indicator variables); In Model B, the within-profile variance profile-varying and covariances are constrained to be 0 in all profiles (i.e., unequal variances across profiles and no covariances among indicator variables); In Model C, the within-profile variance is constrained to be profile-invariant and covariances are constrained to be equal in all profiles (i.e., equal variances and covariances across profiles); In Model D, the within-profile variance is constrained to be profile-invariant and covariances are profile-varying (i.e., equal variances across profiles and unequal covariances across profiles); In Model E, the within-profile variances are profile-varying and covariances are constrained to be equal in all profiles (i.e., unequal variances across profiles and equal covariances across profiles); In Model F, the within-class variance and covariances both profile-varying (i.e., unequal variances and covariances across profiles).

Statistical indicators, as well as theoretical considerations, were combined to select the optimal latent profile model from the 30 models. As statistical indicators, we used the Akaike Information Criterion (AIC), Consistent Akaike Information Criterion (CAIC), Bayesian information criterion (BIC), and the sample-size-adjusted Bayesian Information Criterion (SABIC). Note that our main criterion is the BIC which is recommended for continuous data and for deciding on the number of latent profiles. In addition, we used the Lo-Mendell-Rubin likelihood ratio test, bootstrapped likelihood-ratio test and the entropy value as a measure of classification accuracy. As content-related indicators, we relied on the principle of parsimony, which states that the more parsimonious solution should be selected if the additional profile in a k profile model represents only a slight variation on a profile found in a k − 1 profile model, along with theoretical considerations and the interpretability considerations for the classes.

In the second step, MLPA with k = 1 to k = 4 classes on the person level was conducted based on the selected latent profile models at the day level (see Figure 2). In order to retain the identified latent profiles at the day level, final parameter estimates from the single-level LPA were fixed for the measurement model parameters of the MLPA. The number of latent classes at the person level was determined using the group-based BIC (see Lukočienė et al., 2010), in conjunction with content-related indicators (parsimony principle, theoretical considerations and the interpretability of the classes). Additionally, AIC, group-based SABIC and group-based CAIC were assessed. In the last step, we included outcome variables (i.e., need satisfaction for autonomy, competence, relatedness, and emotional exhaustion) on the day level and on the person level to test for mean differences between latent profiles (i.e., day level) and latent classes (i.e., person level) on the outcome variable.

Figure 2. Path diagram for the multilevel latent profile model with need satisfaction for autonomy, competence, relatedness, and emotional exhaustion regressed on the categorical latent variable CW at the day level and the categorical latent variable CB at the person level. Note. The filled circles represent random intercepts for Profiles 1, 2, and 3 of the categorical latent variables CW, which has four latent profiles. The random intercepts are referred to as CW#1 to CW#3 on the person level.

We conducted all analyses using the statistical software Mplus Version 8.4 (Muthén & Muthén, 1998–2017) and relying on the maximum likelihood estimation method with robust standard errors (MLR). Five hundred random sets of starting values with 50 initial stage iterations and 50 final stage optimizations were requested. In the case of model non-convergence, the random set of starting values, initial stage iterations and final stage optimizations were gradually increased until model convergence or the maximum number of starting values (100,000) was reached. To ensure that the estimation process found the global solution, we inspected the results to check whether the highest log-likelihood was replicated in at least two final stage optimizations.

Results

Single-level latent profile analyses for days of support exchange behaviour

As can be seen in Table 2, results of the single-level LPA showed that models based on variance-covariance structures A, B, C, D and F converged with the highest log-likelihood value replicated. Models based on variance-covariance structure E, however, showed convergence problems and were discarded from further consideration. The Lo-Mendell-Rubin likelihood ratio test was statistically not significant for most of the models based on variance-covariance structures B and F. In addition, these models often resulted in relatively small classes (n = 8, 11 or 12). Hence, these models were discarded from further consideration. In the remaining three variance-covariance structures, BIC indicated that the model with five latent classes in variance-covariance structure A, the model with four latent profiles in variance-covariance structure C, and the model with three profiles in variance-covariance structure D should be chosen. The three-profile model based on variance-covariance structure D had the lowest BIC among these models. However, the three profiles only differ in their overall level of providing and receiving support. Likewise, the profiles in the five-profile model based on variance-covariance structure A only resulted in profiles with different overall levels of support. Given that the result did not show different patterns of providing and receiving support, these models were discarded from further consideration. The four-profile model based on variance-covariance structure C, on the other hand, resulted in profiles with different patterns of providing and receiving support. Moreover, the LMR-LRT and the BLRT were statistically significant. In addition, classification accuracy according to entropy was acceptable. Of the four profiles, Profile 1 (n = 70, 7.83%), low-reciprocating days, represents days with low providing and low receiving of support. Profile 2 (n = 39, 4.36%), under-reciprocating days, are days with relatively high providing support but relatively low receiving support. Profile 3 (n = 66, 7.38%), over-reciprocating days, are days with relatively low providing support but relatively high receiving support. Profile 4 (n = 719, 80.43%), high-reciprocating days, are days with relatively high providing and high receiving support (see Table 3 for descriptive information and Figure 3, Panel A for a graphical representation).

