the impact of mass shootings on gun policy jpe - hbs.edu files/16-126_23dbdd9e-2135... · abstract...
TRANSCRIPT
Electronic copy available at: http://ssrn.com/abstract=2776657
The Impact of Mass Shootings on Gun Policy
Michael Luca Deepak Malhotra Christopher Poliquin
Working Paper 16-126
Electronic copy available at: http://ssrn.com/abstract=2776657
Working Paper 16-126
Copyright © 2016 by Michael Luca, Deepak Malhotra, and Christopher Poliquin
Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.
The Impact of Mass Shootings on Gun Policy
Michael Luca Harvard Business School
Deepak Malhotra Harvard Business School
Christopher Poliquin Harvard Business School
1
TheImpactofMassShootingsonGunPolicy
MichaelLucaHarvardBusinessSchool
DeepakMalhotraHarvardBusinessSchool<[email protected]>
ChristopherPoliquinHarvardBusinessSchool<[email protected]>
October2016
Abstract There have been dozens of high-profile mass shootings in recent decades. This paperpresentsthreemainfindingsabouttheimpactofmassshootingsongunpolicy.First,massshootingsevokelargepolicyresponses.Asinglemassshootingleadstoa15%increaseinthe number of firearm bills introducedwithin a state in the year after amass shooting.Second,massshootingsaccountforasmallportionofallgundeaths,buthaveanoutsizedinfluence relative tootherhomicides.Ourestimates suggest that theper-death impactofmassshootingsonbills introduced isabout80 timesas largeas the impactof individualgunhomicides in non-mass shooting incidents. Third,when looking at enacted laws, theimpactofmass shootingsdependson theparty inpower.Amass shooting increases thenumber of enacted laws that loosen gun restrictions by 75% in states with Republican-controlled legislatures. We find no significant effect of mass shootings on laws enactedwhenthereisaDemocrat-controlledlegislature.
WethankJosephHallandJessicaLiforexcellentresearchassistance.
2
1.Introduction
Recentdecadeshavewitnessedaseriesofhigh-profilemassshootingsthroughout
theUnitedStatesintownsrangingfromNewtown,CTtoKilleen,TX.Whilemosthomicides
receive little attention from the general public, mass shooting incidents are extremely
salient.Nonetheless, a common and frequently articulated view is that despite extensive
discussionaboutmassshootings,theyhavelittleinfluenceonpolicymaking.
Shouldwe expect policymakers to proposenew legislation in thewake of amass
shooting?Giventhatthevastmajorityofgundeathsdonotresultfrommassshootings,it
would be difficult to reconcile large responses to mass shootings with basic models of
optimalpolicyaimedexclusivelyatreducinggunviolence.However,massshootingsmay
have another effect – bringing attention to the issue of gun violence. Mass shootings
potentiallyleadtopolicychangesbyfocusingattentionongunviolence,eveniftheydonot
providenewinformationorchangepoliticians’preferences(whicharegenerallystaticand
alignedwithpartypreferences).Politicalscientistshavenotedthefactthatissuestendto
riseandfallwithinapolicyagenda,creatingperiodsinwhichspecificpolicieschangevery
rapidlyandotherperiodsinwhichtheydonotchangeatall(BaumgartnerandJones1993,
Kingdon 1984). In the context of gun violence, events like the Columbine shooting have
lead to both calls for new restrictions on guns and vehement reaction from gun rights
groupsopposedtosuchchanges(Goss2006,Spitzer2012).Moregenerally,massshootings
may create “policy windows” during which legislatures become receptive to change –
potentiallyduetoshiftsintheattentionofconstituents.Buttheextenttowhichthisoccurs,
andthedirectionofresultingchangesareempiricalquestions.
3
Inthispaper,weexploretheimpactofmassshootingsongunpolicy,constructinga
dataset of all U.S. gun legislation andmass shootings over a period of twenty-five years
(1989-2014)–combiningdatafromavarietyofmediaandgovernmentsources.Webegin
by looking at the extent of deaths resulting from mass shootings relative to other gun
deaths.Overall, therearemorethan30,000gunrelatedfatalities intheUnitedStatesper
year. Roughly 56%of these are suicides and40%are homicides. The remaining 4%are
accidentsorincidentsofundeterminedintent.Massshootingsaccountedforabout0.13%
ofallgundeathsand0.34%ofgunmurdersbetween1989and2014.
Becausemassshootingsaresalientandplausiblyrandomoccurrences,weareable
to implementadifference-in-differencesstrategyaround the timingofmassshootings to
estimate theircausal impactongunregulation.Specifically,wecomparegun lawsbefore
andaftermassshootings,instateswheremassshootingsoccurrelativetoallotherstates.
Wethenpresentthreemainfindingsabouttheimpactofmassshootingsonpolicy.
First, mass shootings evoke large policy responses. A single mass shooting leads to an
approximately15%increaseinthenumberoffirearmbillsintroducedwithinastateinthe
yearafteramassshooting.Thiseffect is largestaftershootingswiththemost fatalities–
andholdsforbothRepublican-controlledandDemocrat-controlledlegislatures.
Second,althoughmassshootingsaccountforasmallportionofallgundeaths,they
haveanoutsizedinfluencerelativetootherhomicides.Ourestimatessuggestthattheper-
deathimpactofmassshootingsonbillsintroducedisabout80timesaslargeastheimpact
ofgunhomicidesinnon-massshootingincidents.
Third,whenlookingatenactedlaws,theimpactofmassshootingsdependsonthe
party in power. Amass shooting increases the number of enacted laws that loosen gun
4
restrictions by 75% in states with Republican-controlled legislatures. We find no
significanteffectofmassshootingson lawsenactedwhenthere isaDemocrat-controlled
legislature.
These findingscontribute to theempirical literature thatusesapolitical economy
lens to explore the determinants of policymaking (Makowsky and Stratmann 2009,
Bardhan and Mookherjee 2010). Our results show that salient events – such as mass
shootings–can lead tosignificantpolicyresponses.Moreover,policymakersseemtouse
massshootingsasanopportunitytoproposebillsthatareconsistentwiththeir ideology.
This helps to shed light on the role of attention and salience in shaping policy and the
interactionbetweenissuesalienceandexistingpoliticalpreferencesinshapingthedegree
anddirectionofenactedpolicies.
2.BackgroundandData
Asdescribedabove,outoftheroughly30,000annualgundeathsintheUnitedStates,
fewerthan100occurinmassshootings.Forthepurposeofthispaper,wedefinea“mass
shooting” as an incident in which 4 or more people, other than the perpetrator(s), are
unlawfullykilledwithafirearminasingle,continuousincidentthatisnotrelatedtogangs,
drugs, or other criminal activity. This definition closelymatches theoneusedbyKrouse
andRichardson(2015) and the FBI’s definition of “massmurder” as 4 ormoremurders
“occurring during the same incident, with no distinctive time period between the
murders… typically involv[ing] a single location” (Morton and Hilts2008). We further
restrict our analysis to cases where at least three of the fatalities were individuals
unrelatedto,andnotromanticallyinvolvedwith,theshooter(s).Weincludespreemurders
5
–homicidesatmultiple locationswithout a significantpausebetween incidents– if they
resultinfourormoredeaths.
We assemble a list of mass shootings since 1989 from a variety of government and
media sources because there is no single, comprehensive government database ofmass
murders.Weextractallgun-relatedmassmurders(fourormoredead)thatarenotfelony
relatedfromtheFBIsupplementaryhomicidereports(SHR).Wethenverifyeachincident
intheSHRusingmediaaccounts;theSHRmaycontainerrorsinwhichseparatehomicides
in amonth are reported as a single incident, which is why it is necessary to verify the
incidentswithmedia coverage. Participation in the SHRprogram is voluntary andmany
lawenforcementagenciesdonotreportdetaileddatatotheFBI.Wethereforesupplement
the FBI data with mass shootings gathered from media accounts or compiled by other
researchers and journalists interested in the topic.We combine the SHRdatawithmass
shootingscollectedbytheMassShootingsinAmerica(MSA)projectatStanfordUniversity
(Stanford Geospatial Center and Stanford Libraries2015) and a list created by USA
Today(2013).Foreachshooting,wedeterminetheeventlocationaswellasthenumberof
victimfatalitiesandinjuries.Wealsoclassifyshootingsbasedontherelationship(ifany)
betweentheallegedshooter(s)andvictims.Previousworkonmassshootings(Duwe2007;
KrouseandRichardson2015)distinguishesbetweenpublicmass shootings thatoccur in
places frequented by the public, felony-related murders, and familicide. We categorize
shootingsbywhethertheyarepubliceventsorprimarilyrelatedtodomesticconflicts,and
we focuson incidents inwhichat least threepeoplenotrelatedorromantically involved
with the shooter died. This restriction filters out family-killings in residences aswell as
6
family-related murders in public places.1Figure 1 shows the number of incidents and
fatalitiesinmassshootingsbyyear.Thedatashowaslightupwardtrendinthenumberof
incidentsandfatalitiesovertime,butbothincidentsandfatalitiesvarysubstantiallyfrom
yeartoyear.
