1 Determinants of the need for employees' assistance programs (EAPs): An empirical examination among Canadian teachers Aaron Cohen and Hanit Schwartz Department of political science University of Haifa, Haifa, Israel Please address correspondence to Dr. Aaron Cohen, E-mail: ACOHEN@POLI.HAIFA.AC.IL 2 Summary. The goal of this study was to propose and test a model of the determinants of the need for employees' assistance programs. This need was defined as multidimensional. The research model proposed six independent variables related to employees' assistance programs: organizational support, personal coping, negative spillover, occupational commitment, job satisfaction, and tenure. The model proposed that the relationship between the independent variables and the dependent ones, the needs for employees' assistance programs, are not direct but mediated by work-nonwork conflict and job stress. Questionnaires were mailed to employees of a school district in Western Canada, and 300 usable questionnaires were returned, a response rate of 42 percent. The research model received modest support from the data. Some of the fit indices were not as strong as expected. On the other hand, the conceptual model was supported. There is a need for more conceptual research on different issues regarding employees' assistance programs. 3 Determinants of the need for employees' assistance programs (EAP): An empirical examination among Canadian teachers Introduction There is growing interest among scholars and practitioners in employees' assistance programs. Employees' assistance programs are a formal intervention system that identifies and assists organizational members with a wide range of personal problems that may affect their jobrelated behaviors (Blum & Roman, 1988). Interest in such programs began when employers realized that assisting employees in coping with their personal problems such as alcoholism and drug abuse could improve their productivity and was less costly than terminating their employment (Powell & Kotschessa, 1995). While the first employees' assistance programs dealt mostly with alcoholism problems among employees (Masi, 1984) the content of employees' assistance programs in the subsequent programs was expanded to drug abuse and family and domestic problems (Reed, 1983; Gerstein, 1991) and even to problems in the workplace (Trice & Sonnenstuhl, 1988). Conceptually, employees' assistance programs represent an integration of work and nonwork domains by the organization. The goal of these programs was to lower stress and tension that may result from problems outside and inside the workplace. Such an approach was intended to improve the physical and mental health of employees, increase their satisfaction with work, and reduce turnover and absenteeism (Googins & Burden, 1987). Employees’ assistance programs were further expanded to cover employees' family, domestic, and workplace problems, and have become a part of the programs of human resources departments in many organizations as evidence has accumulated on the cost benefits of such programs in reducing costs and increasing employees' productivity and satisfaction (Berridge & Cooper, 1993). Employers realized that expecting employees to cope with their nonwork difficulties by themselves without any involvement by the organization was not effective. Evidence showed that employees who participated in employees' assistance programs improved their functioning in the organization. This was demonstrated in less absenteeism, fewer health insurance claims, and 4 less termination of employment of problematic employees. The rationale to use employees' assistance programs is entirely financial. In each organization, 10 to 12 percent of employees have personal problems that affect their productivity negatively (Csiernik, 1990). In fact, research has shown that decreases in productivity because of personal problems reach 25% (Masi, 1984). The costs of employees' assistance programs were found to be significantly less than the alternative direct and indirect costs (Wolfe, Parker, & Napier, 1994). For example, terminating the employment of employees with problems, including the direct and indirect costs of hiring new employees, was found to be more costly than assisting these employees in better functioning in their personal lives. As a result, more and more organizations have chosen employees' assistance programs as their strategy in dealing with a variety of employees' problems (Gerstein, 1991; Park, 1992). Despite promising evidence about the benefits of employees' assistance programs, organizations have experienced difficulty in encouraging employees to participate in these programs. This may be due to the sensitive nature of the issues involved in personal use of employees’ assistance programs and problems of confidentiality. Employees’ assistance programs can be used as a basis for supervisors to confront problematic employees and suggest that they seek assistance (i.e., supervisory referral) or as a resource for employees to use at their own initiative for problems they perceive (i.e., self-referral). Since participation in employees' assistance programs is clearly voluntary, participation depends on attitudes and motivation to participate. This study investigated the propensity of employees to self-refer, a subject that has been under-researched compared with the attention given to supervisory referrals (Milne, Blum, & Roman, 1994). Employees' tendency to participate in an employees' assistance programs can result from recommendations and suggestions of family members, friends, and co-workers, who are closer to them and are aware of their problems. It can also result from their own rationale when they realize that they cannot cope with their problems by themselves (Sonnenstuhl, Staudenmeier, & Trice, 1988). The way employees perceive employees' assistance programs can determine 5 participation in the programs and success (Milne, et al., 1994). Management should be interested in understanding employees' tendency to use an employees' assistance programs because low participation can cause failure of these programs and prevent the organization from utilizing their benefits. It is the interest of the organization to detect employees who need assistance and will be more willing to participate in these programs. From the employees' point of view the need for an employees' assistance programs is a subjective perception that reflects an objective need. From the organization's point of view, understanding of the determinants of the need to participate will facilitate its detection of employees who need assistance