Figure 3. Latent profiles at the day level (Panel A) and latent classes at the person level (Panel B) based on the relative frequency of profiles at the day level.

Table 2. Single-level latent profile analyses at the day level.

Model	# Profiles	# Par.	LL	LL Replicated	AIC	CAIC	BIC	SABIC	LMR-LRT	BLRT	Entropy	Number of days in each profile
												1	2	3	4	5	6
A: Class Invariant ${\theta }_{mm}$, diagonal ${\mathrm{\Sigma }}_k$
	2	7	−2587.24	Yes	5188.47	5229.04	5222.04	5199.81	<.001	<.001	.812	108	786
	3	10	−2541.04	Yes	5102.08	5160.04	5150.04	5118.28	<.001	<.001	.682	301	92	501
	4	13	−2523.83	Yes	5073.67	5149.01	5136.01	5094.73	.003	<.001	.734	57	67	507	263
	5	16	−2505.64	Yes	5043.28	5136.02	5120.02	5069.20	<.001	<.001	.727	59	68	347	274	146
	6	19	−2500.50	Yes	5039.01	5149.13	5130.13	5069.78	.465	<.001	.7373	9	59	343	146	68	269
B: Class Varying ${\theta }_{mm}$, diagonal ${\mathrm{\Sigma }}_k$
	2	9	−2584.08	Yes	5186.15	5238.31	5229.31	5200.73	<.001	<.001	.694	155	739
	3	14	−2526.95	Yes	5081.89	5163.03	5149.03	5104.57	.346	<.001	.827	704	56	134
	4	19	−2516.16	Yes	5070.31	5180.43	5161.43	5101.09	.046	<.001	.856	52	696	12	134
	5	24	−2504.26	Yes	5056.52	5195.62	5171.62	5095.40	.219	-	.875	54	10	692	18	120
	6	29	−2489.63	Yes	5037.25	5205.33	5176.33	5084.23	.843	-	.861	53	10	16	33	127	655
C: Class Invariant ${\theta }_{mm}$, Class Invariant Unrestricted ${\mathrm{\Sigma }}_k$
	2	8	−2548.15	Yes	5112.31	5158.67	5150.67	5125.26	<.001	<.001	.808	93	801
	3	11	−2530.82	Yes	5083.63	5147.39	5136.39	5101.45	<.001	<.001	.657	510	97	287
	4	14	−2514.90	Yes	5057.79	5138.93	5124.93	5080.47	.013	<.001	.734	70	39	66	719
	5	17	−2503.93	Yes	5041.87	5140.40	5123.40	5069.41	.338	<.001	.685	70	509	248	28	39
	6	20	−2497.19	Yes	5034.37	5150.29	5130.29	5066.77	.104	.033	.643	61	68	462	23	69	211
D: Class Invariant ${\theta }_{mm}$, Class Varying Unrestricted ${\mathrm{\Sigma }}_k$
	2	9	−2537.68	Yes	5093.36	5145.52	5136.52	5107.94	.173	<.001	.918	26	868
	3	13	−2510.70	Yes	5047.40	5122.75	5109.75	5068.46	<.001	<.001	.786	81	765	48
	4	17	−2491.31	Yes	5016.63	5115.15	5098.15	5044.17	.542	<.001	.727	85	70	44	695
	5	21	−2484.59	Yes	5011.18	5132.89	5111.89	5045.20	.002	-	.774	70	73	704	3	44
	6	25	−2485.93	Yes	5021.87	5166.76	5141.76	5062.36	.281	-	.640	23	77	72	43	165	514
E: Class Varying ${\theta }_{mm}$, Class Invariant Unrestricted ${\mathrm{\Sigma }}_k$
	2	10	−2535.72	Yes	5091.45	5149.40	5139.40	5107.64	<.001	<.001	.898	75	819
	3	15	−2496.03	Yes	5022.06	5108.99	5093.99	5046.35	.002	-	.856	71	786	37
	4	20	−2480.15	No	5000.30	5116.22	5096.22	5032.70	<.001	-	.892	10	43	38	803
	5	25	−2458.07	No	4966.13	5111.02	5086.02	5006.63	<.001	-	.900	14	34	16	793	37
	6	30	−2449.40	No	4958.79	5132.66	5102.66	5007.39	.246	-	.914	31	6	17	4	801	35
F: Class Varying ${\theta }_{mm}$, Class Varying Unrestricted ${\mathrm{\Sigma }}_k$
	2	11	−2534.44	Yes	5090.87	5154.63	5143.63	5108.69	.005	<.001	.917	65	829
	3	17	−2494.77	Yes	5023.53	5122.06	5105.06	5051.07	.095	-	.852	63	783	48
	4	23	−2478.08	Yes	5002.15	5135.45	5112.45	5039.41	1.000	-	.889	11	49	785	49
	5	29	−2470.51	Yes	4999.02	5167.09	5138.09	5046.00	.317	-	.841	17	49	8	772	48
	6	35	−2460.83	Yes	4991.66	5197.51	5159.51	5048.35	.132	-	.789	11	52	59	727	19	26

