Gangs, labor mobility, and development

#672 | 2025
Gangs, labo mobili y, and
de elopmen
NIKITA MELNIKOV, CARLOS SCHMIDT-PADILLA, MARÍA
MICAELA SVIATSCHI
Submi ed o Econome ica
GANGS, LABOR MOBILITY, AND DEVELOPMENT
NIKITA MELNIKOV
No a School o Business and Economics
CARLOS SCHMIDT-PADILLA
Goldman School o Public Policy, Uni e si y o Cali o nia, Be keley
MARÍA MICAELA SVIATSCHI
Depa men o Economics, P ince on Uni e si y
We s udy how c iminal o ganiza ions a ec economic de elopmen . We exploi a na u al
expe imen in El Sal ado , whe e hese c iminal o ganiza ions eme ged due o an exogenous
shi in Ame ican immig a ion policy ha led o he depo a ion o gang leade s om he
Uni ed S a es o El Sal ado . Using a spa ial eg ession discon inui y design ha ocuses on
he gang-c ea ed sys em o bo de s, we ind ha indi iduals in gang-con olled neighbo -
hoods ha e less ma e ial well-being, income, and educa ion han indi iduals li ing only 50
me e s away bu ou side o gang e i o y. None o hese discon inui ies exis ed be o e he
a i al o he gangs. A key mechanism behind he esul s is ha gangs es ic indi iduals’
mobili y, a ec ing hei labo -ma ke op ions by p e en ing hem om commu ing o o he
pa s o he ci y. The esul s a e no de e mined by high a es o selec i e mig a ion, di e -
en ial exposu e o ex o ion and iolence, o di e ences in public goods p o ision.
KEYWORDS: gangs, de elopmen , mobili y, c ime.
1. INTRODUCTION
How do nons a e a med ac o s a ec economic de elopmen ? On he one hand, hey can
impede he s a e om p o iding public goods, en o cing p ope y igh s and con ac s, and
Niki a Melniko : niki a.melniko @no asbe.p
Ca los Schmid -Padilla: [email p o ec ed]
Ma ía Micaela S ia schi: [email p o ec ed]
We hank Alicia Adse a, Ce a Gi ay Aksoy, Albe o Alesina, So ia Ama al, O iana Bandie a, Samuel Bazzi,
Ch is Bla man, Leah Bous an, Timo hy Besley, Eli Be man, E han Bueno de Mesqui a, Filipe Campan e, Do is Chi-
ang, Abby Có do a, Raúl Sanchez de la Sie a, Melissa Dell, Jenni e Doleac, Pa icio Dominguez, John J. Donohue,
Oeind ila Dube, Thad Dunning, S e ano Fio in, Thomas Fujiwa a, Ta ek Ghani, Edwa d Glaese , Je G ogge , Guy
G ossman, Se gei Gu ie , Gau a Khanna, Asim Khwaja, Tom Ki chmaie , Ilyana Kuziemko, Leah Lakdawala, Ho-
acio La eguy, Benjamin Lessing, Nicola Limodio, Sa ah Lowes, S ephen Machin, Luis Ma ínez, A i Mian, Magne
Mogs ad, Ch is Neilson, Sam No is, Ben Olken, Daniel O ega, Emily Owens, Rohini Pande, Paolo Pino i, Osca
Pocasang e, Nishi h P akash, S ephen Redding, James Robinson, Ma k Rosenzweig, Ma eo Sandi, Rogé io San a -
osa, Jacob Shapi o, San iago Tobón, Daniel T eisman, Oli e Vanden Eynde, Juan Va gas, Leona d Wan chekon,
Aus in W igh , Na haniel Young, Eka e ina Zhu a skaya, Owen Zida , Fab izio Zilibo i, and he pa icipan s o sem-
ina s and con e ences a he AEA, AL CAPONE, APPAM, APSA, Be keley, Bocconi, CERP, Con e ence on he
Economics o C ime and Jus ice, he EBRD, ESOC, Ha a d, he IDB, i o Ins i u e, LACEA, LSE, MIEDC, MIT,
NBER SI, PSE, P ince on, Sciences Po, S an o d, Uni e si y o Chicago, Uni e si y o Connec icu , Uni e si y o
Munich, Uni e si y o Passau, Wo ld Bank G oup, and Yale o help ul commen s and sugges ions. We also hank
he In e na ional C isis G oup o helping us access pa s o he da a. Ca los Aguila , B uno Jiménez, Paulo Ma os,
Sa i a O é Quispe, G aciela Saca, and Édga Sánchez-Cue as p o ided excellen esea ch assis ance.
2
p e en ing iolence (Acemoglu e al.,2001,Michalopoulos and Papaioannou,2013). On he
o he hand, i he s a e is weak and unable o con ol pa s o i s e i o y, nons a e a med
ac o s may ake on he ole o he s a e in ul illing essen ial ins i u ional unc ions, po en ially
enabling economic g ow h (Tilly,1985,Olson,1993,Ba es e al.,2002,Ibáñez e al.,2019,
De la Sie a,2020) and compe ing o he “hea s and minds” o ci ilians (Ibáñez e al.,2019,
De la Sie a,2020,Bla man e al.,2022). O e all, how and why nons a e a med ac o s a ec
de elopmen emains an open ques ion.
In his pape , we s udy how a speci ic ype o nons a e a med ac o —namely, c iminal
o ganiza ions—a ec s socioeconomic de elopmen . In u ban a eas in he de eloping wo ld,
millions o people li e unde some o m o c iminal go e nance (Lessing,2021,Bla man e al.,
2022). C iminal o ganiza ions unc ion mainly in u ban cen e s, o en con olling pa s o he
ci y, while o he pa s a e con olled by he s a e. In pa icula , his pape analyzes how wo
o he wo ld’s mos p oli ic gangs—MS-13 (Ma a Sal a ucha) and 18 h S ee (Ba io 18)—
a ec ed socioeconomic de elopmen in El Sal ado .1
We exploi a na u al expe imen ha ook place in El Sal ado . Be o e he mid-1990s, El
Sal ado had no signi ican c iminal o ganiza ions. Howe e , in 1996, a e a shi in Ame ican
immig a ion policy ha made i easie o depo indi iduals—especially hose wi h c iminal
backg ounds—back o hei coun y o o igin, many Sal ado an mig an s who we e membe s
o Cali o nia-based gangs (speci ically, MS-13 and 18 h S ee ) we e depo ed back o El Sal-
ado . These depo ed gang membe s ees ablished hei gangs in El Sal ado and quickly
gained con ol o e ce ain pa s o he coun y. To p o ec hei e i o y om ou side s, he
gangs also e-c ea ed a sys em o bo de s and checkpoin s ha hey used o es ablish e i o ial
dominance in Cali o nia (Nuño and Magui e,2021), esul ing in he di ision o u ban a eas
be ween he gangs and he s a e.
To es ima e he e ec s o gangs’ e i o ial con ol, we use he bounda ies o gang-con olled
neighbo hoods in El Sal ado ’s capi al, San Sal ado , o implemen a spa ial eg ession discon-
inui y design. These e i o ial dema ca ions we e o med soon a e he gang leade s a i ed
in 1996, and hey oughly coincide wi h exis ing na u al ba ie s, such as boule a ds and high-
ways. We measu e he ou come a iables using he 2007 census and ou own geocoded su ey,
which we conduc ed in 2019 in bo h gang and nongang neighbo hoods.
Ou esul s indica e ha esiden s o gang-con olled neighbo hoods in San Sal ado ha e
wo se dwelling condi ions, less income, and a lowe p obabili y o owning du able goods com-
pa ed o indi iduals li ing jus 50 me e s away bu ou side o gang e i o y. They a e also
less likely o wo k in la ge i ms. Fo ins ance, we ind ha esiden s o gang a eas ha e $350
less mon hly household income ( he sample mean is $625) compa ed o indi iduals li ing in
neighbo ing nongang loca ions and ha e a 12-pe cen age-poin lowe p obabili y o wo king
1Bo h MS-13 and 18 h S ee also ha e a majo p esence in Hondu as, Gua emala, and pa s o I aly, Mexico,
Spain, and he Uni ed S a es. Simila c iminal o ganiza ions a e also p esen in many o he coun ies (e.g., B azil,
Colombia, Jamaica, and Sou h A ica).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 3
in a i m wi h a leas 100 employees. The esul s a e highly obus o he choice o empi ical
speci ica ions.
These di e ences in li ing s anda ds did no exis be o e he gang leade s a i ed. We epli-
ca e he eg ession discon inui y design wi h da a om he 1992 census o show ha , be o e
he gangs eme ged, a eas on bo h sides o he gang bo de s had simila socioeconomic and ge-
og aphic cha ac e is ics, as well as simila le els o c ime. These esul s a e consis en wi h he
ac ha he bounda ies o gang e i o y we e no o med based on p eexis ing socioeconomic
di e ences, bu a he on he a ailabili y o na u al ba ie s (i.e., majo oads). We also show
ha he na u al ba ie s a e no associa ed wi h di e ences in socioeconomic condi ions when
hey do no de e mine gang e i o ial con ol.
An impo an mechanism h ough which gangs a ec socioeconomic de elopmen in he
neighbo hoods hey con ol is ela ed o es ic ions on indi iduals’ mobili y. The gangs’ long-
e m su i al depends on hei abili y o secu e he bo de s o hei e i o y and p e en he
police and i al gang membe s om a es ing o killing hem. The e o e, o main ain secu e
con ol o e hei e i o y, bo h MS-13 and 18 h S ee ins i u ed a sys em o checkpoin s,
no allowing indi iduals o eely en e o lea e gang-con olled neighbo hoods (ICG, 2018).
The secu i y o hei e i o y also allows he gangs o use i as a b idgehead om which hey
conduc ex o ion aids o neighbo ing a eas.
Using he da a om ou geocoded su ey, we pe o m a spa ial eg ession discon inui y de-
sign o documen he p esence o es ic ions on indi iduals’ mobili y. We show ha esiden s
o gang a eas a e 50 pe cen age poin s mo e likely o wo k in gang e i o y compa ed o indi-
iduals li ing only 50 me e s away bu on he nongang side o he bounda ies. They a e also
mo e likely o say ha gang-imposed bo de s p e en ed hem om ge ing jobs in la ge i ms in
o he pa s o he ci y, less likely o say ha he e is eedom o mo emen in he neighbo hood
whe e hey li e, and less likely o ha e been o places ou side o San Sal ado . Howe e , hose
indi iduals do no ha e lowe le els o mobili y pe se. Using cell phone ping da a, we show
ha , while esiden s o gang-con olled neighbo hoods a e la gely con ined in hei mo emen s
o gang e i o y, hey a el he same dis ance as hei pee s on he o he side o he gang
bounda ies.
These mobili y es ic ions a ec labo -ma ke ou comes: esiden s o gang e i o y end
up wo king in smalle i ms and ea ning lowe wages because hey canno commu e o he
a eas whe e he la ges and bes -paying i ms a e loca ed. No ably, labo -ma ke condi ions do
no change di ec ly a he bounda ies o gang e i o y (i.e., he e is no change in i m size,
wages, p o i abili y, o he numbe o business es ablishmen s). Ins ead, we show ha , a e he
eme gence o he gangs, new business es ablishmen s inc easingly opened in a eas a away
om gang e i o y. None heless, esiden s o nongang neighbo hoods close o he bounda ies
we e able o ake ad an age o hese new labo -ma ke oppo uni ies by commu ing o pa s
o he ci y whe e he la ges i ms a e loca ed, whe eas indi iduals li ing in gang a eas we e
p e en ed om doing so by he es ic ions on hei mobili y.
4
Ano he ac o limi ing socioeconomic de elopmen in gang-con olled neighbo hoods is
ela ed o educa ional a ainmen . Using school census da a, we show ha he annual school
d opou a e is 2 pe cen age poin s highe in gang e i o y han in nongang a eas. The di e -
ences in educa ional a ainmen con ibu e o u he widening he income gap be ween gang
and nongang e i o ies.
We also examine o he po en ial de e minan s o lowe socioeconomic de elopmen in gang-
con olled neighbo hoods, bu we ind ha , in his con ex , hey canno explain he esul s.
In pa icula , we demons a e ha indi iduals and i ms on bo h sides o he bounda ies a e
equally exposed o ex o ion and o he iolen c imes. This esul is explained by he ac ha ,
since gang membe s a e no subjec o he same mobili y es ic ions as he o he people li ing
on hei e i o y, hey conduc egula aids in o neighbo ing a eas ou side hei immedia e
con ol. This esul is ully consis en wi h he inding ha labo -ma ke condi ions do no
change di ec ly a he bounda ies o gang e i o y.
Simila ly, we ind no di e ences in he a ailabili y and quali y o public goods p o ision
(e.g., schools and hospi als), consis en wi h he quali a i e e idence sugges ing ha he go -
e nmen has been willing o p o ide public goods in gang a eas o a oid os acizing he es-
iden s o hose loca ions.2In u n, because he gangs bene i om public goods p o ision in
hei neighbo hoods, hey ha e been willing o allow he go e nmen o p o ide (nonpolice-
ela ed) se ices in he a eas hey con ol.3Finally, we show ha he esul s a e no d i en by
highe le els o unemploymen (o in o mal employmen ) in gang-con olled neighbo hoods
and ha selec i e mig a ion o indi iduals ac oss he bounda ies o gang e i o y can explain
no mo e han 14% o he gap in socioeconomic de elopmen be ween he gang and nongang
neighbo hoods.
Finally, we use da a om all o El Sal ado o pe o m a di e ence-in-di e ences design ha
analyzes how gang p esence a ec ed he spa ial alloca ion o economic ac i i y in he coun y.
We ind ha , a e he a i al o he gangs, municipali ies leas exposed o gang ac i i y expe-
ience signi ican ly mo e openings o new business es ablishmen s, as well as highe g ow h in
nigh ime ligh densi y and household income. These esul s highligh how he economic cos s
o mobili y es ic ions inc eased o e ime: as employmen oppo uni ies imp o ed in places
2In addi ion, he go e nmen and o he poli ical ac o s a e mo i a ed by elec o al conside a ions: wi hou p o iding
public goods in gang-con olled neighbo hoods, poli ical pa ies would likely ha e been unable o campaign in hose
a eas (e.g., see Có do a,2019). This s ems om he clien -b oke ela ionship be ween he poli ical pa ies and he
gangs, pa icula ly du ing elec ions. To campaign in gang-con olled neighbo hoods, poli ical pa ies need o p o ide
public goods in hose a eas.
3We ind ha he gangs hemsel es p o ide e y limi ed public se ices, he p obabili y o which does no change
a he bounda ies o gang e i o y. This esul may be di e en in o he se ings whe e nons a e ac o s ha e he
esou ces and incen i es o co-op he popula ion unde hei con ol (e.g., Magaloni e al.,2020b,Bla man e al.,
2022). In pa icula , in San Sal ado , he gangs migh no p o ide mo e public se ices in hei e i o ies because he
go e nmen has been willing o p o ide hem. Sal ado an gangs also ha e limi ed inancial esou ces (Ma ínez e al.,
2016), making i di icul o hem o compe e o hea s and minds. Howe e , in se ings whe e he go e nmen is
no p esen (e.g., in u al a eas) and c iminal o ganiza ions ha e he esou ces o p o ide se ices o he public (e.g.,
d ug ca els), e i o ial con ol by nons a e ac o s may esul in mo e public goods p o ision.

