Discussion

This paper asks a practical primary prevention question: how are attitudes in different domains associated with attitudes that normalise, minimise or excuse violence against women? Using the weighted Victorian NCAS 2021 sample, the analysis tests a theoretically specified, cross-sectional model linking attitudes towards gender inequality (AGIS), understanding of violence against women (UVAWS), psychological distancing, and violence-supportive attitudes (AVAWS). SEM allows these relationships, including proposed indirect pathways, to be examined simultaneously. It does not establish the temporal or causal sequence through which attitudes form or change.

Among the variables included in the model, AGIS has the strongest association with AVAWS after accounting for demographics and the proposed mediators. This is consistent with the centrality of gender inequality in primary prevention frameworks (1, 2, 7, 10). Practically, the result supports addressing beliefs about gender inequality alongside education or messaging about violence. It does not show that changing gender-inequality attitudes alone will shift violence-supportive attitudes at scale or reduce violence.

Within AGIS, Deny Inequality has the strongest association with AVAWS, followed by Reinforce Gender Roles. One plausible interpretation is that denial of inequality can function as a justificatory frame: if gender inequality is assumed not to exist, claims about gendered violence may be more readily dismissed as exaggerated, unfair or ideologically motivated. Combined with endorsement of rigid gender roles, this worldview may weaken the perceived legitimacy of women’s experiences and collective responsibility for prevention. This interpretation is consistent with, but not directly tested by, the model.

The gender pattern is more nuanced than a simple ‘men hold more harmful attitudes’ account. Men have lower AVAWS scores on average. In reduced models, the AGIS–AVAWS association is slightly stronger for the combined women and non-binary group than for men, contrary to the hypothesised direction; however, the interaction is not statistically significant in the full model. The result should therefore be treated as tentative. Measurement explanations, including ceiling effects or differences in how the scales operate across groups, remain plausible. Testing measurement invariance would strengthen future comparisons, and the combined category does not support separate conclusions about non-binary respondents.

UVAWS partially mediates the AGIS–AVAWS relationship in the statistical model: stronger recognition of violence and its gendered nature accounts for part, but not all, of the association between attitudes supporting gender inequality and violence-supportive attitudes. Because the data are cross-sectional, this pattern is consistent with a possible indirect pathway but does not establish causal mediation. It nevertheless aligns with prevention frameworks in which community knowledge is important but insufficient in isolation. For Victorian practice, the finding supports testing violence-literacy initiatives alongside explicit work to address denial of inequality and rigid gender-role endorsement, rather than assuming knowledge change will automatically translate into broader attitude change.

The model also identifies a distinct psychological distancing pattern. People with more gender-unequal attitudes are less likely to acknowledge violence as a problem in their own suburb or town, and local acknowledgement accounts statistically for part of the AGIS–AVAWS association. This is consistent with research describing how violence against women can be framed as something that happens ‘elsewhere’ or among ‘other’ groups through cultural othering, stereotyping and moral boundary drawing (14–16), as well as social identity processes that protect a positive image of one’s own community (11). The association is smaller than that involving UVAWS, and Distancing is measured by a single reverse-coded item administered to a subset of respondents. Its persistence identifies local relevance as a promising area for further measurement and intervention testing, rather than demonstrating that place-based messaging is effective.

The association between UVAWS and Distancing suggests that narrower recognition of violence can co-occur with a tendency to see violence as ‘not here’; it does not establish that one causes the other. Both are also associated with attitudes supporting gender inequality. Taken together, the model suggests that responses to violence may involve both definitional questions – what people recognise as violence – and psychological distance from the issue. This provides a rationale for testing whether efforts to broaden recognition of coercive control and other non-physical forms of violence are strengthened when prevention is also made locally relevant and socially shared.

Limitations

This analysis uses secondary, cross-sectional self-report data from the 2021 NCAS. It provides a point-in-time account of associations and cannot establish temporal order, causal mediation, intervention effects or behaviour change. The data were collected during the COVID-19 period and should not be treated as current estimates of Victorian attitudes. Participation, survey coverage, item non-response, social-desirability bias and topic sensitivity may also have influenced responses. The use of validated observed composite scale scores is appropriate to the analysis but does not remove potential conceptual overlap or shared-method effects between measures.

The Distancing construct is represented by one reverse-coded item administered to a subset of the sample and may capture local awareness or perceived prevalence as well as psychological distance. Gender comparisons are constrained by the available categories and sample sizes, including the combination of women and non-binary respondents; separate conclusions about gender-diverse respondents cannot be drawn. Although the models incorporated survey weights, they include only selected NCAS attitudes and demographics and do not comprehensively account for other individual, relational, community and structural factors relevant to violence and its prevention, including alcohol and other drug use, economic stress, employment conditions, housing insecurity and other social determinants. The findings should therefore be used to generate and refine prevention hypotheses, not as evidence that changing any single attitude will reduce violence.