Note. N = 894 days from 124 employees; ^c = Bootstrapped likelihood-ratio test could not be computed; #Par. = Number of estimated parameters; LL = Log-Likelihood; LL Replicated = Log-Likelihood replicated?; AIC = Akaike information criterion; CAIC = consistent Akaike information criterion; BIC = Bayesian information criterion; SABIC = sample-size-adjusted Bayesian information criterion; LMR-LRT = Lo-Mendell-Rubin likelihood ratio test; BLRT = Bootstrapped likelihood-ratio test.

Table 3. Latent profile analysis with four profiles at the day level: Mean and standard deviation on the indicator variables for each profile.

Variable	Latent Profile
	Low-reciprocating days (n = 70)		Under-reciprocating days (n = 39)		Over-reciprocating days (n = 66)		High-reciprocating days (n = 719)
	M	SD	M	SD	M	SD	M	SD
Providing support	2.03	1.07	5.14	1.07	3.19	1.07	4.95	1.07
Receiving support	1.65	1.10	2.59	1.10	5.51	1.10	4.90	1.10

Note. N = 894 days from 124 employees; M = Arithmetic mean; SD = Standard deviation.

Multilevel latent profile analyses for employees of support exchange behaviour

Results of the MLPA showed that the k = 4 class model had the lowest AIC, group-based BIC, and group-based SABIC, whereas the k = 3 class model had the lowest group-based CAIC (see Table 4). However, the model with k = 4 had convergence problems with the highest log-likelihood not replicated. Thus, we chose the k = 3 class model, with a good classification accuracy according to the entropy value: Class 1 (n = 14), over-reciprocators, included employees with a relatively high percentage of over-reciprocating days, a medium percentage of high-reciprocating days and a relatively low percentage of low-reciprocating days and under-reciprocating days. Class 2 (n = 20), low-reciprocators, comprised employees with a relatively high percentage of low-reciprocating days and a relatively low percentage of under-reciprocating days and over-reciprocating days, and a low percentage of high-reciprocating days. Class 3 (n = 90), high-reciprocators, included employees with a very high percentage of high-reciprocating days, a very low percentage of low-reciprocating and under-reciprocating days, and non-existing over-reciprocating days (see Table 5 for descriptive information and Figure 3, Panel B for a graphical representation).

Table 4. Multilevel latent profile analyses with different numbers of latent classes at the person level.

# Classes Between Level	#Par.	LL	LL Replicated	AIC	Group-Based CAIC	Group-Based BIC	Group-Based SABIC	Entropy	Number of employees in each class
									1	2	3	4
1	6	−2514.90	Yes	5039.74	5064.72	5058.72	5039.74	1.000	124
2	10	−2408.91	Yes	4837.81	4876.02	4866.02	4834.40	.851	29	95
3	14	−2382.05	Yes	4792.11	4845.59	4831.59	4787.32	.857	14	20	90
4	18	−2371.36	No	4778.72	4847.48	4829.48	4772.56	.869	12	23	2	87

Note. N = 894 days from 124 employees; #Par. = Number of estimated parameters; LL = Log-Likelihood; LL Replicated = Log-Likelihood replicated?; AIC = Akaike information criterion; CAIC = consistent Akaike information criterion; BIC = Bayesian information criterion; SABIC = sample-size-adjusted Bayesian information criterion.

Table 5. Multilevellatent Multilevel latent profile analysis with three classes at the person level: Percentage of profiles at the day level for each class.

Variable	Latent Class
	Over-reciprocators (n = 14)	Low-reciprocators (n = 20)	High-reciprocators (n = 90)
Low-reciprocating days	4.48%	65.93%	0.16%
Under-reciprocating days	2.24%	17.04%	1.60%
Over-reciprocating days	61.94%	13.33%	0.00%
High-reciprocating days	31.34%	3.70%	98.24%

N = 894 days from 124 employees.

Mean differences between day-level profiles in need satisfaction for autonomy, competence and relatedness, and emotional exhaustion

The results of the MLPA investigating mean differences between profiles at the day level are reported in Table 6.

Table 6. Multilevel latent profile analyses results at the day level: Mean differences between profiles at the day level on outcome variables.