GANGS, LABOR MOBILITY, AND DEVELOPMENT 5
wi hou gang ac i i y, i became inc easingly impo an o be able o commu e o wo k in hose
a eas.
Ou pape is ela ed o se e al s ands o he exis ing li e a u e. Fi s , i con ibu es o he li -
e a u e s udying he o igins and consequences o o ganized c ime and o he nons a e a med ac-
o s (e.g., Gambe a,1996,F ye and Zhu a skaya,2000,Bandie a,2003,Daniele and Ma ani,
2011,Acemoglu e al.,2013,Daniele and Geys,2015,Buonanno e al.,2015,2016,Dell,2015,
Pino i,2015,Daniele and Dipoppa,2017,De Feo and De Luca,2017,Acemoglu e al.,2019,
Alesina e al.,2019,De la Sie a,2020,Mu phy and Rossi,2020,Mi enda e al.,2022,S i-
a schi,2022a,b). Mos o his li e a u e has ocused on iolence, o he po en ial he eo , as
he channel behind he e ec s o o ganized c ime on poli ics, in es men , mig a ion, and o he
aspec s o socioeconomic de elopmen .
We complemen his li e a u e by p esen ing no el e idence on one speci ic aspec o c im-
inal o ganiza ions ha is inc easingly p e alen in he de eloping wo ld: e i o ial con ol in
u ban se ings. By looking a u ban a eas whe e he e i o y is di ided be ween he s a e and
he gangs, we documen a p e iously igno ed mechanism h ough which c iminal o ganiza ions
a ec socioeconomic de elopmen : es ic ions on mobili y. As Glaese and Sims (2015) poin
ou , li le is known abou he consequences o c ime in he u banized, de eloping wo ld. In
hese con ex s, because c iminal o ganiza ions cons an ly ace he po en ial o e i o ial chal-
lenges bo h om i al c iminal g oups and om he s a e, hey need o implemen s ingen
secu i y measu es o p o ec he bo de s o he neighbo hoods hey con ol (e.g., imposing e-
s ic ions on indi iduals’ mobili y). As a esul , esiden s o hese neighbo hoods end up ha ing
signi ican ly wo se labo -ma ke ou comes because o hei inabili y o wo k in o he pa s o
he ci y.
Second, ou pape is ela ed o he li e a u e on c iminal go e nance and he o ganiza ional
s uc u e o c iminal en e p ises (Le i and Venka esh,2000,Ska bek,2011,Ca alho and
Soa es,2016,Ibáñez e al.,2019,Lessing and Willis,2019,Magaloni e al.,2020a,Lessing,
2021,Bla man e al.,2022). Much o he exis ing li e a u e has shown how nons a e a med
ac o s eme ge o ill he oid le by he s a e and p o ide secu i y and o he public goods o he
local popula ion in exchange o poli ical in luence (e.g., Bla man and Miguel,2010), axa ion
(e.g., Olson,1993,De la Sie a,2020), and he oppo uni y o conduc hei illegal ac i i ies.
Ou pape analyzes how hese ela ionships a e al e ed in an u ban con ex , whe e he p oximi y
o he s a e, on he one hand, poses a h ea o he gangs’ e i o ial con ol bu , on he o he
hand, allows he gangs o ely on he p o ision o mos public goods by he go e nmen .4
Thi d, ou pape con ibu es o he li e a u e s udying he causes and consequences o he
o ma ion o ex ac i e ins i u ions, which can ha e a long-las ing impac on socioeconomic
4In pa icula , while he li e a u e on s a iona y bandi s would imply ha a med ac o s ha e incen i es o max-
imizing esiden s’ incomes—including h ough some public goods p o ision— o maximize ex o ion en s in he
e i o y hey con ol (e.g., Olson,1993,De la Sie a,2020), we p o ide no el e idence ha his incen i e can be
unde mined in an u ban con ex whe e labo -ma ke mobili y is needed o maximize income.
6
de elopmen (e.g., Acemoglu e al.,2001,2002,Dell,2010,Michalopoulos and Papaioannou,
2013,Dell e al.,2018,Dell and Olken,2020,Lowes and Mon e o,2021). Speci ically, we show
how he depo a ion o c iminal leade s om he Uni ed S a es o El Sal ado has esul ed in
hei es ablishing ex o iona y gangs ha signi ican ly limi ed socioeconomic de elopmen in
El Sal ado . I also con ibu es o a long-s anding deba e on whe he indi idual leade s—in his
case, gang leade s—a ec economic g ow h in de eloping coun ies (Jones and Olken,2005).
Finally, ou wo k is ela ed o he li e a u e analyzing he economic e ec s o ba ie s o
geog aphical mobili y. The exis ing li e a u e has ocused on he e ec s o in e na ional bo -
de s (e.g., Clemons e al.,2008,McKenzie e al.,2010,Me go,2016,Calì and Miaa i,2018,
Alsawady e al.,2022) and he absence o anspo a ion in as uc u e (e.g., Donaldson,2018,
Ashe and No osad,2020). We complemen his wo k by showing how gang-imposed es ic-
ions on indi iduals’ eedom o mo emen can signi ican ly a ec socioeconomic de elop-
men , e en wi hin an in eg a ed me opoli an a ea and in he absence o di ec anspo a ion
cos s and legal bo de s. Gi en he global p e alence o simila in acoun y ba ie s o mobil-
i y, ou esul s p o ide impo an policy implica ions o many de eloping coun ies. In pa -
icula , nons a e a med ac o s es ic indi iduals’ eedom o mo emen in B azil, Colombia,
Gua emala, and Hondu as (e.g., Ibáñez e al.,2019,Magaloni e al.,2020a); many o he coun-
ies, oo, expe ience a ious o ms o mobili y es ic ions (e.g., see Wal he e al.,2020).
The es o his pape is s uc u ed as ollows. Sec ion 2desc ibes he ise and o ganiza ion
o c iminal g oups in El Sal ado . Sec ion 3desc ibes he main da a sou ces. Sec ion 4p esen s
he iden i ica ion s a egy and he main esul s. Sec ion 5examines he mechanisms d i ing he
esul s. Sec ion 6analyzes he agg ega e e ec s o gang p esence. Sec ion 7concludes.
2. HISTORICAL BACKGROUND
In his sec ion, we p esen an o e iew o how MS-13 and 18 h S ee de eloped in Sal ado-
an mig an communi ies in he Uni ed S a es and how c iminal capi al was expo ed om hese
communi ies o El Sal ado ollowing a shi in Ame ican immig a ion policy in 1996. We hen
desc ibe how, once in El Sal ado , he gangs quickly ees ablished hei c iminal s uc u es,
began ec ui ing, and gained e i o ial con ol o e many u ban neighbo hoods h oughou he
coun y, mos no ably in he capi al, San Sal ado . We also p o ide quali a i e e idence on how
he bounda ies o gang e i o y we e o med soon a e he a i al o he c iminal depo ees,
based on he sys em o e i o ial con ol ha he gangs had de eloped in he Uni ed S a es.
2.1. The O igins o MS-13 and 18 h S ee
Sou he n Cali o nia, especially Los Angeles, became home o housands o Sal ado ans
leeing he coun y’s descen in o ci il wa in he 1980s (S anley,1987). Lacking an es ab-
lished suppo ne wo k, Sal ado an mig an s li ed in poo , o e c owded neighbo hoods and
o en aced disc imina ion om o he mig an g oups (B e ell,2011). In a ypical amily, bo h
pa en s wo ked, o en lea ing he child en unsupe ised (Sa enije,2009).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 7
Le on hei own and acing p ejudice om o he mig an g oups and hei gangs, some Sal-
ado an you h o med he p ecu so s o MS-13—sel -de ense g oups ha we e ini ially be e
known o pe y c imes and o hei a ini y o cannabis and hea y me al, a he han o b u-
al iolence—while o he s joined 18 h S ee , an exis ing Mexican gang (Dunn,2007,C uz,
2010,Ma ínez and Ma ínez,2018). As membe ship in MS-13 and 18 h S ee g ew ac oss
Sal ado an immig an communi ies, he gangs became known o he local au ho i ies. Some o
hei membe s we e sen o p ison, whe e hey gained c iminal capi al and social connec ions
ha helped hem solidi y hei s uc u es (Wome and Bunke ,2010,Ma ínez and Ma ínez,
2018). By he mid-1980s, bo h MS-13 and 18 h S ee had de eloped independen iden i ies,
o ganiza ional s uc u es e ol ing a ound e i o y-based cliques (clicas), and a ie ce i al y
ha con inues o his day (Wa d,2013).
Many gangs in 1980s Los Angeles sha ed a no ewo hy ai : hey p ecisely dema ca ed hei
e i o y, which g ea ly con ibu ed o hei iden i y and de elopmen (Coughlin and Venka esh,
2003). Fo example, hey used g a i i o de ine he e i o ies unde hei con ol and o p ojec
au ho i y o e hei i als and he local popula ion (Ti a e al.,2005,A sy,2018). This de-
ma ca ion had a p o ound impac on he mobili y and decisions o indi iduals li ing in gang
e i o ies: “One o he eally impo an hings o hink abou is how he in isible bo de s [...]
add cos s we o en don’ hink abou . I I’m a young pe son g owing up in a pa icula neigh-
bo hood [in Los Angeles] and he closes mo ie hea e o he closes shopping mall is claimed
by a i al gang, [...] I’m going o ha e o spend mo e ime on a bus, pu mo e gas in my ca ,
o a el o o he a eas” (A sy,2018).
In an obse a ional s udy o inca ce a ed MS-13 gang membe s in Los Angeles Coun y,
Nuño and Magui e (2021) highligh how “mos MS-13 membe s a e in ol ed in cliques ha
claim ce ain u o e i o y (96.3%) and would be willing o use iolence o de end i agains
o he s (92.6%),” elying on g a i i and ou pos s o ma k and con ol hei e i o ies.5This ace
o gang cul u e became a undamen al ai o gang s uc u es in El Sal ado .
2.2. Ame ican Immig a ion Policy and he Eme gence o Gangs in El Sal ado
In 1996, o educe c ime in u ban a eas and add ess he su ge in i egula mig a ion, he
Uni ed S a es passed he Illegal Immig a ion Re o m and Immig a ion Responsibili y Ac
(IIRIRA) (Chacón,2009,Ab ego e al.,2017). IIRIRA d as ically inc eased immig a ion en-
o cemen , c ea ing p ocedu es o expedi ed emo al, adding new g ounds o depo a ion, and
inc easing he numbe o bo de pa ol agen s. This shi in Ame ican policy had a p o ound
impac on El Sal ado . Du ing he i s wa e o depo a ions in 1996, o e 500 Sal ado an gang
membe s we e depo ed om he Uni ed S a es, leading o de as a ing changes in Sal ado an
communi ies (S ia schi,2022b).
5The e i o ial iden i y is so impo an ha , when MS-13 and 18 h S ee expanded o El Sal ado , many o he
cliques he e adop ed names ha e e enced he loca ions whe e hei gang leade commenced hei illici ca ee s in
he Uni ed S a es (e.g., Hollywood Locos Sal a uchos).
8
Gi en ha hey did no ha e c iminal eco ds in El Sal ado , he epa ia ed gang membe s—
many o whom we e se ing o had p e iously se ed sen ences in he Uni ed S a es—gained
hei eedom a e e u ning o hei home coun y (Wa d,2013). El Sal ado was s ill eco e -
ing om i s ci il wa , which ended in 1992, and he Sal ado an s a e did no ha e he esou ces
o p e en he gangs om expanding. The 1992 Peace Acco ds manda ed he c ea ion o a new
police o ce— he Na ional Ci il Police (Policía Nacional Ci il, PNC)—and a he ime o he
epa ia ions, he s uc u e o he PNC was s ill being de ined (e.g., no u al police uni s exis ed
un il 2004). The epa ia ed gang leade s exploi ed his low le el o s a e capaci y and expanded
hei ope a ions o many u ban a eas.
Mos o he epa ia ed MS-13 and 18 h S ee gang membe s had li ed in he Uni ed S a es
since a young age and knew li le abou hei home coun y. Fo his eason, mos o hem
e u ned o hei bi h municipali ies, elying on hei amily ne wo ks o ese le in a new en-
i onmen (DeCesa e,1998,S ia schi,2022b). Seeking social accep ance and s a us, he gang
depo ees banded oge he and apped in o local you h g oups o eplica e he gang s uc u es
hey had in Cali o nia. E en hough only a ew hund ed gang membe s we e epa ia ed om
he Uni ed S a es in 1996, hey quickly expanded hei anks, ec ui ing new membe s om
he local popula ion. Many locals we e a ac ed by he cama ade ie and espec ha he gangs
o e ed, o he s sough mo e angible ma e ial gains such as money and d ugs (C uz and Po illo
Peña,1998,Ma ínez and Ma ínez,2018). S ia schi (2022b), in pa icula , shows how, a e
he MS-13 and 18 h S ee gang membe s a i ed, and began ec ui ing adolescen s o join hei
s uc u es, El Sal ado expe ienced an immedia e inc ease in gang- ela ed ac i i ies. Acco ding
o he local au ho i ies, by he end o 1996, a leas 20 housand indi iduals had joined he wo
gangs (C uz and Po illo Peña,1998).
2.3. The Fo ma ion o Gang Te i o y in El Sal ado
Taking ad an age o he pos wa en i onmen and widesp ead des i u ion, bo h MS-13 and
18 h S ee quickly expanded hei in luence o e many neighbo hoods, pa icula ly in he cap-
i al, San Sal ado , and o he u ban a eas, “gain[ing] comple e con ol o [ce ain] locali ies”
(Zoe hou ,2015). This apid o ma ion and en o cemen o bounda ies was possible due o
ou main ac o s: (i) he gangs’ expe ience in implemen ing a sys em o e i o ial con ol in
Cali o nia, (ii) he impo ance o e i o ial con ol o he gangs’ iden i y and long- e m su -
i al, (iii) he gangs’ abili y o ec ui new membe s om he local popula ion, and (i ) El
Sal ado ’s low s a e capaci y in he 1990s.
The sys em o e i o ial con ol buil upon he s a egy he gangs honed in Cali o nia, whe e
dema ca ion, la gely based on na u al ba ie s, spli u ban a eas in o small geog aphical con-
ines known as cliques (Miguel C uz,2010). In El Sal ado , he gangs also de ined hei e i-
o y based on na u al ba ie s such as majo oads and boule a ds (Teno io,2002,Vega,2015).
We iden i y and ake ad an age o h ee such majo oads (see Figu e 1)—Bule a Venezuela,
49 A enida Su , and Au opis a Comalapa, all o which exis ed in 1996— ha la gely de e mined
GANGS, LABOR MOBILITY, AND DEVELOPMENT 15
whe e he loca ions o he bounda ies we e de e mined by he p esence o na u al ba ie s ha
p e en ed he gangs om expanding u he . We hen use hese na u al bounda ies o gang
e i o y o e i y ha ou esul s a e no d i en by he po en ial endogenei y o some o he
o he bounda ies. The second assump ion is ha esiden s o gang e i o y did no selec i ely
mig a e om hose a eas o neighbo ing loca ions in he con ol g oup. Subsec ion 4.3 and Ap-
pendix Subsec ion A.2 p o ide a de ailed discussion o his assump ion, showing ha selec i e
mig a ion can explain no mo e han 14% o he socioeconomic gaps be ween gang and nongang
a eas.
4.2. Main Resul s
Table Ip esen s he esul s o es ima ing Speci ica ion (1) using he 2007 census da a. I
shows ha , a e expe iencing gang ule, indi iduals li ing in gang-con olled neighbo hoods
ha e signi ican ly wo se dwelling condi ions, lowe le els o educa ion, and a e less weal hy
han hei pee s on he o he side o he bounda ies. Fo ins ance, esiden s o gang e i o y a e
es ima ed o ha e a 21-pe cen age-poin lowe p obabili y o owning a ca , a 15-pe cen age-
poin lowe p obabili y o ha ing a high school deg ee, and a 5-pe cen age-poin lowe p oba-
bili y o li ing in a house wi h conc e e walls han indi iduals li ing less han 50 me e s away
bu no unde he con ol o gangs.20 The esul s o he o he measu es o socioeconomic de-
elopmen p esen he same pa e n.
Figu e 2illus a es he indings om Table I o he i s p incipal componen s o he
dwelling, household, and indi idual cha ac e is ics. The e ical axis ep esen s he a e age
alue o he ou comes a iables; he ho izon al axis ep esen s dis ance (in me e s) o he bound-
a ies o gang e i o y. A eas o he le o he dashed line a e loca ed ou side o gang e i o y;
a eas o he igh a e con olled by he gangs. Fo all he ou come a iables, he e is a clea
discon inui y a he bounda ies o gang-con olled neighbo hoods.21
O e all, he esul s sugges ha gangs ha e had a signi ican nega i e e ec on socioeco-
nomic de elopmen in he neighbo hoods hey con ol. To es ima e he o al mone a y cos o
his e ec , we conside a a iable ha po en ially agg ega es all he e ec s o li ing unde gang
con ol in o one—household income— he da a o which come om he 2019 su ey. The le
pa o Appendix Figu e A.4 p esen s he eg ession discon inui y plo o his a iable. The
esul s sugges ha esiden s o gang neighbo hoods ea n app oxima ely $350 less each mon h
compa ed o esiden s o nongang a eas. Gi en ha he a e age mon hly income in ou sample
is $625, his discon inui y implies a subs an ial educ ion in ea nings. Table A.I in he Ap-
20In he indi idual-le el eg essions, he sample consis s o he en i e popula ion. The esul s a e e y simila i ,
ins ead, we analyze jus he adul popula ion.
21In he Appendix, we illus a e he esul s o all he o he ou come a iables om Table I. Figu e A.1 p esen s
he esul s o dwelling cha ac e is ics, Figu e A.2 o indi idual cha ac e is ics, and Figu e A.3 o household cha -
ac e is ics.

