Differentiating Perceptions of Intimate Partner Violence

Purpose. Women's empowerment is essential to understanding widespread gender inequalities that perpetuate violence and stall development. Perception of domestic violence is frequently used as an indicator of women’s empowerment (Williams, 2005). It is most frequently constructed as a binary variable, where women either justify wife beating or they do not.  In the Nepal Demographic Health Survey (NDHS, 2006-2007), empowerment is considered high if the woman answers no (i.e., does not justify wife beating) to the following hypothetical circumstances: burning food, refusing sex, going out without permission from husband, neglecting children and arguing with husband.  If the women answers “yes” to any of the circumstances empowerment is considered to be low.  While these questions attempt to tap empirically supported constructs in gauging level of empowerment, a "yes" to one question may not always mean that the respondent justified wife beating in another instance. This study develops a Latent Class Analysis (LCA) model to determine if there are categories of women who endorse wife beating in some circumstances but not others. Differentiating women's perceptions of violence will thus increase understanding of the nuances of women's empowerment and lead to the development of improved measures.

Method. Latent Class Analysis (LCA) using data from the 2006-2007 Nepal Demographic and Health Survey (N=10,793) was conducted. The five items used for the analysis ask women’s perception of wife beating in the five circumstances listed above. The different class solutions to be tested explored all variations of the responses to these items in order to determine whether certain women justified wife beating in all, some, or no circumstances.  Mplus 6.11 was used for this analysis.

Results. Latent class analysis with a four class solution was chosen as the best fitting among the two through five class solutions examined. The four class solution was the only one that had a non-significant likelihood ratio, suggesting that the model is a good fit for the data (X²=9.8, df=7, p=0.19). The four class solution had smaller AIC and BIC compared to the five class solution. Class 1 (2.1%) and four (83.4%) were very distinct such that women in Class 1 endorsed wife beating for almost all the reasons and women in Class 4 did not endorse wife beating for any of the reasons. These two classes can be thought of as the most and least tolerable to domestic violence. Women in Class 2 (8.1%) slightly endorsed all of the reasons whereas women in Class 3 (6.5%) endorsed wife beating for “Going outside without informing her husband”, “neglecting children”, and “arguing”. However, they did not endorse wife beating for “refusing sex” and “burning food”.

Implications.  It is apparent that Nepali women justify domestic violence differently. Simplifying these differences by using a categorical variable for justification of wife beating as an indicator of women’s empowerment can lead to biased results. The finer gradation of perception of domestic violence can lead to better understanding of women's empowerment.