2.2.GunLegislation
Stategovernmentsaretheprimaryregulatorsoffirearms.Federallawsestablisha
minimum level of gun control,which is thenaugmented to varyingdegreesby state and
local policies. Federal government has limited commerce, the possession of guns by
potentially dangerous individuals, and some types of firearms and ammunition. States
decideavarietyofgunpoliciesrangingfromwhocanpurchaseandpossessaguntowhat
typesof gunsareallowed indifferent situations tohowguns shouldbe storedandwhat
typesof trainingshouldbeundertakenbygunowners.Localordinancescanalsorestrict
firearmpossessionanduse,butstatestatutesenactedinthepastfewdecadeshavelimited
theimportanceoflocalgovernmentinthisarenabypre-emptinglocalregulations.
Wecreateacomprehensivedatasetofgunlegislationinallfiftystatesusingthebill
tracking reports service from LexisNexis, which includes all bills introduced in state
legislatures since at least 1990with a synopsis and timelineof eachbill’s progress.This
allows us to determinewhether bills pass the legislature and become law. We identify
firearm bills by searching for the firearm-related terms “firearm”, “handgun”, “pistol”,
“revolver”,“rifle”,“shotgun”,“long-gun”,and“assaultweapon.”Weidentify20,409firearm
billsand3,199lawsbetween1990and2014.Inotherwords,therewere20,409proposals
1A2006shootingatachurchinLouisianaisoneexample.Amankilledhiswifeandin-lawswhileabductingherandtheirchildrenfromachurch.Onlythewife’sfamilywaspresentatthechurchduringtheshooting.
7
introduced and 3,199 laws passed in the twenty-five year sample period across all fifty
states.Thisincludeslawsthatloosenortightengunrestrictions,andmanythatdoneither
orboth.Weexcluderesolutions,executiveorders,andballotinitiativesfromtheanalysis.2
Toexplorewhetherguncontrol is tightenedor loosenedaftermassshootings,we
hired eightpeople tomanually code the summaryof bills that became law.Coderswere
giveninstructionsexplaininghowtocodelegislation,butwereotherwiseblindtothetopic
and design of the study. We presented bill summaries from LexisNexis to coders in
randomly chosen groups of 50. Twopeople coded each summary, andno coder saw the
same summarymultiple times. For each summary, coders decidedwhether the bill was
tightening (stricter gun control), loosening (weaker gun control), uncertain (insufficient
information),bothtighteningandloosening,orneithertighteningnorloosening(neutral).
There were therefore five possible labels for bills: tighten, loosen, both, neutral, or
uncertain.AppendixAshowsexamplebillsummariesandtheirexpectedlabels.
Tocross-validate(andincentivize)thebillcoding,wecodedasmallfractionofbills
ourselvesasabaselinecomparisonpoint.Forthisprocess,weblindedourselvesfromany
informationaboutwhenorwherethebillwasproposed.Wethenusedourscorestoassess
thequalityofcoders.Specifically,eachgroupof50billsgiventoacodercontainedfivebills
thatwehadalsocoded(theydidnotknowwhichbillswereandweren’tcoded,anddidnot
haveaccesstoanyofourassessmentsofwhetherabillwaslooserortighter).Coderswere
paiduptoa50%bonusbasedontheextenttowhichtheircodingmatchedours(whichwe
simplytoldthemwasa“goldstandard”ofknowncodes).
2Legislatorsinsomestatesfirstsubmitideasforbillsintheformofadraftrequestorsimilardocument.Weexcludethesefromouranalysesbecausetheyresultindoublecountingsomelegislation.Weinsteadfocusonlyonactualbills.
8
Across all five categories, coders agreed with each other 52% of the time (the
agreementratewouldbe20%bychance)andagreedwith thegoldstandard71%of the
time. Coders performedworst on the neutral category, andbest on the tighten-only and
loosen-onlycategories;whenabill tightensguncontrol (accordingto thegoldstandard),
codersagreeon tightening67%of the time, andwhenabill loosensgun control, coders
agreeonloosening60%ofthetime.
Most importantly for the purposes of our analysis: when coders agreewith each
otherontightening,theyalsoagreewithourcoding93%ofthetime;whencodersagreeon
loosening, they are consistent with our scores 91% of the time. When analyzing the
direction of policy change, we leverage this high degree of reliability by restricting our
analysistobillsonwhichcodersagreedthatthelawwasdesignedtotightenorloosengun
control.Becausestatescanpasseither,none,orbothtypesoflawsinayear,ourdependent
variableisthecountoflawsineachdirection.
Figure2showsmeanbills introduced, lawsenacted,and tighteningand loosening
lawsbypoliticalcontrolofthestatelegislature.Republicansenactmorelawslooseninggun
control, and fewer laws tightening gun control, thanDemocrats.Republican,Democratic,
andsplit legislaturesenactasimilarnumberof totalgun laws.Thecoderswhoclassified
the legislationwere only given summaries of each bill; theywere not providedwith the
state,year,oranyinformationonpoliticalaffiliation.
2.3.ControlVariables Whileourempiricalstrategyallowsustocontrolforalltimeinvariantfactorsthat
mayaffectgunlegislation,wealsoaddtimevaryingcontrols.Theseincludeeconomicand
demographic factors such as unemployment, divorce rates, and rates ofmilitary service.
9
Wealsocontrolforinstitutionaldifferencesbetweenlegislatures.First,wecontrolforthe
numberof lawmakersasameasureof legislature size.Larger legislaturesconsidermore
bills.Second,wecreateadummyforlegislaturesthatheldaregularsessioninagivenyear
becausenotalllegislaturesmeetannually.Third,wecontrolforwhetherbillsineachyear
carryover intosubsequentsessions; somechambersallow forcarryoverwhileotherskill
all unpassed bills at the end of each session. Fourth,we control for years inwhich bills
were restricted to specific topics; seven states restrict the scope of legislation
(e.g.appropriationsonly)inspecificyears.Fifth,wecontrolforthepoliticalpartyinpower.
Table1containssummarystatisticsforallvariablesusedintheanalyses.
3.TheImpactofMassShootingsonGunPolicy
3.1IdentificationStrategy We implement a difference-in-differences strategy that compares gun lawsbefore
andaftermassshootings,instateswheremassshootingsoccurrelativetoallotherstates.
Our dependent variables are counts of bills or enacted laws at the state-year level. We
studytheeffectofmassshootingsusingPoissonregressionswithconditionalmean:
E 𝑦!,! 𝛼!, 𝜆! , 𝑆ℎ𝑜𝑜𝑡!,!!!,𝑋!,! = exp 𝛼! + 𝜆! + 𝛽!𝑆ℎ𝑜𝑜𝑡!,!!! + 𝛾!𝑋!,!
where𝑦!,!isacountofbillsintroducedorlawsenactedinstate𝑠andyear𝑡;𝛼!and𝜆!are
stateandyearfixedeffects;𝑆ℎ𝑜𝑜𝑡!,!!!iseitheranindicatorforstateswithamassshooting
or the fatality count in mass shootings, and𝑋!,! is a vector of time-varying political,
economic,anddemographicfactors.Weestimatetheparametersviamaximumlikelihood
byconditioningonthesumof𝑦!,!withinstatesandincludingyearindicators.
10
Our identificationallowsus tomeasure the impactofamassshootingwithin that
state, controlling forotherchanges thatarehappeningat thenational level.Because this
identification strategy does not identify national responses to mass shootings (which
would be absorbed by our year effect), our estimates of changes in gun policy may
underestimatethetotalimpactofamassshooting.Wecanaccountforpotentialspillovers
into neighboring states, but do not see significant spillover effects; these results are
presentedinAppendixB.