16
TABLE I
SOCIOECONOMIC CONDITIONS AFTER EXPOSURE TO GANG CONTROL
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.047 0.026 -0.050 -0.079 0.006 -0.131
(0.015)*** (0.010)** (0.021)** (0.021)*** (0.002)*** (0.029)***
[0.017]*** [0.010]** [0.027]* [0.027]*** [0.003]** [0.039]***
Mean o dep. a . 0.932 0.028 0.941 0.108 0.005 0.181
Obse a ions 72,087 60,675 62,169 62,169 62,169 59,776
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.013 -0.207 -0.135 -0.021 -0.173 -0.693
(0.006)** (0.046)*** (0.033)*** (0.006)*** (0.035)*** (0.195)***
[0.005]*** [0.057]*** [0.040]*** [0.008]** [0.045]*** [0.203]***
Mean o dep. a . 0.033 0.429 0.697 0.952 0.346 3.093
Obse a ions 59,096 60,045 60,168 60,384 60,020 62,169
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.032 -0.153 -0.121 -0.036 -0.089 -0.101
(0.007)*** (0.029)*** (0.026)*** (0.012)*** (0.019)*** (0.020)***
[0.008]*** [0.033]*** [0.030]*** [0.013]*** [0.024]*** [0.023]***
Mean o dep. a . 0.928 0.449 0.208 0.952 0.378 0.522
Obse a ions 208,416 202,935 202,935 60,675 58,293 202,935
No e: *** p<0.01, ** p<0.05, * p<0.1. A e expe iencing gang con ol, gang-con olled a eas ha e wo se socioeconomic condi ions
han neighbo ing a eas ha we e no unde he con ol o gangs. The able p esen s he esul s o es ima ing Speci ica ion (1) o he a iables
om he 2007 census. The uni o obse a ion is a dwelling, household, o indi idual, depending on which cha ac e is ics a e being conside ed.
In he indi idual-le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols include a linea end in dis ance o he
bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e
bins, deno ing he dis ance o gang e i o y (sepa a ely o each side o he bounda ies). S anda d e o s in b acke s a e adjus ed o allow o
spa ial co ela ion wi hin a 100 me e adius (Conley co ec ion).
pendix p esen s he eg ession es ima es o household income and he o he socioeconomic
cha ac e is ics om he 2019 su ey.
4.3. Add essing Iden i ica ion Challenges
In his subsec ion, we analyze he assump ions ha need o be sa is ied o he es ima es in
Table I o ep esen he causal e ec o gang con ol on socioeconomic de elopmen .
Condi ions be o e he a i al o he gangs. To ensu e ha nongang a eas close o he
bounda ies o gang e i o y a e he app op ia e coun e ac ual o gang-con olled neighbo -
hoods, we check whe he , be o e he a i al o he gangs, hose loca ions had any p eexis ing
di e ences in geog aphy, socioeconomic de elopmen , o c ime.
Fi s , we es ima e Speci ica ion (1) o po en ially impo an neighbo hood cha ac e is ics
(e.g., ele a ion, access o wa e ways, oad densi y) and he socioeconomic cha ac e is ics om
GANGS, LABOR MOBILITY, AND DEVELOPMENT 17
FIGURE 2.—Socioeconomic Condi ions A e 10 Yea s o Gang Con ol
No e: By 2007, socioeconomic condi ions had become signi ican ly wo se in gang-con olled a eas. The igu e illus a es he esul s o
he 1s p incipal componen s o he dwelling, household, and indi idual cha ac e is ics om Table I. All he a iables come om he 2007
census. The uni o obse a ion is a dwelling, a household, and an indi idual, depending on he speci ica ion. All he a iables a e no malized
o a y be ween ze o and one wi h highe alues ep esen ing be e ou comes. The e ical axis ep esen s he a e age alue o he ou comes
a iable; he ho izon al axis—dis ance (in me e s) o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed
ou side o gang e i o y; a eas o he igh a e con olled by he gangs. The do s ep esen he a e age alue o he ou come a iable in ha 30
me e bin.
he 1992 census (e.g., dwelling condi ions, ha ing a TV).22 Columns 1–24 o Table II p esen
he esul s. The e a e no discon inui ies in any o he a iables, con i ming he no ion ha ,
ini ially, he loca ions on opposi e sides o he bounda ies we e no di e en om one ano he .
Appendix Figu e A.5 illus a es hese esul s o he i s p incipal componen s o dwelling,
household, and indi idual cha ac e is ics.23
Nex , we es ima e Speci ica ion (1) o he le el o c ime p io o he a i al o he gangs,
measu ed by he numbe o people inca ce a ed in di e en pa s o he ci y. Using he inca ce -
a ion eco ds om San Sal ado ’s p isons, we geocode he esiden ial add esses o he 4,726
indi iduals who had been inca ce a ed p io o 1997. Then, we calcula e he numbe o inca -
ce a ions pe squa e kilome e o each 10-me e bin, deno ing he dis ance o he bounda ies o
gang e i o y (sepa a ely o each side o he bounda ies).24 Columns 25–30 o Table II p esen
22Some neighbo hood cha ac e is ics (e.g., ele a ion o access o wa e ways) a e ime-in a ian . O he neighbo -
hood cha ac e is ics may change o e ime. Fo all he a iables excep o oad densi y, we use he da a om ei he
be o e he a i al o he gangs o soon a e hei a i al. Fo oad densi y, he da a e lec 2020 in as uc u e, making
he p e ea men balance es o his a iable alid only unde he assump ion ha oad densi y is p ac ically ime-
in a ian . Howe e , gi en he di icul y o cons uc ing new oads in he cen e o a la ge ci y, his assump ion is likely
o be sa is ied. We desc ibe he da a in de ail in he Supplemen a y Ma e ials.
23Figu es S.1–S.4 in he Supplemen a y Ma e ials p esen he esul s o each o he neighbo hood, dwelling,
household, and indi idual cha ac e is ics om Table II.
24We pe o m he calcula ion as ollows. Fi s , we di ide he map o San Sal ado in o zones, deno ing e e y 10
me e s om he bounda ies o gang e i o y, sepa a ely o gang and nongang a eas (all nongang loca ions wi hin 10
me e s o he bounda ies o gang e i o y, all nongang loca ions 10–20 me e s away om gang e i o y, and so on).
Then, o each o he zones, we calcula e he numbe o geocoded add esses wi hin i and di ide ha numbe by he
a ea o he zone. We employ he same p ocedu e o o he ou come a iables wi h he same uni o analysis.
18
TABLE II
GEOGRAPHIC AND SOCIOECONOMIC CHARACTERISTICS BEFORE THE ARRIVAL OF THE GANGS
Neighbo hood cha ac e is ics
U ban e i o y Road densi y Has access o Ele a ion Te i o y used o T ee co e age
he wa e ways co ee p oduc ion
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.018 -0.571 0.027 -0.193 0.008 -0.005
(0.060) (0.953) (0.069) (16.599) (0.019) (0.026)
[0.052] [1.849] [0.095] [17.439] [0.023] [0.027]
Mean o dep. a . 0.814 17.84 0.326 720.4 0.049 0.029
Obse a ions 476 476 476 476 476 476
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Sha ed ba h oom
o conc e e in as uc u e ligh ing and cooking
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.015 -0.003 -0.032 -0.036 -0.007 0.021
(0.036) (0.028) (0.047) (0.039) (0.017) (0.032)
[0.035] [0.030] [0.046] [0.030] [0.013] [0.029]
Mean o dep. a . 0.813 0.010 0.816 0.182 0.030 0.142
Obse a ions 64,899 64,899 64,899 64,899 64,899 64,899
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a blende Numbe o ooms
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.004 -0.049 -0.030 0.009 0.014 -0.069
(0.009) (0.051) (0.054) (0.019) (0.032) (0.170)
[0.007] [0.043] [0.049] [0.019] [0.034] [0.172]
Mean o dep. a . 0.034 0.285 0.320 0.860 0.625 2.670
Obse a ions 64,899 64,899 64,899 64,899 64,899 64,899
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(19) (20) (21) (22) (23) (24)
Gang e i o y -0.000 -0.014 -0.019 -0.005 -0.016 -0.013
(0.011) (0.028) (0.017) (0.031) (0.030) (0.018)
[0.009] [0.028] [0.017] [0.031] [0.026] [0.018]
Mean o dep. a . 0.904 0.314 0.112 0.863 0.525 0.380
Obse a ions 234,749 227,281 227,281 64,899 64,899 227,281
Numbe o inca ce a ions pe km2p io o 1997:
All c imes Homicide Robbe y Sex c imes Assaul O he iolen c imes
(25) (26) (27) (28) (29) (30)
Gang e i o y -2.096 1.548 -0.640 -1.404 -0.823 -1.777
(18.200) (1.291) (3.979) (1.321) (3.400) (1.873)
Mean o dep. a . 114.6 4.476 21.61 6.147 19.78 9.275
Obse a ions 86 86 86 86 86 86
No e: *** p<0.01, ** p<0.05, * p<0.1. Be o e he a i al o he gangs, loca ions on ei he side o he bounda ies o gang e i o y
had simila geog aphic and socioeconomic cha ac e is ics. The able p esen s he esul s o es ima ing Speci ica ion (1) o he neighbo hood
cha ac e is ics and he a iables om he 1992 census. The uni o obse a ion is a census ac , dwelling, household, o indi idual, depending
on which cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols
include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in
pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
S anda d e o s in b acke s a e adjus ed o allow o spa ial co ela ion wi hin a 100 me e adius (Conley co ec ion). In columns 25–30, he
Conley s anda d e o s a e no epo ed because he e he loca ion o he obse a ions is no de ined ( he uni o obse a ion is a 10 me e bin,
deno ing he dis ance o he bounda ies o gang e i o y).
he esul s o es ima ing Speci ica ion (1) o di e en ypes o c imes, showing ha loca ions
on bo h sides o he bounda ies had simila le els o c ime p io o he a i al o he gangs.25
25As we explain in oo no e 18, we canno epo Conley s anda d e o s in hese speci ica ions, because he uni
o analysis includes a eas om di e en pa s o he ci y.
GANGS, LABOR MOBILITY, AND DEVELOPMENT 19
O e all, be o e he mid-1990s, gang and nongang loca ions had simila le els o socioeco-
nomic de elopmen and c ime, allowing us o conclude ha nongang a eas close o he bound-
a ies a e he app op ia e coun e ac ual o gang neighbo hoods in he absence o gang con ol.
Bounda ies o gang e i o y om geog aphical ba ie s. To add ess any emaining con-
ce ns ega ding he po en ial endogenei y o he bounda ies, we pe o m he ollowing analysis.
We iden i y h ee majo mul ilane oads—Bule a Venezuela,49 A enida Su , and Au opis a
Comalapa—which oge he o m mo e han 45 kilome e s o na u al ba ie s ha la gely de e -
mined he sou he n and wes e n bounda ies o gang e i o y.26 Table III epo s he esul s o
es ima ing Speci ica ion (1) using hese h ee oads, a he han he ac ual bounda ies o gang
e i o y, o p edic he loca ion o he bo de s. The esul s emain highly signi ican , demon-
s a ing ha hey a e no d i en by he po en ial endogenei y o some gang- e i o y bounda ies.
We also pe o m a placebo analysis in which we use majo mul ilane oads ha did no de ine
he bounda ies o gang e i o y o ensu e ha hese geog aphical ba ie s did no a ec socioe-
conomic de elopmen h ough ac o s un ela ed o he gang bounda ies. The analysis ocuses
on a se ies o consecu i e oads, anging om Redondel Mas e e in he wes o A enida Inde-
pendencia in he eas , ha spli San Sal ado in o wo simila -size pa s (see Figu e 1). We hen
es ima e whe he he le el o socioeconomic de elopmen changes a he placebo bounda ies.27
Appendix Table A.II p esen s he esul s, con i ming he no ion ha majo oads do no a ec
de elopmen ou comes h ough ac o s un ela ed o he gang bounda ies.
S abili y o he bounda ies o gang e i o y. A po en ial conce n is ha he bounda ies
o gang e i o y may no ha e emained s able be ween he ime hey we e o med (soon a e
he gangs eme ged) and 2015, when EDH published he map o gang e i o y. I he EDH map
does no accu a ely e lec which a eas we e con olled by he gangs in 2007, he es ima es
in Table Iwould be biased owa d ze o (i.e., agains inding an e ec ).28 Thus, he esul s in
Table Ishould be in e p e ed as he lowe bound o he e ec s o gang con ol.
26To ensu e compa abili y o he census ac s on bo h sides o he eg ession discon inui y h eshold, we exclude
25% o he la ges census ac s, which a e disp opo iona ely p esen ou side gang e i o y, and include dummies o
he h ee emaining qua iles o he census ac size dis ibu ion. Table S.II in he Supplemen a y Ma e ials epo s
he esul s o es ima ing he same eg ession speci ica ion wi hou excluding he la ges census ac s and, ins ead,
including dummies o all ou qua iles o he census ac size dis ibu ion.
27Speci ically, we es ima e he eg ession speci ica ion de ined below, whe e no h is a dummy a iable o a census
ac being o he no h o he placebo bounda ies. The coe icien o in e es is ω4, which es ima es he change in
socioeconomic condi ions a he placebo bounda ies. Simila ly o he o he eg ession speci ica ions, we limi he
sample o obse a ions wi hin 420 me e s o he (placebo) discon inui y h eshold. In addi ion, simila ly o Table III,
we exclude 25% o he la ges census ac s, which a e p edominan ly p esen ou side gang e i o y, and include
dummies o he h ee emaining qua iles o he census ac size dis ibu ion. Table S.III in he Supplemen a y
Ma e ials epo s he esul s o es ima ing he same eg ession speci ica ion wi hou excluding he la ges census
ac s, and ins ead, including dummies o all ou qua iles o he census ac size dis ibu ion.
yic =ω0+ω1dis ancec+ω2dis ancec→no hc+ω3gang e i o yc+ω4no hc+εic.(2)
28Fo ins ance, i , in eali y, he gangs con olled mo e neighbo hoods han sugges ed by he map, hen, unde he
assump ion ha he gangs ha e a homogeneous e ec on socioeconomic de elopmen in all he a eas hey con ol,
li ing condi ions in he con ol g oup would be unde es ima ed. In u n, he di e ence in li ing condi ions be ween
he gang and nongang a eas would also be unde es ima ed. Simila ly, i he gangs ac ually con olled ewe neighbo -
20
TABLE III
BOUNDARIES OF GANG TERRITORY FROM GEOGRAPHICAL BARRIERS
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.096 0.047 -0.064 -0.226 0.004 -0.287
(0.014)*** (0.009)*** (0.014)*** (0.056)*** (0.002)** (0.029)***
[0.021]*** [0.012]*** [0.022]*** [0.102]** [0.002]** [0.101]***
Mean o dep. a . 0.947 0.021 0.966 0.050 0.002 0.097
Obse a ions 7,424 6,312 6,348 6,348 6,348 6,056
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.021 -0.606 -0.385 -0.045 -0.556 -2.061
(0.009)** (0.052)*** (0.032)*** (0.014)*** (0.062)*** (0.321)***
[0.013]* [0.155]*** [0.053]*** [0.014]*** [0.118]*** [0.398]***
Mean o dep. a . 0.033 0.305 0.671 0.957 0.256 2.814
Obse a ions 6,021 6,080 6,098 6,119 6,086 6,348
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.167 -0.406 -0.233 -0.071 -0.246 -0.266
(0.039)*** (0.032)*** (0.054)*** (0.009)*** (0.025)*** (0.028)***
[0.038]*** [0.049]*** [0.107]** [0.014]*** [0.059]*** [0.051]***
Mean o dep. a . 0.926 0.406 0.146 0.964 0.335 0.486
Obse a ions 21,488 20,722 20,722 6,312 5,933 20,722
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1), using he loca ions o majo oads
and boule a ds (geog aphical ba ie s) as he p edic ed bounda ies o gang e i o y. To ensu e compa abili y o he census ac s on bo h
sides o he eg ession discon inui y h eshold, we exclude 25% o he la ges census ac s, which a e disp opo iona ely p esen ou side gang
e i o y. We also include dummies o he h ee emaining qua iles o he census ac size dis ibu ion. Table S.II in he Supplemen a y
Ma e ials epo s he esul s o es ima ing he same eg ession speci ica ion wi hou excluding he la ges census ac s and, ins ead, including
dummies o all ou qua iles o he census ac size dis ibu ion. All he a iables come om he 2007 census. The uni o obse a ion is a
dwelling, household, o indi idual, depending on which cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample
consis s o he en i e popula ion. Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions
on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang
e i o y (sepa a ely o each side o he bounda ies). S anda d e o s in b acke s a e adjus ed o allow o spa ial co ela ion wi hin a 100 me e
adius (Conley co ec ion).
Ne e heless, in Appendix Subsec ion A.1, we demons a e ha he gang- e i o y bound-
a ies ha e emained s able since hey we e i s o med. Speci ically, we exploi he ac ha
mos gang- ela ed homicides ake place p ecisely a he bounda ies o gang e i o y because
o people a emp ing o en e o lea e gang-con olled neighbo hoods wi hou pe mission.29 As
a esul , by showing ha , h oughou he yea s, gang- ela ed homicides consis en ly ake place
igh a he bounda ies om he EDH map, we a e able o con i m he alidi y o ha map and
demons a e he s abili y o hose bounda ies.
In addi ion, in 2023, we conduc ed a new su ey o indi iduals om gang and nongang
neighbo hoods, in which, among o he ques ions, he esponden s we e asked whe he hei
hoods han sugges ed by he map, hen li ing condi ions in he ea men g oup would be o e es ima ed, which would
also lead o a smalle di e ence be ween he ea men and con ol g oups.
29This phenomenon has also been documen ed o he 1970s h ough he 1990s in gang neighbo hoods in Los
Angeles, whe e mos o he iolence ook place igh a he en ance o hese neighbo hoods (A sy,2018).

GANGS, LABOR MOBILITY, AND DEVELOPMENT 21
neighbo hood had been con olled by gangs 20 yea s ago, du ing he p esidency o F ancisco
Flo es Pé ez (P esiden o El Sal ado in 1999–2004). As shown in Appendix Figu e A.6, he
sha e o esponden s answe ing in he a i ma i e signi ican ly inc eases a he bounda ies o
gang e i o y, sugges ing ha he bo de s ha e emained s able o e ime.
Selec i e mig a ion: in-sample mig a ion. Ano he assump ion ha needs o be sa is ied
o ou es ima es o be in e p e ed as causal is ha he e has been no selec i e mig a ion o
indi iduals ac oss he eg ession discon inui y h eshold. Selec i e mig a ion can a ec ou
esul s in wo ways. The i s is wha we call in-sample mig a ion: indi iduals mo ing om
a neighbo hood on one side o he bounda ies o an a ea on he o he side o he bounda ies
while emaining in San Sal ado and, hus, in ou sample. This ype o mig a ion would be
a di ec h ea o iden i ica ion because i would imply ha indi iduals can manipula e hei
ea men s a us. The second is wha we call ou -o -sample mig a ion: indi iduals mo ing om
San Sal ado o a di e en municipali y in El Sal ado o ab oad. This ype o mig a ion does
no in alida e he iden i ica ion s a egy, bu i changes he in e p e a ion o he mechanism
h ough which he gangs a ec local socioeconomic condi ions (i.e., ha gang con ol makes
weal hy, educa ed indi iduals lea e San Sal ado ).
In his subsec ion, we conside he di ec h ea o iden i ica ion ha comes om in-sample
mig a ion. To show ha in-sample mig a ion is no d i ing ou indings, we le e age ou 2019
su ey, whe e, among o he ques ions, we asked indi iduals whe he hey had li ed in he exac
same place hei en i e li e: 77% o esponden s said hey had. This in o ma ion allows us o
compa e he esul s o he ull sample and o he subsample o esponden s o whom we
know he ex an e ea men s a us (i.e., ha hey li ed in he loca ion be o e he a i al o he
gangs). In he absence o in-sample mig a ion, he wo se s o esul s would be qui e simila ,
whe eas, i he esul s a e de e mined by in-sample mig a ion, he discon inui ies would appea
only in he ull sample.
No ably, his exe cise also allows us o de e mine ha he esul s a e no d i en by weal hy
and educa ed newcome s choosing o se le in nongang pa s o San Sal ado . By es ic ing he
sample o indi iduals who ha e li ed in he same neighbo hood hei en i e li e, by de ini ion,
we exclude all newcome s.
When we limi he sample his way, he esul s o he eg ession discon inui y analysis a e
p ac ically unchanged. Appendix Figu e A.4 illus a es his ac by showing he wo eg ession
discon inui y plo s o household income. The le -hand side o he igu e p esen s he esul s
o he ull sample; he igh -hand side p esen s he subsample o ne e -mo e s. The wo plo s
a e qui e simila , sugges ing ha he esul s a e no d i en by selec i e in-sample mig a ion.
Table A.I in he Appendix epo s he eg ession es ima es o he socioeconomic cha ac e is ics
om ou 2019 su ey, bo h o he ull sample and o he sample o ne e -mo e s; Figu e S.5
in he Supplemen a y Ma e ials illus a es hese esul s.30
30In he 2007 census, indi iduals we e also asked whe he hey had li ed in he same municipali y hei en i e li e.
Since indi iduals who answe ed in he a i ma i e could s ill ha e mo ed wi hin he municipali y, his ques ion is
22
Fo a de ailed discussion o ou -o -sample mig a ion (i.e., indi iduals mo ing om San Sal-
ado o a di e en municipali y o ab oad), see Appendix Subsec ion A.2.
Absence o p e ends. In Sec ion 6, we also demons a e he absence o p e ends in socioe-
conomic de elopmen be ween a eas wi h and wi hou gang p esence. Speci ically, we pe o m
a di e ence-in-di e ences analysis using nigh ime ligh densi y, household income, and i m
openings o show ha hese a iables only s a ed o change a e he depo a ion o he gang
leade s om he Uni ed S a es o El Sal ado .
O he obus ness checks. In Appendix Subsec ion A.3, we also p esen a wide ange o
addi ional obus ness checks o ensu e ha he es ima es in Table I ep esen he causal e ec
o gang con ol on socioeconomic de elopmen .
5. MECHANISMS
In his sec ion, we explo e he mechanisms behind he nega i e e ec s o gangs’ e i o-
ial con ol on de elopmen ou comes. In pa icula , we p o ide no el e idence on how gang-
imposed mobili y es ic ions a ec indi iduals’ labo -ma ke choices by p e en ing hem om
commu ing o a eas ou side o gang e i o y, whe e he la ges and bes -paying i ms a e lo-
ca ed. We also show ha he di e ences in educa ional a ainmen be ween gang and nongang
a eas can be explained by highe d opou a es in gang-con olled neighbo hoods. Finally, we
in es iga e al e na i e mechanisms and ind ha he eg ession discon inui y esul s canno be
explained by di e ences in c ime (i.e., homicides, ex o ion) o he composi ion o i ms a he
bounda ies o gang e i o y.
In Appendix Subsec ion A.2, we show ha ou esul s a e no d i en by selec i e mig a ion o
indi iduals ou o gang e i o y. Speci ically, we es ima e he a es o selec i e ou -o -sample
mig a ion by conside ing he ela ionship be ween household weal h and he p obabili y o
a amily membe mig a ing ab oad om 1997 h ough 2007, inding ha selec i e mig a ion
accoun s o no mo e han 14% o he gaps in socioeconomic de elopmen be ween gang and
nongang a eas.31 In Appendix Subsec ions A.4 and A.5, we also demons a e ha he eg ession
less p ecise a de e mining he ex an e ea men s a us o he esponden s. Coinciden ally, howe e , he 2007 su ey
ound ha he sha e o popula ion ha had always li ed in San Sal ado municipali y was 77%, he same pe cen age
as he sha e o popula ion ha had always li ed in he same loca ion acco ding o he 2019 su ey. Thus, i appea s
ha , in his con ex , indi iduals p ima ily mo e ac oss municipali ies and no wi hin he same municipali y. Unde
his assump ion, we es ima e Speci ica ion (1) o he a iables om he 2007 census o he subsample o indi iduals
who had always li ed in he same municipali y. Table S.IV in he Supplemen a y Ma e ials p esen s he esul s,
which a e e y simila o hose p esen ed in Table I, con i ming ha in-sample mig a ion is no likely o be d i ing
he esul s. In addi ion, o he wo main ou come a iables ha a e po en ially a ec ed by in-sample mig a ion (i.e.,
he i s p incipal componen s o household and indi idual cha ac e is ics), we pe o m a es in he spi i o Lee
(2009). Focusing on indi iduals li ing wi hin 100 me e s o he gang bounda ies, we calcula e he bounds o he
ea men e ec s unde he e y s ong assump ion ha all indi iduals who ha e e e changed hei municipali y
o esidence did so in a way ha biases ou es ima es. E en unde his s ong assump ion, we ind ha he e ec s o
gang con ol a e bounded be ween -0.127 and -0.07 o household cha ac e is ics and -0.124 and -0.048 o indi idual
cha ac e is ics. In he 2019 su ey, a simila exe cise bounds he e ec s on household income be ween -497.7 and
-299.1. All he bounds a e s a is ically di e en om ze o.
31In Panel B o Appendix Table A.XIV, we also show ha ou esul s a e no de e mined by posi i e selec i e
in e nal mig a ion wi hin El Sal ado . Speci ically, we ind ha indi iduals who ha e p e iously li ed in San Sal ado
GANGS, LABOR MOBILITY, AND DEVELOPMENT 23
discon inui y esul s canno be explained by di e ences in public goods p o ision o occupa-
ional s uc u e (i.e., unemploymen , in o mal employmen , o hou s wo ked), espec i ely.32
5.1. Res ic ions on Mobili y
The p esence o mobili y es ic ions. To documen he p esence o es ic ions on indi-
iduals’ mobili y, we es ima e Speci ica ion (1) o mobili y ques ions om h ee di e en
sou ces: he 2019 su ey (columns 1–5), a ollow-up su ey ha we conduc ed in 2023 (col-
umn 6), and cell phone ping da a om ea ly 2022 (columns 7–8). Table IV p esen s he esul s.
The es ima es in column 1 sugges ha he sha e o popula ion wo king in gang-con olled
neighbo hoods d ama ically inc eases by almos 50 pe cen age poin s ( om 5.7% o 55.2%)
a he bounda ies o gang e i o y. Residen s o gang e i o y a e also mo e likely o wo k
in he same neighbo hood whe e hey li e and a e less likely o ha e a eled ou side o San
Sal ado : he sha e o indi iduals who ha e e e been o he beach o isi ed San a Ana depa -
men , which a e bo h 30 o 60 kilome e s away, discon inuously dec eases a he bounda ies o
gang e i o y.
To u he demons a e he salience o es ic ions on indi iduals’ mobili y, we show ha es-
iden s o gang a eas acknowledge he p esence o hese es ic ions. Fi s , in 2019, hey we e
signi ican ly less likely o say ha he e is eedom o mo emen in he neighbo hood whe e
hey li e (column 5 o Table IV). Second, in 2023, we conduc ed a new su ey in San Sal-
ado , in which he esponden s we e asked whe he , in he pas , he gangs’ “in isible bo de s”
p e en ed hem om inding jobs in la ge i ms in o he pa s o he ci y. While indi iduals ou -
side o gang e i o y we e also a ec ed (e.g., due o gang a eas blocking he ou es be ween
some nongang pa s o he ci y), he impac was signi ican ly s onge o he esiden s o gang
neighbo hoods (column 6 o Table IV).
Finally, in columns 7 and 8 o Table IV, we use cell phone ping da a om ea ly 2022 o
illus a e ha esiden s o gang e i o y a e no gene ally less mobile han indi iduals li ing in
o he pa s o he ci y, bu ha hei mo emen s a e con ined o gang-con olled a eas. We begin
wi h di iding he map o San Sal ado in o 100→100-me e g id cells and using he p e alence
o pings du ing he nigh hou s ( om 9 p.m. o 7 a.m.) o iden i y he g id cell whe e an
indi idual li es. Then, o each indi idual, we calcula e he sha e o pings inside gang e i o y
du ing he day ime ( om 9 a.m. o 7 p.m.), excluding pings in hei home g id cells.33 Simila ly,
we calcula e he a e age dis ance ha an indi idual a els away om hei home du ing he
day ime. Columns 7 and 8 o Table IV p esen he esul s o es ima ing Speci ica ion (1) o
bu now eside in a di e en municipali y ha e simila socioeconomic cha ac e is ics o indi iduals li ing in gang-
con olled neighbo hoods o San Sal ado (see columns 5 and 7 o Appendix Table A.XIV).
32No ably, he e is no discon inui y in he p obabili y o being employed. The esul s o es ima ing Speci ica ion (1)
sugges ha esiden s o gang e i o y aged be ween 18 and 65 a e only 0.2 pe cen age poin s less likely o be
employed han indi iduals om nongang a eas (wi h he s anda d e o o 1.1 pe cen age poin s).
33Home g id cells a e in en ionally excluded o ensu e ha he esul s a e no d i en by indi iduals spending ime
a home. I home g id cells a e no excluded, he es ima e in column 7 o Table IV becomes h ee imes la ge .
24
hese ou come a iables. They con i m ha esiden s o gang-con olled neighbo hoods spend
a subs an ially la ge sha e o hei ime in gang e i o y han indi iduals on he o he side o
he eg ession discon inui y cu o . Howe e , wi hin he a eas o which hey a e con ined, bo h
g oups o indi iduals a el he same dis ance h oughou he day, sugges ing ha esiden s o
gang neighbo hoods do no ha e a lowe capaci y o a el away om home.
Figu e 3p esen s he eg ession discon inui y plo s o he ou mos impo an a iables in
Table IV: he sha e o people wo king in gang e i o y, he sha e o ime indi iduals spend in
gang e i o y, he sha e o people who hink he e is eedom o mo emen in he a ea whe e
hey li e, and he sha e o people who say ha gang-imposed es ic ions on mobili y p e en ed
hem om inding jobs in la ge i ms in o he pa s o he ci y.
TABLE IV
GANG CONTROL AND RESTRICTIONS ON INDIVIDUALS’MOBILITY
Wo ks in Wo ks in neighbo hood Has been o Has been o
gang e i o y whe e hey li e San a Ana he beach
(1) (2) (3) (4)
Gang e i o y 0.495 0.111 -0.277 -0.064
(0.039)*** (0.031)*** (0.043)*** (0.031)**
[0.042]*** [0.050]** [0.052]*** [0.032]**
Mean o dep. a . 0.334 0.302 0.495 0.872
Obse a ions 1,738 2,071 2,314 2,314
F eedom o Gang bo de s p e en ed you Sha e o ime spen in Mean dis ance away
mo emen om ge ing jobs in la ge i ms gang a eas, excluding om home du ing
whe e hey li e in o he pa s o he ci y ime spen a home he day (in me e s)
(5) (6) (7) (8)
Gang e i o y -0.097 0.100 0.213 -52.82
(0.039)** (0.041)** (0.039)*** (154.82)
[0.039]** [0.046]** [0.034]*** [147.55]
Mean o dep. a . 0.811 0.407 0.222 1955.62
Obse a ions 2,314 2,313 9,605 9,605
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he ques ions ela ed o mobili y.
The ou come a iables in columns 1–5 come om he 2019 su ey, he ou come a iable in column 6— om he 2023 su ey, and he ou come
a iables in columns 7–8 a e based on cell phone ping da a om SDK. San a Ana is a neighbo ing municipali y, which is app oxima ely
60 kilome e s away om San Sal ado . The beach is app oxima ely 30 kilome e s away om San Sal ado . The uni o obse a ion is an
indi idual. Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he
bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely
o each side o he bounda ies). S anda d e o s in b acke s a e adjus ed o spa ial co ela ion wi hin a 100 me e adius (Conley co ec ion).
We addi ionally in es iga e whe he he esul s in Table IV can be de e mined by esiden s
o gang a eas ha ing less access o pe sonal anspo a ion, such as ca s and mo o cycles. To
add ess his ques ion, we conside he he e ogenei y o he e ec s o gang con ol on people
who ge o wo k by ca o mo o cycle and people who ge o wo k in ano he way, con olling
GANGS, LABOR MOBILITY, AND DEVELOPMENT 31
hey should be in e p e ed o deno e he e ec s o gangs’ e i o ial con ol and accompanying
es ic ions on mobili y.
In Sec ion 6, we discuss his la e implica ion in much mo e de ail and quan i y he e ec s
o exposu e o o e all gang p esence, no jus gang e i o ial con ol.
6. AGGREGATE EFFECTS OF GANG PRESENCE
As we p e iewed a he end o Subsec ion 5.3, he eg ession discon inui y esul s la gely
ep esen he socioeconomic cos s o ull e i o ial con ol by he gangs and no necessa ily
o he o ms o gang ac i i y. In his sec ion, we use da a o all o El Sal ado o analyze he
b oade consequences o gang p esence on economic ac i i y in he coun y.
6.1. Theo e ical F amewo k
To cla i y he dis inc ion be ween gang p esence and gang e i o ial con ol, we begin wi h
p esen ing a simple concep ual amewo k o he in e p e a ion o ou esul s. Appendix Sub-
sec ion A.6 p o ides a mo e de ailed discussion o he model behind his concep ual ame-
wo k, as well as an analysis o wo coun e ac ual scena ios: he emo al o es ic ions on
indi iduals’ mobili y and he ull emo al o gang p esence.
We conside a one-dimensional ci y on a uni in e al, whe e loca ions a e cha ac e ized by
hei p oximi y o he gangs. O e all, he ci y is di ided in o h ee quali a i ely di e en a eas.
Places in [0,b]a e ully con olled by he gangs, and indi iduals li ing he e canno wo k in
o he pa s o he ci y. These loca ions a e he equi alen o gang e i o y in he eg ession
discon inui y design. Places in [b, b +ϑ](ϑ>0) a e no con olled by he gangs. Indi iduals
li ing he e a e ee o wo k in any nongang pa o he ci y, bu i ms in [b, b +ϑ]a e s ill
exposed o ex o ion and o he gang- ela ed ac i i ies. Toge he , places in [0,b+ϑ]comp ise
wha we e e o as a eas wi h gang p esence. Finally, places in [b+ϑ,1] do no ha e any gang
p esence.
The di e ences in labo -ma ke condi ions be ween hese h ee a eas a e de e mined by he
p oduc ion echnology used by i ms in ha loca ion. All i ms can choose be ween wo op ions:
a simple echnology ha does no equi e any in es men om he i ms and a mo e p oduc i e
echnology ha equi es an ini ial in es men a a ixed cos . In he absence o gangs, i ms
bene i om he adop ion o he p oduc i e echnology. Howe e , p oduc i e i ms in a eas
wi h gang p esence ace a high isk o hei ou pu being ex o ed, which makes hem choose he
simple echnology ins ead. Thus, only i ms in a eas wi hou gang p esence choose o inc ease
hei p oduc i i y. In u n, as we show in Appendix Subsec ion A.6, unde a ealis ic se o
pa ame e s, his esul s in inc eased employmen and highe wages in hose i ms.
Despi e labo -ma ke condi ions only being be e in a eas wi hou gang p esence, indi iduals
li ing in [b, b +ϑ]a e s ill able o ake ad an age o hem because o hei abili y o commu e
o [b+ϑ,1], whe eas people li ing in [0,b]canno do so. This pa o he mechanism highligh s
he impo ance o es ic ions on mobili y o people li ing in gang e i o y (i.e., [0,b]). A he

