Work-family conflict, self-efficacy, and emotional exhaustion: A test of longitudinal effects

According to Zapf, Dormann, and Frese (1996) longitudinal studies represent a breakthrough since they allow casualty and directionality analysis to be made among variables, something that is not possible in cross-sectional studies, which are more frequent in psychological research. Across time it might be possible that a reciprocal influence between two variables generate spiral relationships between them. For instance, with regard to the relationship between work-family conflict and emotional exhaustion, Hall et al. (2010) supported the hypothesis that emotional exhaustion increases the perception of work-family conflict, namely reverse causality and quote various studies in this direction. Thompson, Kirk, and Brown (2005) pointed out that occupational stress influences family environment from the emotional exhaustion in their members so that the most exhausted ones also experience more obstacles to fit work and family together. Mikkelsen and Burke (2004) , in a study with 766 police agents also found a reverse relationship, so that exhaustion and cynicism arose as powerful predictors of work-family conflict. Furthermore, Innstrand, Langballe, Espnes, Falkum, and Aasland (2008) in a survey of 2,235 people from eight different professions reported results in line with the bi-directionality of the relationship between stress derived from work-family relations and emotional exhaustion. Nevertheless, more research is needed with regard to reverse causality. COR Theory ( Hobfoll, 2002 ), suggests that a reverse causal effect is possible between higher levels of burnout, so that burnout employees could perceive their jobs as increasingly stressful ( Melamed, Armon, Shirom, & Shapira, 2011 ). In a longitudinal study with 257 police officers, Hall et al. (2010) using structural equations found some support to the simultaneous reciprocal effects as the most complete model that fits best, in contrast with the normal causal model and the reverse causality model. Carlson, Ferguson, Hunter, and Witten (2012) , in their study on abusive supervision of 328 workers, conclude that relationships between work-family conflict and burnout are reciprocal and point out at a loss spiral, as COR theory postulates.

The Demands and Resources Model (JD-R) by Bakker and Demerouti (2008) is one of the most relevant explanatory models of the stress influence on work and occupational health. Current meta-analytical studies the relevance of the re-formulated JD-R model to predict burnout ( Crawford, Le Pine, & Rich, 2010; Lizano & Barak, 2012; Nahrgang, Morgeson, & Hofmann, 2010 ). The JD-R model consider various individual as well as organizational variables as resources, and self-efficacy is one of the most d.

From the Social Cognition Theory, self-efficacy is defined as an individual's belief in his/her own skills to organize and execute the d course of action so that the aims can be accomplished ( Bandura, 1997 ). On the other hand, professional self-efficacy ( Cherniss, 1993 ), one of the self-efficacy facets, is defined as an individual's belief in his/her own capability to perform his/her work roles properly.

Direct relationships between self-efficacy and burnout have been confirmed in several meta-analyses ( Aloe, Amo, & Sanan, 2014 ), but self-efficacy seems to play a more complex role in the stress process than has been assumed so far. In recent literature, empirical evidence has been found for several possible self-efficacy roles: (a) as an indirect predictor of strain via stress as mediator ( Wang, Hall, & Rahimi, 2015 ), (b) as a mediator between stress and strain ( Yu, Wang, Zhai, Dai, & Yang, 2014 ), and (c) as a moderator of the stress and strain relationship (buffering effect; Schwarzer & Hallum, 2008).

In the military context, the buffering effect of self-efficacy has been confirmed with regard to several demands and various health outcomes ( Stetz, Stetz, & Bliese, 2006 ) and performance ( Brusso, Orvis, Bauer, & Tekleab, 2012 ). Specifically, Stetz et al. (2006) showed the importance of self-efficacy when high levels of stress are combined with a low level of support from the supervisor. In an earlier paper, Jex and Bliese (1999) found that both self-efficacy and collective-efficacy buffered the relationship between several stressors and the health in the expected direction.

In this paper we focus on the analysis of the roles of self-efficacy in the causation paths of work-family conflict and burnout. This approach is in line with the contribution of Michel, Michelson, Pichler, and Cullen (2010) , who d similar questions concerning the role of social support, in an attempt to shed some light on the relationships between self-efficacy, work-family conflict, and emotional exhaustion, testing four different models: independence, mediation, antecedent, and independence plus antecedent models. The testing of these models will complement the analysis of the directionality of longitudinal relationships between demands, resources, and outcomes (normal causal model, reversed causal model, and reciprocal causal models).

To test the directionality models presented above the following hypotheses are formulated:

Hypothesis 1a (H 1a): Work-family conflict at Time 1 will positively predict emotional exhaustion at Time 1 and Time 2 (normal causal model).

Hypothesis 1b (H 1b): Professional self-efficacy at Time 1 will negatively predict emotional exhaustion at Time 1 and Time 2 (normal causal model).

Hypothesis 2a (H 2a): Emotional exhaustion at T1 will positively predict work-family conflict at T2 (reversed causal model).

Hypothesis 2b (H 2b): Emotional Exhaustion at T1 will negatively predict professional self-efficacy at T2 (reversed causal model).