and will ensure a higher participation rate. The goal of this study was to propose and test a model of the determinants of the need for employees' assistance programs. This need was defined as multi-dimensional. One dimension was the need to address personal problems and the second, workplace problems. This model is intended to advance the understanding of employees' assistance programs. Much previous research was practitioner-oriented, with few attempts to develop a conceptual framework for these programs. Research model -------------------------------------Insert Figure 1 about here -------------------------------------- The research model is presented in Figure 1. It should be noted that the path coefficients in the Figure are the results of this study and will be described in the findings section. Figure 1 shows six independent variables related to employees' assistance programs that may be grouped in two categories. The first three represent variables that reflect the way one manages the work-nonwork interface. These are: organizational support, personal coping, and negative spillover. The second category represents variables that reflect one's work experience. They are occupational commitment, job satisfaction, and tenure. As can be seen in Figure 1, the relationship between the independent variables and the dependent ones, the need for 6 employees' assistance programs, is not direct but is mediated by two variables, namely worknonwork conflict and job stress. The two mediators have a reciprocal relationship. The three independent variables that represent managing the work-nonwork interface are related to worknonwork conflict, and the three independent variables that represent work experience are related to job stress. Finally, each of the mediators is related to the two dependent variables, namely the need of employees' assistance programs for personal issues and for workplace issues. The two dependent variables are correlated as shown in Figure 1. The following is a conceptual clarification of the proposed model. How variables that represent managing the work-nonwork interface are related to work-nonwork conflict Growing demands from either work or nonwork domains, including time-related and psychological pressures, have been shown to intensify work-nonwork conflict and increase the incidence of stress symptoms (e.g., Parasuraman & Alutto, 1984; Klitzman, House, Israel, & Mero, 1990; Cohen & Kirchmeyer, 1995; Parasuraman, Purohit, Godshalk, & Beutell, 1996). Work-nonwork conflict refers to an incompatibility of work demands with those of nonwork, whereby participation in one domain is made more difficult by virtue of participation in another (Greenhaus & Beutell, 1985). High demands from one domain can drain personal resources of time, energy, and commitment that are available for activities in other domains. Employer-based supports for nonwork such as flexible work schedules typically are implemented to help employees deal with the many demands of multiple domains. As illustrated in Figure 1, this form of interdomain management was hypothesized to act on work-nonwork, consistent with Cohen and Kirchmeyer's (1995) findings with nurses. In another study, Kirchmeyer (1995) also found only a weak correlation between the organizational support perceived by male and female managers and nonwork-work conflict. Frone, Russell, and Cooper (1992) suggested that employees hold their organizations responsible for work-nonwork conflict and accept less responsibility themselves. Therefore, the idea that interdomain management stemming from employers' efforts will affect mostly the interference from work seems justifiable. 7 For individuals who are active in multiple domains, effective interdomain management seems to be critical for enhancing domain experiences and reducing the conflict between domains (Hall, 1972; Beutell & Greenhaus, 1983; Kirchmeyer, 1993; Cohen & Kirchmeyer, 1995). These relations are consistent with the idea in the stress literature of personal control and active coping being the key to handling job demands and other potential stressors (for review see Ganster & Schaubroeck, 1991). Thus, this study expected that the extent to which certain personal coping strategies such as establishing priorities and scheduling role activities carefully are used by employees will correlate significantly with work-nonwork conflict. This expectation supports arguments about individuals having greater personal control over nonwork than work and being ultimately responsible for managing nonwork demands (Schonfeld, 1990; Gutek, Searle, & Klepa, 1991; Frone et al., 1992). Therefore, in our model, use of personal coping strategies was hypothesized to affect work-nonwork conflict. Kirchmeyer (1992) argued that resource depletion from nonwork involves decreasing not only the individual's capacity to meet work demands and the employee’s value to the employer, but the employee’s sense of personal competence as well. Attitudes about both work and nonwork also have been associated with certain personal strategies that individuals use to manage the work-nonwork interface (Hall, 1972; Kirchmeyer, 1993). Negative spillover may influence work-nonwork conflict through its negative effect upon domain experiences and as well decrease one's personal competence, a factor associated with work-nonwork conflict. All the above leads to the expectation that organizational support and personal coping will be positively related to work-nonwork conflict and negative spillover will be negatively related to it. How variables that represent work experiences are related to job stress. Symptoms of stress, including tiredness, nervousness, feeling blue, and lack of enthusiasm for life, have been associated with work experiences in several studies of men and women (Parasuraman & Alutto, 1984; Jamal, 1990). Concerns have been raised that both physical and psychological health might be influenced by job attitudes. The more favorable the individual's work experience, the more the employee identifies with, is involved in, and enjoys the 8 job and belonging to the employing organization. It can be expected that a person with high occupational commitment will experience less job stress. A person with higher occupational commitment will have a more positive perception of the job and its obligations, and this will reduce that person's tensions. Individuals who dislike their jobs may experience adverse health outcomes. These outcomes include both physical symptoms and psychological problems. Some indications of how people's job satisfaction can influence their physiology come from job