32
same ime, es ic ions on mobili y only ma e due o highe economic g ow h in a eas wi hou
gang p esence: i i ms in all pa s o he ci y we e he same, he e would ha e been no need
o commu e o [b+ϑ,1]. Based on his conclusion, we now analyze whe he , a e he a i al
o he gangs, loca ions wi hou gang p esence indeed expe ienced mo e economic g ow h han
places exposed o gang ac i i y.
6.2. Di e ence-in-Di e ences: Empi ical S a egy
To analyze he agg ega e impac o gang ac i i y, we use da a om all o El Sal ado o
pe o m a di e ence-in-di e ences analysis, compa ing he e olu ion o economic ac i i y in
a eas wi h a ying le els o gang ac i i y a e 1996. Ou analysis exploi s wo sou ces o
a ia ion: he iming o gang membe s’ depo a ion om he Uni ed S a es, which led o he
eme gence o gangs in El Sal ado , and he geog aphic di e ences in exposu e o o ganized
c ime. Ou hypo hesis is ha p io o 1996, he yea o he i s wa e o depo a ions om he
Uni ed S a es, loca ions ha would la e expe ience di e en le els o gang ac i i y had simila
a es o economic de elopmen . Howe e , a e 1996, we expec o see highe a es o economic
g ow h in a eas wi h low le els o gang p esence.
We exploi he ac ha , a e being depo ed, many gang membe s who we e bo n in El Sal-
ado e u ned o hei municipali y o bi h (S ia schi,2022b). Thus, we use he municipali ies
o bi h o known gang leade s as a ea men a iable o whe he he municipali y became ex-
posed o gang ac i i y.43 We hen es ima e he ollowing e en s udy model (Speci ica ion 3) o
measu e he e ec o gang p esence on economic g ow h.
Econ. g ow hi, =gi+ϖ +!
j→=1995
ϱjgang p esencei→1{Yea =j} +εi, .(3)
Econ. g ow h ep esen s a ious measu es o economic g ow h in municipali y ia ime .
gang p esence is a dummy o whe he a gang leade was bo n in municipali y i;giand ϖ
ep esen municipali y and yea ixed e ec s, espec i ely. S anda d e o s a e es ima ed using
Conley s anda d e o s wi h spa ial co ela ion wi hin a 5 km adius. The coe icien s o in e es
a e ϱj, which ep esen he di e ences in economic g ow h be ween loca ions wi h and wi hou
gang p esence ela i e o 1995— he yea be o e he change in he Uni ed S a es immig a ion
policy.
We use h ee ou come a iables o measu e municipali y-le el g ow h in economic ac i i y.
The i s one is he opening o new business es ablishmen s. Speci ically, we use da a om
43We e i y he ele ance o his ea men a iable by showing ha gang leade s’ bi h municipali ies we e 80
pe cen age poin s mo e likely o expe ience a gang- ela ed homicide in 2003–2004, he i s yea s o which geocoded
homicide da a a e a ailable.
GANGS, LABOR MOBILITY, AND DEVELOPMENT 33
he 2005 economic census, which includes in o ma ion on when he i ms we e opened.44 The
second ou come a iable is nigh ime ligh densi y (o luminosi y) which ecen s udies ha e
ound o be a good p oxy o local-le el economic ac i i y (Chen and No dhaus,2011,Hen-
de son e al.,2012). Finally, we use da a on household income, which is based on annual
household su eys conduc ed in 1992–2007 by DIGESTYC. In all h ee cases, he ou comes
a e measu ed in pe cen age poin s, no malized o be 100 pe cen in 1995–1996, bo h in a eas
wi h and wi hou gang p esence. In addi ion, gi en ha he gangs we e p ima ily a ac ed o
la ge ci ies, o a oid he compa ison be ween u ban and u al loca ions, we limi ou analysis
o u ban municipali ies.
6.3. Di e ence-in-Di e ences: Resul s
Figu e 4p esen s he esul s o es ima ing Speci ica ion (3) o he h ee ou come a iables.45
I shows ha , be o e he 1996 change in Uni ed S a es immig a ion policy, a eas wi h and wi h-
ou u u e gang p esence expe ienced simila g ow h in economic ac i i y. Howe e , a e he
a i al o he gangs in 1996–1997, municipali ies wi h gang p esence expe ienced signi ican ly
lowe economic g ow h.
FIGURE 4.—Gang P esence and Economic Ac i i y
No e: The igu e p esen s an e en s udy g aph o he di e ences in economic g ow h be ween municipali ies wi h and wi hou gang
p esence. Fo all h ee ou come a iables, he da a a e in pe cen age poin s, no malized o be equal o 100 pe cen in 1995–1996, be o e he
change in he Uni ed S a es immig a ion policy.
The magni ude o he e ec s is subs an ial. Fo example, by 2005, municipali ies wi hou
gang p esence had expe ienced a 105 pe cen age poin highe a e o new business openings.
Addi ionally, a e 1997, on a e age, hese a eas had an 82 pe cen age poin highe g ow h
in nigh ime ligh densi y and a 28.5 pe cen age poin highe g ow h in household income.46
44Gi en ha he economic census was conduc ed in 2005, o be included in ou sample, he i m mus no ha e
closed un il ha ime. Thus, ou esul s should be in e p e ed as illus a ing an e ec on he opening o mode a ely
success ul business es ablishmen s.
45The co esponding eg ession coe icien s a e epo ed in Table A.XII in he Appendix.
46Acco ding o Hende son e al. (2012), a one pe cen age poin change in luminosi y co esponds o a 0.28 pe cen -
age poin change in GDP. Thus, be ween 1998 and 2013, on a e age, a eas wi hou gang p esence had app oxima ely
34
O e all, hese esul s con i m he no ion ha , a e he a i al o he gangs, mos economic
g ow h has aken place in a eas a away om gang e i o y, plausibly due o business owne s’
desi e o a oid ex o ion and o he o ms o gang ac i i y.
We also complemen he di e ence-in-di e ences esul s by using household income da a
om ou 2019 su ey in San Sal ado and pe o ming a back-o - he-en elope calcula ion ha
compa es loca ions wi hou gang p esence sepa a ely o ully gang-con olled neighbo hoods
and places wi h only some gang ac i i ies.47 We ind ha , a e 1997, a eas wi h no gang p es-
ence expe ienced app oxima ely 50 pe cen age poin s highe g ow h in household income han
he o me and app oxima ely 9 pe cen age poin s highe g ow h han he la e . Thus, while
p oximi y o places wi h he highes g ow h o employmen oppo uni ies posi i ely a ec ed
indi iduals’ ea nings, i was esiden s o gang-con olled neighbo hoods who we e pa icula ly
nega i ely a ec ed due o hei inabili y o commu e ac oss he bounda ies o gang e i o y.
We also analyze whe he he e ec s o gang p esence a e di e en in he la ges u ban cen-
e s (e.g., San Sal ado ) and in he es o he coun y. To add ess his ques ion, we ollow
A khangelsky e al. (2021) and implemen wo ypes o syn he ic di e ence-in-di e ences anal-
yses. The i s one de ines he ea men a iable in he same way as in he baseline di e ence-
in-di e ences es ima ion. The second one na ows he ea men g oup o he ou la ges ci ies
in El Sal ado , all o which ha e had a subs an ial gang p esence since he la e 1990s: San Sal-
ado , Soyapango, San a Ana, and San Miguel. Appendix Figu e A.9 p esen s he wo se s o
esul s. In gene al, we ind he wo speci ica ions o be qui e simila , sugges ing ha he la ges
ci ies we e no di e en ially a ec ed compa ed o o he places wi h gang p esence.
7. CONCLUDING REMARKS
O e all, he esul s p esen ed in his pape indica e ha , ia a combina ion o es ic ions
on indi iduals’ mobili y and displacemen o economic ac i i y, nons a e a med ac o s can
ha e a conside able nega i e impac on socioeconomic de elopmen . These indings ha e b oad
policy implica ions, shedding ligh on he long- e m consequences o depo ing indi iduals
wi h c iminal eco ds o a coun y wi h low s a e capaci y, sugges ing ha imp o emen s in
s a e capaci y can signi ican ly imp o e economic g ow h, and highligh ing he impo ance o
eedom o mo emen o socioeconomic de elopmen .
82→0.28 = 23 pe cen age poin s highe GDP g ow h compa ed o a eas wi h gang ac i i y. This es ima e closely
aligns wi h he one o household income.
47Acco ding o he 2007 census, 60% o San Sal ado ’s popula ion li ed in gang-con olled neighbo hoods. Thus,
he a e age household income in San Sal ado is equal o A=0.4GC +0.6GP , whe e GC and GP ep esen
household income om he 2019 su ey in ully gang-con olled pa s o he ci y and places wi h only some gang
p esence, espec i ely. Then, wi hou gang p esence, San Sal ado would ha e been expec ed o ha e an a e age
household income o 1.285A, which, in u n, allows o a compa ison wi h GC and GP .
GANGS, LABOR MOBILITY, AND DEVELOPMENT 35
REFERENCES
ABREGO,LEISY,MAT COLEMAN,DANIEL MARTÍNEZ,CECILIA MENJÍVAR,AND JEREMY SLACK (2017): “Mak-
ing immig an s in o c iminals: Legal p ocesses o c iminaliza ion in he pos -IIRIRA e a,” Jou nal on Mig a ion
and Human Secu i y, 5, 694–715. [7]
ACEMOGLU,DARON,GIUSEPPE DEFEO,AND GIACOMO DAVIDE DELUCA (2019): “Weak S a es: Causes and
Consequences o he Sicilian Ma ia,” Re iew o Economic S udies, 87, 537–581. [5]
ACEMOGLU,DARON,SIMON JOHNSON,AND JAMES A. ROBINSON (2001): “The Colonial O igins o Compa a i e
De elopmen : An Empi ical In es iga ion,” Ame ican Economic Re iew, 91, 1369–1401. [2,6]
——— (2002): “Re e sal o Fo une: Geog aphy and Ins i u ions in he Making o he Mode n Wo ld Income Dis i-
bu ion,” Qua e ly Jou nal o Economics, 117, 1231–1294. [6]
ACEMOGLU,DARON,JAMES A. ROBINSON,AND RAFAEL J. SANTOS (2013): “The Monopoly o Violence: E i-
dence om Colombia,” Jou nal o he Eu opean Economic Associa ion, 11, 5–44. [5]
ALESINA,ALBERTO,SALVATORE PICCOLO,AND PAOLO PINOTTI (2019): “O ganized c ime, iolence, and poli-
ics,” Re iew o Economic S udies, 86, 457–499. [5]
ALSAWADY,YOSEPH A., M. KABIR HASSAN,AND ARJA H. TURUNEN-RED (2022): “Limi ing The Mobili y o
Mig an Wo ke s in The Saudi Labo Ma ke ,” In e na ional Jou nal o Economics and Managemen , 16, 45–64.
[6]
ARKHANGELSKY,DMITRY,SUSAN ATHEY,DAVID A. HIRSHBERG,GUIDO W. IMBENS,AND STEFAN WAGER
(2021): “Syn he ic Di e ence-in-Di e ences,” Ame ican Economic Re iew, 111, 4088–4118. [34]
ARTSY,AVISHAY (2018): “Gang bo de s c ea e in isible walls in Los Angeles,” Tech. ep., KCWR,
h ps://www.kc w.com/cul u e/shows/design-and-a chi ec u e/gang-bo de s-c ea e-in isible-walls-in-los-angeles.
[7,20]
ASHER,SAM AND PAUL NOVOSAD (2020): “Ru al Roads and Local Economic De elopmen ,” Ame ican Economic
Re iew, 110, 797–823. [6]
BANDIERA,ORIANA (2003): “Land Re o m, he Ma ke o P o ec ion, and he O igins o he Sicilian Ma ia: Theo y
and E idence,” Jou nal o Law, Economics, and O ganiza ion, 19, 218–244. [5]
BATES,ROBERT,AVNER GREIF,AND SMITA SINGH (2002): “O ganizing iolence,” Jou nal o Con lic Resolu ion,
46, 599–628. [2]
BLATTMAN,CHRISTOPHER,GUSTAVO DUNCAN,BENJAMIN LESSING,AND SANTIAGO TOBON (2022): “Gang
ule: Unde s anding and Coun e ing C iminal Go e nance,” NBER Wo king Pape 28458.[2,4,5]
BLATTMAN,CHRISTOPHER AND EDWARD MIGUEL (2010): “Ci il wa ,” Jou nal o Economic li e a u e, 48, 3–57.
[5]
BOERMAN,THOMAS AND ADAM GOLOB (2020): “Gangs and mode n-day sla e y in El Sal ado , Hondu as and
Gua emala: A non- adi ional model o human a icking,” Jou nal o Human T a icking, 1–17. [9,10]
BRETTELL,CAROLINE B (2011): “Expe iencing e e yday disc imina ion: A compa ison ac oss i e immig an pop-
ula ions,” Race and Social P oblems, 3, 266–279. [6]
BROWN,ZACH Y., EDUARDO MONTERO,CARLOS SCHMIDT-PADILLA,AND MARÍA MICAELA SVIATSCHI
(2020): “Ma ke S uc u e and Ex o ion:E idence om 50,000 Ex o ion Paymen s,” NBER Wo king Pape .[73]
BUONANNO,PAOLO,RUBEN DURANTE,GIOVANNI PRAROLO,AND PAOLO VANIN (2015): “Poo Ins i u ions,
Rich Mines: Resou ce Cu se in he O igins o he Sicilian Ma ia,” Economic Jou nal, 125, F175–F202. [5]
BUONANNO,PAOLO,GIOVANNI PRAROLO,AND PAOLO VANIN (2016): “O ganized c ime and elec o al ou comes.
E idence om Sicily a he u n o he XXI cen u y,” Eu opean Jou nal o Poli ical Economy, 41, 61–74. [5]
CALÌ,MASSIMILIANO AND SAMI H. MIAARI (2018): “The Labo Ma ke Impac o Mobili y Res ic ions: E i-
dence om he Wes Bank,” Labou Economics, 51, 136–151. [6]
CARVALHO,LEANDRO S. AND RODRIGO R. SOARES (2016): “Li ing on he edge: You h en y, ca ee and exi in
d ug-selling gangs,” Jou nal o Economic Beha io & O ganiza ion, 121, 77–98. [5]
CHACÓN,JENNIFER M. (2009): “Managing mig a ion h ough c ime,” Columbia Law Re iew, 109, 135–148. [7]
CHEN,XIANDWILLIAM D. NORDHAUS (2011): “Using luminosi y da a as a p oxy o economic s a is ics,” PNAS,
108, 8589–8594. [33]
CLEMONS,MICHAEL,CLAUDIO MONTENEGRO,AND LANT PRITCHETT (2008): “The g ea disc imina ion: bo -
de s as a labo ma ke ba ie ,” Mimeo.[6]
CÓRDOVA,ABBY (2019): “Li ing in Gang-Con olled Neighbo hoods: Impac s on Elec o al and Nonelec o al Pa -
icipa ion in El Sal ado ,” La in Ame ican Resea ch Re iew, 54, 201–221. [4,11,44]
COUGHLIN,BRENDA CAND SUDHIR ALLADI VENKATESH (2003): “The u ban s ee gang a e 1970,” Annual
Re iew o Sociology, 41–64. [7]
CRUZ,JOSÉ MIGUEL (2010): “Cen al Ame ican ma as: F om you h s ee gangs o ansna ional p o ec ion acke s,”
Global C ime, 11, 379–398. [7,9]
CRUZ,JOSÉ MIGUEL AND NELSON PORTILLO PEÑA (1998): Solida idad y iolencia en las pandillas del g an San
Sal ado , UCA Edi o es. [8]
36
DANIELE,GIANMARCO AND GEMMA DIPOPPA (2017): “Ma ia, elec ions and iolence agains poli icians,” Jou nal
o Public Economics, 154, 10–33. [5]
DANIELE,GIANMARCO AND BENNY GEYS (2015): “O ganised C ime, Ins i u ions and Poli ical Quali y: Empi ical
E idence om I alian Municipali ies,” The Economic Jou nal, 125, F233–F255. [5]
DANIELE,VITTORIO AND UGO MARANI (2011): “O ganized c ime, he quali y o local ins i u ions and FDI in
I aly: A panel da a analysis,” Eu opean Jou nal o Poli ical Economy, 27, 132–142. [5]
DEFEO,GIUSEPPE AND GIACOMO DAVIDE DELUCA (2017): “Ma ia in he Ballo Box,” Ame ican Economic
Jou nal: Economic Policy, 9, 134–67. [5]
DELASIERRA,RAÚL (2020): “On he O igins o he S a e: S a iona y Bandi s and Taxa ion in Eas e n Congo,”
Jou nal o Poli ical Economy, 128, 32–74. [2,5]
DECESARE,DONNA (1998): “The child en o wa s ee gangs in el sal ado ,” NACLA Repo on he Ame icas, 32,
21–29. [8]
DELL,MELISSA (2010): “The Pe sis en E ec s o Pe u’s Mining Mi a,” Econome ica, 78, 1863–1903. [6]
——— (2015): “T a icking Ne wo ks and he Mexican D ug Wa ,” Ame ican Economic Re iew, 105, 1738–1779.
[5]
DELL,MELISSA,NATHAN LANE,AND PABLO QUERUBIN (2018): “The His o ical S a e, Local Collec i e Ac ion,
and Economic De elopmen in Vie nam,” Econome ica, 86, 2083–2121. [6]
DELL,MELISSA AND BENJAMIN OLKEN (2020): “The De elopmen E ec s o he Ex ac i e Colonial Economy:
The Du ch Cul i a ion Sys em in Ja a,” Re iew o Economic S udies, 87, 164–203. [6]
DONALDSON,DAVE (2018): “Rail oads o he Raj: Es ima ing he Impac o T anspo a ion In as uc u e,” Ame i-
can Economic Re iew, 108, 899–934. [6]
DUNN,WILLIAM CARL (2007): The Gangs o Los Angeles, iUni e se. [7]
FRYE,TIMOTHY AND EKATERINA ZHURAVSKAYA (2000): “Racke s, Regula ion, and he Rule o Law,” Jou nal o
Law, Economics, & O ganiza ion, 16, 478–502. [5]
GAMBETTA,DIEGO (1996): The Sicilian Ma ia: The Business o P i a e P o ec ion, Camb idge MA: Ha a d Uni-
e si y P ess. [5]
GANONG,PETER AND SIMON JÄGER (2018): “A Pe mu a ion Tes o he Reg ession Kink Design,” Jou nal o he
Ame ican S a is ical Associa ion, 113, 494–504. [14]
GLAESER,EDWARD AND HELEN SIMS (2015): “Con agion, c ime, and conges ion: o e coming he downsides o
densi y,” IGC G ow h B ie .[5]
HENDERSON,J.VERNON,ADAM STOREYGARD,AND DAVID N. WEIL (2012): “Measu ing Economic G ow h
om Ou e Space,” Ame ican Economic Re iew, 102, 994–1028. [33]
IBÁÑEZ,ANA MARÍA,JULIAN ARTEAGA,JUAN CAMILO CÁRDENAS CAMPO,ANA ARJONA,AND PATRICIA
JUSTINO (2019): “The E ec s o Wa ime Ins i u ions on Households’ Abili y o Cope wi h Shocks: E idence o
Colombia,” WIDER Wo king Pape №2019/84.[2,5,6]
INSIGHT CRIME AND CLALS (2018): “MS13 in he Ame icas. How he Wo ld’s Mos No o ious Gang De ies Logic,
Resis s Des uc ion,” Tech. ep., Insigh C ime and The Cen e o La in Ame ican & La ino S udies (CLALS). [10]
INTERNATIONAL CRISIS GROUP (2017): “El Sal ado ’s Poli ics o Pe pe ual Violence,” La in Ame ica Repo №64.
[9,11]
——— (2018): “Li e Unde Gang Rule in El Sal ado ,” La in Ame ica & Ca ibbean Commen a y.[3,9,10,11]
JONES,BENJAMIN FAND BENJAMIN AOLKEN (2005): “Do leade s ma e ? Na ional leade ship and g ow h since
Wo ld Wa II,” The Qua e ly Jou nal o Economics, 120, 835–864. [6]
KULISH,NICHOLAS (2018): “Wha I Cos s o Be Smuggled Ac oss he U.S. Bo de ,” The New Yo k Times.[40]
LEE,DAVID S. (2009): “T aining, Wages, and Sample Selec ion: Es ima ing Sha p Bounds on T ea men E ec s,”
The Re iew o Economic S udies, 76, 1071–1102. [22]
LESSING,BENJAMIN (2021): “Concep ualizing C iminal Go e nance,” Pe spec i es on Poli ics, 19, 854–873. [2,5]
LESSING,BENJAMIN AND GRAHAM DENYER WILLIS (2019): “Legi imacy in C iminal Go e nance: Managing a
D ug Empi e om Behind Ba s,” Ame ican Poli ical Science Re iew, 113, 584–606. [5]
LEVITT,STEVEN AND SUDHIR VENKATESH (2000): “An Economic Analysis o a D ug-Selling Gang’s Finances,”
Qua e ly Jou nal o Economics, 115, 755–789. [5]
LIBREROS,VICTORIA,JULIÁN ANTONIO,AND ROBERTO CARBAJAL (2010): El PAEBA de El Sal ado . La expe-
iencia y el apo e de la Uni e sidad F ancisco Ga idia, San Sal ado : UFG - Edi o es. [28]
LOWES,SARA AND EDUARDO MONTERO (2021): “Concessions, iolence, and indi ec ule: e idence om he
Congo F ee S a e,” The Qua e ly Jou nal o Economics, 136, 2047–2091. [6]
MAGALONI,BEATRIZ,EDGAR FRANCO VIVANCO,AND VANESSA MELO (2020a): “Killing in he Slums: Social
O de , C iminal Go e nance, and Police Violence in Rio de Janei o,” Ame ican Poli ical Science Re iew, 114,
552–572. [5,6]
MAGALONI,BEATRIZ,GUSTAVO ROBLES,AILA MMATANOCK,ALBERTO DIAZ-CAYEROS,AND VIDAL
ROMERO (2020b): “Li ing in Fea : he Dynamics o ex o ion in Mexico’s D ug Wa ,” Compa a i e Poli ical
S udies, 53, 1124–1174. [4]

