![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
The attitudinal effects of mass shootings and terrorist attacks are well documented, but their electoral consequences have not been studied systematically in the American context. We derive a set of expectations about mass violence and electoral behavior from terror management theory and issue ownership theory. We then examine the effects of mass shootings and terrorist attacks on United States presidential elections from 1976 through 2016. We find mass shootings increase turnout in the counties where they occur. Furthermore, terrorist attacks reduce the Democratic presidential candidate's shares of the two-party vote at the county level. Both of these effects decay sharply as the time between the event and Election Day increases. These findings illustrate the power of mass violence to influence American electoral behavior, conditional on the type of violent event and its proximity to Election Day.
Acknowledgements
We thank Scott Althaus, Andrew Bean, Brian Gaines, Peter Nardulli, Daron Shaw, and the editor and anonymous referees of the Journal of Elections, Public Opinion and Parties for their helpful comments and suggestions.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Others make more sweeping claims about the societal impact of war and violence. For example, Turchin (Citation2016) contends wars change human psychology at a deep level, leading to more complex forms of social organization.
2 Blair and Schweit (Citation2014) estimate 1,043 non-perpetrator fatalities across 160 incidents between 2000 and 2013; Krouse and Richardson (Citation2015) count 1,554 non-perpetrator fatalities across 317 incidents.
3 Some argue mass shootings have increased starkly in recent decades (e.g. Cohen, Azrael, and Miller Citation2014), while others dissent (e.g. Fox Citation2015).
4 Similar patterns of issue ownership manifest for the Conservative and Labour parties in the United Kingdom (Budge and Farlie Citation1983).
5 Throughout this study, “counties” refers to counties, independent cities, parishes in Louisiana, and the District of Columbia. These are the most local units for which we can obtain comprehensive electoral data for the continental United States. Our theory is not predicated on the selection of counties in lieu of some other comparable areal unit, nor does it rely on the salience of some individual-level “county identity.” Rather, county boundaries represent a convenient, albeit rough, index for where mortality salience stemming from mass shootings and terrorist attacks will likely be greatest.
6 This asymmetry precedes the current spate of mass shootings. Ronald Reagan opposed gun control laws and ended any push for such legislation on Capitol Hill as president. George H. W. Bush continued in the footsteps of his predecessor and issued veto threats against gun control bills. The Republican Party’s stance on gun-related issues paved the way for Bill Clinton to campaign on and fully steal the issue (Holian Citation2004).
7 We exclude shootings in which fewer than four non-perpetrators were killed and terrorist attacks in which no non-perpetrators were killed. Four fatalities is an arbitrary cutoff point, but it is the arbitrary cutoff point commonly used by academics and journalists for defining mass shootings (Follman Citation2012; Fox and DeLateur Citation2014). Twelve counties were coded as having suffered both a mass shooting and a terrorist attack (sometimes representing a single event) in the same term; excluding these overlapping cases from our analyses does not substantively change our results.
8 The Mass Shootings in America database is available at http://library.stanford.edu/projects/mass-shootings-america. The Guide to Mass Shootings in America is available at http://www.motherjones.com/politics/2012/07/mass-shootings-map.
9 The Global Terrorism Database is available at http://www.start.umd.edu/gtd.
10 The Atlas of U.S. Presidential Elections is available at http://www.uselectionatlas.org.
11 The SEER Program data are available at http://www.seer.cancer.gov/popdata.
12 Breusch-Pagan and Hausman tests indicated fixed effects models were more appropriate than random effects models or models with neither fixed nor random effects. In instances where counties merged with or split off from one another during the timespan of our study (such as Alleghany County, Virginia, or Broomfield County, Colorado), we treated the pre- and post-change counties as distinct for the purpose of calculating fixed effects. Due to a lack of geographical precision in the Global Terrorism Database and SEER data, we were forced to treat the counties comprising New York City (Bronx, Kings, New York, Queens, and Richmond) as a single county for the purposes of our analyses. Removing this “super-county” from our data set (which also effectively removes the September 11 attacks in New York) does not substantively change our results.
13 Because log(0) is undefined, we code the “days before” variables as 1 for events on Election Day, as 2 for events on the day before Election Day, etc.
14 We do not include a main effect for logged number of days because the variable is undefined for county-years without a mass shooting or terrorist attack, and therefore the effect estimate would have no practical interpretation with respect to our models.
15 States were defined as battleground or non-battleground according to the “pivotal state” criteria developed by Gerber et al. (Citation2009). County-level unemployment figures were derived from Bureau of Labor Statistics estimates, available at http://www.bls.gov/lau/. We declined to control for county-level crime rates in our models because the best available data – from the FBI’s Uniform Crime Reporting Program – are too incomplete to serve as a reliable control, especially considering our study’s timespan (Loftin and McDowall Citation2010; Maltz and Targonski Citation2002).
16 We put little stock in this implication of our model for a hypothetical Election-Day mass shooting. Many citizens would likely be deterred from voting, either preoccupied with the event itself and its immediate aftermath or waylaid by unexpected work and family obligations. As with many such abstractions, our model may not merit substantive interpretation at the extremes.
17 As was the case with our model’s predictions about mass shootings and turnout (and for the same reason as before), we are reluctant to attach much credence to this estimated effect size of a hypothetical Election-Day terrorist attack.
18 Naturally, any decrease in the Democratic share of the two-party vote is by definition an increase of equal magnitude in the Republican share of the two-party vote.
19 It is possible different tests of the electoral effects of mass shootings and terrorist attacks would have revealed some capacity of mass shootings to influence Democratic vote share or terrorist attacks to influence turnout. We consider several alternative models – the results of which are reported in the online appendix – but fail to observe any such effects.