3.2TheEffectofMassShootingsonGunBillIntroductions Table2 shows that a mass shooting leads to a 15% increase in firearm bills
introduced.Fortheaveragestate,thisamountstoanadditional2.5firearmbillsintroduced
in the year following a mass shooting. Mass shootings with more deaths lead to larger
effects.Onaverage,eachadditionaldeathinamassshootingleadstoa2.5%increaseinthe
number of gun bills introduced. This result holds both for Republican-controlled and
Democrat-controlledlegislatures.3
3.3ComparingMassShootingsandNon-MassShootings
Table 3 shows that fatalities resulting from mass shootings lead to much larger
increasesingunbillintroductionsthangunhomicidesinnon-massshootingincidents.We
estimatethemodelsinthistableusingmassshootingfatalitiesandordinarygunhomicides
per 100,000 people to facilitate comparison between the two types ofmurder. Itwould
take approximately80peopledying in individualgunhomicide incidents to have asmuch
impactonbillsintroducedaseachpersonwhodiesinamassshooting.Ourestimatesimply
3Resultsonbillsproposedbrokendownbypoliticalaffiliationareavailableuponrequest.
11
that, on average, a single mass shooting has as much impact on the number of bills
proposedaswoulda270%increaseinthenumberofgunhomicidesinastate.Giventhe
average number of gun homicides per year is roughly 260 per state, this would be
equivalenttoanadditional448homicidesperstateyear.
3.4TheRoleofPoliticalPartyonLawsEnacted Asmentionedpreviously,thetwomajorpoliticalpartiesintheUnitedStatesdiffer
dramaticallyintheirstancesonhowrestrictivegunpolicyshouldbe,withtheRepublican
Party favoring fewer gun restrictions.4To look at the impact of political parties on gun
policy,werestrictouranalysistoenactedlaws,allofwhichwerecodedforwhetherthey
loosenedortightenedgunrestrictions(seedatadescriptionformoredetails).
Table4 shows the effect of mass shootings interacted with Democrat and
Republican control of state government (divided government, inwhich the legislature is
not controlledby a singleparty, is theomittedgroup).The results show thatDemocrats
andRepublicansresponddifferentlytomassshootings.
When there is a Republican-controlled legislature, mass shootings lead to more
firearm laws that loosenguncontrol.Amassshooting in thepreviousyear increases the
number of enacted laws that loosen gun restrictions by 75% in stateswith Republican-
controlled legislatures.When there is a Democrat-controlled legislature, mass shootings
lead to a statistically insignificant reduction in laws that loosen gun control.We find no
significanteffectsofmassshootingsonlawsthattightengunrestrictions,buttheestimates
are imprecise.Summingacrossall legislatures (Republican,Democrat,andsplit), there is
4See,forexample,https://www.gop.com/platform/andhttps://www.democrats.org/party-platform.
12
roughly a 10% increase in laws enacted after a mass shooting, but this estimate is
impreciseandstatisticallyinsignificant(AppendixC).
3.5RobustnessChecks
Inthissection,wepresentthreesetsofrobustnesschecks.First,weprovidesupport
fortheexogeneityofmassshootings.Second,weshowthatourmainresultsarerobustto
the inclusion of state-specific time trends. Third, we perform a falsification exercise in
whichweuserandomlygeneratedplaceboshootingsinsteadofactualshootings;weshow
therearenoeffectsusing theplacebo shootings,providing support forour identification
strategy.
3.5.1DeterminantsofMassShootings
Our ability to identify the causal impact ofmass shootings on policy rests on the
assumptionthattheyareplausiblyexogenoustootherfactorsthatwoulddriveguncontrol
inagivenyear.Giventheerraticnatureofmassshootings,thisisaplausibleassumption.
Nonetheless,onemightbe concerned thatbothmass shootingsandgunpolicyarebeing
drivenbya thirdvariable.Toprovidesupport forourassumptionand interpretation,we
regressan indicator forwhetheramass shootingoccursoneconomic,demographic, and
policyvariables.
Consistentwiththeassumptionthatmassshootingsareexogenouswithrespectto
potentialconfounds,theresultsinAppendixDshowthat,outof32variablesweconsider,
onlyunemploymentissignificantlyassociatedwithahigherprobabilityofmassshootings.
Becausehigherunemploymentisalsoassociatedwithareductioningunbillintroductions
(Table2), thepotentialbiasof thiswouldwork in theoppositedirectionofour finding–
13
makingitunlikelythatthisisdrivingourresults.Tofurthersupportourinterpretation,we
control forunemployment in allmodels. Importantly, bills introduced, lawsenacted, and
majorgunpoliciesdonotpredictfuturemassshootings(AppendixD).
3.5.2State-SpecificTimeTrends
Anotherpotentialconcernisthatstateshavedifferentialtrendsinmassshootings,
and that these trends correlate with gun regulations, which would violate the parallel
trends assumption. As a robustness check, we run our main specifications with state-
specifictrends.AppendixEshowstheresultsofre-estimatingthemodelsinTables2and3
withstatespecific timetrends.The inclusionofstatespecific trendsdoesnotchangeour
main results fromTables 2 and 3.We are unable to estimatemodelswith state specific
trendsforouranalysesoftighteningandlooseninglawsbecausethelikelihoodfunctionis
discontinuouswhen including the additional parameters due to some states having very
few laws that we can identify as tightening or loosening. We can, however, conduct a
placeboanalysistoaddressanyresidualconcerns.
3.5.3PlaceboTests
Asafinalrobustnesscheck,weperformafalsificationexercisebasedontheinsights
ofBertrand,Duflo,andMullainathan(2004)andDonaldandLang(2007).Specifically,we
randomly assign placebo mass shootings to state-years in which no actual shooting
occurredwithprobabilityequaltoeachstate’sfrequencyofshootings,andrandomlydraw
afatalitycountfromtheempiricaldistributionoffatalities.AppendixFshowspercentiles
of the test statistic based on 1,000 repetitions of this procedure and our actual test
14
statistics from Tables 2 and 4. The results suggest our tests do not over-reject the null
hypothesisthatmassshootingshavenoeffectongunpolicy.
4.Discussion
Massshootingsaccountforasmall fractionofgundeaths intheUnitedStates,but
haveasignificantimpactongunpolicy.Moregunlawsareproposedintheyearfollowinga
mass shooting, a result that holds for both Republican- and Democrat-controlled
legislatures.Notably,massshootingshavemuchlargereffectsonpolicy,perfatality,than
ordinarygunhomicides.
However, we also find evidence that Democrat- and Republican-controlled
legislaturesdiffersignificantlywhenitcomestoenactinggunlaws.Republicansaremore
likelytoloosengunlawsintheyearafteramassshooting.TheeffectforDemocrats,which
tends toward less loosening of gun restrictions after a mass shooting, is statistically
insignificant.Thisisconsistentwithsurveyevidencesuggestingthatevenwhenamajority
supportsaguncontrolproposal,thoseopposedtoincreasedguncontrolaremorelikelyto
take actions likewriting a letter or donatingmoney to support their side (Schuman and
Presser1981).
Our results are consistent with qualitative research that has hypothesized the
possibility of mass shootings precipitating change. For example, Godwin and
Schroedel(1998) argue that the Stockton schoolyard massacre in 1989 led to the
enactmentof California’s assaultweaponsban.We find empirical evidence that sporadic
eventssuchasmassshootingscanleadtomajorpolicychanges.Thisraisesthequestionof
other factors thatmight drive policy, and conditions underwhichwemight expect such
15
effects. For example, Egan andMullin (2012) show that extreme temperatures influence
beliefs about global warming in the short-term. Might we expect a greater impact of
randomeventsinsomepolicycontexts(e.g.,theeffectofaterroristattack)thaninothers
(e.g.,theeffectofanEbolaoutbreak)?
Ourfindingsraiseanumberofadditionalquestions,andsuggestseveraldirections
forfutureresearch. First,ourestimatesfocusontheimpactonpolicywithinthestatein
which each shooting took place. Somemass shootings get national media attention and
potentially affect policy nationwide, which would not be identified by our fixed effects
model. One direction for future research is to develop strategies to identify national
responses. With respect to our findings, this suggests that the total impact of mass
shootingsongunpolicymaybeevenlargerthanourestimates.
Second,futureresearchmightfurtherexploretheroleofsalienceinshapingpolicy
byexaminingtheconditionsunderwhicheventsaremoreorlessinfluential.Forexample,
sometypesofevents(e.g.,schoolshootings)mayhavelargereffectsthanothers,something
wecouldnottestgiventherelativeinfrequencyofsuchevents.Salienteventsmighthavea
greater impact if they occur at a time when few other events are competing for media
attention(e.g.,EisenseeandStrömberg2007),orduringelections,whenpublicattentionis
more focused on such issues (e.g., Bouton, Conconi, and Pino2015). Finally, the role of
interestgroupsthattrytopromotetheirpreferredpoliciesintheaftermathofsuchevents
deservesfurtherexploration.