GANGS, LABOR MOBILITY, AND DEVELOPMENT 37
MARTÍNEZ,JUAN JOSÉ AND ÓSCAR MARTÍNEZ (2018): El niño de Hollywood, Penguin Random House G upo
Edi o ial. [7,8]
MARTÍNEZ,OSCAR (2016): “Li ing Wi hin he Bounda ies o El Sal ado ’s Gang ‘Wa ’,” Tech. ep., Insigh C ime.
[10]
MARTÍNEZ,OSCAR,EFREN LEMUS,CARLOS MARTÍNEZ,AND DEBORAH SONTAG (2016): “Kille s on a
Shoes ing: Inside he Gangs o El Sal ado ,” The New Yo k Times.[4,11]
MCCRARY,JUSTIN (2008): “Manipula ion o he unning a iable in he eg ession discon inui y design: A densi y
es,” Jou nal o Econome ics, 142, 698–714. [40]
MCKENZIE,DAVID,STEVEN STILLMAN,AND JOHN GIBSON (2010): “How Impo an is Selec ion? Expe imen al
s. Non-Expe imen al Measu es o he Income Gains om Mig a ion,” Jou nal o he Eu opean Economic Associ-
a ion, 8, 913–945. [6]
MELNIKOV,NIKITA,CARLOS SCHMIDT-PADILLA,AND MARIA MICAELA SVIATSCHI (2024): “Gang C ack-
downs and De elopmen ,” Tech. ep., P ince on Uni e si y. [25,48]
MERGO,TEFERI (2016): “The E ec s o In e na ional Mig a ion on Mig an -Sou ce Households: E idence om
E hiopian Di e si y-Visa Lo e y Mig an s,” Wo ld De elopmen , 84, 69–81. [6]
MICHALOPOULOS,STELIOS AND ELIAS PAPAIOANNOU (2013): “P e-Colonial E hnic Ins i u ions and Con empo-
a y A ican De elopmen ,” Econome ica, 81, 113–152. [2,6]
MIGUEL CRUZ,JOSÉ (2010): “Cen al Ame ican ma as: om you h s ee gangs o ansna ional p o ec ion acke s,”
Global C ime, 11, 379–398. [8]
MIRENDA,LITTERIO,SAURO MOCETTI,AND LUCIA RIZZICA (2022): “The Economic E ec s o Ma ia: Fi m
Le el E idence,” Ame ican Economic Re iew, o hcoming. [5]
MURPHY,TOMMY E. AND MARTÍN A. ROSSI (2020): “Following he poppy ail: O igins and consequences o
Mexican d ug ca els,” Jou nal o De elopmen Economics, 143. [5]
NUÑO,LIDIA E. AND EDWARD R. MAGUIRE (2021): “The Na u e and S uc u e o MS-13 in Los Angeles Coun y,”
C iminal Jus ice Re iew, 07340168211029990. [2,7]
OLSON,MANCUR (1993): “Dic a o ship, Democ acy, and De elopmen ,” Ame ican Poli ical Science Re iew, 87,
567–576. [2,5]
PALMA,EDGARD (1999): “Pandillas y Delincuencia en El Sal ado ,” Asociacíon Sal ado eña de Socioloía.[10]
PINOTTI,PAOLO (2015): “The Causes and Consequences o O ganised C ime: P elimina y E idence Ac oss Coun-
ies,” Economic Jou nal, 125, F158–F174. [5]
SAVENIJE,WIM (2009): Ma as y ba as: pandillas y iolencia ju enil en los ba ios ma ginales de Cen oamé ica,
FLACSO—P og ama El Sal ado . [6]
SKARBEK,DAVID (2011): “Go e nance and P ison Gangs,” Ame ican Poli ical Science Re iew, 105, 702–716. [5]
STANLEY,WILLIAM DEANE (1987): “Economic mig an s o e ugees om iolence? A ime-se ies analysis o
Sal ado an mig a ion o he Uni ed S a es,” La in Ame ican Resea ch Re iew, 22, 132–154. [6]
SVIATSCHI,MARÍA MICAELA (2022a): “Making a Na co: Childhood Exposu e o Illegal Labo Ma ke s and C im-
inal Li e Pa hs,” Econome ica, o hcoming. [5,27]
——— (2022b): “Sp eading Gangs: Expo ing US C iminal Capi al o El Sal ado ,” Ame ican Economic Re iew,
112, 1985–2024. [5,7,8,27,32,41]
TENORIO,OSCAR (2002): “Tie a de odos, ie a de nadie,” Tech. ep., El Dia io de Hoy. [8]
TILLY,CHARLES (1985): “Wa Making and S a e Making as O ganized c ime,” in Collec i e Violence, Con en ious
Poli ics, and Social Change, New Yo k: Rou ledge, 121–139. [2]
TITA,GEORGE E., JACQUELINE COHEN,AND JOHN ENGBERG (2005): “An ecological s udy o he loca ion o
gang “se space”,” Social p oblems, 52, 272–299. [7]
VEGA,MARIO (2015): “El mapa de las pandillas,” Tech. ep., El Dia io de Hoy,
h ps://his o ico.elsal ado .com/his o ico/145152/el-mapa-de-las-pandillas.h ml. [8]
WALTHER,OLIVIER J., LAWALI DAMBO,MOUSTAPHA KONÉ,AND MICHIEL VAN EUPEN (2020): “Mapping
a el ime o assess accessibili y in Wes A ica: The ole o bo de s, checkpoin s and oad condi ions,” Jou nal o
T anspo Geog aphy, 82, 10–1016. [6]
WARD,THOMAS W (2013): Gangs e s wi hou bo de s: An e hnog aphy o a Sal ado an s ee gang, Ox o d: Ox-
o d Uni e si y P ess. [7,8,9]
WOMER,SARAH AND ROBERT J. BUNKER (2010): “Su eños gangs and Mexican ca el use o social ne wo king
si es,” Small Wa s & Insu gencies, 21, 81–94. [7]
ZOETHOUT,MARGRIET ANTOINETTE (2015): “Reco e ing Go e nmen Con ol o e Ma a Sal a ucha e i o y:
Analysis based on he ‘San a Tecla, a Municipali y F ee o Violence’ Ag eemen ,” Police and Public Secu i y
Jou nal, 5, 179–246. [8,11]
——— (2016): “Consolida e S a e con ol in ‘The Ma a Sal a ucha (Sal a ucha gang)’ e i o y: Lessons om
communi y wo k in high- isk communi ies,” Police and Public Secu i y Jou nal, 6, 115–204. [44]
38
ONLINE APPENDIX
A. ADDITIONAL INFORMATION AND ROBUSTNESS CHECKS
A.1. S abili y o he Bounda ies o Gang Te i o y
To he bes o ou knowledge, he bounda ies o gang e i o y ha e emained s able h ough-
ou he sample pe iod. In pa icula , we con ac ed he PNC, inqui ing abou his issue, and
mul iple PNC o icials con i med ha he bounda ies o gang e i o y had had no signi ican
changes since hey we e ini ially o med in he la e 1990s and ea ly 2000s. This in o ma ion
has also been con i med by in o mal con e sa ions wi h esiden s o San Sal ado .
To p o ide addi ional e idence ha he bounda ies o gang e i o y did no change in ime,
we ake ad an age o he ollowing ac . As desc ibed in Subsec ion 2.4, bo h MS-13 and 18 h
S ee conside ou side s a h ea o hei secu i y. Thus, a disp opo iona e numbe o gang-
ela ed homicides ake place a he bounda ies o gang e i o y (bo h be ween he gangs and he
s a e and be ween he wo gangs) because o ou side s a emp ing o en e gang neighbo hoods
wi hou pe mission (Ma ínez, 2016). Le e aging his ac , we conside geocoded da a on all
gang- ela ed homicides ha we e commi ed in San Sal ado in 2003-2014 and spli i in o
wo subsamples: hose ha ook place in he i s six yea s o he sample pe iod (2003-2008)
and hose ha ook place in he la es six yea s o he sample pe iod (2009-2014). Fo each o
he homicides, we iden i y whe he i ook place in a gang loca ion and calcula e he dis ance
o he bounda ies o gang e i o y (ei he be ween he gang and he s a e o be ween he wo
gangs). Panel A o Appendix Figu e A.10 p esen s he numbe o gang- ela ed homicides ha
ook place in 2003-2008 by 10-me e bins on ei he side o he bounda ies o gang e i o y;
Panel B o Figu e A.10 p o ides a simila illus a ion o gang- ela ed homicides in 2009-2014.
In bo h cases, he numbe o homicides was pa icula ly high in a eas close o he bounda ies
o he gang neighbo hoods om he EDH map, con i ming ha he map co ec ly iden i ies
he bounda ies o gang e i o y in he wo pe iods.48 In u n, he ac ha he highes numbe
o gang- ela ed homicides ook place in he same loca ions bo h in 2003-2008 and 2009-2014
sugges s ha he bounda ies o gang e i o y ha e emained s able du ing his pe iod.
A.2. Selec i e Mig a ion: Ou -o -Sample Mig a ion
In Subsec ion 4.3, we demons a ed ha ou main esul s a e no d i en by selec i e in-
sample mig a ion: indi iduals mo ing o o om gang-con olled neighbo hoods, while e-
maining in San Sal ado municipali y. Ano he ype o selec i e mig a ion ha can po en ially
a ec he in e p e a ion o ou esul s is ou -o -sample mig a ion: indi iduals mo ing om San
48No ably, as shown in Figu e A.10, he e a e mul iple gang- ela ed homicides ou side o gang e i o y. We p o ide
a de ailed discussion o his ac in Sec ion 5. Also, as we show in Sec ion A.3, he esul s in Table Ia e obus o
excluding obse a ions om neighbo hoods close o he eg ession discon inui y cu o (see Table A.III). Thus, while
he loca ion o he gang- ela ed homicides allows us o alida e he bounda ies o gang e i o y om he EDH maps,
he esul s in Table Ia e no d i en by a eas wi h he highes numbe s o gang- ela ed homicides.
GANGS, LABOR MOBILITY, AND DEVELOPMENT 39
Sal ado o a di e en municipali y o ab oad. In pa icula , i ich, educa ed indi iduals who
ini ially li ed in gang-con olled neighbo hoods we e mo e likely o mo e ou o San Sal ado
han poo and uneduca ed indi iduals om he same a eas, i could imply ha he esul s in
Table Ia e pa ly de e mined by his change in he composi ion o he popula ion. We analyze
his mechanism in he ollowing ways.
Fi s , we calcula e he a es o selec i e ou -o -sample mig a ion om gang-con olled neigh-
bo hoods ha would be equi ed o gene a e he discon inui ies om Table I. Fo each o he
bina y household-le el cha ac e is ics, we de ine a household o be “ ich” i i has ha cha ac-
e is ic (e.g., a phone, a compu e ) and “poo ” i i does no . The only excep ion is he a iable
o no ha ing a ba h oom, which is de ined in he opposi e way. Simila ly, o each o he
indi idual-le el cha ac e is ics, we de ine an indi idual o be “educa ed” i hey ha e ha cha -
ac e is ic (e.g., a high school deg ee, a uni e si y deg ee) and “uneduca ed” i hey do no . We
make he conse a i e assump ion ha ou side o gang e i o y, he p obabili y o mo ing ou
o San Sal ado is he same o all indi iduals and ha in gang neighbo hoods, poo and uned-
uca ed indi iduals mig a e ou o sample wi h p obabili y ϱ.49 Then, o a gi en ϱ, we calcula e
he sha e o ich households and educa ed indi iduals om gang e i o y ha needed o mo e
ou o San Sal ado o gene a e he discon inui ies o each o he ou come a iables.
We use he example o he sha e o households wi h a compu e o show how hese a es we e
calcula ed. F om he eg ession ou pu , we ge he p edic ed sha e o households wi h a com-
pu e o obse a ions ze o me e s away om he bounda ies o gang e i o y, sepa a ely o
loca ions inside and ou side o gang e i o y. We deno e hose numbe s as Gand NG, espec-
i ely. We u he deno e he numbe o “ ich” households (i.e., hose ha ha e a compu e ) in
gang-con olled a eas be o e any mig a ion ook place as xand he sha e o “poo ” households
(i.e., hose ha do no ha e a compu e ) as 1↓x. Nex , we assume ha a ac ion ωo he
“ ich” households and a ac ion ϱo he“poo ” households mig a ed ou o sample. Thus, in
he da a, we obse e he ollowing ela ionship.
(1 ↓ω)x
(1 ↓ω)x+(1↓ϱ)(1 ↓x)=G.(4)
Then, assuming di e en alues o ϱ, we calcula e he alue o ω ha would make his ela-
ionship hold i , in he absence o mig a ion, he e would no ha e been any di e ence in he
ou come a iable be ween gang and nongang loca ions (i.e., x=NG).
Appendix Table A.XIII p esen s he esul s o hese calcula ions o ϱequal o 0%, 10%,
and 20%. E en i we un ealis ically assume ϱ=0% (i.e., ha poo and uneduca ed indi iduals
om gang a eas do no ha e a chance o mo e ou o San Sal ado ), on a e age, he a e o
ou -o -sample mig a ion o ich, educa ed indi iduals would ha e o be as high as 51.7% o
gene a e he discon inui ies om Table I. Fo highe alues o ϱ, his a e is e en highe .
49I ich, educa ed indi iduals om nongang a eas a e mo e likely o mig a e ou o sample, ha would make he
equi ed a es o selec i e ou -o -sample mig a ion om gang e i o y e en highe .
40
Can he a e o ou -o -sample mig a ion o ich indi iduals be ha high? To add ess his
ques ion, we ake ad an age o he ac ha , un il he mid-2010s, in e na ional mig a ion o
en i e amilies had been e y a e.50 In e na ional mig a ion is expensi e: e.g., he cos s o mi-
g a ing om El Sal ado o he Uni ed S a es— he mos popula des ina ion among Sal ado an
mig an s—a e app oxima ely $12,500 (Kulish,2018). In u n, he a e age mon hly household
income in San Sal ado is only $625. Thus, e en o send one amily membe ab oad, Sal ado-
an households ha e o sa e up o a long ime, and mig a ion o en i e amilies is inc edibly
a e. This ac allows us o es ima e he a e o ou -o -sample mig a ion by conside ing whe he
a household has a amily membe who mo ed ab oad in 1997-2007 ( he 2007 census con ains
his in o ma ion). In addi ion, by looking a he co ela ion be ween he p obabili y o a amily
membe mo ing ab oad and he i s p incipal componen o he household cha ac e is ics, we
a e able o es ima e he ex en o which indi iduals om ich households we e mo e likely o
mig a e ou o San Sal ado .
Appendix Table A.XIV p esen s he esul s o es ima ing Speci ica ion (1) o he p obabili y
o a household ha ing a amily membe who mo ed ab oad in 1997-2007. On a e age, only
6% o he households ha e a amily membe who mo ed ab oad, and his a e does no change
a he bounda ies o gang e i o y. We also ind ha ich households bo h inside and ou side o
gang e i o y a e mo e likely o ha e a amily membe li ing ab oad. Howe e , he co ela ion
be ween weal h and ou -o -sample mig a ion in gang and nongang a eas a e no s a is ically
di e en om one ano he . Mo eo e , al hough ich households a e mo e likely o ha e a amily
membe who mo ed ab oad, he magni ude o ha e ec is much smalle han he a es o
selec i e ou -o -sample mig a ion om Appendix Table A.XIII ha a e equi ed o gene a e he
discon inui ies. In gang e i o y, an inc ease in he i s p incipal componen o he household
cha ac e is ics om ze o o one (i.e., he di e ence be ween he poo es and iches household)
inc eases he p obabili y o he household ha ing a amily membe mo e ab oad by only 7.1%,
whe eas he es ima es om Table A.XIII sugges ha , e en unde he un ealis ic assump ion o
ϱ=0%, he a e o ou -o -sample mig a ion o ich households needs o be a leas 51.7% o
explain he discon inui ies. The e o e, ou -o -sample mig a ion can accoun o no mo e han
100 →7.1/51.7 =13.7 pe cen o he e ec s in Table I.5152
50Fo ins ance, acco ding o Uni ed S a es Cus oms and Bo de P o ec ion, in 2012, he numbe o app ehensions
o indi iduals in amily uni s cons i u ed less han 3% o all app ehensions o Sal ado an ci izens a he Sou hwes
bo de o he Uni ed S a es. In p e ious yea s, ha numbe was e en smalle .
5113.7% should be in e p e ed as he uppe bound o he sha e o he esul s ha can be explained by ou -o -
sample mig a ion o he ollowing easons. Fi s , he 7.1% numbe assumes ha he e is no selec i e ou -o -sample
mig a ion ou side o gang e i o y. I he e is selec i e ou -o -sample mig a ion om nongang a eas, as sugges ed
by he esul s in Table A.XIV, hen his numbe should be lowe . Second, i is possible ha some households wi h
a amily membe ab oad ha e inc eased hei weal h because o ha ac (e.g., because o ecei ing emi ances). I
ha is he case, he esul s om Table A.XIV would o e es ima e he p obabili y o indi iduals om ich households
mig a ing ou o sample. Finally, he 50% numbe equi ed o gene a e he discon inui ies in Table Iis calcula ed
unde he assump ion ha poo indi iduals a e unable o mig a e ou o sample a all. I poo indi iduals also ha e a
chance o mig a ing ou o sample, his numbe should be highe .
52We also pe o m a es in he spi i o McC a y (2008) o check whe he , a he bounda ies o gang e i o y,
he e is a discon inuous change in popula ion densi y o a ious g oups o he popula ion. I indi iduals om gang-
GANGS, LABOR MOBILITY, AND DEVELOPMENT 47
The solu ion o he p o i maximiza ion p oblem o hese i ms yields
L(wl)="ω¯
K
wl#
1
1→ω
.(10)
ς(wl)=wl
ω
ω→1¯
K1
1→ω$ωω
1→ω↓ω1
1→ω%↓c↓ . (11)
The ee en y condi ion implies ha , in equilib ium, he i ms’ p o i s a e equal o ze o.
w↑↑
l
ω
ω→1¯
K1
1→ω$ωω
1→ω↓ω1
1→ω%↓c↓ =0 =↗w↑↑
l
ω
1→ω=
¯
K1
1→ω$ωω
1→ω↓ω1
1→ω%
c+ =↗
(12)
=↗w↑↑
l=