Profile	Need satisfaction for autonomy		Need satisfaction for competence		Need satisfaction for relatedness		Emotional exhaustion
	M	SD	M	SD	M	SD	M	SD
1 Low-reciprocating days (n = 70)	5.25	1.54	4.75	1.46	4.79	1.34	2.71	1.67
2 Under-reciprocating days (n = 39)	4.15	1.86	3.88	1.25	5.17	1.15	4.03	2.12
3 Over-reciprocating days (n = 66)	6.49	0.48	5.25	1.06	6.26	0.64	1.95	0.87
4 High-reciprocating days (n = 719)	5.78	0.89	5.31	1.06	5.93	0.78	3.09	1.49
Comparison	p	Cohen’s d	p	Cohen’s d	p	Cohen’s d	p	Cohen’s d
Profile 1 vs 2	.157	0.66	.071	0.63	.222	0.30	.105	0.72
Profile 1 vs 3	.022	1.07	.267	0.39	<.001	1.39	.197	0.57
Profile 1 vs 4	.363	0.55	.172	0.51	<.001	1.35	.517	0.25
Profile 2 vs 3	<.001	1.96	.003	1.21	.001	1.26	.002	1.42
Profile 2 vs 4	.003	1.69	<.001	1.34	.001	0.95	.148	0.62
Profile 3 vs 4	.003	0.82	.890	0.06	.213	0.43	<.001	0.79

Note. N = 894 days from 124 employees.

Need satisfaction for autonomy, competence and relatedness

The results showed that on over-reciprocating days higher need satisfaction for autonomy is reported than on low-reciprocating days (p = .022, d = 1.07). Moreover, on over-reciprocating days (p < .001, d = 1.96) and high-reciprocating days (p = .003, d = 1.69), higher need satisfaction for autonomy is reported than on under-reciprocating days. Finally, on over-reciprocating days higher need satisfaction for autonomy is reported than on high-reciprocating days (p = .003, d = 0.82).

The results showed that on high-reciprocating days (p < .001, d = 1.34) and on over-reciprocating days (p = .003, d = 1.21), higher need satisfaction for competence is reported than on under-reciprocating days.

The results showed that on over-reciprocating days (p < .001, d = 1.39) and on high-reciprocating days (p < .001, d = 1.35), higher need satisfaction for relatedness is reported than on low-reciprocating days. Moreover, on over-reciprocating days (p = .001, d = 1.26) and on high-reciprocating days (p = .001, d = 0.95), higher need satisfaction for relatedness is reported than on under-reciprocating days.

Emotional exhaustion

The results showed that on over-reciprocating days lower emotional exhaustion is reported than on under-reciprocating days (p = .002, d = 1.42). Moreover, on high-reciprocating days higher emotional exhaustion is reported than on over-reciprocating days (p < .001, d = 0.79).

Mean differences between person-level classes in need satisfaction for autonomy, competence and relatedness, and emotional exhaustion

The results of the MLPA investigating mean differences between classes at the person level are reported in Table 7.

Table 7. Multilevel latent profile analyses results at the person level: Mean differences between classes at the person level on outcome variables.

Class	Need satisfaction for autonomy		Need satisfaction for competence		Need satisfaction for relatedness		Emotional exhaustion
	M	SD	M	SD	M	SD	M	SD
1. Over-reciprocators (n = 14)	6.20	0.47	5.60	0.66	6.29	0.59	2.96	1.25
2. Low-reciprocators (n = 20)	5.14	1.43	5.24	0.98	5.82	0.86	3.31	1.51
3. High-reciprocators (n = 90)	5.49	1.07	5.50	1.04	5.96	0.78	3.36	1.23
Comparison	p	Cohen’s d	p	Cohen’s d	p	Cohen’s d	p	Cohen’s d
Class 1 vs 2	.005	0.93	.228	0.42	.089	0.62	.546	0.25
Class 1 vs 3	<.001	1.00	.628	0.10	.079	0.44	.387	0.32
Class 2 vs 3	.364	0.04	.339	0.25	.568	0.18	.895	0.04

Note. N = 894 days from 124 employees.

Need satisfaction for autonomy, competence and relatedness

The results showed that low-reciprocators (p = .005, d = 0.93) and high-reciprocators (p < .001, d = 1.00) report lower need satisfaction for autonomy than over-reciprocators. The results did not show any statistically significant mean differences between classes in need satisfaction for competence. The results did not show any statistically significant mean differences between classes in need satisfaction for relatedness.

Emotional exhaustion

The results did not show any statistically significant mean differences between classes in emotional exhaustion

Discussion

This study is the first to apply MLPA to identify qualitatively different support exchange profiles that occur in employees’ daily lives and to explore whether employees differ in their propensity to engage in those different day-level profiles and thus form person-level classes. Using this “person-oriented” (vs “variable-oriented”) approach, we not only identified patterns of day-specific receiving and providing social support (i.e., day-level profiles) but also described differences in employees’ general tendency to employ certain day-level profiles of support exchange behaviour over time (i.e., person-level classes). Further, building on resource-gaining, resource-draining and resource-conserving considerations of COR theory (Hobfoll, 1989), the present study set out to investigate how the joint engagement in receiving and providing social support is related to employees’ basic need satisfaction and emotional exhaustion at work.

Summary and interpretation of results

At the day level, we identified four latent profiles that differed in varying levels of low vs high receiving support and low vs high providing support: (1) low-reciprocating days, when equally low levels of support are received and provided; (2) under-reciprocating days, when higher levels of support are provided than received; (3) over-reciprocating days, when higher levels of support are received than provided; and (4) high-reciprocating days, when equally high levels of support are received and provided. Accordingly, our profiles that refer to latent subpopulations at the level of days align with past research that chose the traditional “variable-oriented” approach (e.g., Buunk et al., 1993; Nahum-Shani et al., 2011; Tomprou et al., 2020).