Third, future research might directly explore the preferences of politicians. Do
RepublicanlegislaturesloosengunrestrictionsbecauseRepublicanpoliticiansthemselves
prefer looser restrictions or due to pressure from constituents or interest groups? For
16
example,ifconstituentpreferencesaredrivingresults,wemightexpectthatresultsdiffer
in areaswith high versus low rates of gun ownership. To provide exploratory evidence,
Appendix G shows the results from adding a proxy for gun ownership to the models.
Following Cook and Ludwig (2006), we calculate the percentage of suicides that are
firearm related as a proxy for gun ownership and interact this variable with the mass
shootingindicator.5Thecoefficientonthisvariable isnotsignificanteither in isolationor
when added to the specification with all political interactions. This suggests that the
tendencyofRepublicans to loosenguncontrol isnotentirelydrivenbyhighratesofgun
ownership (andpresumablyhigh ratesof support for lessgun control), but representsa
distincteffectofpoliticalaffiliation.
Fourth, there is a large literatureon the impact of gunpolicies on crime (Duggan
2001, Ludwig andCook2000, Ludwig andCook2003,Abrams2012),whichhas yielded
mixedresults.Therelationshipwefindbetweenmassshootingsandgunpolicyraisesthe
possibilityofusingmassshootingsasaninstrumentalvariabletostudytheimpactofgun
laws on gun deaths. Unfortunately, in our sample, mass shootings are not a sufficiently
stronginstrumenttoestimatetheeffectsofgunpolicyongundeaths,duetotheirrelative
infrequency.(AppendixHpresentsresultsofthisanalysis.)Thisleavesopenthepossibility
of using salient and plausibly random events to instrument for policy changes in future
research.
Our findings suggest that while much attention has been rightfully devoted to
understanding the impact of policy, there is a lot to be learned from exploring the
5Wesmooththepercentageofsuicidesthatarefirearm-relatedbytakinganaverageoftheprevious5-yearstoreducetheinfluenceofshort-termfluctuationsingunsuicides.Theresultsarequalitativelysimilarwithoutthisadjustment.
17
determinants of policy change aswell.We find that even random and infrequent events
that account for a relatively small portion of total societal harm in a domain might
nonethelessbecrucialleversforpolicyconsiderationandchange.Thisdoesnotimplythat
politicians and policymakers are over-reacting; itmay be that on issueswhere there is
usually political deadlock, salient events create opportunities for change that has been
sought all along. Whether these changes reflect appropriate responses to the problem
remainsanopenquestion.
18
References
Abrams, David S. 2012. “Estimating the Deterrent Effect of Incarceration Using SentencingEnhancements.”AmericanEconomicJournal:AppliedEconomics4(4):32-56.
Andrews, DonaldW. K.,Marcelo J.Moreira, and James H. Stock. 2006. “Optimal Two-SidedInvariantSimilarTestsforInstrumentalVariablesRegression.”Econometrica74(3):715-52.
Bardhan, Pranab, and Dilip Mookherjee. 2010. “Determinants of Redistributive Politics: AnEmpiricalAnalysisofLandReformsinWestBengal,India.”AmericanEconomicReview100(4):1572–1600.
Baum, Christopher F., Mark E. Schaffer, and Steven Stillman. 2007. “Enhanced Routines forInstrumental Variables/generalized Method of Moments Estimation and Testing.” TheStataJournal7(4):465–506.
Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2004. “How Much Should WeTrust Differences-in-Differences Estimates?”TheQuarterly Journal of Economics 119 (1):249-75.
Baumgartner,FrankR.andBryanD.Jones.1993.AgendasandInstability inAmericanPolitics.Chicago,IL:UniversityofChicagoPress.
Bouton,Laurent,PaoloConconi,FranciscoPino,andMaurizioZanardi.2015.“GunsandVotes,”workingpaper.
Cook,PhillipJ.,andJensLudwig.2006.“TheSocialCostsofGunOwnership.”JournalofPublicEconomics90(1–2):379–391.
Donald,StephenG,andKevinLang.2007.“InferencewithDifference-in-DifferencesandOtherPanelData.”ReviewofEconomicsandStatistics89(2):221-33.
Duggan, Mark. 2001.”More Guns, More Crime.” Journal of Political Economy 109 (5):1086-1114.
Duwe,Grant,2007.MassMurderintheUnitedStates.NorthCarolina:McFarland&Company,Inc.
Egan,PatrickJ.andMeganMullin.2012.“TurningPersonalExperienceintoPoliticalAttitudes:The Effect of Local Weather on Americans’ Perceptions about Global Warming.” TheJournalofPolitics74(3):796–809.
Eisensee,ThomasandDavidStrömberg.2007.“NewsDroughts,NewsFloods,andU.S.DisasterRelief.”TheQuarterlyJournalofEconomics,122(2):693–728.
Finlay,Keith,andLeandroM.Magnusson.2009.“ImplementingWeak-InstrumentRobustTestsforaGeneralClassofInstrumental-VariablesModels.”TheStataJournal9(3):398–421.
19
Godwin,MarciaL.,andJeanReithSchroedel.1998.“GunControlPoliticsinCalifornia.”InTheChanging Politics of Gun Control, edited by John M. Bruce and Clyde Wilcox, 88–110.Lanham,MD:Rowman&LittlefieldPublishers,Inc.
Goss,KristinA.2006.Disarmed:TheMissingMovementforGunControlinAmerica.Princeton,NJ:PrincetonUniversityPress.
Kingdon,JohnW.1984.Agendas,Alternatives,andPublicPolicies.Boston,MA:Little,Brown&Company.
Krouse, William and Daniel J. Richardson. 2015.Mass Murder with Firearms: Incidents andVictims, 1999-2013 (CRS Report No. 7-5700). Washington, DC: Congressional ResearchService,2015.https://fas.org/sgp/crs/misc/R44126.pdf.
Levitt, Steven D. 1996. “The Effect of Prison Population Size on Crime Rates: Evidence fromPrisonOvercrowdingLitigation.”TheQuarterlyJournalofEconomics111(2):319–51.
Ludwig,Jens,andPhilipJ.Cook.2000.“HomicideandSuicideRatesAssociatedwithImplementationoftheBradyHandgunViolencePreventionAct.”JAMA284(5):585-91.
Ludwig,Jens,andPhilipJCook.2003.EvaluatingGunPolicyEffectsonCrimeandViolence.Washington,D.C.:BrookingsInstitutionPress.
Makowsky,MichaelD.,andThomasStratmann.2009.“PoliticalEconomyatAnySpeed:WhatDeterminesTrafficCitations?”TheAmericanEconomicReview99(1):509–27.
Moreira,MarceloJ.2003.“AConditionalLikelihoodRatioTestforStructuralModels.”Econometrica71(4):1027-48.
Morton,RobertJ.andMarkA.Hilts,eds.2008.“SerialMurder:Multi-disciplinaryperspectivesfor investigators.” Serial Murder Symposium, San Antonio, TX, August 29-September 2,2005. Washington, DC: Federal Bureau of Investigation, http://www.fbi.gov/stats-services/publications/serial-murder/serial-murder-july-2008-pdf.
Schuman,HowardandStanleyPresser.1981.“TheAttitude-ActionConnectionandtheIssueofGunControl.”TheANNALSoftheAmericanAcademyofPoliticalandSocialScience455(1):40–47.
Spitzer,RobertJ.2012.ThePoliticsofGunControl.5thEd.Boulder,CO:ParadigmPublishers.
StanfordGeospatialCenterandStanfordLibraries.2015.“StanfordMassShootingsinAmerica,”July31release,https://library.stanford.edu/projects/mass-shootings-america.
Stock, James H., and Motohiro Yogo. 2005. “Identification and Inference for EconometricModels.” In Identificationand Inference forEconometricModels, editedbyDonaldW.K.AndrewsandJamesH.Stock,80–108.NewYork:CambridgeUniversityPress.
20
USA Today. 2013. “Behind the Bloodshed: The Untold Story of America’s Mass Killings,”http://www.gannett-cdn.com/GDContent/mass-killings/index.html.
21
Figures
Figure1Massshootingincidentsandfatalitiesbyyear,1989-2014.Theupperpanelshowsthenumberoffatalitiesinmassshootingsinwhichatleast3peoplenotrelatedorromanticallyconnectedtotheshooterwerekilled.Thebottompanelshowsthenumberoftheseincidents.Washington,D.C.isnotincludedinthesample.