¯
K1
1→ω$ωω
1→ω↓ω1
1→ω%
c+ 

1→ω
ω
=↗L↑↑ =ω1
1→ω(c+ )1
ω
¯
K1
ω(ωω
1→ω↓ω1
1→ω)1
ω
.
Empi ically we obse e ha i ms a away om gang e i o y (i.e., in (b+ϑ,1]) pay highe
wages and hi e mo e wo ke s. Fo he model o ma ch hese empi ical indings, he ollowing
condi ions need o be sa is ied.


¯
K1
1→ω$ωω
1→ω↓ω1
1→ω%
c+ 

1→ω
ω
>&ωω
1→ω↓ω1
1→ω
c'
1→ω
ω
=↗¯
K>(1+
c)1↓ω(13)
ω1
1→ω(c+ )1
ω
¯
K1
ω(ωω
1→ω↓ω1
1→ω)1
ω
>ω1
1→ωc1
ω
(ωω
1→ω↓ω1
1→ω)1
ω
=↗¯
K<1+
c(14)
Toge he , hese wo condi ions yield ¯
K↔(1+
c1↓ω,1+
c).65 In ui i ely, condi ion (13)
s a es ha ¯
Kneeds o be su icien ly high o he p oduc i i y gains om ¯
K o o se he
dec ease in demand due o he necessi y o paying o adop he echnology. In u n, condi ion
(14) s a es ha i ms will only hi e mo e wo ke s i ¯
Kis su icien ly low ha he inc ease in
wages does no nega e he inc ease in labo demand due o highe p oduc i i y.
Finally, o ensu e ha he desc ibed si ua ion ep esen s an equilib ium in his ma ke , we
need o ensu e ha i ms a away om gang e i o y (i.e., in (b+ϑ,1]) ha e incen i es o
pu chase he p oduc i i y-enhancing equipmen ins ead o using he labo -only echnology.
Gi en (12), his condi ion equi es ha
w↑↑
l
ω
ω→1$ωω
1→ω↓ω1
1→ω%↓c<0=↗¯
K>(1+
c)1↓ω
.(15)
65Fo ins ance, his condi ion would be sa is ied o ¯
K=2, =2,c=1,and ϖ=0.5.

48
Thus, (13) gua an ees ha he i ms do no ha e incen i es o change hei p oduc ion beha io .
Coun e ac ual analysis. Using he heo e ical amewo k ou lined abo e, we now p o ide
a coun e ac ual analysis o wo hypo he ical scena ios: he elimina ion o gang-imposed e-
s ic ions on indi iduals’ mobili y and he ull emo al o gang p esence.
In he se ing o ou model, unde he o me scena io, i ms in [0,b+ϑ]would s ill be unwill-
ing o adop he p oduc i e echnology. As a esul , bo h agg ega e p oduc ion and agg ega e
employmen in p oduc i e i ms, which is de e mined by he ee en y condi ion, would emain
he same. Indi iduals li ing in [0,b]would s ill bene i om hei abili y o commu e o wo k
in high-p oduc i i y i ms, bu gi en he ze o-sum na u e o his economy, hese bene i s would
come a he expense o wo ke s li ing in [b, 1].
On he o he hand, i gang p esence we e ully elimina ed, i ms in [0,b+ϑ]would ind i
p o i able o adop he p oduc i e echnology, leading o bo h highe agg ega e p oduc ion and
highe demand o labo . Consequen ly, bene i s would acc ue o indi iduals li ing in all pa s
o he ci y, al hough esiden s o o me gang-con olled neighbo hoods would gain he mos
due o hei p e ious inabili y o wo k in high-p oduc i i y i ms.
In mid-2022, El Sal ado wi nessed unp eceden ed go e nmen -led c ackdowns on c iminal
ac i i y, leading o he de ac o implemen a ion o his la e scena io. In a sepa a e p ojec ,
we p o ide a de ailed analysis o he consequences o hese c ackdowns, showing how o me
gang-con olled neighbo hoods expe ienced signi ican imp o emen s in mobili y (Melniko
e al.,2024).
ONLINE APPENDIX TABLES
GANGS, LABOR MOBILITY, AND DEVELOPMENT 49
TABLE A.I
SOCIOECONOMIC CHARACTERISTICS FROM THE 2019 SURVEY
Has a high Has a uni e si y Household Wo ks in a i m wi h Wo ks in a i m wi h
school deg ee deg ee income ↘100 employees ↘200 employees
(1) (2) (3) (4) (5)
Panel A: All su ey esponden s
Gang e i o y -0.311*** -0.254*** -352.60*** -0.123*** -0.115***
(0.057) (0.062) (112.22) (0.019) (0.028)
Mean o dep. a . 0.508 0.180 625.0 0.169 0.133
Obse a ions 2,275 2,275 2,314 2,071 2,071
Panel B: Responden s who ha e li ed in he same loca ion hei en i e li e
Gang e i o y -0.281*** -0.173*** -271.05** -0.114*** -0.104**
(0.061) (0.056) (118.14) (0.033) (0.041)
Mean o dep. a . 0.474 0.149 602.3 0.155 0.123
Obse a ions 1,757 1,757 1,787 1,589 1,589
No e: *** p<0.01, ** p<0.05, * p<0.1. A e yea s o gang con ol, gang-con olled a eas ha e wo se socioeconomic condi ions han
neighbo ing a eas ha we e no unde he con ol o gangs. The able p esen s he esul s o es ima ing Speci ica ion (1) o he a iables
om he 2019 su ey. Panel A p esen s he esul s o he ull sample; Panel B— o he subsample o esponden s who ha e always li ed in
he same loca ion. Fo household income, he uni o obse a ion is a household; o all he o he a iables—an indi idual. Omi ed con ols
include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in
pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
50
TABLE A.II
PLACEBO:EFFECTS OF MAJOR ROADS THAT DID NOT DEFINE THE BORDERS OF GANG TERRITORY
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Placebo ea men g oup -0.147 -0.018 0.026 0.033 -0.008 0.010
(0.151) (0.044) (0.036) (0.059) (0.028) (0.076)
Mean o dep. a . 0.883 0.033 0.979 0.064 0.012 0.074
Obse a ions 6,716 5,623 5,714 5,714 5,714 5,374
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Placebo ea men g oup 0.007 -0.048 -0.091 -0.022 -0.016 -0.130
(0.014) (0.147) (0.150) (0.035) (0.106) (0.377)
Mean o dep. a . 0.026 0.222 0.528 0.936 0.171 2.427
Obse a ions 5,308 5,411 5,434 5,475 5,419 5,714
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Placebo ea men g oup -0.034 0.022 0.007 -0.035 -0.009 0.001
(0.028) (0.080) (0.059) (0.081) (0.054) (0.050)
Mean o dep. a . 0.892 0.299 0.0920 0.930 0.313 0.360
Obse a ions 19,130 18,563 18,563 5,623 5,238 18,563
Demog aphic cha ac e is ics Neighbo hood cha ac e is ics
Female Age U ban e i o y Road densi y Ele a ion T ee co e age
(19) (20) (21) (22) (23) (24)
Placebo ea men g oup 0.016 -0.741 -0.002 -9.353 -51.593 -0.010
(0.017) (1.920) (0.062) (9.426) (52.059) (0.016)
Mean o dep. a . 0.541 29.63 0.954 19.54 676.8 0.008
Obse a ions 20,967 20,967 47 47 47 47
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (2), using he loca ions o majo oads
ha did no con ibu e o he o ma ion o he bounda ies o gang e i o y as a placebo. Simila ly o Table III, we exclude 25% o he la ges
census ac s, which a e p edominan ly p esen ou side gang e i o y. We also include dummies o he h ee emaining qua iles o he census
ac size dis ibu ion. Table S.III in he Supplemen a y Ma e ials epo s he esul s o es ima ing he same eg ession speci ica ion wi hou
excluding he la ges census ac s, and ins ead, including dummies o all ou qua iles o he census ac size dis ibu ion. The uni o
obse a ion is a dwelling, household, indi idual, o census ac , depending on which cha ac e is ics a e being conside ed. In he indi idual-
le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols include a dummy o gang e i o y as well as a linea end in
dis ance o he placebo bounda ies, sepa a ely o loca ions on each side o he placebo bounda ies and on each side o he bounda ies o gang
e i o y. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he placebo bounda ies (sepa a ely o each
side o he bounda ies).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 51
TABLE A.III
EXCLUDING OBSERVATIONS WITHIN 100 METERS OF THE BOUNDARIES OF GANG TERRITORY
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.067*** 0.054*** -0.092*** -0.101*** 0.002 -0.176***
(0.019) (0.011) (0.028) (0.017) (0.003) (0.024)
Mean o dep. a . 0.936 0.026 0.943 0.116 0.004 0.194
Obse a ions 50,183 42,287 43,258 43,258 43,258 41,726
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.032*** -0.288*** -0.204*** -0.036*** -0.239*** -1.006***
(0.008) (0.048) (0.052) (0.008) (0.045) (0.235)
Mean o dep. a . 0.034 0.456 0.708 0.954 0.362 3.179
Obse a ions 41,205 41,911 41,964 42,108 41,860 43,258
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.040*** -0.208*** -0.163*** -0.058*** -0.126*** -0.136***
(0.009) (0.029) (0.028) (0.013) (0.022) (0.021)
Mean o dep. a . 0.931 0.464 0.223 0.955 0.388 0.533
Obse a ions 144,977 141,210 141,210 42,287 40,651 141,210
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he a iables om he 2007
census a e excluding obse a ions wi hin 100 me e s o he bounda ies o gang e i o y. The uni o obse a ion is a dwelling, household,
o indi idual, depending on which cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e
popula ion. Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he
bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely
o each side o he bounda ies).
52
TABLE A.IV
CONTROLLING FOR 300→300 METER FIXED EFFECTS
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.052* 0.023*** -0.073*** -0.097*** 0.006*** -0.160***
(0.030) (0.007) (0.026) (0.025) (0.002) (0.028)
Mean o dep. a . 0.932 0.028 0.941 0.108 0.005 0.180
Obse a ions 72,087 60,675 62,169 62,169 62,169 59,776
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.010* -0.224*** -0.135*** -0.019 -0.190*** -0.641***
(0.006) (0.047) (0.032) (0.011) (0.037) (0.207)
Mean o dep. a . 0.033 0.428 0.697 0.952 0.346 3.093
Obse a ions 59,096 60,045 60,168 60,384 60,020 62,169
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.031*** -0.137*** -0.101*** -0.040** -0.100*** -0.089***
(0.006) (0.031) (0.032) (0.017) (0.021) (0.023)
Mean o dep. a . 0.928 0.449 0.208 0.952 0.378 0.522
Obse a ions 208,416 202,935 202,935 60,675 58,293 202,935
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he a iables om he 2007
census, con olling o 300↔300 me e ixed e ec s. The uni o obse a ion is a dwelling, household, o indi idual, depending on which
cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols include
300↔300 me e ixed e ec s and a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he
bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely
o each side o he bounda ies).

