By identifying and combining variables with similar question wording across DHS, MICS, and PMA, we created three comparable compound variables that measure physical, psychological, and sexual IPV in the last 12 months.

To achieve comparability, we adjusted the subpopulation of women included. In PMA, only currently married or cohabiting women were asked IPV questions, whereas DHS and MICS asked ever-partnered women. DHS and MICS asked about violence ever experienced, with follow-up questions about the frequency in the past 12 months, while PMA only asked whether the behavior occurred in the past 12 months by the current partner. With a bit of coding, we easily imposed a consistent subpopulation and time frame. To assist other researchers, we have made our code available on the IPUMS Global Health GitHub page [7].

Another comparability issue arises when surveys offer summary measures that group abusive behaviors in different ways. PMA opted for a short set of physical IPV questions, combining multiple behaviors into a single query, while DHS and MICS asked separate questions about various violent behaviors. We created a new summary measure for physical IPV, whose components matched across the surveys.

For psychological IPV, DHS and MICS asked, among other questions, whether the partner “threaten[ed] her or someone close to her with harm”, while PMA did not. We therefore excluded this threatening behavior when grouping responses to create a comparable compound measure of psychological IPV.

The sexual IPV variables posed the most significant challenges to comparability. PMA, DHS, and MICS surveys differ in whether they asked about sexual acts obtained through physical force versus coercion without force. We could make only one comparable sexual abuse variable: physical forcing of sex in the past 12 months.