Figure 2Comparisonofannual legislation introducedbypoliticalparty.Points represent themeanand linesare95%confidenceintervals.Legislaturecontrolmeansonepoliticalpartyincludesbothchambersofthelegislature.Thecountsoftighteningandlooseninglawsarebasedonlawswithcoderagreement(seesection2.2oftext).
0
20
40
60
80
Fata
litie
s
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
Mass Shooting Fataliites
2
4
6
8
10
12
Num
ber
of S
ho
otin
gs
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
Year
Mass Shooting Incidents
●
●
●
●●
●
●
●
●
●
●
●
5
10
15
20
0.25
0.50
0.75
1.00
Bills Introduced Laws Passed Tightening Laws Loosening Laws
Mea
n Le
gisl
atio
n pe
r Yea
r
Legislature●
●
●
Republican
Democrat
Split
22
Tables
Table1:SummaryStatistics
mean sd p5 p10 p50 p90 p95 N
Legislation BillsIntroduced 16.33 22.04 0 1 10 38 53 1,250LawsEnacted 2.56 3.35 0 0 1 6 9 1,250TighteningLaws 0.70 1.29 0 0 0 2 3 1,250LooseningLaws 0.25 0.62 0 0 0 1 1 1,250GunViolence MassShooting 0.12 0.32 0 0 0 1 1 1,250Fatalities 0.72 2.40 0 0 0 4 5 1,250GunHomicideRate 3.76 2.55 0.72 0.98 3.42 7.40 8.65 1,250PoliticalControls DemocraticLegislature 0.42 0.49 0 0 0 1 1 1,250RepublicanLegislature 0.34 0.47 0 0 0 1 1 1,250RepublicanGovernor 0.53 0.50 0 0 1 1 1 1,200InstitutionalControls RegularSession 0.94 0.24 0 1 1 1 1 1,250BillCarryover 0.27 0.44 0 0 0 1 1 1,250LimitedLeg.Topic 0.06 0.24 0 0 0 0 1 1,250LegislatureSize 148 59.3 62 82.5 144 200 236 1,250DemographicControls Elderly(65+)% 12.9 2.0 9.8 10.7 13.1 15.2 15.7 1,250Under25% 35.1 2.7 31.4 32.2 34.8 38.0 39.5 1,250Black% 10.3 9.5 0.6 0.8 7.4 26.4 30.1 1,250Hispanic% 8.3 9.2 0.8 1.2 5.1 20.3 29.9 1,250Unemployment% 5.7 1.9 3.1 3.5 5.4 8.1 9.3 1,250Incomepercapita 19.1 3.3 14.1 15.0 18.7 23.3 25.8 1,250HighSchool% 85.2 5.2 75.7 78.4 86.1 91.2 92.0 1,250Veteran% 11.8 2.4 7.9 8.8 11.8 15.0 16.1 1,250Divorced% 11.8 1.8 8.9 9.5 11.8 14.1 14.7 1,250
Note: Observations are state-years. Bills Introduced is the number of bills introduced in the legislature; LawsEnactedisthenumberofbillsthatbecamelaw.TighteningandLooseningLawsarenumbersofenactedlawsthattightened and loosened gun control respectively. Mass Shooting is an indicator for state-years with a massshootinginwhich3+peoplenotromanticallyinvolvedwithorrelatedtotheshooter(s)werekilled.Fatalitiesisthetotalnumberofdeathsinmassshootingsinastate-year.DemocraticandRepublicanLegislatureareindicatorsforpartycontrolofthestatelegislature.RepublicanGovernorisanindicatorforRepublicangovernors.RegularSessionindicateswhetherthelegislatureconvenedaregular(asopposedtospecial)sessiontoconsiderbills;somestatelegislatures onlymeet every other year. Bill Carryover is proportion of chambers in which bills are eligible forcarryovertothenextsession.LimitedLeg.Topicisanindicatorforlegislativesessionsduringwhichbillsarelimitedtospecifictopics(e.g.appropriations).LegislatureSizeisthenumberoflawmakersservinginthestatelegislature.Incomepercapitaismeasuredinthousandsof1987U.S.dollars;otherdemographicvariablesarepercentages.
23
Table2:TheEffectofMassShootingsonGunBillIntroductionsDependentvariable:numberoffirearm-relatedbillsintroducedinthestatelegislature.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
MassShooting 0.154* 0.087 0.155** 0.147**
(0.088) (0.074) (0.065) (0.062)
Fatalities 0.023* 0.020** 0.023*** 0.023***
(0.013) (0.009) (0.007) (0.007)
Dem.Legislature -0.136* -0.134*
(0.074) (0.076)
Rep.Legislature 0.067 0.071
(0.064) (0.064)
Rep.Governor -0.022 -0.014
(0.054) (0.053)
InstitutionalControls No
No
Yes
Yes
No
No
Yes
YesDemographicControls No No No Yes No No No Yes
StateFixedEffects Yes Yes Yes Yes Yes Yes Yes YesYearFixedEffects No
Yes
Yes
Yes
No
Yes
Yes
Yes
N 1,250
1,250
1,250
1,250
1,250
1,250
1,250
1,250Note:Robust standarderrors clusteredby state in parentheses. Stars following coefficients representp-values less
than.10(*),.05(**)and.01(***).SeenotetoTable1forvariabledefinitions.
24
Table3:ComparingMassShootingsandNon-MassShootingsDependentvariable:numberoffirearm-relatedbillsintroducedinthestatelegislature.
(1)
(2)
(3)
(4)
MassShootingFatalities/100,000 1.678*** 1.332*** 1.303*** 1.316***
(0.428)
(0.240)
(0.223)
(0.202)
OrdinaryGunHomicides/100,000 -0.007
0.017
0.014
0.017
(0.024)
(0.032)
(0.032)
(0.035)
PoliticalControls No No No Yes InstitutionalControls No No Yes Yes DemographicControls No No No Yes StateFixedEffects Yes
Yes
Yes
Yes
YearFixedEffects No
Yes
Yes
YesN 1,250
1,250
1,250
1,250
Note:Robust standarderrors clusteredby state in parentheses. Stars following coefficients representp-values lessthan.10(*),.05(**)and.01(***).MassShootingFatalities/100,000isthenumberofdeathsinmassshootingsper100,000stateresidents.OrdinaryGunHomicides/100,000isthenumberofgunhomicidesnotinmassshootingsper100,000stateresidents.ControlvariablesaredefinedasinTable1.
25
Table4:MassShootingsandEnactedLawsDependentvariable:numberoffirearm-relatedlawsenacted(i.e.billsthatbecamelaw). TighteningLaws LooseningLaws (1) (2) (3) (4) (5) (6) (7) (8) MassShooting -0.030 0.246 (0.101) (0.181) Fatalities 0.015* 0.008 (0.008) (0.022) Rep.Leg.×Shooting -0.017 0.733*** (0.238) (0.255) Dem.Leg.×Shooting 0.037 -0.250 (0.140) (0.402) SplitLeg.×Shooting -0.216 0.168 (0.262) (0.340) Rep.Leg.×Fatalities 0.018 0.152*** (0.050) (0.031)Dem.Leg.×Fatalities 0.014 -0.047 (0.015) (0.054)SplitLeg.×Fatalities 0.015 -0.018 (0.013) (0.018)Dem.Legislature 0.110 0.105 0.070 0.105 -0.318 -0.303 -0.253 -0.259 (0.147) (0.148) (0.171) (0.165) (0.200) (0.198) (0.229) (0.222)Rep.Legislature 0.165 0.180 0.134 0.178 0.494*** 0.496*** 0.402* 0.318 (0.137) (0.137) (0.147) (0.142) (0.187) (0.190) (0.213) (0.202)Rep.Governor -0.046 -0.045 -0.047 -0.045 -0.113 -0.111 -0.093 -0.084 (0.086) (0.085) (0.085) (0.083) (0.168) (0.167) (0.166) (0.167)InstitutionalControls Yes Yes Yes Yes Yes Yes Yes Yes DemographicControls Yes Yes Yes Yes Yes Yes Yes Yes StateFixedEffects Yes Yes Yes Yes Yes Yes Yes Yes YearFixedEffects Yes Yes Yes Yes Yes Yes Yes Yes N 1,250 1,250 1,250 1,250 1,175 1,175 1,175 1,175 Note:Robuststandarderrorsclusteredbystateinparentheses.Starsfollowingcoefficientsrepresentp-valueslessthan .10 (*), .05 (**) and .01 (***). Models 3-4 and 7-8 show the effect of mass shootings in Republican,Democratic,andsplitlegislatures;theomittedgroupinthesemodelsisstateswithoutamassshooting.AllvariablesaredefinedasinTable1.