At the person level, we identified three latent classes of individuals who differed in their general propensity to employ certain support exchange profiles over time: (1) over-reciprocators, who predominantly displayed over-reciprocating days and high-reciprocating days; (2) low-reciprocators, who predominantly displayed low-reciprocating days and under-reciprocating days; and (3) high-reciprocators, who almost entirely displayed high-reciprocating days. Accordingly, while it seems that most employees followed the norm of reciprocity when it comes to receiving and providing support, some employees depicted instead a more heterogeneous composition of support exchange profiles that both follow and violate reciprocity norms, which reflects the complex and dynamic nature of daily support exchange behaviour between co-workers.

Furthermore, our findings highlight how basic need satisfaction and emotional exhaustion differ as a function of latent profile and class membership and are in alignment with our theoretical assumptions (Hobfoll, 1989) that resource-gaining, resource-draining and resource-conserving experiences matter in support exchange behaviour. Below, we discuss our results by comparing mean differences between profiles/classes with the highest and lowest mean values, respectively, of the outcome variables we examined.

In terms of need satisfaction for autonomy, it seems that upon gaining resources on over-reciprocating days, employees may experience higher levels of volition, which plays to their day-specific need satisfaction for autonomy (Ryan & Solky, 1996). These results show that employees can still satisfy their autonomy need even when depending on others, as long as they act willingly and with a rationale that is consistent with their values (Van den Broeck et al., 2010). In contrast, upon draining resources on under-reciprocating days, employees report a lesser sense of autonomy.

In terms of need satisfaction for competence, it seems that conserving resources on high-reciprocating days can facilitate the satisfaction of competence needs, while depleting resources on under-reciprocating days hinder the day-specific satisfaction of the competence need (Van den Broeck et al., 2010). This result may appear surprising, as one could argue that being able to provide support should concord with feeling competent. Thus, we speculate that the actual satisfaction of competence needs may be dependent on whether a competence-satisfying environment is available that allows for optimal challenges to extend one’s skills (Legault, 2017), and helping out co-workers does not seem to lead employees to see themselves as effective in and progressing on their tasks.

In terms of need satisfaction for relatedness, it seems that gaining resources on over-reciprocating days may establish a greater sense of intimacy because it implies self-disclosure and mutual reliance, and thereby increases day-specific feelings of need satisfaction for relatedness (Hobfoll et al., 1990). In contrast, low-reciprocating days, on which mutually low levels of support is exchanged seem to be the least related to experiencing a sense of workplace belonging and connection with co-workers.

Finally, emotional exhaustion was particularly high on under-reciprocating days on which resources are lost or invested and thereby become depleted (Halbesleben & Wheeler, 2015; Halbesleben et al., 2013). In contrast, levels of emotional exhaustion were lower on over-reciprocating days, indicating that receiving support without having to invest in providing support in return may build up resources and mitigate stress reactions.

Regarding the outcomes at the person level, we did neither find statistically significant person-level effects of classes on need satisfaction for competence nor relatedness nor emotional exhaustion. Only in terms of person-level autonomy, it seems that employees characterized by often displaying predominantly over-reciprocating days were particularly equipped with a sense of choice.

When comparing our day-level findings with person-level findings in terms of how profiles and classes relate to the outcomes of interest, it is apparent that our findings vary across analytical levels. This speaks to the MLPA approach of our study in that it shows that effects at the day level (profiles) are not the same when investigated at the person level (classes). These differences may result from the fact that several day-level profiles are intermingled at the person level, resulting in effects of “blended” day-level profiles, while “pure” day-level profiles are investigated at the day level. In other words, person-level classes are based on the proportion of different day-level profiles people engage in; thus, effects of person-level classes on our outcomes of interest are formed by a blend of day-level classes. To illustrate, the day-level profile over-reciprocating days consists of 100% over-reciprocating days, while the person-level class over-reciprocators consists of people who mainly experience over-reciprocating days (61.94%), but also some high-reciprocating days (31.34%), and few low-reciprocating days (4.48%) and under-reciprocating days (2.24%).

Theoretical implications

Our findings have implications for support exchanges in organizations. First and foremost, the use of MLPA extends the identification and analysis of latent profiles by identifying support exchange patterns on different hierarchical levels at the same time (Mäkikangas et al., 2018). With this multi-level method, it was possible to identify qualitatively different days (i.e., latent profiles) and groups of employees (i.e., latent classes). We identified four types of day-specific support exchange behaviour that were in close alignment with past “variable-oriented” research (e.g., Tomprou et al., 2020). Moreover, MLPA allowed the identification of three Level 2 latent classes, showing that the number of classes does not simply reflect the number of day-level profiles found. These results suggest that people did not consistently engage in just one of the Level 1 profiles over time. Rather, people differed in the specific combinations of Level 1 profiles they showed over time.