GANGS, LABOR MOBILITY, AND DEVELOPMENT 53
TABLE A.V
RESTRICTIONS ON INDIVIDUALS’MOBILITY AND LABOR-MARKET OUTCOMES,BY CAR AVAILABILITY
Wo ks in Wo ks in same Has been o Has been o F eedom o
gang e i o y neighbo hood San a Ana he beach mo emen
whe e hey li e whe e hey li e
(1) (2) (3) (4) (5)
Gang e i o y →0.238*** 0.102** -0.228*** -0.090** -0.105*
→Ge s o wo k by ca o mo o cycle (0.059) (0.044) (0.052) (0.042) (0.055)
Gang e i o y →0.532*** 0.062* -0.214*** -0.044 -0.105**
→Ge s o wo k in ano he way (0.041) (0.035) (0.044) (0.034) (0.040)
Mean o dep. a . 0.334 0.302 0.511 0.876 0.809
Obse a ions 1,738 2,071 2,071 2,071 2,071
Household Wo ks in a Wo ks in a Gang bo de s p e en ed
income i m wi h i m wi h you om inding jobs
↘100 employees ↘200 employees in la ge i ms in
o he pa s o he ci y
(6) (7) (8) (9) (10)
Gang e i o y →-401.78** -0.144*** -0.153*** 0.132**
→Ge s o wo k by ca o mo o cycle (165.87) (0.037) (0.047) (0.050)
Gang e i o y →-257.37*** -0.080*** -0.073** 0.091**
→Ge s o wo k in ano he way (92.81) (0.024) (0.031) (0.042)
Gang e i o y →Has a ca 0.089*
(0.048)
Gang e i o y →Does no ha e a ca 0.105**
(0.046)
Mean o dep. a . 629 0.169 0.133 0.407 0.407
Obse a ions 2,071 2,071 2,071 2,313 2,313
No e: *** p<0.01, ** p<0.05, * p<0.1. Columns 1–9 o he able p esen he esul s o es ima ing Speci ica ion (1) o mobili y and
labo -ma ke ou comes wi h he dummy o gang e i o y eplaced wi h dummies o indi iduals in gang e i o y who ge o wo k by ca o
mo o cycle and indi iduals in gang e i o y who ge o wo k in some o he way. In Column 10, he dummy o gang e i o y is eplaced wi h
simila dummies o indi iduals in gang e i o y who own a ca and indi iduals in gang e i o y who do no . The sample consis s only o
su ey esponden s who ha e a job. San a Ana is a neighbo ing municipali y, which is app oxima ely 60 kilome e s away om San Sal ado .
The beach is app oxima ely 30 kilome e s away om San Sal ado . The uni o obse a ion is an indi idual. Omi ed con ols include a linea
end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies, as well as a dummy o ge ing o
wo k by ca o mo o cycle (in Column 10, a dummy o owning a ca ). S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing
he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
54
TABLE A.VI
RESTRICTIONS ON INDIVIDUALS’MOBILITY AND LABOR-MARKET OUTCOMES,BY EDUCATION
Wo ks in F eedom o Gang bo de s p e en ed Household Wo ks in a Wo ks in a
gang e i o y mo emen you om inding jobs income i m wi h i m wi h
whe e hey li e in la ge i ms in ↘100 employees ↘200 employees
o he pa s o he ci y
(1) (2) (3) (4) (5) (6)
Gang e i o y →0.323*** -0.072* 0.115** -378.77*** -0.085*** -0.087**
→High school deg ee (0.043) (0.041) (0.042) (113.63) (0.028) (0.036)
Gang e i o y →0.577*** -0.106** 0.083* -158.74 -0.052* -0.057*
→No high school deg ee (0.035) (0.043) (0.047) (106.20) (0.028) (0.030)
Mean o dep. a 0.330 0.812 0.407 628.3 0.170 0.133
Obse a ions 1,707 2,275 2,313 2,275 2,033 2,033
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he mobili y and labo -ma ke
ques ions by educa ion. The uni o obse a ion is an indi idual. Omi ed con ols include a dummy o ha ing a high school deg ee and a linea
end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en heses a e
clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
TABLE A.VII
CONSEQUENCES OF LOW LABOR MOBILITY
Household income Wo ks in a i m wi h Wo ks in a i m wi h
↘100 employees ↘200 employees
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Li es in gang e i o y -352.60 -429.99 -235.09 -0.123 -0.210 -0.105 -0.115 -0.187 -0.102
(112.22)*** (127.82)*** (112.56)** (0.019)*** (0.022)*** (0.023)*** (0.028)*** (0.025)*** (0.030)***
[84.97]*** [98.80]*** [81.33]*** [0.042]*** [0.046]*** [0.041]*** [0.035]*** [0.038]*** [0.035]***
Li es in gang e i o y, 167.64 85.39 0.182 0.129 0.152 0.110
wo ks in nongang e i o y (32.69)*** (30.23)*** (0.026)*** (0.025)*** (0.027)*** (0.026)***
[37.08]*** [38.73]** [0.025]*** [0.024]*** [0.024]*** [0.023]***
Has a high school deg ee 89.11 0.124 0.088
(19.90)*** (0.021)*** (0.018)***
[26.78]*** [0.020]*** [0.019]***
Has a uni e si y deg ee 445.46 0.148 0.132
(76.96)*** (0.029)*** (0.027)***
[62.62]*** [0.032]*** [0.030]***
Mean o dep. a . 625.00 634.70 638.90 0.169 0.169 0.170 0.133 0.132 0.132
Obse a ions 2,314 1,738 1,707 2,071 1,738 1,707 2,071 1,738 1,707
No e: *** p<0.01, ** p<0.05, * p<0.1. The able shows ha he discon inui y in income and i m size is signi ican ly smalle o
nonexis en o indi iduals li ing in gang e i o y bu wo king ou side o gang e i o y. All he a iables come om he 2019 su ey. Fo
household income, he uni o obse a ion is a household; o he o he a iables—an indi idual. Omi ed con ols include a linea end in
dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. In he columns ha include a a iable
o wo king ou side o gang e i o y, we addi ionally include a dummy o whe he he indi idual is cu en ly employed (in he su ey,
unemployed indi iduals we e asked o desc ibe hei mos ecen wo k expe ience). S anda d e o s in pa en heses a e clus e ed by 30 me e
bins, deno ing he dis ance o he bounda ies o gang e i o y, sepa a ely o each side o he bounda ies. S anda d e o s in b acke s a e
adjus ed o allow o spa ial co ela ion wi hin a 100 me e adius (Conley co ec ion).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 55
TABLE A.VIII
RESTRICTIONS ON INDIVIDUALS’MOBILITY AND LABOR-MARKET OUTCOMES,BY GENDER
Wo ks in F eedom o Gang bo de s p e en ed Household Wo ks in a Wo ks in a
gang e i o y mo emen you om inding jobs income i m wi h i m wi h
whe e hey li e in la ge i ms in ↘100 employees ↘200 employees
o he pa s o he ci y
(1) (2) (3) (4) (5) (6)
Gang e i o y →Male 0.454*** -0.077* 0.119** -370.07*** -0.138*** -0.116***
(0.042) (0.043) (0.050) (114.52) (0.034) (0.037)
Gang e i o y →Female 0.520*** -0.107** 0.084** -332.33*** -0.108*** -0.110***
(0.045) (0.041) (0.041) (107.53) (0.019) (0.030)
Mean o dep. a 0.334 0.811 0.407 625 0.169 0.133
Obse a ions 1,738 2,314 2,313 2,314 2,071 2,071
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he mobili y and labo -ma ke
ques ions by gende . The uni o obse a ion is an indi idual. Omi ed con ols include a dummy o being emale and a linea end in dis ance
o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed by 30
me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
TABLE A.IX
FIRMS’LOCATION,PROFITS,REVENUE,AND COSTS
Log o i ms pe km2:
Log o he i m’s: Census: Google Maps:
P o i s Re enue Cos s Employees Cos s pe All i ms All i ms Ca es & G oce y Pha macies
employee es au an s s o es
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Gang e i o y -0.198 -0.027 0.094 -0.019 0.133 -0.303 0.034 -0.197 0.113 -0.056
(0.362) (0.332) (0.330) (0.121) (0.239) (0.406) (0.103) (0.182) (0.144) (0.148)
Mean o dep. a . 9.767 10.97 10.44 1.756 8.679 5.062 4.860 2.394 1.882 1.251
Obse a ions 5,631 6,118 6,083 6,120 6,083 156 86 86 86 86
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he numbe o business es ab-
lishmen s and hei cha ac e is ics. The esul s in columns 1–6 a e based on he supplemen o he 2005 economic census. In columns 1–5,
he uni o obse a ion is a i m; in columns 6—a sec o , he analogue o he census ac in he economic census. The da a on he numbe o
business es ablishmen s in columns 7–10 come om Google Maps. In hese eg essions, he uni o obse a ion is a 10 me e bin, deno ing
dis ance o he bounda ies o gang e i o y, weigh ed by he size o he a ea o he dis ance bins. Omi ed con ols include a linea end in
dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed
by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
56
TABLE A.X
GANG CONTROL AND DROPOUT RATES
D opou a e
Subsample: All obs. Yea ↑2007 Yea >2007 Male Female All obs.
(1) (2) (3) (4) (5) (6)
Gang e i o y 0.019*** 0.021** 0.018*** 0.021*** 0.019***
(0.004) (0.008) (0.004) (0.006) (0.003)
Gang e i o y →S anda d p og am 0.019***
(0.004)
Gang e i o y →P og am o adul s 0.038**
(0.018)
Mean o dep. a . 0.020 0.021 0.019 0.023 0.016 0.020
Obse a ions 3,199 684 2,515 3,088 3,186 3,377
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he d opou a es o schools in
San Sal ado . The da a come om he annual census o schools. In columns 1–5, he uni o obse a ion is a school in a yea . In hese esul s,
omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies.
In column 6, he uni o obse a ion is he ype o p og am (s anda d o o adul s) in a school in a yea . In hese esul s, omi ed con ols
include a dummy o he p og am being o adul s and linea ends in dis ance o he bounda ies o gang e i o y, sepa a ely o each ype o
p og am on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies
o gang e i o y (sepa a ely o each side o he bounda ies). S anda d e o s in b acke s a e adjus ed o allow o spa ial co ela ion wi hin a
100 me e adius (Conley co ec ion).
TABLE A.XI
GANG CONTROL AND EXAM SCORES
Ma h Na u al sciences Social sciences Languages & li e a u e
Subsample: All obs. Yea ↑2007 All obs. Yea ↑2007 All obs. Yea ↑2007 All obs. Yea ↑2007
(1) (2) (3) (4) (5) (6) (7) (8)
Gang e i o y -0.835** -0.801** -0.652** -0.603** -0.666*** -0.686** -0.712*** -0.649**
(0.337) (0.331) (0.248) (0.250) (0.234) (0.278) (0.240) (0.252)
Mean o dep. a . 5.434 5.511 5.776 5.901 6.432 6.382 6.151 5.960
Obse a ions 1,284 436 1,284 436 1,284 436 1,284 436
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he a e age exam sco es in San
Sal ado schools. The da a come om he schools’ adminis a i e eco ds in 1999-2001 and 2005-2017. The uni o obse a ion is a school
in a yea . Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he
bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely
o each side o he bounda ies).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 63
FIGURE A.4.—Household Income A e 22 Yea s o Gang Con ol
No e: The igu e illus a es he esul s o household income om Appendix Table A.I. The le -hand side o he igu e p esen s he esul s
o he ull sample (Panel A o Appendix Table A.I), he igh -hand side— o he subsample o indi iduals who ha e li ed in he same loca ion
all hei li e (Panel B o Appendix Table A.I). The esul s a e e y simila . The e ical axis ep esen s he a e age alue o household income;
he ho izon al axis—dis ance (in me e s) o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side
o gang e i o y; a eas o he igh a e con olled by he gangs. The do s ep esen he a e age alue o he ou come a iable in ha 30 me e
bin.
FIGURE A.5.—Socioeconomic Condi ions Be o e he Gangs’ A i al: 1s P incipal Componen s o he Dwelling,
Household, and Indi idual Cha ac e is ics
No e: The igu e illus a es he esul s o he 1s p incipal componen s o he dwelling, household, and indi idual cha ac e is ics om
Table II. All he a iables come om he 1992 census. The uni o obse a ion is a dwelling, a household, and an indi idual, depending on
he speci ica ion. All he a iables a e no malized o a y be ween ze o and one wi h highe alues ep esen ing be e ou comes. The e ical
axis ep esen s he a e age alue o he ou comes a iable; he ho izon al axis—dis ance (in me e s) o he bounda ies o gang e i o y.
Neighbo hoods o he le o he dashed line a e loca ed ou side o gang e i o y; a eas o he igh a e con olled by he gangs. The do s
ep esen he a e age alue o he ou come a iable in ha 30 me e bin.

64
FIGURE A.6.—S abili y o he Bounda ies o Gang Te i o y
No e: The ou come a iable comes om he 2023 indi idual su ey. The e ical axis ep esen s he a e age sha e o su ey esponden s
who ag ee wi h he s a emen ha he gangs con olled hei neighbo hood 20 yea s ago, du ing he p esidency o F ancisco Flo es Pé ez
(P esiden o El Sal ado om 1999 o 2004); he ho izon al axis—dis ance (in me e s) o he bounda ies o gang e i o y. Neighbo hoods o
he le o he dashed line a e loca ed ou side o gang e i o y; a eas o he igh a e con olled by he gangs. The do s ep esen he a e age
alue o he ou come a iable in ha 30 me e bin.
FIGURE A.7.—Excluding Obse a ions Wi hin 100 Me e s o he Bounda ies o Gang Te i o y
No e: The igu e illus a es he eg ession discon inui y plo s o he 1s p incipal componen s o he dwelling, household, and indi idual
cha ac e is ics om he 2007 census a e excluding obse a ions wi hin 100 me e s o he bounda ies o gang e i o y. The uni o obse a ion
is a dwelling, a household, and an indi idual, depending on he speci ica ion. All he a iables a e no malized o a y be ween ze o and one
wi h highe alues ep esen ing be e ou comes. The e ical axis ep esen s he a e age alue o he ou comes a iable; he ho izon al axis—
dis ance (in me e s) o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side o gang e i o y; a eas
o he igh a e con olled by he gangs. The do s ep esen he a e age alue o he ou come a iable in ha 30 me e bin.
GANGS, LABOR MOBILITY, AND DEVELOPMENT 65
FIGURE A.8.—Fi m Cha ac e is ics, by Dis ance o he Bounda ies o Gang Te i o y
No e: The igu e illus a es he esul s o i m cha ac e is ics om columns 1, 4, and 5 o Table A.IX and column 2 o Table V. The i s
h ee a iables come om he 2005 economic census; he ou h one— om he 2015 su ey o i ms conduc ed by FUSADES. The uni o
obse a ion is a i m. The e ical axis ep esen s he a e age alue o he ou comes a iable; he ho izon al axis—dis ance (in me e s) o he
bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side o gang e i o y; a eas o he igh a e con olled
by he gangs. The do s ep esen he a e age alue o he ou come a iable in ha 30 me e bin.
66
FIGURE A.9.—Syn he ic Di e ence-in-Di e ences
No e: The igu e illus a es he esul s o he syn he ic di e ence-in-di e ences analysis. The plo s on he le side o he igu e ep esen
he e ec s o gang p esence when he ea men g oup is de ined in he same way as in he baseline di e ence-in-di e ences analysis. The
plo s on he igh side o he igu e p esen he same analysis wi h he ea men g oup na owed down o he ou la ges ci ies in El Sal ado ,
all o which ha e had a subs an ial gang p esence since he la e 1990s.
GANGS, LABOR MOBILITY, AND DEVELOPMENT 67
FIGURE A.10.—Gang- ela ed Homicides, by Dis ance o he Bounda ies o Gang Te i o y
Panel A
Panel B
No e: The igu e illus a es he numbe o gang- ela ed homicides in 2003-2008 (Panel A) and 2009-2014 (Panel B), by dis ance o
he bounda ies o gang e i o y. In bo h cases, he la ges numbe o he homicides ook place igh a he bounda ies o gang e i o y.
The e ical axis ep esen s he a e age alue o he ou comes a iable; he ho izon al axis—dis ance (in me e s) o he bounda ies o gang
e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side o gang e i o y; a eas o he igh a e con olled by he gangs. The
do s ep esen he a e age alue o he ou come a iable in ha 10 me e bin.
68
S. SUPPLEMENTARY MATERIALS: NOT FOR PUBLICATION
S.1. Addi ional Da a Sou ces
Homicides and obbe ies. The da a on gang- ela ed homicides come om he PNC and
co e 2003 o 2014. Fo each obse a ion, we ob ained in o ma ion abou he ime and day
i occu ed, whe he he pe pe a o was a membe o a gang, and he add ess o occu ence.
Using hese add esses, we manually geocoded he da a o ob ain he la i ude and longi ude o
he homicides ca ied ou by gang membe s. The PNC also sha ed wi h us he da a on gang-
ela ed homicides in 2000, bu hese da a a e a ailable only a he municipali y le el.
The da a on obbe ies come om he Me opoli an Planning O ice o San Sal ado (O ic-
ina de Plan icación del Á ea Me opoli ana de San Sal ado , OPAMSS) and co e 2014 o
2015. They con ain in o ma ion abou he ime, da e, and loca ion o all obbe ies, including
hei la i ude and longi ude.
Inca ce a ion da a. The da a on inca ce a ions come om he Gene al Di ec o a e o P is-
ons (Di ección Gene al de Cen os Penales, DGCP) and ep esen he uni e se o all indi idu-
als inca ce a ed in El Sal ado since he mid 1980s. The eco ds con ain in o ma ion abou he
c imes he indi idual commi ed, he da e o inca ce a ion, he municipali y o bi h, and he
las known add ess. Fo inma es who en e ed p ison be o e 1997 and whose las known add ess
is in San Sal ado municipali y (4,726 indi iduals), we manually geocoded he esiden ial ad-
d esses o ob ain he p ecise geocoo dina es used in he analyses. Gi en ha geocoded c ime
da a p io o 2003 a e una ailable, he inma es’ esiden ial add esses ep esen he bes measu e
o c iminal ac i i y in he p e ea men pe iod.
Household su eys. To analyze household income da a beyond San Sal ado , we ob ained
yea ly household su eys om DIGESTYC spanning 1992 o 2007. These su eys sample
o e 10,000 households and include a b oad ange o ques ions. In ou analysis, we ocus on
he ques ion ela ed o household income which was asked h oughou he yea s.
2023 su ey. In he summe o 2023, we conduc ed a ollow-up su ey in gang and non-
gang neighbo hoods in San Sal ado , ollowing a simila sampling p o ocol as ou 2019 su ey
desc ibed in Sec ion 3. This su ey included in-pe son in e iews wi h ques ions abou indi id-
uals’ mobili y and employmen . Due o imp o ed secu i y in he coun y, we we e also able o
ask abou gang ac i i y in hei neighbo hoods, including ques ions ela ed o he gangs’ sys em
o bo de s.
Cell phone GPS pings. We pu chased cell phone ping da a om Quexopa, a da a agg ega o
specializing in La in Ame ica, co e ing Janua y and Feb ua y o 2022.66 The da ase includes a
unique cell phone iden i ie , ime o cap u e, ope a ing sys em, GPS coo dina es, and accu acy
me ics.
66Al hough we sough da a om 2019, his was he ea lies pe iod a ailable. Due o he la ge olume o da a, i is
common p ac ice o dele e eco ds olde han one yea o s o age easons.