26
AppendixA:CodingGunLaws
id summary tighten loosen uncertain
1 Createsanewfelonyforfiringagunwithin1,000feetofaneducationalfacility. 1 0 0
2 Reducestheagelimitforpurchaseofahandgunfrom21to18. 0 1 0
3 Allowsparoleofficerstocarryaloadedfirearmwhilecommutingtoandfromwork. 0 0 0
4 Relatestotheuseoffirearmsinstateparksandcampgrounds. 0 0 1
5
Requiresalicensetooperateagunshow.Eliminatesthewaitingperiodforfirearmsalesifthepurchaserhasavalidpermittocarryaconcealedweapon.
1 1 0
Note: Table shows examples of coding gun laws based on bill summaries. Coders were given a full manual toexplain themeaning of “tighten”, “loosen”, “neutral,” and “uncertain” alongwith the following examples. ThistablemimicstheappearanceoftheExcelworkbooksusedbythecoders.Thefirstbillcreatesanewcrimerelatedto firearms. It tightens restrictions on firearms. The second bill makes it easier for people to acquire guns; itloosensrestrictionsonfirearms.Thethirdbillisexclusivelyaboutparoleofficers;itisneutralbecauseitdoesnotaffectthegeneralpublic.Thefourthbill isuncertainbecausethesummaryisagenericdescriptionthatdoesnotspecifywhether the law tightens or loosens restrictions on firearms. The fifth bill both tightens and loosens; itregulatesgunshows,butalsoeliminatesarestrictiononfirearmpurchasers.
27
AppendixB:EffectofShootingsinNeighboringStates
TableB1:MassShootingsinNeighboringStates,BillsandLawsDependentvariable:numberoffirearm-relatedbillsintroducedorlawsenacted BillsIntroduced LawsEnacted (1) (2) (3) (4) (5) (6) (7) (8) MassShooting 0.147** 0.158** 0.097 0.087 (0.064) (0.073) (0.067) (0.080) NeighborShooting -0.052 0.028 (0.041) (0.048) Cen.DivisionShooting -0.018 0.020 (0.042) (0.049) Fatalities 0.023*** 0.022*** 0.014* 0.011 (0.007) (0.008) (0.008) (0.009)NeighborFatalities -0.002 0.002 (0.004) (0.004) Cen.DivisionFatalities 0.002 0.004 (0.005) (0.004)PoliticalControls Yes Yes Yes Yes Yes Yes Yes Yes InstitutionalControls Yes Yes Yes Yes Yes Yes Yes Yes DemographicControls Yes Yes Yes Yes Yes Yes Yes Yes StateFixedEffects Yes Yes Yes Yes Yes Yes Yes Yes YearFixedEffects Yes Yes Yes Yes Yes Yes Yes Yes N 1,250 1,250 1,250 1,250 1,250 1,250 1,250 1,250 Note:Robuststandarderrorsclusteredbystateinparentheses.Starsfollowingcoefficientsrepresentp-valueslessthan.10(*),.05(**)and.01(***).Neighborvariablesrefertostateswithasharedborder;Cen.DivisionreferstostateswithinthesameCensusdivision.AllothervariablesaredefinedasinTable1.
28
TableB2:MassShootingsinNeighboringStatesandDirectionsofPolicyChangeDependentvariable:numberoffirearm-relatedlawsenacted TighteningLaws LooseningLaws (1) (2) (3) (4) Rep.Leg.×Shooting -0.015 -0.104 0.732*** 0.809*** (0.223) (0.260) (0.257) (0.275)Dem.Leg.×Shooting 0.072 0.082 -0.294 -0.169 (0.129) (0.122) (0.379) (0.416)SplitLeg.×Shooting -0.242 -0.057 0.174 0.090 (0.249) (0.224) (0.331) (0.383)Rep.Leg.×NeighborShooting 0.208 -0.059 (0.135) (0.236) Dem.Leg.×NeighborShooting -0.315** 0.370 (0.131) (0.246) SplitLeg.×NeighborShooting -0.157 -0.095 (0.202) (0.234) Rep.Leg.×Cen.DivisionShooting 0.166 -0.153 (0.154) (0.232)Dem.Leg.×Cen.DivisionShooting -0.071 -0.141 (0.112) (0.175)SplitLeg.×Cen.DivisionShooting -0.258 0.146 (0.170) (0.258)PoliticalControls Yes Yes Yes Yes InstitutionalControls Yes Yes Yes Yes DemographicControls Yes Yes Yes Yes StateFixedEffects Yes Yes Yes Yes YearFixedEffects Yes Yes Yes Yes N 1,250 1,250 1,250 1,250 Note:Robuststandarderrorsclusteredbystateinparentheses.Starsfollowingcoefficientsrepresentp-valueslessthan.10(*),.05(**)and.01(***).ModelsshowtheeffectofmassshootingsinRepublican,Democratic,andsplitlegislatures;theomittedgroupinthesemodelsisstateswithoutamassshooting.Neighborvariablesrefertostateswithasharedborder;Cen.DivisionreferstostateswithinthesameCensusdivision.AllothervariablesaredefinedasinTable1.
29
TableB3:FatalitiesinNeighboringStatesandDirectionsofPolicyChangeDependentvariable:numberoffirearm-relatedlawsenacted TighteningLaws LooseningLaws (1) (2) (3) (4) Rep.Leg.×Fatalities 0.014 0.002 0.151*** 0.150*** (0.047) (0.049) (0.031) (0.043)Dem.Leg.×Fatalities 0.016 0.023 -0.046 -0.018 (0.014) (0.014) (0.053) (0.054)SplitLeg.×Fatalities 0.013 0.029** -0.022 -0.023 (0.012) (0.014) (0.019) (0.029)Rep.Leg.×NeighborFatalities 0.017 -0.005 (0.013) (0.022) Dem.Leg.×NeighborFatalities -0.025** 0.015 (0.012) (0.017) SplitLeg.×NeighborFatalities -0.008 -0.035** (0.022) (0.018) Rep.Leg.×Cen.DivisionFatalities 0.020* 0.002 (0.011) (0.029)Dem.Leg.×Cen.DivisionFatalities -0.009 -0.031** (0.008) (0.015)SplitLeg.×Cen.DivisionFatalities -0.015 0.007 (0.021) (0.023)PoliticalControls Yes Yes Yes Yes InstitutionalControls Yes Yes Yes Yes DemographicControls Yes Yes Yes Yes StateFixedEffects Yes Yes Yes Yes YearFixedEffects Yes Yes Yes Yes N 1,250 1,250 1,250 1,250 Note:Robuststandarderrorsclusteredbystateinparentheses.Starsfollowingcoefficientsrepresentp-valueslessthan.10(*), .05(**)and.01(***).Modelsshowtheeffectofmassshootingfatalities inRepublican,Democratic,andsplitlegislatures;theomittedgroupinthesemodelsisstateswithoutamassshooting.Neighborvariablesrefertostateswithasharedborder;Cen.DivisionreferstostateswithinthesameCensusdivision.AllothervariablesaredefinedasinTable1.
30
AppendixC:LawsEnactedTableC1:LawsEnactedandMassShootingsDependentvariable:numberoffirearm-relatedlawsenacted(i.e.billsthatbecamelaw).
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
MassShooting 0.074 0.020 0.079 0.098
(0.118) (0.108) (0.075) (0.067)
Fatalities 0.011 0.007 0.011 0.014*
(0.013) (0.011) (0.008) (0.008)
Dem.Legislature -0.023 -0.020
(0.082) (0.085)
Rep.Legislature 0.269*** 0.275***
(0.087) (0.087)
Rep.Governor -0.002 -0.000
(0.051) (0.050)
InstitutionalControls No
No
Yes
Yes
No
No
Yes
YesDemographicControls No No No Yes No No No Yes
StateFixedEffects Yes Yes Yes Yes Yes Yes Yes YesYearFixedEffects No
Yes
Yes
Yes
No
Yes
Yes
Yes
N 1,250
1,250
1,250
1,250
1,250
1,250
1,250
1,250Note:Robust standarderrors clusteredby state in parentheses. Stars following coefficients representp-values less
than .10 (*), .05 (**) and .01 (***). Variables are identical to those in Table 2, except for the dependent variable,whichisthenumberoffirearm-relatedlawsenactedinsteadofthenumberofbillsintroduced.SeenotetoTable1forvariabledefinitions.
