Academic journal article Journal of Marriage and Family

Comparing One-Partner and Couple Data on Sensitive Marital Behaviors: The Case of Marital Violence

Academic journal article Journal of Marriage and Family

Comparing One-Partner and Couple Data on Sensitive Marital Behaviors: The Case of Marital Violence

Article excerpt

Data on sensitive marital relationship issues such as marital violence, marital conflicts, or sexuality are often subject to considerable reporting bias. This reporting bias can affect validity and prevalence estimates, as well as correlations between the sensitive issue under consideration and other variables. Couple data render it possible to assess some (though certainly not all) reporting bias effects on results. The following analyses focus on marital violence but are applicable to research on other sensitive topics.

Many surveys dealing with sensitive marital behaviors, including the 1975 and 1985 National Family Violence Surveys (NFVS), lack empirical evidence on concurrent validity (Egley, 1991; Straus, 1992a). Studies comparing spouses' responses about marital violence have relied on small and often nonrepresentative samples (Jouriles & O'Leary, 1985; Malone, 1992; Szinovacz, 1983) or on clinical populations (Browning & Dutton, 1986; Edleson & Brygger, 1986; Langhinrichsen-Rohling & Vivian, 1994). Results from these studies indicate considerable disagreement between spouses on whether and what types of violence occurred. Larger community samples of premarital and recently married couples have revealed moderate disagreement between spouses (McLaughlin, Leonard, & Senchak, 1992; O'Leary et al., 1989). Relations of one-partner and couple data to predictors of violence appeared to be similar. However, this latter finding may be due to the small samples used for these studies. Given the NFVS' exclusive reliance on one-partner data, further clarification of these issues seems essential. This article reassesses these issues on the basis of a nationally representative data set, the National Survey on Families and Households (NSFH).

ONE-PARTNER AND COUPLE DATA

One-partner data are based on responses from only one marital partner. Spouses can give valid information about the other partner for some objective characteristics such as socioeconomic background and marital history. However, they usually cannot provide valid data on their partner's attitudes or perceptions of marital relations and interactions, because their attributions to the other partner may differ from responses the partner would have given (Thompson & Walker, 1982). Similar discrepancies may occur in reporting socially disapproved behaviors. In contrast, studies relying on couple data obtain information about the same variables from both partners or from interactions between partners. Couple data thus allow comparing spouses' answers to the same questions and enable the researcher to assess differences in spouses' perceptions of the marital relationship ("his" and "her" marriage). In addition, couple data permit a dyadic approach to data analysis: Relevant background and attitudinal variables available from both spouses can be used in the analyses. Moreover, couple data are usually necessary to measure and/or construct relationship characteristics such as spouses' agreement. Finally, couple data are not only of theoretical relevance, they also serve methodological purposes. Couple data render it possible to estimate underreporting of socially undesirable behaviors, to assess which variables are associated with underreporting, and to show whether associations between explanatory and dependent variables are affected by selective reporting. The focus of this article is on the methodological uses of couple data, applied to the measurement of marital violence.

DISCREPANCIES BETWEEN SPOUSES' RESPONSES

Generally, discrepancies between spouses' answers derive from three sources: random measurement error, systematic measurement error, and differences between spouses' perceptions of the relationship. In studies relying on self-reports, random measurement error is typically caused by ambiguities in the questions or scale items. For example, if a scale item refers to several behaviors (e.g., hit or throw things at the other), one spouse may answer in regard to the first behavior (hitting), whereas the other spouse may refer to both (hitting and throwing). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.