Predicting Employee Turnover from Friendship Networks

Abstract

Employees (n=40) at a fast-food restaurant were surveyed about characteristics of their position and their level of satisfaction. Employees were then asked to report with whom they regularly communicated inside and outside the workplace and to indicate how close they were to employees with whom they were linked. Employee turnover was measured after three months had elapsed. A goal of the research was to replicate a model of employee turnover that predicts employees more central in their social network to be less likely to leave, and to test a social support explanation of the centrality model. The results indicated that employees who reported a greater number of out-degree links with friends were less likely to leave. The number of in-degree links with friends did not significantly predict turnover, and neither did network links with peers. Friendship prestige, measured by the number of in-degree links, was strongly correlated with relational closeness and amount of time spent with employees outside the workplace.

Keywords:

• Turnover
• Centrality
• Networks
• Friendship
• Peer
• Social Support

Notes

1. Research has shown (e.g., Feeley & Barnett, 1997; Mossholder et al., 2005) that measures of centrality, while conceptually distinct (Freeman, 1979), are highly correlated. In the current study, friendship out-degree was correlated significantly with betweenness (r=.74) and closeness (r=.45); friendship in-degree was correlated significantly with betweenness (r=.65), closeness (r=.99), and friendship out-degree (r=.39). Betweenness was measured using Borgatti, Everett, and Freeman's (1992) operationalization, which considers the number of pairs of links an individual goes between. Borgatti et al.'s measure of closeness was also used; closeness measures the minimum number of links required to get to all others in the network (see Freeman, 1979, for conceptualization; and Freeman et al., 1979, for operationalization).

2. The reciprocity rate for the current study was .30, as determined by UCINET. This indicates a low level of reciprocity between out-degree and in-degree links. Thus, in almost 70% of reported links, the focal other failed to report the same level of relationship as the ego. It is necessary to examine in- and out-degree networks separately, as overall degree typically “forces” reciprocation (e.g., Feeley, 2000; Monge et al., 1983). The authors acknowledge an anonymous reviewer for this methodological improvement.

3. Two additional logistic regressions were run using closeness and betweenness measures of centrality in friendship networks. High betweenness suggests that an individual has a high level of control of communication in the workplace, while closeness represents efficiency in communication, as one who has greater closeness can get and send information more quickly to others. A person with high betweenness may be considered a liaison, a gatekeeper, or a connector or linchpin. This individual fills a structural hole (Burt, 1992); that is, s/he connects two or more clusters or regions of density in the network. In this case, s/he has friends in different groups of people. Thus, it would be expected that individuals with high betweenness would be likely to give and receive social support and, therefore, remain an organizational member. Closeness as thus defined is unrelated to the measure of relational closeness used in the employee survey; network closeness examines how close one is to all others in the social network. Both analyses used hierarchical logistic regression with individual-level factors included as the first block (age, job satisfaction, job position, and race) and betweenness or closeness as the second block. Closeness centrality did not predict turnover, R ²Δ=.02, n.s. Betweenness centrality in friendship networks significantly predicted turnover, R ²Δ=.22, p<.05, OR=.90 (95% CI=.82, .99).

Additional information

Notes on contributors

Thomas Hugh Feeley

Thomas Feeley is a faculty member in Communication and Family Medicine at the University at Buffalo (UB), of the State University of New York

Jennie Hwang

Jennie Hwang is a doctoral candidate in Communication at UB

George A. Barnett

George Barnett is Professor of Communication at UB