stress research done in Sweden (Johansson, 1981). This researcher assessed catecholamine levels as people worked on machine-paced and self-paced jobs. The results showed that catecholamine, which is associated with stress and negative emotions, increased when people did machine-paced jobs and decreased when they did self-paced jobs. The researchers speculated that performing jobs that caused distress led to increase in stress. It seems reasonable to speculate that being in a dissatisfying job would have a similar effect. Models of job experience propose that employees’ reactions (e.g., job stress) are likely to vary systematically with tenure. The main theme guiding these models is that tenure affects the manner in which work environment features combine to influence job stress (Bedeian, Ferris, & Kacmar, 1992). Veteran employees are more experienced in the organization and in many aspects of their job. Their job security is probably stronger than that of younger employees. All of this should combine to reduce their job stress in comparison with less veteran employees. All the above leads to the expectation that occupational commitment, job satisfaction, and tenure will be negatively related to job stress. Reciprocal relationship As illustrated in Figure 1, a positive reciprocal relation between work-nonwork conflict and job stress was hypothesized. A similar hypothesis was advanced by Frone, et al. (1992), and was supported by structural equation modeling. As these authors argued, if the demands of one domain interfere with one's ability to fulfill obligations in another domain, then these unfulfilled obligations may begin to interfere with one's day-to-day functioning in the first domain. 9 Therefore, conflict between work and nonwork should be examined within the context of a reciprocal relation in order to fully understand the experience. The possibility of a reciprocal relation is also supported by the moderate positive correlation between the two directions that is reported consistently (Klitzman et al., 1990; Gutek et al., 1991; Cohen & Kirchmeyer, 1995). This reciprocity merits specific testing with a second sample of workers. The mediators and the need for employees' assistance programs. The expectation is that work-nonwork conflict and job stress will be positively related to the two sources of need for employees' assistance programs. The belief that interdomain conflict has serious unfavorable consequences for individuals and their organizations is common throughout the literature. Frone, et al. (1992), who tested the two directions of conflict, nonworkto-work conflict and work-to-nonwork conflict, found significant correlations with depression in both cases, but only nonwork-work conflict was significantly related directly to depression in the modeling. Given the large body of literature suggesting that work-induced stress leads to health problems (for review see Ganster & Schaubroeck, 1991), these latest findings seem surprising. In our model, based on the conceptual support and longitudinal evidence relating job conditions to mental health in teachers (Schonfeld, 1992), both work-nonwork conflict and job stress were hypothesized to directly affect the need for employees' assistance programs. Employees who report higher work-nonwork conflict and job stress will probably face more difficulties in their private as well as their working lives. As a result these employees will demonstrate stronger needs for employees' assistance programs that should assist them in coping with problems in their private and working lives. Method Participants. All 720 employees of a school district in Western Canada were surveyed. Questionnaires were sent by mail and 300 usable questionnaires were returned, a response rate of 42 percent. Of the sample, 33 percent were females. The mean age of the respondents was 42.1 years and mean tenure in the local unit (mostly schools), school district and occupation were 7.6, 11.2, and 14.2 years respectively. Sixty-seven percent of the sample had completed 10 their university education. Of the university graduates, 16.4 percent possessed an MA or PhD degree. Seventy percent of the sample was in teaching positions, 10 percent in administrative positions, and 20 percent in clerical, maintenance, or caretaking positions. Measures. Personal coping. This form of interdomain management was measured with eight items developed by Kirchmeyer (1993). Her factor analysis found these strategies to have a common underlying dimension and to represent a coping theme of having good personal organization and an appropriate attitude. Examples include "Overlap different roles whenever possible (such as participating in a sport that my kids enjoy and doing community service that really uses my professional skills)" and "Develop attitudes which put role demands in a positive light". A fivepoint scale ranging from (1) not typical of me to (5) very typical was applied. The resultant Cronbach's alpha of .72 was slightly lower than earlier values. Organizational support. Kirchmeyer's (1995) four-item unidimensional scale of an employer's respect for workers' nonwork participation was used to measure this form of interdomain management. Examples include "Is flexible about employees' work schedules (such as providing Flextime and not rigid about keeping certain hours)" and "Accommodates employees' special nonwork need (such as arranging time off for an employee to participate in a community event important to the employee)". The scale was anchored with five points ranging from (1) not typical of my organization to (5) very typical. The resultant Cronbach's alpha of .77 was consistent with earlier values. Negative spillover. Kirchmeyer's (1993) eight-item measure of negative nonwork-to-work spillover was used. Items began with the phrase "Being involved in nonwork activities". Examples of the continuation of the statement include "demands time from me that could be spent on my job", "makes me so irritable that I take it out on the people at work", and "makes it hard to adjust back to the way I must act at work". The instructions provided a list of nonwork activities for respondents to consider, including those related to hobbies and recreation, family and friends, and community and political associations. A five-point scale of agreement was 11 employed for the negative spillover scales. The resultant Cronbach's alpha of .81 was within the range of values reported by Kirchmeyer (1993). Occupational commitment This study utilized a definition and measurement of occupational commitment proposed and tested by Cohen (1993). Occupational commitment was