In the days since the mass shooting at Emanuel AME church, elected officials from across the political spectrum have called for reflection, conversation and solidarity.

A number of politicians, however, have condemned overtly political responses to the shooting. Sen. Lindsey Graham (R-S.C.) said on CNN, “Really, the last thing on my mind right now is a political debate.” Graham’s fellow presidential candidate Ben Carson decried liberals for seeking to make “political hay” of the tragedy.

But the political consequences associated with mass shootings are very real. And, at the local level, they may weaken support for the Republican Party, the party that — for the most part — stands for expanding citizens’ access to guns. Scholars have repeatedly demonstrated that the media and members of the general public inevitably lose interest in even the most violent tragedies, but members of the communities where violence happens may not move on so quickly.

In Connecticut, for instance, the 2012 Sandy Hook shooting plausibly had localized electoral ramifications. The same two politicians faced off in the state’s 2010 gubernatorial election (two years before the shooting) and 2014 gubernatorial election (two years after the shooting).

Across Connecticut’s 169 towns, the vote share of Tom Foley, the Republican candidate for governor, decreased by an average of 0.16 percent between 2010 and 2014. In Newtown, the town in which the Sandy Hook shooting occurred, Foley’s vote share decreased by 7.65 percent, according to data provided by the Connecticut Secretary of State’s office. This was the single largest drop Foley experienced in any of the 169 towns.

The enduring local impact associated with mass shootings is not limited to Sandy Hook.

Arguably the most infamous mass shooting in American history is the 1999 Columbine High School killings. Here too, Republican Party vote share was substantially depressed in the localities closest to the shooting.

Across the entire state of Colorado, GOP vote share increased between 1996 and 2000. Denver County, the geographic center of which is just 20 miles away from the location of the shooting, was the county in the state with the smallest increase; Republican support in other nearby counties was similarly depressed.

Given the substantial media coverage of these shootings, it is perhaps unsurprising that voter behavior was significantly altered in the subsequent major election. As President Obama pointed out in his recent press conference, however, mass shootings occur in the United States with a frequency unparalleled among other developed nations.

Most of these shootings do not generate the national response that the Sandy Hook or Columbine attacks provoked. Nonetheless, my research found a strong correlation between public gun crimes and a decline in support for local Republican office-seekers.

An analysis of the most comprehensive dataset of American mass shootings (courtesy of the Stanford Geospatial Center) reveals that a mass shooting in the four years preceding a presidential election is associated with a significantly depressed Republican vote share in the counties near which it occurred.

In total, 125 mass shootings occurred between 1980 and the 2012 presidential elections. (Stanford’s dataset extends back to 1966, but in no election cycle prior to 1984 were there more than five shootings.) Empirically, counties with shootings experienced a decrease in GOP vote share relative to counties without shootings.

In 2008, for instance, 26 counties witnessed mass shootings; the average GOP vote share change in these counties was more than 2.5 percent lower than in all other counties.

As the graph below illustrates, Republican vote share in all counties whose geographic centers were within 25 miles of the location of a mass shooting decreased by an average of 1.6 percentage points in the subsequent presidential election.

The impact associated with mass shootings is not limited to geographic effects. Counties within which there were shootings within 100 days of a presidential election witnessed a significantly greater decrease in GOP vote share than counties with shootings that were further in the past. Similarly, counties within which shootings occurred with 10 or more fatalities experienced a much sharper decrease in GOP vote share than counties where the shooting caused fewer than 10 victim fatalities.

Of course, there are other instances in which tragedy potentially affected voter behavior. One study found that the 9/11 terrorist attacks more significantly affected the political behavior of voters who lived close by than those who did not. Another study found that a third of respondents in a community with a deadly public shooting reported that the attack had affected their lives considerably.

The results I found were significant, but they were often small and highly localized. Nevertheless, dozens of local, state, and federal elections are decided by just a handful of percentage points every election cycle. Connecticut’s Democratic governor won reelection after the Sandy Hook shooting by less than 30,000 votes out of 1.1 million cast. (It is also worth recalling that the 2000 presidential election ultimately came down to several hundred votes in a few Florida counties.) If the data show a real effect, the chances of a tragedy affecting an election with national implications are low. Nonetheless, this is a possibility.

It is very possible that Charleston voters will not punish Republicans in the upcoming elections. GOP vote share did not decrease following every shooting in the dataset; many other factors influence elections. Moreover, shootings do not occur at random. The counties in which they occur may also be those that are already predisposed to penalize Republican politicians. Still, the data I have examined indicate that mass shootings may disrupt political equilibria and decrease local GOP vote share.

Eric W. Stern graduated from Yale University in 2015 with a simultaneous BA-MA in political science.