Emotions Running High: Political Stress and Domestic Violence Reports

Article by Anya Syed

Prior studies show that when local NFL teams lose, especially when unexpected, reports of at-home violence by men against their partners increase by 10% (Card and Dahl 2011). Research also shows that Americans increasingly treat politics as a sport where they identify heavily with the wins and losses of their team. We investigate whether election results also lead to changes in domestic violence as political frustration is vented against vulnerable members of a household.

To understand whether political frustration leads to domestic violence, we collected local data from 911 calls in Charleston County, South Carolina following the 2016 US Presidential elections. We matched this data with political preferences within one politically divided metropolitan area. By comparing the levels of domestic violence between areas that voted for Trump and Clinton, we shed light on whether political frustration leads to an increased frequency of domestic violence calls.

We found a decline in 911 domestic violence calls in census tracts supporting the electoral victor. In areas that favored Trump, Republican voters are less frustrated, hence domestic violence calls were lower relative to census tracts that voted predominantly for the losing candidate. There was a decrease domestic violence reports in Republican leaning areas after the election.

During the 2016 election, Charleston County voted for Clinton with a slim majority of 50.6%. The county has a population of 134,637, and it is 70% white. The map below shows the election results with the percentage of individuals in each tract who voted for Trump.

We first analyzed how domestic violence reports change around elections as public events, which can trigger strongly competitive emotions and loyalties to partisan teams.

The map below depicts the pre-election and post-election 911 calls in census tracts surrounding Charleston, South Carolina.

We found that domestic violence 911 calls decrease after elections, as shown by the majority orange-colored census tracts. Using a moving average model, our analysis of daily data, which ranged from May 5, 2016, to May 13, 2017, indicated a positive trend before elections, and a negative trend in the aftermath of elections, showing a decrease in domestic violence reports in
the aftermath of elections. The theory could be that elections fuel tensions when most contentious, and, after, things begin to settle. The decline manifested more in areas supporting the victor, which corroborates the sports case study. Therefore, this demonstrates that political stress is different and longer lasting than momentary emotions from sports.

For the second part of the analysis, we explored the temporal effect patterns of elections and whether partisan loyalties behave in the same way as team attachments in sports, leading to the increase in domestic violence calls that Card and Dahl had associated with team loss.

We aggregated our daily 911 calls into weekly data with the dependent variable being the percent change in 911 calls and the main independent variable being the votes cast for the winning candidate, Donald Trump. While domestic violence data was available at the census tract level, election data was provided only at the voting precinct level. To overcome the problem, we matched voting precincts to census tracts by calculating the area overlaps between these two partitions.

An analysis of weekly changes in 911 calls in the pre- and post-election period, at weeks 1, 2, 3, 4, 5 and 6, revealed a negative and statistically significant effect at week 4 and 5.

The decline in domestic violence reports peaks within a month after the elections. The effect of elections, however, starts to fade away in the sixth week after the elections. This finding suggests that political stress has a lasting effect and dissipates over time longer than negative emotions associated with sports, which are sudden and short-lived. The results align with the hypothesis.
In the census tracts where Trump won, Trump voters are presumably less frustrated, hence there are fewer domestic violence calls.

For further analysis, it would be necessary to expand the geographic coverage of 911 call data. Future studies could also consider how alcohol sales fluctuate during particularly contentious elections and this influence on domestic violence 911 calls – a similar model as Ivandic, Kirchmaier, and Blas.

Warren Roberts contributed the calculation of precinct-tract overlaps.