defined as an affective attachment to the organization, expressed in one or more of the following three dimensions: Identification, adoption of the goals and values of the commitment objects as one's own; Affiliation, feelings of belonging to the commitment objects, being part of them; Moral involvement, internalization of the roles of the commitment objects, demonstrated in feelings of care and concern for them. The items for the Identification dimension were: "I find it very easy to identify with the objectives of this occupation", "Most of the values I believe in are demonstrated by this occupation", "There are a lot of similarities between my personal goals and the goals of this occupation". The items for the Affiliation dimension were: "I talk to my friends about how great it is to work in/belong to this occupation", "I am proud to work for/be a member of this occupation", " I feel myself a part of this occupation". The items for the Moral Involvement dimension were: "I take personally any problem that occurs in the occupation", " I become upset when things are not working out as they should in my occupation", "I really care about everything that happens in my occupation". A seven-point scale of agreement was employed ranging from (1) strongly disagree to (7) strongly agree. The variable of years in the organization was measured as an interval variable. The resultant Cronbach's alpha was .77. Job satisfaction. Respondents were asked to indicate how satisfied they were with six aspects of their employment: current job, co-workers, supervisors, current salary, opportunities for promotion, work in general. The scale for these questions ranged from (1) very dissatisfied to 5 very satisfied. This measure was developed by Schriesheim and Tsui (1980). Note that the reliability in this sample (.77) is slightly higher than that reported by Tsui, Egan, and O'Reilly (1992) in an American sample (.73). Finally, number of years in the school was used as an indicator for tenure. 12 Mediators Work-nonwork conflict. Shamir's (1983) six-item conflict measure was used. Items refer to the disruption of nonwork by work, and include "One of the difficulties of my job is that I am not home enough" and "Work often prevents me from participating in leisure activities that take place at the same time". Shamir (1983) reported a Cronbach's alpha of .78 and the one obtained in this ample was .80. Symptoms of stress. Physical symptoms such as headache and tiredness, and mood changes represent the most commonly reported symptoms of stress among workers, including teaching faculty (e.g., Brown, Bond, Gerndt, Krager, Krantz, Lukin, & Prentice, 1986). Three such symptoms, namely feelings of depression, tiredness, and nervousness, were examined in this study. Respondents were asked to indicate the frequency of each symptom according to a five-point scale. Reliability analysis revealed moderate consistency across the three symptoms (Cronbach's alpha = .78). Recent research indicates that such self-reports of health do not yield inflated estimated of stress effects despite earlier contentions that they may do so (James, Lovaho, & Khoo, 1994). The need for employees’ assistance programs Following Reed (1983), we divided the need for employees’ assistance programs into two categories: personal problems and workplace problems. The specific issues in each of the categories were also adopted from Reed. The respondents were asked to indicate their attitudes towards statements describing several personal and work issues in which organizations or other agencies could offer counseling and assistance programs to help employees to cope with their busy lives. Specifically they were asked to indicate how much they thought there was a need for a counseling and assistance programs on that issue in their school/workplace and district. The nine issues in personal problems were: (1) financial, e.g., bankruptcy, financial difficulties relating to inflationary costs, poor financial management; (2) legal, e.g., family and tenant landlord conflicts; (3) alcohol abuse; (4) drug abuse (5) mental health, e.g., depression; (6) familial, e.g., parent-child and parent-relative relationships; (7) marital, e.g., difficulties with communication 13 and problem resolution; (8) personal, e.g., difficulties with identity, relationships, and sex; (9) physical health. The nine issues for the workplace problems were: (1) compensation and benefits, e.g., sick days, health insurance coverage, vacation, overtime, sick leave; (2) retirement; (3) performance, e.g., disputes over the content of regular performance appraisal to specific disciplinary actions; (4) transfers and promotions; (5) policies and procedures, e.g., requests for familiarization with specific written corporate policies; (6) interpersonal relations, e.g., conflicts with supervisors or other employees; (7) career counseling; (8) discrimination; (9) working conditions. The scales ranged from (1) there is no need for counseling and assistance programs to (5) there is a very strong need for counseling and assistance programs. Reliability for the need for employees’ assistance programs for personal problems was .92 and for workplace problems .89. Data analysis LISREL (version 7.20; Joreskog & Sorbom, 1993) was appropriate for testing the proposed model, which was nonrecursive due to the reciprocal relation between the worknonwork conflict and job stress. The structural model as depicted in Figure 1 was evaluated using a 10x10 covariance matrix as input. Because a single-indicator approach was taken, we had to correct for random measurement error in the analysis (Netemeyer, Johnston, & Burton, 1990). That is, random error variance for each construct was calculated "by taking one minus its reliability times its observed variance" (Joreskog & Sorbom, 1993. p. 165). Testing for the hypothesized relations among the variables was based on the signs and statistical significance of the estimates of the model’s parameters. The overall fit of the model to the data was evaluated by five indicators. Experts generally advise that multiple fit criteria be examined. First, the chi-square goodness-of-fit statistic (Χ2) is used routinely and compares the predicted and observed covariance matrices, with a nonsignificant value indicating good fit. Second, the chi-square/degrees-of-freedom ratio Χ2/df is a common adhoc measure and is thought to be less susceptible to sample size effects. Although there is no consensus regarding what ratio constitutes an acceptable fit, a ratio below 14 2.0 falls within the most conservative estimates (Bollen, 1989). Next, LISREL's Goodness