Four Measures of Partner Violence: Construct Similarity and Classification Differences

Abstract

Although measures of physical are commonly used to classify individuals as sustained violence or having sustained violence, little is known about the interrelationships among measures and whether variation across measures is random or systematic. In this study, 224 female undergraduates completed four different measures. Confirmatory factor analysis indicated that a one-factor model best fit the data, but that the measures did not represent the construct equally well (i.e., the congeneric model was significantly better than tau-equivalent and parallel measurement models). Frequency measures were more strongly associated with the latent construct, partner violence than were severity measures. Some variation across measures appears to be systematic; participants reporting milder and more infrequent were classified most inconsistently. Single measures may classify individuals unreliably. Physical between partners has been measured in many different ways. Although most measures rely, to some degree, on the reports o victims or perpetrators, differences among kit of reports may lead to variations in what is identified as a physical assault (cf. Koss et al., 1994; Straus, 1990; Walker, 1984). For example, there has been considerable discussion of the merits and liabilities of using a predetermined set of violent acts when asking respondents about in their relationships (e.g., Straus, 1990). In addition to asking respondents about a set of behaviors. other common strategies include asking them to make a summary judgment about their total experience of (e.g., Bergman, 1992; Leonard & Blane, 1992) and asking for detailed descriptions of specific incidents (e.g., Dobash & Dobash, 1984; Frieze & McHugh, 1992; Walker, 1984). Major reviews of the field (e.g., Browne, 1993; Hotaling & Sugarman, 1986; Koss et al., 1994) often discuss methodological issues but accept the premise that each of the varying approaches to measuring refers to the same underlying construct (although not necessarily equally well). Olson (1977) has provided a framework that helps to highlight same of the differences found in measures. His framework has two dimensions, frame of reference (insider or outsider) and type of data (subjective or objective). Insider reports are self-reports, and one of the advantages of his model is that he identifies two types of these. The Conflict Tactics Scales (CTS) and similar measures (e.g., Hudson & McIntosh, 1991; Shepard & Campbell, 1992), which focus respondents' attention on a specific set of behaviors, fall into Olson's insider category. These types of behavioral self-reports have been the subject of both praise and criticism in the family literature (cf. Straus, 1990). Another common approach involves asking about the overall aggregate frequency of violence. These self-reports typically allow respondents to determine what acts constitute violence. Critics of behavioral self-reports believe this approach, called subjective insider by Olson, will be more inclusive because unusual kinds of can be measured, but some loss of specificity occurs. Pure outsider reports, of course, are difficult, if not impossible, to obtain in research on because of the practical and ethical problems related to observing such acts directly, especially in any systematic manner. Thus, most measures that try to adopt more of an outsider's perspective are typically hybrids consisting of an outsider's judgment of participants' descriptions of violent episodes provided in interviews or questionnaires. In the literature, these judgments have typically been codified in some way (e.g., Frieze & McHugh, 1992) and, thus, belong on the objective side of the continuum. The most important advantages of outsider approaches to research are an increase in the amount of contextual information obtained from respondents and an attempt to reduce some of the potential biases of pure self-reports. …