31
AppendixD:PredictingMassShootings
TableD1:LinearProbabilityModelforMassShootingusingControlVariablesDependentvariable:indicatorforstate-yearwithamassshooting. (1)
(2)
(3)
(4)
(5)
(6)
(7)
LagBillsIntroduced
0.000
(0.001)
LagLawsEnacted
0.000
(0.005)
LagTighteninglaws
-0.000
(0.009)
LagLooseninglaws
0.016
(0.014)
Dem.Legislature
0.006
0.011
0.010
0.010
0.010
0.012
(0.044)
(0.043)
(0.045)
(0.046)
(0.045)
(0.045)
Rep.Legislature
-0.030
-0.030
-0.027
-0.027
-0.027
-0.028
(0.036)
(0.035)
(0.037)
(0.037)
(0.037)
(0.038)
Rep.Governor
-0.012
-0.010
-0.010
-0.010
-0.010
-0.010
(0.020)
(0.020)
(0.021)
(0.021)
(0.022)
(0.021)
RegularSession
0.121
0.122* 0.122* 0.121
0.128*
(0.077)
(0.072)
(0.067)
(0.073)
(0.073)
BillCarryovera
0.057** 0.056* 0.055** 0.055** 0.056**
(0.026)
(0.029)
(0.027)
(0.027)
(0.027)
LimitedLeg.Topic
-0.044
-0.062
-0.062
-0.062
-0.065
(0.062)
(0.059)
(0.060)
(0.059)
(0.060)
LegislatureSize
0.002
0.002
0.002
0.002
0.002
(0.001)
(0.002)
(0.002)
(0.002)
(0.002)
LogPopulation -0.191 -0.131 -0.140 -0.210 -0.211 -0.210 -0.210 (0.279) (0.265) (0.273) (0.297) (0.298) (0.297) (0.299)Elderly(65+)% 0.001
-0.001
-0.002
0.004
0.004
0.004
0.003
(0.025)
(0.024)
(0.024)
(0.024)
(0.024)
(0.025)
(0.025)Under25% -0.002
0.001
0.001
0.002
0.003
0.003
0.002
(0.020)
(0.019)
(0.019)
(0.020)
(0.020)
(0.020)
(0.020)Black% -0.009
-0.009
-0.008
0.000
-0.000
-0.000
0.002
(0.015)
(0.015)
(0.015)
(0.017)
(0.018)
(0.018)
(0.017)Hispanic% -0.001
-0.003
-0.002
-0.001
-0.001
-0.001
-0.001
(0.015)
(0.015)
(0.015)
(0.016)
(0.016)
(0.016)
(0.016)Unemployment% 0.025** 0.025** 0.025** 0.024** 0.024** 0.024** 0.024**
(0.011)
(0.011)
(0.011)
(0.011)
(0.011)
(0.011)
(0.012)
Incomepercapita 0.013
0.012
0.013
0.013
0.013
0.013
0.013
(0.013)
(0.013)
(0.014)
(0.014)
(0.014)
(0.014)
(0.014)
HighSchool% -0.004
-0.003
-0.002
-0.003
-0.003
-0.003
-0.003
(0.006)
(0.006)
(0.006)
(0.006)
(0.006)
(0.006)
(0.006)
Veteran% -0.003
-0.001
-0.001
-0.004
-0.004
-0.004
-0.004
(0.012)
(0.012)
(0.012)
(0.013)
(0.013)
(0.012)
(0.012)
Divorced% -0.004
-0.003
-0.002
0.001
0.001
0.001
0.001
(0.010)
(0.010)
(0.011)
(0.011)
(0.011)
(0.011)
(0.011)
Constant 3.072
2.024
1.700
2.612
2.633
2.621
2.637
(4.410)
(4.218)
(4.221)
(4.627)
(4.636)
(4.620)
(4.645)
StateFixedEffects Yes
Yes
Yes
Yes
Yes
Yes
YesYearFixedEffects Yes
Yes
Yes
Yes
Yes
Yes
Yes
N 1,250
1,250
1,250
1,200
1,200
1,200
1,200aThereisnoapriorireasontothinkbillcarryoverwouldberelatedtomassshootings;thiscorrelationisinsignificantwhen
Virginia,whichunlikemoststates,allowscarryoverinevenyears,isdroppedfromthesample.FourofVirginia’ssixmassshootingshappenedinevenyears.
32
TableD2:LinearProbabilityModelforMassShootingusingPolicyVariablesDependentvariable:indicatorforstate-yearwithamassshooting.
(1)
(2)
HandgunWaitingPeriod(days) 0.002
0.002
(0.005)
(0.005)
Long-gunWaitingPeriod(days) -0.005
-0.005
(0.019)
(0.020)
Age18+forTransaction 0.031
0.026
(0.026)
(0.028)
Age21+forTransaction -0.067
-0.080
(0.057)
(0.056)
HandgunPermitSystem -0.137
-0.136
(0.094)
(0.100)
BackgroundCheck,AllHandgunSales -0.066
-0.066
(0.080)
(0.085)
BackgroundCheck,AllFirearmSales 0.019
-0.005
(0.116)
(0.127)
AssaultWeaponsBan 0.042
0.050
(0.051)
(0.053)
ShallIssueConcealedCarry -0.010
-0.006
(0.038)
(0.039)
NoPermitNeededConcealedCarry 0.162
0.181
(0.188)
(0.184)
LogPopulation -0.494 -0.445 (0.325) (0.310)PoliticalControls No
Yes
DemographicControls Yes YesStateFixedEffects Yes YesYearFixedEffects Yes
Yes
N 1,250
1,250Note:HandgunWaitingPeriodisthenumberofdayspurchasersmustwaitbeforeacceptingdeliveryofahandgun.
Long-gunWaiting Period is similarly defined for long-guns (e.g. rifles and shotguns).Age 18+ Transaction is anindicator for laws that prevent vendors from selling handguns to minors or prevent minors from purchasinghandguns. Age 21+ Transaction is defined the same way for persons under 21.Handgun Permit System is anindicator for states that require permits to purchase a handgun. Background Check, All Handgun Sales is anindicator for requiring a background check for all handgun transactions (including private sales). BackgroundCheck, All Firearm Sales is an indicator for requiring a background check for all firearm transactions (includingprivatesales).AssaultWeaponsBanisanindicatorforstatesthatbansometypesofassaultriflesorpistols.ShallIssueConcealedCarry isanindicatorforstatesthatrequirethepermittingauthoritytograntalicensetoanyonemeetingtheminimumstatutoryqualifications(i.e.donotpermit lawenforcementdiscretion in issuingpermits).NoPermitNeededConcealedCarryisanindicatorforstatesthatallowconcealedcarrywithoutapermit.
33
AppendixE:State-SpecificTimeTrendsTableE1:TheEffectofMassShootingsonGunBillIntroductionsDependentvariable:numberoffirearm-relatedbillsintroducedinthestatelegislature.
(1)
(2)
(3)
(4)
(5)
(6)
MassShooting 0.074 0.151** 0.157**
(0.075) (0.069) (0.063)
Fatalities 0.020** 0.024*** 0.024***
(0.010) (0.008) (0.007)
InstitutionalControls No
Yes
Yes
No
Yes
YesPoliticalControls No No Yes No No Yes
DemographicControls No No Yes No No YesStateFixedEffects Yes Yes Yes Yes Yes YesState-SpecificTrends Yes Yes Yes Yes Yes YesYearFixedEffects Yes
Yes
Yes
Yes
Yes
Yes
N 1,250
1,250
1,250
1,250
1,250
1,250Note:Robust standarderrors clusteredby state in parentheses. Stars following coefficients representp-values less
than.10(*),.05(**)and.01(***).VariablesareidenticaltothoseinTable2anddefinedinthenotetoTable1.
TableE2:MassShootings,OrdinaryGunHomicides,andBillIntroductionsDependentvariable:numberoffirearm-relatedbillsintroducedinthestatelegislature.
(1) (2) (3) MassShootingFatalities/100,000 1.504*** 1.481*** 1.409***
(0.323)
(0.261)
(0.195)
OrdinaryGunHomicides/100,000 0.010
0.005
0.007
(0.058)
(0.055)
(0.049)
InstitutionalControls No Yes Yes PoliticalControls No No Yes DemographicControls No No Yes StateFixedEffects Yes Yes Yes State-SpecificTrends Yes
Yes
Yes
YearFixedEffects Yes
Yes
YesN 1,250
1,250
1,250
Note:Robust standarderrors clusteredby state in parentheses. Stars following coefficients representp-values lessthan.10(*),.05(**)and.01(***).VariablesareidenticaltothoseinTable3anddefinedinthenotetoTable1.