Hypothesis 3 (H 3): The causal reciprocal model integrating the relationships hypothesised in the normal causal model and the reversed causal model will present a significantly better fit than the ones presented by each of the models independently.

Moreover, to test the role of professional self-efficacy in the cross-sectional and longitudinal relationships between work-family conflict and emotional exhaustion, four alternative models depicted in Figure 1 will be tested. No specific hypothesis is formulated for these analyses given their exploratory characteristics: independence model, mediation model, antecedent model, and independence plus antecedent model.

Figure 1

The roles of self-efficacy in the relationship between work family conflict and emotional exhaustion.

Note . These models in our study will be tested paying attention to the cross-sectional as well as longitudinal relationships and testing the causal reciprocal relationships.

Table 1 provides the means, standard deviations, correlations, and internal consistency (Cronbach's alpha) for all variables. As can be seen, reliability coefficients are above .70, similar to the ones obtained in other meta-analytical studies ( Aguayo, Vargas, de la Fuente, & Lozano, 2011; Wheeler, Vassar, Worley, & Barnes, 2011 ).

Correlations significant relationships between variables in the same direction of the formulated hypotheses, WFC and emotional exhaustion at both times (T1 and T2). Correlations between professional self-efficacy and WFC and emotional exhaustion are negative at both times.

As mentioned earlier, in order to test hypotheses as well as to identify the model that best explains the directionality of the relationships between variables, two sets of four models have been compared. The goodness of fit indexes is shown in Table 2 and Table 3.

In general, four causal models showed a good fit, since both CFI and NFI yield a value close or above .95, RMSEA is lower than .08, and the ratio between chi-square and degrees of freedom is lower than 3.

Causal Model (M1) fitted better than Stable Model (M0),

??2 = 6.159 (1), p = .01, and ?CFI = .020. This difference was considered significant since ?CFI > .01. These results pointed to the importance of including cross-lagged effects from WFC at T1 to explain emotional exhaustion at T2 With regard to Reverse Model (M2), its adjustment was better than Stable Model (M0), ??2 = 3.852 (1), p = 05 and ?CFI = .011, but worse than Causal Model (M1), ??2 = 2.307 (1), p = 20 y ?CFI = .009. According to these results the model including cross-lagged effects fitted better than the model which d temporal stability and synchronic effects.

On the other hand, Reciprocal Model (M3), that depicts reciprocal and simultaneous relationships over time, fitted better than Stable Model (M0), ??2 = 10.011 (2), p = .01 and ?CFI = .031, Causal Model (M1), ??2 = 3.852 (1), p = .01 and ?CFI = .011, and Reverse Model (M2), ??2 = 6.159 (1), p = .01 and ?CFI = .020. In line with these results, the model that best fitted these empirical data is the Reciprocal Model (M3).

With regard to the relationship between self-efficacy, work-family conflict and emotional exhaustion, four models were compared ( Figure 1).

From the fit indexes assessed, only the Independence Model (M1) and Independence plus Antecedent Model (M4) showed adequate results. On the other hand, Model 4 adjusts better than M1, ?? 2 = 3.507 (0), p = .05 and ?CFI = .014, M2, ?? 2 = 29.401 (2), p = .001, and ?CFI = 0.105, and M3, ?? 2 = 42.624 (2), p = .001 and ?CFI = .156. A double role of self-efficacy both as a preventive antecedent of work family conflict and as a strong direct reducer of emotional exhaustion was confirmed.

As a result of the comparative analyses of these two series of models, the final solution that best fits the data is the Reciprocal Model (M3) depicted in Figure 2.

Figure 2

The final model.

*p < .05, **p < .01, ***p < .001.

According to the formulated hypotheses, stress caused by family-work conflict shows direct positive effects on emotional exhaustion levels both synchronous at T1 and T2 and cross-lagged (? = .16, t = 2.48, p = .013). From these results hypothesis 1a can be accepted. With regard to emotional exhaustion at T1, this variable shows a positive cross-lagged effect on work-family conflict at T2 (? = .12, t = 2.00, p = .04), confirming hypothesis 2a and the results obtained also partially support the reciprocal and simultaneous relationships between the variables WFC and emotional exhaustion as posited by hypothesis 3 . That is, the reciprocity effects ( hypothesis 3 ) are partially proved because they are not significant longitudinal causal relations of professional self-efficacy on emotional exhaustion and of emotional exhaustion in T1 on self-efficacy in T2. About hypotheses 1b y 2b , self-efficacy plays a double role both as a preventive antecedent of WFC at T1 (? = ?.16, t = -2.61, p = .009) and at T2 (? = ?.18, t = -2.96, p = .003) and as a strong direct antecedent that reduces WFC and through this variable has a negative effect on emotional exhaustion at T1 (? = ?.28, t = -5.14, p = .001) and at T2 (? = ?.17, t = -3.14, p = .002). However, the hypothesized direct (? = ?.09, t = 1.31, p = .18) and reverse (? = ?.05, t = ?.74, p = .45) cross-lagged effects of self-efficacy have not been confirmed as depicted in the Model 3 ( Figure 2 ). That is, the effects are partially proved because the longitudinal relationship of self-efficacy on emotional exhaustion and the relationship between emotional exhaustion in T1 and self-efficacy in T2 are not significant.