of Fit Index (GFI) and Adjusted Goodness of Fit Index (AGFI) measure the relative amount of variances and covariances jointly accounted for by the model. These indices range from zero to one, with values greater than .90 indicating acceptable overall fit. Fifth, the Comparative Fit Index (CFI) was recommended by Medsker, Williams, and Holahan (1994) in their recent review as the best approximation of the population value for a single model. It is normed to the range of 0 to 1, with values above .90 being acceptable. Normed Fit Index (NFI) was proposed by Bentler and Bonett (1980). Importantly, NFI is additive for nested model comparison. The closer its value to 1 the better the fit. Note that the reciprocal relationship with its identification constraints prevented the testing of an alternative model where all exogenous variables affected both work-nonwork conflict and job stress. Such a model could have served as a comparison for the proposed model. However, to ensure full identification, Frone, et al. (1992) advise that determinants be related to only one of the reciprocally related variables and that each such variable have at least two associated determinants. In this study, an adequately identified model to serve for comparison purposes would involve switching the determinants of one mediator with those of the other mediator. Because such an alternative model was supported neither theoretically nor by initial bivariate diagnostics (see results of correlational analysis below), it was not tested with the LISREL procedure. The magnitude of the path coefficients The fit of a given model to the data is an important criterion of the quality of the model but it does not necessarily imply that this model is the correct model (Saris & Stronkhorst, 1984). The path coefficients, their significance, and their magnitude provide an important criterion for model evaluation, termed the plausibility criterion. The plausibility of a model refers to a judgment made about the theoretical argument underlying the specified model (Saris & Sronkhorst, 1984). According to this criterion the decision regarding the correct model should be based on the theoretical correctness of the model demonstrated by its path coefficients. Accordingly, a model that fits the data well, but many of whose theoretical 15 paths do not support the theoretical arguments of the model, cannot be defined as a correct one. There has to be some balance between the fit indices and the theoretical predictions regarding the relationships among research variables. The accuracy of the theoretical predictions can be tested by the path coefficients in each of the models. Findings Descriptive statistics and bivariate correlations are presented in Table 1. There were several significant relations among the exogenous variables, although the correlation coefficients did not indicate concerns about multicollinearity. For example, organizational support correlated with both work-nonwork conflict (r = -.25 at p < .01) and job stress (r = -.25 at p < .01). Tenure failed to correlate significantly with any of the research variables, and job satisfaction correlated significantly with work-nonwork conflict and job stress (r = -.16 at p < .05, r = -.36 at p < .001). Relations between personal coping and negative spillover with the two mediating variables were more complex. That is, personal coping had a nonsignificant correlation with work-nonwork conflict and a significant one with job stress (r = -.23 at p < .01). Occupational commitment had a significant correlation with job stress (r = -.15 at p < .05) but not with work-nonwork conflict. Also consistent with earlier findings was the positive correlation between work-nonwork conflict and job stress (r = .33 at p < .01). Regarding significant relations between the mediators and the outcomes, work-nonwork conflict correlated with need of employees' assistance programs for personal issues (r = .18 at p < .01) and need of employees' assistance programs for workplace problems (r = .24 at p < .01). Job stress was also correlated with the need of employees' assistance programs for personal issues (r = .13 at p < .05) the need of employees' assistance programs for workplace issues (r = .20 at p < .01). _____________________ Insert Table 1 about here _____________________ As for the research model presented in Figure 1, according to the goodness-of-fit estimates, the proposed model provided a sufficient but not on outstanding fit to the data [X 2 (df=17) = 67.52, p ≤ .001, X2∕df = 3.97, GFI = .95, AGFI = .83, NFI=.82, CFI = .85]. The 16 difference between predicted and observed matrices did yield a significant X2, and X2∕df was above the conservative level of 2. Parameter estimates for the model are presented in Figure 1. Of the 12 relations proposed in the model, eight were significant in the expected direction. All the three nonwork domain determinants predicted work-nonwork conflict significantly. Organizational support was negatively related to this conflict (-.38; p<.05) meaning that less organizational support is associated with increases in work-nonwork conflict. Personal coping scores also predicted this conflict, indicating that more effective coping is associated with increases work-nonwork conflict (.27; p<.05). This direction of the relationship was not expected because we hypothesized that more successful coping would decrease work-nonwork conflict. Negative nonwork-to-work spillover was also a significant determinant of work-nonwork conflict. The path coefficient here was relatively high (.69; p<.05). Job satisfaction predicted job stress. As for the three-workplace determinants, only one was found to be a significant predictor of job stress. Lower job satisfaction was associated with higher job stress (-.43; p<.05). Occupational commitment and job tenure were not significantly related to job stress. Contrary to the hypothesis concerning reciprocity, only one of the paths between the two mediators proved to be significant, namely the path from work-nonwork conflict to job stress (.43; p<.05). The path from job stress to work-nonwork conflict was not significant. As regards the hypothesized effects of the two mediators on the need of employees' assistance programs, three of the four predicted paths were supported by the parameter estimates. Work-nonwork conflict was a significant determinant of both forms of the need of employees' assistance programs, for the personal issues (.22; p<.05) and for the workplace ones (.21; p<.05), whereas job stress was a significant predictor of the need of employees' assistance programs for workplace issues (.17; p<.05) but not for personal issues. In conclusion, the overall fit of the proposed model was not fully consistent with the observed data. The LISREL program calculates a "modification index" for every fixed parameter in a model. This index reflects the minimum reduction in the chi-square statistics if the parameter 17 is changed from fixed to free. The model tested here was revised on the basis of these modification indexes. Breckler (1990) and Cudeck and Browne (1983) argued that crossvalidation should be conducted whenever an initial model is modified on the basis of the data, namely the modified model should be assessed by using different data. Otherwise, they should be cautiously interpreted. Thus the model revised on the basis of the modification indices should be treated with caution. The modification indices of this paper may serve as an additional indicator for the fit of the model tested: higher modification indices in a model indicate a poor fit. The modification indices showed that the model would be significantly improved if the following paths were included: from occupational commitment to work-nonwork conflict (a reduction of 13.3 in X2); from job satisfaction to work-nonwork conflict (a reduction of 12.5in X2); from negative spillover to job stress (a reduction of 18.1 in X2); and from coping strategies to job stress (a reduction of 2.5 in X2). On the basis of these indices the model was revised. The rationale behind the revised model was first that there was no reciprocity between the moderators, and only one path, from work-nonwork conflict to job stress, was allowed. Second, the determinants were allowed to correlate with the two mediators. This becomes possible when the constraint of reciprocacy was freed. The goodness-of-fit estimates of the revised model provided better fit to the data than those of the first model [X2 (df=12) = 40.61, p ≤ .001; X2/df =3.38, GFI = .97, AGFI = .86, NFI=.89, CFI = .91]. However, in the revised model too the difference between predicted and observed matrices did yield a significant X2 and X2/df was above the conservative level of 2.0. The main problem with the revised model was its parameter estimates. Of the 17 relations proposed in the model, only seven were significant. Therefore, the conclusion is that the proposed conceptual model is superior to the revised one despite its somewhat lower fit indices. This is because the conceptual model had more significant paths relatively to the nonsignificant one. Also, in the conceptual model there was some relationship between the two mediators and more significant paths between the mediators and the need for employees' assistance programs. 18 Discussion The goal of this research was to propose and test a model of the relationship between variables that represent work and nonwork domains and employees' assistance programs. This under research area is important because the success of employees' assistance programs depends on the employees' willingness to participate in them. With the extraordinary increase in employees' assistance programs during the past decade has come a call for needs assessments for these programs (Grisson, Baldadian & Swisher, 1988) and evaluation of these program themselves. Stephen and Prentice (1978) were among the first to stress the importance of needs assessment in helping a company to set the direction and focus of its programs. This research may provide a conceptual framework for understanding the determinants of the need for employees' assistance programs. Very little is known about willingness for using employees' assistance programs and what can affect it. One of the few studies that have examined this is Milne, et al.'s (1994) paper, which differs from the present work in several aspects. First, these authors examined propensity to use employees' assistance programs, while our paper tested perceived need for it. We felt that a more indirect approach adopted to measuring perceived need could result in higher participation in the research itself, as well in more honest and unaffected responses. Second, our model suggested that employees' assistance programs are a multi dimensional concept: one dimension is the need of employees' assistance programs for personal problems and the second for workplace problems. Third, many of the determinants proposed here represent nonwork domains, while Milne et al. concentrated on workplace determinants. Fourth, we proposed a mediated model, as opposed to the direct one advanced by Milne, et al. In short, our design was somewhat more complex than that of Milne, et al. The research model received modest from the data. Some of the fit indices were not as strong as we expected. On the other hand, the conceptual relationship was supported quite strongly, as demonstrated in eight significant paths out of the twelve that were advanced. This finding shows the usefulness of the proposed model, which may be used as a conceptual basis 19 for developing and testing alternative models. An important finding of this research is the support for the multi-dimensionality of employees' assistance programs. This distinction, suggested by Reed (1983), shows that employees' assistance programs should not be treated as one concept and that assistance for problem solving related to one group of issues may be more fruitful than assistance for another. The distinction is important for both conceptual and practical reasons. Here we found that the model better predicted expressed need for employees' assistance programs in personal problems. It could be that in another sample, with a different occupational group, need for employees' assistance programs in workplace problems will be better predicted. Grisson, et al. (1988) found that employees' needs of employees' assistance programs vary considerably across firms. While more research is required to further clarify the nature of the relationships, the findings here provide some interesting contributions that should be considered in future studies. The main finding of this study is that it emphasizes the importance of nonwork domains on the need for employees' assistance programs. A closer look at the paths of the model readily shows that all the nonwork domain paths were significant. The three nonwork domain determinants were significantly related to work-nonwork conflict, and work-nonwork conflict was related to job stress and to the two dimensions of the perceived need for employees' assistance programs. The paths that were not significant were the workplace ones. This finding shows that nonwork affects are an important concept for understanding perceptions of the