34
AppendixF:PlaceboMassShootingAnalysis
TableF1:PlaceboAnalysisforBillIntroductions(MirrorsTable2)
PercentilesofPlaceboTestStatistic
Actual 1st 5th 10th 90th 95th 99th
ShootingIndicator(model4) 2.37 -3.69 -2.80 -2.33 0.71 1.06 1.90ShootingFatalities(model8) 3.29 -4.01 -2.81 -2.40 0.94 1.35 2.32
TableF2:PlaceboAnalysisforEnactedLaws(MirrorsTable4)
PercentilesofPlaceboTestStatistic
Actual 1st 5th 10th 90th 95th 99thTighteningLaws(models1-4)
PooledShooting -0.30 -2.19 -1.40 -1.04 1.82 2.29 3.02PooledFatalities 1.88 -2.39 -1.64 -1.24 1.80 2.28 3.44Rep.Leg.×Shooting -0.07 -2.93 -1.71 -1.18 1.63 2.18 3.28Dem.Leg.×Shooting 0.26 -2.47 -1.63 -1.33 1.45 1.96 2.60SplitLeg.×Shooting -0.82 -2.65 -1.46 -0.98 2.13 2.60 4.14Rep.Leg.×Fatalities 0.36 -2.80 -1.86 -1.33 1.68 2.46 3.72Dem.Leg.×Fatalities 0.93 -2.81 -1.80 -1.47 1.52 2.02 3.49SplitLeg.×Fatalities 1.15 -2.79 -1.65 -1.09 2.26 2.94 4.82
LooseningLaws(models5-8)PooledShooting 1.36 -3.19 -2.28 -1.89 0.84 1.26 2.11PooledFatalities 0.36 -3.04 -2.25 -1.82 0.93 1.46 2.58Rep.Leg.×Shooting 2.87 -2.98 -2.27 -1.97 0.68 1.05 1.69Dem.Leg.×Shooting -0.62 -2.69 -1.79 -1.40 1.23 1.66 2.80SplitLeg.×Shooting 0.49 -3.10 -2.12 -1.70 1.40 1.86 2.89Rep.Leg.×Fatalities 4.90 -2.80 -2.24 -1.89 0.93 1.35 2.22Dem.Leg.×Fatalities -0.87 -2.89 -1.90 -1.48 1.47 1.94 2.87SplitLeg.×Fatalities -1.00 -3.22 -2.12 -1.75 1.47 2.04 3.32
Notes:We randomly assign placebo mass shootings to state-years in which no actual shooting occurred withprobabilityequal toeach state’s frequencyof shootings,and randomlydrawa fatality count fromtheempiricaldistributionof fatalities.Wethenre-runthemodelsandcalculatetheteststatistic fortheplaceboshootingandfatalitycoefficients.Theabovepercentilesarebasedon1,000replications.The“Pooled”rowsinTableE2mirrormodels1-2and5-6ofTable4(themodelswithoutinteractioneffects).Thelegislatureeffectsmirrormodels3-4fortighteninglawsand7-8forlooseninglaws.
35
AppendixG:GunOwnership,MassShootingsandEnactedLaws
TableG1:EffectsofMassShootings,withGunOwnershipProxyDependentvariable:numberoffirearm-relatedlawsenacted(i.e.billsthatbecamelaw).
TighteningLaws LooseningLaws (1) (2) (3) (4) (5) (6) MassShooting -0.209 -0.469 -0.325 0.189 0.720 0.527 (0.260) (0.584) (0.511) (0.347) (0.786) (0.710)Rep.Leg.×Shooting 0.193 0.169 0.533 0.580 (0.399) (0.385) (0.484) (0.482) Dem.Leg.×Shooting 0.255 0.243 -0.385 -0.382 (0.308) (0.295) (0.467) (0.440) GunSuicidePct.×Shooting 0.005 0.005 -0.010 -0.005 (0.009) (0.010) (0.014) (0.014)GunSuicidePercent 0.037 0.036 0.035 0.106* 0.109* 0.116* (0.027) (0.027) (0.027) (0.060) (0.061) (0.061)Dem.Legislature 0.062 0.062 0.100 -0.255 -0.261 -0.322* (0.171) (0.170) (0.149) (0.229) (0.225) (0.194)Rep.Legislature 0.145 0.149 0.177 0.423** 0.413* 0.510*** (0.143) (0.143) (0.133) (0.216) (0.218) (0.192)Rep.Governor -0.040 -0.046 -0.046 -0.092 -0.082 -0.108 (0.086) (0.088) (0.089) (0.164) (0.162) (0.166)InstitutionalControls Yes Yes Yes Yes Yes Yes DemographicControls Yes Yes Yes Yes Yes Yes StateFixedEffects Yes Yes Yes Yes Yes Yes YearFixedEffects Yes Yes Yes Yes Yes Yes N 1,250 1,250 1,250 1,175 1,175 1,175 Note:Robuststandarderrorsclusteredbystateinparentheses.Starsfollowingcoefficientsrepresentp-valueslessthan.10(*),.05(**)and.01(***).GunSuicidePct.isthe5-yearmovingaverageofthepercentageofsuicidesthatarefirearm-relatedandisusedtoproxyforgunownership(CookandLudwig2006).Models1and4showthatthemain result does not change with adding this control variable; the other models suggest that the response ofRepublicanlegislaturescannotbeexplainedbyratesofgunownership.AllvariablesaredefinedasinTable1.
36
AppendixH:MassShootingsasanInstrumentforGunPolicy
Inthisappendixweusemassshootingsasaninstrumentalvariabletostudytheimpactof
gunlawsongundeaths.Westartwiththefollowingmodel:
ln 𝐷!" = 𝛼! + 𝜃! + 𝛽 𝐺𝑢𝑛 𝐶𝑜𝑛𝑡𝑟𝑜𝑙!" + 𝛿!𝑍!" + 𝜖!"
where𝐷!"isnon-massshootinggundeathsper100,000peopleinstate𝑠andyear𝑡,𝛼!and
𝜃!are stateandyeareffects,𝐺𝑢𝑛 𝐶𝑜𝑛𝑡𝑟𝑜𝑙!"is an index representing the strictnessof gun
policy,and𝑍!"isavectorofcontrols–demographic,political,andeconomicfactors–that
potentiallyaffectgundeaths.WeusethesamevariablesasLevitt (1996)ascontrols,but
also include dummies for Republican and Democratic trifectas or legislatures, and a
dummyforRepublicangovernors.
We do not directly observe𝐺𝑢𝑛 𝐶𝑜𝑛𝑡𝑟𝑜𝑙!"; instead, we observe the enactment of
newlawsthatchangegunpolicy.Therefore,weestimatetheequationinfirstdifferences:
𝛥 ln 𝐷!" = 𝜆! + 𝛽 𝑁𝑒𝑤 𝐺𝑢𝑛 𝐿𝑎𝑤𝑠!" + 𝛥𝑍!"𝛿 + 𝛥𝜖!"
where 𝑁𝑒𝑤 𝐺𝑢𝑛 𝐿𝑎𝑤𝑠!" = 𝛥𝐺𝑢𝑛 𝐶𝑜𝑛𝑡𝑟𝑜𝑙!"isnegativeforlawsthatloosenguncontroland
positiveforlawsthatincreaseguncontrol(accordingtoourcoders,seedatadescription).
Basedonourmainresults,weinstrumentforgunlawsusingthefirstlagsofmassshooting
fatalitiesandtheinteractionof laggedmassshootingfatalitieswithRepublicancontrolof
stategovernment.Theformershouldbepositivelycorrelatedwithnewlawsandthelatter
negativelycorrelatedwithnewlaws.
We estimate the model using Fuller’s (1977) modified LIML with𝛼 = 1(Baum,
Schaffer,andStillman2007).Firststageresultssuggesttheinstrumentsareweakdespite
being jointly significant(𝐹 = 5.98)with the expected sign (Stock and Yogo 2005). The
37
coefficientsontheexogenousinstrumentsinthereducedformequationforfirearmdeaths
are not significant, but also have the expected signs (negative for laggedmass shooting
fatalities and positive for the interaction of lagged fatalities with Republican control of
government).Ourestimate𝛽is−0.016withstandarderror0.013.Aconditionallikelihood
ratiotest(Moreira2003;Andrews,Moreira,andStock2004;FinlayandMagnusson2009)
cannot reject the null hypothesis that𝛽 = 0 (𝑝 = 0.24). We also estimated models that
includeproxiesforgunownership.Includingthepercentageofsuicidescommittedwitha
gun(CookandLudwig2006)doesnotchangeourinference.