Furthermore, in alignment with COR theorization (Hobfoll, 1989), our results indicate that employees may gain, conserve or lose resources as a consequence of certain patterns of receiving and providing support, which has different implications for employees functioning expressed through basic need satisfaction and emotional exhaustion. The most apparent contrast is found when comparing the two support exchange profiles where an actual net gain or loss of resources occurs: Resource-gaining over-reciprocating days, seem to reap the greatest benefits regarding day-level basic need satisfaction in terms of autonomy and relatedness, and day-level emotional exhaustion. In contrast, resource-draining under-reciprocating days, seem to evoke negative consequences for day-level need satisfaction for competence and day-level emotional exhaustion. These findings are of theoretical relevance because they suggest that both reciprocation and over-reciprocation in daily support exchanges play to employees’ need satisfaction for autonomy, competence and relatedness and can make them feel less emotionally exhausted. Our results suggest that the greatest benefits, in terms of person-level need satisfaction for autonomy, occur when employees experience higher levels of resource-gaining days, that is, over-reciprocating days, over time. It is worth noting that, while we subsume both low-reciprocating and high-reciprocating support exchanges under the category of a resource-conserving pattern according to COR theory (Hobfoll, 1989), they are not the same thing. High-reciprocating exchanges indicate resource exchanges in which employees actively invest resources in order to gain resources. In contrast, low-reciprocating exchanges are not characterized by an active exchange of resources. Thus, it appears that on low-reciprocating days, no exchange of resources is actively sought with the aim to build up resources. Mean levels regarding our outcome variables support this notion, as low-reciprocating days were associated with lower levels of need satisfaction for autonomy, competence and relatedness, but also lower levels of emotional exhaustion when compared to high-reciprocating days. Considering that people have various and sometimes competing needs at any given time, we propose that the relative value of one specific resource (e.g., relatedness-need-related resources such as co-worker support) may have been somewhat reduced on low-reciprocating days, while the relative value of other resources (e.g., autonomy-need-related resources such as job control) that employees actively choose to conserve or acquire was greater (Halbesleben et al., 2014).

While we extend knowledge about the facilitating and mitigating nature of resources in support exchanges, our findings diverge from theorizing based on the norm of reciprocity, which contends an advantage for balanced (i.e., high-reciprocal and low-reciprocal) over unbalanced (i.e., under-reciprocal and over-reciprocal) support exchanges (Cropanzano & Mitchell, 2005; Nahum-Shani & Bamberger, 2011). Past “variable-oriented” research has indeed supported the favourable effects of reciprocity in resource exchanges with co-workers (Halbesleben & Wheeler, 2011) and organizations (Liu et al., 2020; Tomprou et al., 2020) on employee outcomes, but did not consider how the give and take of support occurs in conjunction and builds unique combinations of support patterns that employees exhibit day-to-day. As such, in our study, reciprocity in the form of sharing equally high amounts of support seemed to be the go-to approach in support exchanges, as reflected in the most popular day-level profile (i.e., high-reciprocating days accounting for 80.43%) and person-level class (i.e., high-reciprocators accounting for 72.58%). However, our results speak to the possibility that a lack of reciprocity may be, in fact, most beneficial, specifically when employees engage in over-reciprocating days and employ particularly over-reciprocating days over time. Thus, our results indicate that violating the norm of reciprocity by being advantaged or over-benefitted with support has no negative effects on employees at the day level and within the timeframe of two workweeks, as far as we can tell. When comparing over-reciprocating days to high-reciprocating days, it seems that when receiving support is high – which is the case in both profiles – the amount of support that is provided matters. That is, over-benefitting on support is associated with high levels of need satisfaction for autonomy and lower levels of emotional exhaustion when compared to sharing equally high amounts of support. These results are in line with the aforementioned idea of resource-gain processes as part of COR theory (Hobfoll, 1989). However, not everyone may be able to optimize their resource pool at the same time. When employees receive support, there must be a counterpart that provides support. Thus, while not following reciprocity norms seems best for the individual in the short term, from a collective perspective, adopting this strategy would not work in the long term.

Finally, we would like to emphasize that “person-oriented” approaches differ from “variable-oriented” approaches in terms of analytic technique, mindset and interpretation (Casper et al., 2019; Zyphur, 2009). Our “person-oriented” approach using MLPA implies that receiving and providing support occur in different combinations that build qualitatively different support exchange profiles (day level) in which employees engage in to a certain propensity over time (person level). Receiving support and providing support do not occur in isolation but co-occur for most employees. When comparing our results to the results of a “variable-oriented” approach such as regression analysis, the value of examining whether receiving support occurs in conjunction with providing support becomes clear: For one, the number of day-level profiles we found and how they are characterized indicate that receiving and providing support are largely independent of the level of days (i.e., receiving high levels of support on one day does not imply that employees also provide high levels of support on this day; vice versa, receiving low levels of support on one day does not imply that employees also provide low levels of support on this day). This is a major difference to “variable-oriented” research, which uses regression analyses testing for interaction effects (e.g., Tomprou et al., 2020), because a certain degree of independence between the predictor variables that form the interaction term is a precondition for performing those analyses (Baron & Kenny, 1986), but is rarely the subject of investigation. MLPA does not simply estimate the amount of variance in receiving support (or providing support) between employees and their outcome variables, but it allows to identify subgroups of qualitatively different days and employees – based on differing levels of receiving and providing support – and to link them to our outcomes of interest on different hierarchical levels (i.e., day, employee).¹

