pervision, overall performance, and willingness to rehire the employee. Further, this scale successfully distinguished between top and bottom performers for customer relations, supervision, and overall quality ratings.
Another modified violence scale, Interpersonal Cooperation, is included on the Employment Productivity Index (EPI) (London House, 1986), a measure designed to screen in successful employees. In one study, 167 job applicants completed the EPI, and supervisory ratings were then obtained for these employees three months after they were hired ( Joy & Frost, 1987). Significant correlations were obtained between the scores on this modified violence scale and supervisory ratings of customer service, employee relations, compliance with safety rules, employee conduct, and productivity. In addition, this scale was related to being angered or annoyed, trying to get along with coworkers, and arguing with supervisors. Rafilson ( 1988) describes another study in which 247 employed college students completed the EPI and an anonymous questionnaire concerning their work behaviors. Significant correlations were obtained between the modified violence scale and counterproductivity, time deviance, coworker relations, drug abuse, waste and damage of merchandise, and a composite measure of work quality. Thirty-two percent of this sample admitted to having argued at least occasionally with customers, coworkers, or supervisors.
Several different studies have been used to establish the relationship between standardized preemployment violence scales and on-the job violence (Table 13.2). Multiple approaches, using different study designs, are desired in any program of research to ensure that the conclusions from the studies are not due to the same type of method. For example, in one type of design, a contrasted- groups approach was used to show the PSI Nonviolence scale successfully distinguished between a group of convicted felons and drugstore and department store applicants ( Jones & Terris, 1981). In another study ( Moretti, 1980), warehouse and supermarket employees were asked to complete anonymously an inventory describing their damage and waste of company property and merchandise. Substantial validity coefficients were found between their PSI Non- violence scores and the admitted dollar amount of damaged, wasted, and stolen company merchandise or property. In yet another study ( Jones, 1983), the number of times employees argued with supervisors, coworkers, and customers was highly correlated with PSI Nonviolence scores. Using a construct validity approach ( Frost & Joy, 1987), the PSI Nonviolence scores of nursing home employees were found to be related significantly to a number of clinical diagnostic scales, including hostility, projection, and adjustment to others. These studies show that on-the-job violence can be predicted when an appropriate standardized instrument is used.
A review of the literature has shown that on-the-job violence can be predicted. Previous reviews and studies have yielded contrary conclusions, in large part