To sum up, the hypotheses have been partially confirmed. It supports the idea of “ spiral of gains and losses ” or reciprocal and simultaneous relationships between WFC and emotional exhaustion (WFC > emotional exhaustion > WFC). Moreover, self-efficacy plays an antecedent role on WFC and an indirect effect, through WFC, on emotional exhaustion, but its reciprocal and longitudinal effect with emotional exhaustion has not been proved.

This is the first study to research the problem of work-family conflict among military personnel in Spain with a longitudinal approach. It s the relationship between work-family conflict, professional self-efficacy, and emotional exhaustion. The lack of longitudinal studies on this topic bears out this interest.

According to our findings, soldiers that suffered high stress caused by work-family conflict have shown higher emotional exhaustion, and this effect remains one year after the first data gathering took place (hypothesis 1a). This confirms the causal model (Zapf et al., 1996 ), widely supported by research about the relationship between work-family conflict and burnout ( Bridger et al., 2008; Hall et al., 2010; Lizano & Barak, 2012 ).

Moreover, a reverse effect between emotional exhaustion and perceived work-family conflict has also been confirmed (hypothesis 2a) . That is to say, emotionally exhausted soldiers also tend to undergo more stress from work-family conflict ( Hall et al., 2010; Innstrand, Langballe, Espnes, Falkum, & Aasland, 2008 ). However, only a few longitudinal studies on occupational stress and health have explored inverse and reciprocal causal relationships between these variables, so it was also necessary to explore the possibility of reverse causality ( Melamed et al., 2011 ). Our results corroborate the simultaneous reciprocal relationships between work-family conflict and emotional exhaustion (hypothesis 3) . This finding is the most important contribution of the present work.

The influence of professional self-efficacy on emotional exhaustion, in line with the Social Cognitive Theory ( Bandura, 1997 ), has been partially proven. This means that people who perceive themselves as more self-effective better health and less burnout ( Schaufeli & Salanova, 2007). Innstrand et al. (2011) suggest that a combination of target orientation, a strong self-efficacy feeling, and values congruency seems to have a protective effect against burnout and thus supports the hypothesis that soldiers who see themselves as the most effective lower levels of emotional exhaustion. Likewise, an antecedent effect of self-efficacy on work-family conflict was found, underlying the double effect of self-efficacy on emotional exhaustion: one direct and another indirect through its reduction effect on work-family conflict. These findings could shed some light on the role played by self-efficacy in order to explain the relationships between work-family conflict and emotional exhaustion. Although the low amount of variance explained on the emotional exhaustion suggests the influence of other factors, our results are within acceptable limits ( Innstrand et al., 2008; Mikkelsen & Burke, 2004; Zapf et al., 1996 ).

Nevertheless, this study has some limitations. Firstly, most people from our sample are young, mainly male and not living in couple. In order to overcome this limitation that could negatively influence the generalization of these results, further studies must be carried out with more heterogeneous samples, including higher and more disperse scores regarding to the d variables.

Another limitation is that work-family conflict is d as a global measure without distinguishing its bi-directionality: from work to family and from family to work, as it has been proposed by some authors (e.g., Lambert & Hogan, 2010 ). It would also be interesting to design a professional self-efficacy scale adapted to measuring work-family conflict in military contexts. Other antecedents of burnout such as engagement with the profession, job involvement, or ability to control should be introduced in future studies as several meta-analyses have pointed out ( Kenworthy, Fay, Frame, & Petree, 2014; Lee et al., 2011 ). Moreover, the level of education, the working hours, the professional status, and the gender should also be taken into consideration as control variables, because of their relationship with emotional exhaustion ( Lim, Kim, Kim, & Lee, 2010 ). Finally, only using self-report measures could increase common variance due to the evaluation method. Future studies should use other data gathering methods.

Despite the earlier mentioned limitations, the strength of this paper lies in its longitudinal design and the results about reciprocal and simultaneous effects between work-family conflict and emotional exhaustion. Likewise, it contributes to clarifying the role of self-efficacy as a personal resource, influencing stress caused by emotional exhaustion. On the one hand, it confirms, as the majority of studies do, its important effect on emotional exhaustion, one key dimension of burnout, and, on the other, its preventive effect on stress that is related in turn to emotional exhaustion.

Among the implications of these findings from an applied perspective, the convenience of designing stress prevention programs as a way to prevent burnout development should be underlined. Firstly, stress due to work-family conflict should be considered an important stressor within military contexts, thereby conciliation programs should be developed in order to overcome its negative effects. Secondly, when stressful situations caused by work-family conflict are difficult to control, training programs focused on reducing emotional exhaustion could be an adequate strategy. These training programs could be an appropriate strategy in order to improve individual and professional self-efficacy and should be provided on a repetitive base since their effect might only be circumstantial, since self-efficacy could change over time, as pointed by this research.

The authors of this article declare no conflict of interest.