Practical implications

Based on the study’s findings, we reinforce the idea that daily social support in the workplace is not inherently beneficial, but that attention should be paid to the “net benefit” (Jolly et al., 2020) that may result from the interplay of how much support is both received from and provided to co-workers on a given day. At first glance, our results suggest that days on which low levels of support are received and high levels of support are provided (i.e., under-reciprocating days) are especially disadvantageous, whereas days on which high levels of support are received and low levels support are provided (i.e., over-reciprocating days) are especially advantageous when it comes to promoting need satisfaction for autonomy and competence as well as reducing emotional exhaustion. However, our results must be interpreted in reference to daily exchanges (day level) and exchanges that are shown over 10 days (person level), respectively. We do not recommend that employees should try to engage in more over-reciprocating days because it is possible that no support will be received in the long run if everyone were to follow this pattern of exchange where one is over-benefitting. It is also worth noting that the class of over-reciprocators consists of both over-reciprocating and high-reciprocating days that are shown over several days, indicating that not only do employees engage in diverse day-level profiles over time, but there also seems to be an advantage for employees in not only receiving support but also providing support (to a fair extent). Hence, providing social support is not to be understood as a behaviour that should be avoided for the sake of employees’ motivation and well-being; rather, the daily costs for providers warrant more attention. Practising managers should keep in mind that the “doing good, feeling good” rationale may not always apply (Lin et al., 2019) and should be mindful of what to demand and expect from employees whose resources may have become drained throughout the day due to supporting their co-workers. At the same time, we propose that receiving support, and in a broader sense seeking help, is a workplace behaviour that indeed should be encouraged. By establishing formal helping roles that are assigned to specific “expert” employees on specific days, support providers may be able to manage and distribute their own resources more appropriately among each other to avoid resource drain (van der Rijt et al., 2013). Furthermore, to facilitate basic need satisfaction in light of support exchanges at work, managers would be well advised to adopt an autonomy-supportive managerial style, which refers to creating a supportive work climate that includes taking the perspective of employees and is associated with the promotion of autonomous motivation (Baard et al., 2004; Gillet et al., 2013). Employees who are enabled to act in an autonomous fashion may experience more positive states when engaging in different support exchange behaviours (Gagné et al., 2008). In any event, social support should be exchanged in a way that is supportive and respectful, thereby promoting the satisfaction of autonomy, competence and relatedness needs (Ryan & Solky, 1996, p. 262).

Limitations and directions for future research

Despite applying a novel approach to the research of social support in the workplace, this study has several limitations that should be considered. First, while our sample sizes were large enough to apply MLPA models (Park & Yu, 2018), the study’s sample sizes at the day level (894 days) and the person level (124 employees) were rather small. Moreover, across the four latent profiles we identified, day units were not evenly distributed, resulting in 39 under-reciprocating days for the smallest profile, compared to 719 high-reciprocating days for the largest profile. Non-even distribution was also the case for the three latent classes, as the smallest class of over-reciprocators included 14 employees, whereas the largest class of high-reciprocators included 90 employees. Relatedly, because MLPA in particular, and LPA studies in general, are more inductive in nature, an important way of validating final profile and class solutions is to examine their replicability across different samples, as well as contexts and time points (Spurk et al., 2020). In our manuscript, we do not provide a replication of the same number and types of classes and profiles across multiple samples, which would be desirable to mitigate the risk of over-interpreting spurious classes and profiles (Spurk et al., 2020). We thus urge future studies to replicate our findings of multilevel latent profiles and classes with larger and more diverse samples (Grommisch et al., 2020).

Furthermore, future studies with a larger sample size could investigate potential interaction effects between day-level profiles and person-level classes. Based on COR (Hobfoll, 1989), it is conceivable that employees with greater resources are less vulnerable to resource loss on certain days (Hobfoll et al., 2018). For example, high-reciprocators (i.e., employees with a very high percentage of high-reciprocating days) may be able to buffer negative effects of under-reciprocating days and be less impaired in their functioning due to their generally solid resource reservoir.

