Association for Public Policy Analysis & Management

There are over 600,000 incidents of domestic violence in the United States every year, and from 1998-2017, about 12% of homicides were known to be perpetrated by an intimate partner (FBI Uniform Crime Reports). In this paper, I estimate the causal effect of temporary shelter availability on intimate partner homicides. Existing literature documents the effect of income shocks, emotional cues, and police behavior on the incidence of domestic violence. However, little is known about the effect of direct services for victims and the provision of these services is a straightforward, actionable policy choice.

I construct county-level intimate partner homicide rates from 1998-2017 using the FBI UCR Supplementary Homicide Reports. Additionally, County Business Patterns data provides the number of temporary shelters in each county, which explicitly includes but is not limited to emergency domestic violence shelters.

In a two-way fixed effects model (county and year), I use changes in the number of shelters in a county to identify the marginal effect of an additional shelter in a given year on a county’s intimate partner homicide rate in the following year. Using lagged shelter counts allows time for changes in shelter availability to have an effect on whether a victim is able to escape an abusive relationship before the violence would have escalated to the level of a homicide. The effect is identified using within-county variation, with 46% of counties experiencing a change in the number of shelters and 25% of counties experiencing a change in whether they have any shelters at all.

Results suggest that each additional shelter in a county in a given year prevents one woman per 20 million people in that county from being killed by her intimate partner in the following year, about a 1.5% reduction. This effect is robust to the inclusion of time-varying demographic controls for the age distribution and the racial and ethnic makeup of the county. I find no effect on the male intimate partner homicide rate. In a placebo test, I find no effect on other homicides, and no effect of future shelter counts on intimate partner homicides.