Measuring gender differences in partner violence: implications from research on other forms of violent and socially undesirable behavior

Sex Roles: A Journal of Research,  June, 2005  by Sherry L. Hamby

• E-mail
• Print
• Link

Partner violence research has, to date, primarily relied upon fairly brief behavioral checklists, such as the Conflict Tactics Scales (Straus et al., 1996), to identify victims and perpetrators. This is not unusual in epidemiological research (Campbell, 2000)--in fact at 78 items the CTS2 is longer than many epidemiological instruments. Use of such instruments has proved successful in many respects, including establishing partner violence as a major social problem and identifying many risk factors for partner violence. It has proved less successful where more precise estimates are required, including not only the unresolved debate about gender differences in partner violence, but also in determining whether there are racial or other group differences, and in monitoring change following intervention.

Evidence from criminology and delinquency research, as well as from partner violence research, suggests that brief instruments may have fairly high rates of both false negatives and false positives. If many participants are classified inaccurately, then even large gender differences could be obscured, particularly for relatively rare phenomena like partner violence. An appreciation of these sources of error is needed in order to evaluate the quality of partner violence data and to generate suggestions for improving measurement. There is currently more data available for overall estimates of imprecision than there are for gender differences in imprecision. These data indicate that there are numerous sources of error for data on violence, and it is likely that these sources interact with each other in complex ways.

THE FALSE NEGATIVE PROBLEM

False negatives are caused by forgetting, concealing, and other sources of underreporting. This is a significant problem for partner violence research and other research on sensitive topics. For example, drug use, abortions, and same-sex sexual behavior are typically underreported in surveys (Fu, Darroch, Henshaw, & Kolb, 1998; Turner, Ku, Rogers, Lindberg, Pleck, & Sonenstein, 1998).

Failure to Report Known Incidents

Several studies have identified a sample of known victims and assessed whether they would report their victimizations to interviewers who did not tell the respondent why they were recruited into the sample. These reverse-records check studies generally find a great deal of underreporting. One study found that people reported only 29% of assaults to interviewers that they had previously reported to police (Czaja, Blair, Bickart, & Eastman, 1994). This was worse than the overlap for robbery and burglary and suggests that assaults may be particularly hard to assess. An early study for the National Crime Survey (the precursor to the NCVS) found that only 63% of incidents already known to the police were later disclosed to interviewers (Turner, 1972). Two studies of known child sexual abuse victims who were re-interviewed as adults found that more than a third did not report a previously documented sexual victimization (Widom & Morris, 1997; Williams, 1994). The same type of study has also been conducted with reports of offending. It will come as no surprise that many people with documented histories of arrest do not disclose these to researchers (Huizinga & Elliott, 1984; Maxfield, Weiler, & Widom, 2000). One study found that young males failed to disclose anywhere from 33% to 68% of arrests for various offenses. In particular, 57% of males arrested for partner violence did not disclose this to interviewers (Babinski, Hartsough, & Lambert, 2001).