need for employees' assistance programs. That is, it the pressures of coping with the work-nonwork interface seem to be the main factor that confronts employees with problems, resulting in stronger needs for external assistance, namely employees' assistance programs. This is an important finding with important conceptual and practical implications. Conceptually, the findings call for more research on the relationship between variables that represent the interface between work and nonwork and need for employees' assistance programs. All the relationships found here suggest that the need for employees' assistance programs is a result of pressures arising from the work-nonwork interface. Milne, et al. (1994) 20 focused on organizational determinants as predictors of need for employees' assistance programs. Most of their determinants had to do with attitudes of employees at different levels in the organization, such as supervisors and top management, to employees' assistance programs. Our research took a more indirect approach and found that many aspects of nonwork, such as negative spillover, coping, and support, cause work-nonwork conflict and stress, and as a result strengthen the need for employees' assistance programs. Much more research in this direction is needed, which seems to be promising. Some of the specific relationships found here are worth noting. First, the strong relationship of negative nonwork-to-work spillover to work-nonwork conflict (.69; p<.05) emphasizes the strong relationship of nonwork perceptions on work, a consistent finding in this research. The finding that organizational support decreased work-nonwork conflict, and hence the need for employees' assistance programs, suggests that organizations can take an active role in developing mechanisms that will help employees to cope better with the work-nonwork interface, and this can reduce the perceived need for employees' assistance programs. Job satisfaction was the only workplace determinant that was related to job stress and employees' assistance programs. The mediator model advanced here also supported the strong relationship of nonwork domains on perceived need for employees' assistance programs. Work-nonwork conflict was a stronger mediator than job stress. It was related to all three determinants and to the other mediator (job stress), and it affected the two employees' assistance programs dimensions. The magnitude of the paths from work-nonwork conflict were modest (.21 and .22) but significant. This shows that while this particular mediator was related to need for employees' assistance programs there were probably other possible mediators in the relationship. Job stress was found to be a weaker mediator. It was related only to one determinant out of the three, it did not affect the other mediator, and it was related to only one of the two employees' assistance programs dimensions. This finding provided modest support for the model. Practically, the findings suggests that employees who experience domestic problems 21 express a stronger need for employees' assistance programs. There are some of implications from the organization's point of view. First, if organizations emphasize the purpose of employees' assistance programs as a program to assist employees better to manage their work-nonwork interface, more employees may participate in the programs. Second, employees who experience domestic problems that are reported to human resources management departments, such as divorce, can be referred to employees' assistance programs, and they may respond positively to such an invitation. It is to devise ways to detect employees who experience difficulties in managing their work-nonwork interface before these surfaces as problems at work or alcohol and drug abuse. But the possibility that those who need employees' assistance programs will have stronger tendency to participate in these programs presents a challenge that organizations should meet. Several limitations of this paper should be acknowledged. The cross-sectional study relies on a single source derived from self-report methodology. While this is common in organizational research it is nonetheless problematic. The fact that all the items of the questionnaire are being assessed from the same source at the same time may produce the effect of common method variance. While research on employees' assistance programs and their potential use as intended is still highly dependent on self-report measures, the threat of the problems associated with them to the results of studies such as this one is very real. However, much of interest can still be learned from this type of studies, provided that efforts are made to substantiate the validity of the data obtained, as was done in this study, and that their results are interpreted accordingly. 22 References Bedeian, A. G., Ferris G. R., & Kacmar K. M. (1992) Age, tenure, and job satisfaction: a tale of two perspectives. Journal of Vocational Behavior, 40, 33-48. Bentler, P. M., & Bonett, D. G. (1980) Significance tests and goodness-of-fit in the analysis of covariance structures. Psychological Bulletin, 88, 588-606. Berridge, J., & Cooper, C. L. (1993) Stress and coping in U.S. organizations: the role of employee assistance program. Work and Stress, 7, 89-102. Beutell, M. J., & Greenhaus, J. H. (1983) Integration of home and nonhome roles: women's conflict and coping behavior. Journal of Applied Psychology, 68, 43-48. Blum, C. T., & Roman, M. P. (1988) Purveyor organizations and the implementation of employee assistance programs. Journal of Applied Behavioral Science, 24, 397-411. Bollen, K. A. (1989) Structural equation with latent variables. New York: Wiley. Breckler, S. J. (1990) Applications of covariance structure modeling in psychology: cause for concern. Psychological Bulletin, 107, 260-273. Brown, R. D., Bond, S., Gerndt, J., Krager, L., Krantz, B., Lukin, M., & Prentice, D. (1986) Stress on campus: an interactional perspective. Research in Higher Education, 24, 97-112. Cohen, A. (1993) Work commitment in relations to withdrawal intentions and union effectiveness. Journal of Business Research, 26, 75-90. Cohen, A., & Kirchmeyer, C (1995). A multidimensional approach to the relation between organizational commitment and nonwork participation. Journal of Vocational Behavior, 46, 189-202. Csiernik, R (1990). Mediation and the workplace: creating and awareness within employees assistance programs. Mediation Quarterly, 8, 161-167. 