Second, we acknowledge the value of investigating support exchange behaviour in co-worker dyads, as has been done in past research (Halbesleben & Wheeler, 2015), especially when one is interested in bilateral investment processes. However, examining daily support exchange behaviour in co-worker dyads only may provide a rather limited perspective on the broader spectrum of available social support that can be received from and provided to various co-workers in the workplace (Hüffmeier & Hertel, 2011). It would thus seem valuable for future research to advance our findings to the team level and examine support exchange behaviour in team interactions in order to account for resource investment decisions within a broader social network (Halbesleben & Wheeler, 2015). Another interesting research avenue would be to examine support exchange behaviours in hierarchical relationships – i.e., employees and supervisors – in which a certain degree of asymmetry of receiving and providing social support may be considered normal, as employees can expect instruction and guidance from their supervisors to a greater extent (Buunk et al., 1993). Relatedly, it might be fruitful to investigate support exchange behaviours in intimate and private relationships, for example, with family members and friends. It is conceivable that different relationship rules apply in closer personal relationships, where receiving and providing support tends not to be viewed as a transaction, as it might be in more distant work relationships (Clark & Mills, 1993; Cropanzano & Mitchell, 2005; Wieselquist et al., 1999).

Third, because the use of self-report scales is associated with common method bias (P. M. Podsakoff et al., 2003), we measured our study variables at different time points. Particularly regarding received and provided social support, we aimed to separate these constructs in time in order to avoid artefactual covariation among them (P. M. Podsakoff et al., 2003). However, we must also point out that emotional exhaustion was assessed in the evening questionnaire. This measurement decision gives rise to the limitation that after-work experiences may have contaminated the assessment of emotional exhaustion. Thus, future studies may want to assess emotional exhaustion immediately after work. What does speak for the use of self-reports is that all variables under investigation depict employees’ subjective reality of work conditions and thus may be difficult for others to observe (Lawrence et al., 2007). Related to this, the reliabilities of the scale for need satisfaction for competence indicated poor internal consistencies at the within- and between-level (Cronbach’s alphas below the critical cut-off of 0.70), which is likely due to the use of a shortened three-item version of this scale with typically low inter-item correlations (Ziegler et al., 2014).

Finally, other variables could influence the proposed relationships in this study. For example, trust is argued to be an important factor when examining interpersonal relationships in light of norms of reciprocity (Halbesleben & Wheeler, 2015). Accordingly, trustful working relationships may be less prone to follow a rigid “tit-for-tat” rationale when receiving and providing social support (Buunk et al., 1993). Thus, with or without a history of trust among co-workers, under-reciprocating days may be experienced more or less emotionally exhausting and could affect basic need satisfaction differently. Moreover, there are indications that employees’ individual characteristics, such as their level of honesty-humility, extraversion, agreeableness and conscientiousness, influence how much help they receive and provide (Bowling et al., 2004; Williamson & O’hara, 2017; Wingate et al., 2019), which, ultimately, influence employee outcomes. In this context, another personal dimension worth investigating could be employees’ degree in equity-sensitivity that governs reactions in over-benefited or under-benefited situations (O’neill & Mone, 1998). Apart from individual differences, work characteristics that presumably shape employees’ perceptions of their social environment should be considered. For example, task interdependence, job complexity and emotional demands of work tasks are likely to affect proposed patterns of support exchanges (Golden & Gajendran, 2019; Semmer et al., 2008). What is more, employees’ personal costs of helping behaviour should receive more attention: Those who engage in high levels of support invest additional resources (e.g., time and energy) and may become overburdened with responsibilities and activities, which is associated with role overload and job stress (Bolino & Turnley, 2005). Future research that seeks to go beyond a reformulation of our study limitations could additionally investigate other important employee outcomes, such as work attitudes (e.g., job satisfaction) and work behaviours (e.g., job performance), that may result from varying levels of satisfied needs and emotional exhaustion caused by different support exchange profiles and classes. Apart from examining what consequences derive from latent profiles and classes of support exchange behaviour, it would be an interesting addition to this study to look into theoretically grounded antecedents. For example, work engagement (Reijseger et al., 2012) or time pressure (Urbach & Weigelt, 2019) have been shown to influence helping behaviour and thus are likely, if applied to our study, to affect the realization of support exchange profiles and class membership. Another interesting research direction might involve investigating employee mood states that may influence resource investment behaviour throughout the workday. Support exchanges may increase or decrease positive affect in employees, and these feelings, in turn, may increase or decrease helping behaviour (Snippe et al., 2018). Moreover, whether support is actually needed could be a factor worth exploring in the context of support exchange behaviour as recipients may not perceive offered help that is not desired as a resource gain (or conservation) (Beehr et al., 2010).

Conclusion

When employees receive support from co-workers, they, in return, may be motivated to reciprocate support. Past research has highlighted the importance of considering reciprocal processes to better understand the interplay of receiving and providing social support and their consequences for employee outcomes. This study adds to this line of research by showing how days and employees can be clustered according to the daily receiving and providing of social support and by examining how different day-level profiles and person-level classes relate to the satisfaction of autonomy, competence and relatedness needs, as well as emotional exhaustion, at work. While social support is an important and well-studied resource in the workplace, we still have much to learn, and applying MLPA may be a promising avenue for moving the research forward.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The results of the multilevel regression analyses are available from Ada Sil Patterer upon request.