23 Cudeck, R., & Browne, M. W. (1983) Cross-validation of covariance structures. Multivariate Behavioral Research, 18, 147-167. Frone, M. M., Russell, M., & Cooper, M. L. (1992). Antecedents and outcomes of work-family conflict. Journal of Applied Psychology, 77, 65-78. Ganster, D. C., & Schaubroeck, J. (1991) Work stress and employee health. Journal of Management, 17, 235-271. Gerstein, L. H. (1991) Counselling and the bystander-equity model of supervisory helping behavior: directions for EAP research. Journal of Counselling and Development, 69, 241247. Googins, B., & Burden, D. (1987) Vulnerability of working parents: balancing work and home roles. Social Work, 32, 295-300. Greenhaus, J. H., & Beutell, N. J. (1985) Sources of conflict between work and family roles. Academy of Management Review, 10, 76-88. Grisson, G. R., Baldadian, K. C., & Swisher, J. D. (1988) The case for needs assessment: a study in three disparate work settings, Employee Assistance Quarterly, 4, 75-82. Gutek, B. A., Searle, S. S., & Klepa, L. (1991) Rational versus gender role explanations for workfamily conflict. Journal of Applied Psychology, 43, 793-809. Hall, D. T. (1972) A model of coping with role conflict: the role behavior of college educated women. Administrative Science Quarterly, 17, 471-486. Jamal, M. (1990) Relationship of job stress and type A behavior to employees' job satisfaction, organizational commitment, psychosomatic health problems and turnover motivation. Human Relations, 43, 727-738. James, K., Lovaho, C., & Khoo, G. (1994) Social identity correlates of minority workers' health. Academy of Management Journal, 37, 383-396. 24 Johansson, (1981) Psycho-neuroendocrine correlates of unpaced and paced performance. In G. Salvendy & M.J. Smith (Eds.), Machine pacing and occupational stress. London, Taylor & Francis. Joreskog, K. G., & Sorbom, D. (1993) Structural equation modeling with the SIMPLIS command language. Hillsdale NJ: Erlbaum. Kanter, R. M. (1977) Work and family in the United States: a critical review and agenda for research and policy. New York: Russell Sage Foundation. Kirchmeyer, C (1992) Nonwork participation and work attitudes: a test of scarcity vs. expansion models of personal resources. Human Relations, 45, 775-795. Kirchmeyer, C. (1993) Nonwork-to-work spillover: a more balanced view of the experiences and coping of professional women and men. Sex Roles, 28, 531-552. Kirchmeyer, C. (1995) Managing the work-nonwork boundary: An assessment of organizational responses. Human Relations, 48, 515-536. Klitzman, S., House, J. S., Israel, B. A., & Mero, R. P. (1990) Work stress, nonwork stress, and health. Journal of Behavioral Medicine, 13, 221-243. Masi, D. A. (1984) Designing employees’ assistance programs, New York, American Management Association. Medsker, G. J., Williams, L. J., & Holahan, P. J. (1994) A review of current practices for evaluating causal models in organizational behavior and human resources management research. Journal of Management, 20, 239-264. Milne, S. H., Blum, T. C., & Roman, P. M. (1994) Factors influencing employees' propensity to use an employees’ assistance program. Personnel Psychology, 47, 123-145. Netemeyer, R. G., Johnston, M. W., & Burton, S. (1990) Analysis of role conflict and role ambiguity in a structural equation framework. Journal of Applied Psychology, 75, 148157. 25 Parasuraman, S. & Alutto, J. A. (1984) Sources and outcomes of stress in organizational settings: toward the development of a structural model. Academy of Management Journal, 27, 330-350. Parasuraman, S., Purohit, Y. S., Godshalk, & Beutell, N. J. (1996) Work and family variables, entrepreneurial career success, and psychological well-being. Journal of Vocational Behavior, 48: 275-300. Park, A. D. (1992) Client satisfaction evaluation: university EAP. Employee Assistance Quarterly, 8, 15-34. Powell, G. N., & Kotschessa, B (1995) Factors that influence professional employees' willingness to seek counseling. Psychological Reports, 77, 872-874. Reed, D. J. (1983) One approach to employee assistance. Personnel Journal, 62, 648-652. Saris, W. & Stronkhorst, H. (1984) Causal modelling in non experimental research: an introduction to the LISREL approach. Amsterdam: Sociometric Research Foundation. Schonfeld, I. S. (1990) Coping with job-related stress: the case of teachers. Journal of Occupational Psychology, 63, 141-149. Schonfeld, I. S. (1992) A longitudinal study of occupational stressors and depressive symptoms in first-year female teachers. Teaching & Teacher Education, 8, 151-158. Schriesheim, C., & Tsui, A. S. (1980) Development and validation of a short satisfaction instrument for use in survey feedback interventions. Paper presented at the Western Academy of Management Meeting. Shamir, B. (1983) Some antecedents of work-nonwork conflict. Journal of Vocational Behavior, 23, 98-111. Sonnenstuhl, W. J., Staudenmeier, W., & Trice, H. M. (1988) Ideology and referral categories in employees’ assistance programs research. Journal of Applied Behavioral Science, 24, 383-396. 26 Stephen, M. & Prentice, R. (1978) Developing an occupational drug abuse program. Rockville MD: Alchohol, Drug Abuse, and Mental Health Administration. Trice, H. M., & Sonnenstuhl, W. J. (1988) Drinking behavior and risk factors related to the workplace: implications for research and prevention. The Journal of Applied Behavioral Science, 24, 327-346. Tsui, A. S., Egan, T. D., & O'Reilly, C. H. (1992) Being different: relational demography and organizational attachment. Administrative Science Quarterly, 37, 549-579. Wolfe R., Parker, D., & Napier, N. (1994) Employee health management and organizational performance. Journal of Applied Behavioral Science, 30, 22-42. 27 Table 1 Basic statistics and correlation matrix (N=270-300) N Mean S.D. 1 2 3 4 1. Organizational support 292 2.23 .90 (.77) 2. Personal coping 296 3.62 .63 .18** (.72) 3. Negative spillover 291 13.67 5.30 -.08 -.30*** (.81) 4. Work-nonwork conflict 292 19.58 7.80 -.25*** -.05 .39*** (.80) 5. Occupational commitment 300 3.32 .50 .13* -.26*** -.07 .04 (.77) 6. Job satisfaction 294 27.15 5.46 .24*** .12* -.06 -.16** .35*** 7. Tenure 285 7.59 7.10 -.03 -.14* -.00 -.09 -.06 8. Job stress Need for EAP: 9. personal issues 296 10.44 3.01 .25*** .23*** -.37*** -.33*** .15* 289 3.14 .90 .05 .04 .04 .18** .27*** 10. workplace issues 287 3.20 .80 .06 .08 .09 .24*** .14* Note.- Internal reliabilities are shown on the diagonals in parentheses. *p.05 **p.01 *** p.001 5 . 28 Figure 1 Research model Organizational support -0.38* Personal coping 0.27* Work-nonwork conflict 0.22* Need for EAP Personal issues Negative spillover 0.69* 0.21* 0.49* Occupational commitment 0.20 0.52* 0.01 -0.00 Job stress Need for EAP Workplace issues Job satisfaction -0.43* Tenure 0.03 0.17*