GANGS, LABOR MOBILITY, AND DEVELOPMENT 69
U ban e i o y. The da a on u ban densi y come om New Yo k Uni e si y’s A las o
U ban Expansion. The as e map p esen s he u ban a eas in he G ea e San Sal ado egion
in 1999.67 We ans o m he da a in o a bina y as e , equal o one when he loca ion is classi ied
as u ban. Then, o each o he census ac s om he 2007 census, we calcula e he sha e o
census ac s’ e i o y ha is u ban.
Wa e ways. The map o he wa e ways in El Sal ado comes om he Humani a ian Open-
S ee Map Team.68 Then, o each o he census ac s om he 2007 census, we c ea ed a
dummy a iable o whe he he census ac con ains a pa o he wa e way.
Road densi y. The map o he oads in El Sal ado comes om he Humani a ian Open-
S ee Map Team and e lec s he oads in he coun y in Ma ch 2020.69 We hen ans o m he
ea u e-based map in o a bina y as e ile wi h he esolu ion o 1 me e →1 me e , whe e we
eplace he lines o oads wi h g id cells equal o one. A e ha , o each o he census ac s
om he 2007 census, we calcula e oad densi y, measu ed in kilome e s pe squa e kilome e .
Ele a ion. The da a on ele a ion a he esolu ion o 3 a c seconds (app oxima ely 90 me-
e s) come om he CGIAR-Conso ium o Spa ial In o ma ion (CGIAR-CSI).70 Fo each o
he census ac s om he 2007 census, we calcula e he a e age ele a ion inside he census
ac .
Te i o y used o co ee p oduc ion. The map o land use in 1998 (including co ee
p oduc ion) comes om he Minis y o En i onmen and Na u al Resou ces. We con e he
ea u e-based map in o a bina y as e , equal o one o a eas ha a e used o co ee p oduc-
ion. Then, o each o he census ac s om he 2007 census, we calcula e he sha e o hei
e i o y ha is used o co ee p oduc ion.
T ee co e age. The da a on ee co e age in 2000 come om Global Fo es Wa ch.71 The
as e ile p esen s he sha e o e i o y co e ed by ees in each 30 me e →30 me e g id cell.
Fo each o he census ac s om he 2007 census, we calcula e he a e age le el o ee
co e age inside o he census ac .
High school exam sco es. The da a on he schools’ a e age high school exi exam sco es
(P ueba de Ap endizaje y Ap i udes pa a Eg esados de Educación Media, PAES) come om
he Minis y o Educa ion. PAES esul s a e epo ed o ma h, na u al sciences, social sci-
ences, and Spanish language and li e a u e. The da a co e he pe iod om 1999 o 2017, bu
exclude he esul s o 2002-2004 because in hose yea s he Minis y o Educa ion applied a
nondisclosed cu e o he es sco es, p e en ing compa ison wi h he o he yea s.
67The da a can be accessed he e: A las o U ban Expansion: San Sal ado (accessed on May 4, 2020).
68The map o he wa e ways in El Sal ado can be accessed he e: Humani a ian Da a Exchange: El Sal ado
Wa e ways (accessed on May 4, 2020).
69The map o he oads in El Sal ado can be accessed he e: Humani a ian Da a Exchange: El Sal ado Roads
(accessed on May 4, 2020).
70The ele a ion map o El Sal ado can be accessed he e: CGIAR-CSI (accessed on May 4, 2020).
71The da a on ee co e age o El Sal ado can be accessed he e: Global Fo es Wa ch (accessed on May 4, 2020).
70
2020 su ey. In 2020, we conduc ed a su ey o 1,957 indi iduals in San Sal ado o e al-
ua e he ex en o gang- ela ed ex o ion in gang and nongang a eas. The su ey ollowed he
same p ocedu e as he 2019 su ey, excep ha i was conduc ed o e he elephone. The main
eason o conduc ing he su ey o e he elephone is ha , in in-pe son in e iews, ex o ion-
ela ed ques ions could ha e po en ially endange ed he enume a o s. A he beginning o he
su ey, he enume a o s asked he esponden s o hei add ess, and he su ey p oceeded i
he add ess was in one o he census segmen s andomly chosen in he sampling p ocedu e.
2005 economic census. The mic oda a o he 2005 economic census was p o ided by
DIGESTYC.72 A e c ea ing a egis y o all o mal and in o mal i ms in he coun y, DI-
GESTYC ook a andom sample o all he i ms o ask a long- o m ques ionnai e on income
sou ces, p oduc ion and emune a ion cos s, he yea he i m was es ablished, e c. F om hese
ques ions, DIGESTYC calcula ed he i ms’ e enue and cos s. In o al, he egis y includes
179,817 i ms ac oss he coun y, while he long- o m ques ionnai e co e s 46,864 i ms (26%).
In he analysis, we ocus on he long- o m ques ionnai e i ms based in San Sal ado (6,120
i ms).
Loca ions o schools, hospi als, and o he es ablishmen s. The da a on he loca ions o
schools, hospi als, and o he es ablishmen s in San Sal ado come om Google Maps.73 In
Augus 2019, we sc aped he da a om Google Maps o iden i y all he es ablishmen s in
San Sal ado . In o al, we ob ained a da ase wi h 7,732 es ablishmen s. Fo each obse a ion,
Google p o ides a classi ica ion o he ype o es ablishmen (e.g., school, hospi al, pha macy).
Housing en . To ob ain in o ma ion on housing en , in Augus -Sep embe 2018, we
sc aped he da a om he mos popula websi e o en lis ings in El Sal ado , OLX (now
Encuen a24).74 We ocused on noncomme cial lis ings in which he en i e apa men was be-
ing en ed ou (i.e., no a oom in he apa men ). The lis ings included he da a on he la i ude
and longi ude o he loca ion, he en eques ed by he landlo d, as well as in o ma ion abou
he apa men such as he numbe o bed ooms, he numbe o ba h ooms, he numbe o squa e
me e s, and whe he he apa men is being en ed ou by an agency. In o al, he da ase con-
ains 1,537 obse a ions.
I should be no ed ha we canno obse e whe he a pa icula apa men was en ed ou o
no . Howe e , a e wo mon hs, he as majo i y o o e s we e no longe a ailable.
I should also be no ed ha , on a e age, he p ope ies lis ed on OLX a e la ge and mo e
expensi e han he o e all pool o p ope ies in San Sal ado . In pa icula , many o he cheapes
p ope ies may be en ed ou on he in o mal ma ke and a e no lis ed online. I he e a e mo e
72Al hough he census was ca ied ou in 2005, he e e ence yea o all he ques ions was 2004.
73We use he da a on he loca ions o schools and hospi als om Google Maps ins ead o go e nmen eco ds. The
p ima y eason is he accu acy o he da a. Fo ins ance, in he shape ile he go e nmen has p o ided o us, some o
he schools a e loca ed ou side o El Sal ado . Howe e , i we use he da a om go e nmen eco ds, he esul s a e
quali a i ely e y simila .
74The Sal ado an e sion o he websi e can be accessed he e: OLX (now Encuen a24).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 71
such p ope ies in gang-con olled neighbo hoods, ou es ima es would p o ide a lowe bound
on he ac ual d op in housing en a he bounda ies o gang e i o y.
Nigh ime ligh densi y. Annual da a on nigh ime ligh densi y (o luminosi y) come om
he De ense Me eo ological Sa elli e P og am-Ope a ional Linescan Sys em (DMSP-OLS) and
spans he pe iod om 1992 o 2013.75 In pa icula , we use he DMSP-OLS da a, ep esen ing
he a e age s able ligh s om ci ies, owns, and o he si es wi h pe sis en ligh ing. The da a
a e p o ided by he Na ional Cen e s o En i onmen al In o ma ion (NCEI). I o a pa icula
yea , he da a we e a ailable om mo e han one sa elli e, we ake he a e age o he wo.
The esolu ion o he da a on nigh ime ligh densi y is 30 a c seconds→30 a c seconds (i.e.,
app oxima ely 1 kilome e →1 kilome e ). The e o e, he da a a e no su icien ly p ecise o be
used in he eg ession discon inui y design.
Gang leade s’ municipali ies o bi h. The da a on he gang leade s’ municipali ies o
bi h come om El Fa o, an in es iga i e newspape . We use he da a om hei in es iga i e
epo s, ocusing on he gang leade s who we e depo ed om he Uni ed S a es and had been
la e con ic ed o commi ing c imes in El Sal ado . O e all, he sample consis s o 33 gang
leade s bo h om MS-13 and 18 h S ee . We hen manually ma ch he names o he gang
leade s and he c imes hey commi ed o he c iminal eco ds om he Minis y o Jus ice and
Public Secu i y o El Sal ado , which con ain in o ma ion on he o ende ’s municipali y o
bi h.
S.2. Fu he De ails Abou he P ima y Da a Sou ces
2019 su ey. Fo he 2019 su ey, he ollowing sampling p ocedu e was applied. Gi en
he unce ain y abou hei ea men s a us, census ac s wi hin 15 me e s o he bounda ies o
gang e i o y we e excluded om he analysis. Then, sepa a ely o places inside and ou side
o gang e i o y, we spli he census ac s in o 30 me e bins, deno ing he dis ance o he
bounda ies (i.e., 15-44 me e s o he bounda ies, e c.). A e ha we andomly selec ed 10
census ac s om each bin and su eyed 8-10 people in each o hem.76 I he e we e ewe
ha 10 census ac s in ha bin, we su eyed indi iduals in all he census ac s ha we e
a ailable. In o al, he su ey includes 2,314 esponden s.
To ensu e he sa e y o he enume a o s, i he su ey eam was denied en y in o some
o he gang-con olled neighbo hoods, hose census ac s we e eplaced by o he ones om
he same bin. I i was no possible o in e iew 10 indi iduals in a census ac (e.g., because
a e epea ed a emp s nobody answe ed he doo ), addi ional people we e in e iewed in o he
census ac s in he same bin.
Gang bounda ies. The map o gang-con olled neighbo hoods ha we use in his s udy
is based on da a om 2015. To he bes o ou knowledge, maps o gang-con olled a eas o
75The da a and a de ailed desc ip ion o i a e a ailable he e: DMSP-OLS (accessed on May 4, 2020).
76In a eas wi hin 250 me e s o he bounda ies, we su eyed 10 indi iduals pe census ac . In loca ions u he
away om he bounda ies, we su eyed 8 indi iduals pe census ac .
72
ea lie yea s a e nonexis en . Howe e , acco ding o mul iple sou ces in he police depa men
as well as con e sa ions wi h he local popula ion, he bounda ies o gang e i o y in San Sal-
ado ha e emained s able since he la e 1990s when he bounda ies we e o med. I changes
o he bounda ies do occu , i ends o be a p oduc o u wa s (i.e., MS-13 and 18 h S ee ak-
ing o e each o he ’s e i o y); no because o he s a e egaining con ol o e gang e i o ies
o he o he way ound.
The da a on he gang-con olled neighbo hoods in San Sal ado come om EDH and a e
p esen ed in Figu e 1. Howe e , o accu a ely calcula e dis ance o he bounda ies o gang
e i o y, we also complemen hese da a wi h con iden ial maps om he police on he gang-
con olled neighbo hoods ou side o San Sal ado municipali y. Since he eg ession discon i-
nui y design ocuses on he census ac s inside o San Sal ado , his ne e a ec s he ea men
s a us o he census ac (i.e., whe he o no i is loca ed inside o gang e i o y). Howe e , o
he loca ions ou side o gang e i o y, i does some imes a ec he dis ance om hem o he
bounda ies o gang e i o y (i.e., i ha loca ion is close o a gang-con olled loca ion ou side
o San Sal ado ). I should be no ed ha , e en wi h he ex ended map o gang e i o y, we a e
unable o implemen he eg ession discon inui y design ou side o San Sal ado because he
map addi ionally includes only a small numbe o loca ions in he G ea e San Sal ado a ea.
1992 and 2007 censal ca og aphy. I should be no ed ha he bounda ies o he census
ac s in he 1992 and 2007 censuses we e no he same. The e o e, we a e no able o pe o m
a di e ence-in-di e ences analysis a he le el o he census ac s. Howe e , in bo h cases, he
size o he census ac s was qui e simila , allowing us o accu a ely measu e he dis ance om
he census ac o he bounda ies o gang e i o y. Thus, he dis ance be ween a pa icula
loca ion and he bounda ies o gang e i o y is e y simila , ega dless o whe he we use he
2007 o 1992 census ac s.
I should also be no ed ha , al hough DIGESTYC digi ized a map he 1992 census ac s, i
did no ully inish ha wo k. Speci ically, he 1992 map does no ha e he bounda ies o 18.9%
o he census ac s in no hwes e n San Sal ado . Howe e , he as majo i y o hose neigh-
bo hoods a e loca ed mo e han 420 me e s away om gang e i o y and, he e o e, would no
be included in he analysis in any case. In pa icula , nea ly all o gang e i o y (excep o a
ew small “islands”) and he neighbo hoods igh nex o i a e included in he 1992 map. Thus,
i is highly unlikely ha ou es ima es would change i all he census ac s we e included.77
Ex o ion. Ou measu es on i m and household ex o ion paymen s d aw om h ee
sou ces. Fi s , he da a on whe he i ms ha e expe ienced ex o ion come om a su ey o
small and medium-sized en e p ises conduc ed by he Sal ado an Founda ion o Economic
and Social De elopmen (Fundación Sal ado eña pa a el Desa ollo Económico y Social, FU-
SADES). The su ey also asked whe he he i m has expe ienced gang ac i i y in he loca ion
77DIGESTYC also old us ha he wo k on digi izing he map o he census ac s had o s op because o he lack
o unding and ha he e was no speci ic eason why some census ac s we e digi ized and some we e no .
GANGS, LABOR MOBILITY, AND DEVELOPMENT 79
1 oom in apa men , 2018 0.113 0.317 1,537 OLX
2 ooms in apa men , 2018 0.187 0.390 1,537 OLX
3 ooms in apa men , 2018 0.528 0.499 1,537 OLX
4 ooms in apa men , 2018 0.110 0.312 1,537 OLX
5 ooms in apa men , 2018 0.040 0.197 1,537 OLX
6 ooms in apa men , 2018 0.010 0.102 1,537 OLX
7+ ooms in apa men , 2018 0.012 0.108 1,537 OLX
1 ba h oom in apa men , 2018 0.157 0.364 1,537 OLX
2 ba h ooms in apa men , 2018 0.176 0.381 1,537 OLX
3 ba h ooms in apa men , 2018 0.446 0.497 1,537 OLX
4 ba h ooms in apa men , 2018 0.141 0.348 1,537 OLX
5 ba h ooms in apa men , 2018 0.053 0.224 1,537 OLX
6 ba h ooms in apa men , 2018 0.019 0.136 1,537 OLX
7+ ba h ooms in apa men , 2018 0.008 0.092 1,537 OLX
Squa e me e s, 2018 189.38 264.65 1,537 OLX
Ren ed ou by agency, 2018 0.491 0.500 1,537 OLX
Panel J: O he RDD a iables
U ban e i o y, 1999 0.814 0.297 476 NYU A las o U ban Expansion
Road densi y (km pe km2), 2020 17.84 8.81 476 Humani a ian OpenS ee Map
Has access o wa e way 0.326 0.469 476 Humani a ian OpenS ee Map
Ele a ion 720.4 87.91 476 CGIAR SRTM
Te i o y used o co ee p oduc ion 0.028 0.132 476 Minis y o he En i onmen
and Na u al Resou ces
T ee co e age, 2000 0.049 0.116 476 Global Fo es Wa ch
Panel K: Di e ence-in-di e ences a iables
Fi m s ock, 1988-2005 153.61 565.33 1,854 2005 census
Luminosi y, 1992-2013 10.18 14.07 2,288 DMSP-OLS
Household income pe capi a, 1992-2013 103.42 115.52 171,257 Household su eys

80
TABLE S.II
BOUNDARIES OF GANG TERRITORY FROM GEOGRAPHICAL BARRIERS,
WITHOUT EXCLUDING THE LARGEST CENSUS TRACTS
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.078** 0.045** -0.054*** -0.073 0.008* -0.072
(0.036) (0.016) (0.015) (0.052) (0.004) (0.055)
Mean o dep. a . 0.945 0.021 0.969 0.064 0.003 0.124
Obse a ions 10,047 8,418 8,684 8,684 8,684 8,260
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.011 -0.274** -0.184*** -0.031** -0.220** -0.959**
(0.007) (0.106) (0.060) (0.015) (0.091) (0.371)
Mean o dep. a . 0.034 0.366 0.697 0.958 0.291 2.978
Obse a ions 8,183 8,296 8,314 8,355 8,293 8,684
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.080** -0.248*** -0.178** -0.063** -0.104*** -0.168***
(0.035) (0.062) (0.068) (0.028) (0.035) (0.042)
Mean o dep. a . 0.927 0.436 0.171 0.962 0.354 0.505
Obse a ions 29,268 28,195 28,195 8,418 8,063 28,195
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1), using he loca ions o majo
oads and boule a ds (geog aphical ba ie s) as he p edic ed bounda ies o gang e i o y. We addi ionally con ol o dummies o he ou
qua iles o he census ac size dis ibu ion. All he a iables come om he 2007 census. The uni o obse a ion is a dwelling, household,
o indi idual, depending on which cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e
popula ion. Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he
bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely
o each side o he bounda ies).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 81
TABLE S.III
PLACEBO:EFFECTS OF MAJOR ROADS THAT DID NOT DEFINE THE BORDERS OF GANG TERRITORY,
WITHOUT EXCLUDING THE LARGEST CENSUS TRACTS
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Placebo ea men g oup 0.002 -0.042 0.034 0.124 -0.005 0.082
(0.127) (0.036) (0.030) (0.087) (0.016) (0.091)
Mean o dep. a . 0.897 0.030 0.980 0.113 0.010 0.147
Obse a ions 9,441 7,678 7,806 7,806 7,806 7,296
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Placebo ea men g oup 0.029** 0.022 -0.064 -0.012 0.035 0.191
(0.013) (0.127) (0.096) (0.030) (0.099) (0.498)
Mean o dep. a . 0.028 0.321 0.591 0.941 0.249 2.729
Obse a ions 7,175 7,343 7,370 7,410 7,344 7,806
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Placebo ea men g oup -0.009 0.032 0.026 0.043 0.029 0.019
(0.040) (0.061) (0.051) (0.071) (0.051) (0.045)
Mean o dep. a . 0.896 0.355 0.141 0.935 0.355 0.399
Obse a ions 25,509 24,817 24,817 7,678 7,044 24,817
Demog aphic cha ac e is ics Neighbo hood cha ac e is ics
Female Age U ban e i o y Road densi y Ele a ion T ee co e age
(19) (20) (21) (22) (23) (24)
Placebo ea men g oup 0.010 0.945 0.016 -9.439 -47.112 -0.030
(0.017) (1.898) (0.051) (7.260) (48.435) (0.020)
Mean o dep. a . 0.545 31.25 0.938 19.48 688.7 0.0144
Obse a ions 27,865 27,865 63 63 63 63
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (2), using he loca ions o majo oads
ha did no con ibu e o he o ma ion o he bounda ies o gang e i o y as a placebo. We addi ionally con ol o dummies o he ou
qua iles o he census ac size dis ibu ion. The uni o obse a ion is a dwelling, household, indi idual, o census ac , depending on which
cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols include
a dummy o gang e i o y as well as a linea end in dis ance o he placebo bounda ies, sepa a ely o loca ions on each side o he placebo
bounda ies and on each side o he bounda ies o gang e i o y. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he
dis ance o he placebo bounda ies (sepa a ely o each side o he bounda ies).
82
TABLE S.IV
SOCIOECONOMIC CONDITIONS AFTER EXPOSURE TO GANG CONTROL,
SUBSAMPLE OF INDIVIDUALS WHO HAVE ALWAYS LIVED IN SAN SALVADOR
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.047*** 0.026** -0.058** -0.076*** 0.005*** -0.132***
(0.015) (0.010) (0.023) (0.019) (0.002) (0.031)
Mean o dep. a . 0.932 0.028 0.934 0.105 0.005 0.178
Obse a ions 72,087 60,675 38,926 38,926 38,926 37,147
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.019** -0.225*** -0.145*** -0.024*** -0.179*** -0.734***
(0.007) (0.044) (0.033) (0.006) (0.037) (0.186)
Mean o dep. a . 0.036 0.426 0.683 0.955 0.345 3.048
Obse a ions 36,679 37,328 37,414 37,542 37,292 38,926
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.027*** -0.151*** -0.120*** -0.036*** -0.094*** -0.098***
(0.006) (0.029) (0.028) (0.012) (0.019) (0.020)
Mean o dep. a . 0.931 0.445 0.201 0.952 0.374 0.520
Obse a ions 156,627 152,953 152,953 60,675 36,147 152,953
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he subsample o indi iduals
who ha e always li ed in San Sal ado . Fo he dwelling cha ac e is ics, none o he obse a ions a e excluded because all he dwellings ha e
always been loca ed in San Sal ado . Fo he household cha ac e is ics, we limi he sample o hose obse a ions o which he head o he
household has always li ed in San Sal ado . All he a iables come om he 2007 census. The uni o obse a ion is a dwelling, household,
o indi idual, depending on which cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e
popula ion. Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he
bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely
o each side o he bounda ies).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 83
TABLE S.V
TWO-DIMENSIONAL REGRESSION DISCONTINUITY IN LATITUDE AND LONGITUDE
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.051*** 0.009* -0.006 -0.076*** 0.004*** -0.141***
(0.007) (0.005) (0.015) (0.008) (0.001) (0.011)
Mean o dep. a . 0.932 0.028 0.941 0.108 0.005 0.181
Obse a ions 72,087 60,675 62,169 62,169 62,169 59,776
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.007** -0.256*** -0.175*** -0.024*** -0.199*** -0.806***
(0.002) (0.021) (0.017) (0.003) (0.017) (0.087)
Mean o dep. a . 0.033 0.429 0.697 0.952 0.346 3.093
Obse a ions 59,096 60,045 60,168 60,384 60,020 62,169
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.026*** -0.161*** -0.141*** -0.028*** -0.104*** -0.109***
(0.004) (0.012) (0.012) (0.006) (0.009) (0.009)
Mean o dep. a . 0.928 0.449 0.208 0.952 0.378 0.522
Obse a ions 208,416 202,935 202,935 60,675 58,293 202,935
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o he a iables om he 2007
census, using la i ude and longi ude as he o cing a iables. The uni o obse a ion is a dwelling, household, o indi idual, depending on
which cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols
include a linea end in la i ude and longi ude (demeaned), sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en-
heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
84
TABLE S.VI
EXCLUDING 10% OF THE OBSERVATIONS WITH THE HIGHEST VALUES
OF THE 1ST PRINCIPAL COMPONENTS FROM NONGANG AREAS
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.042** 0.023** -0.047** -0.031* 0.005*** -0.064***
(0.016) (0.010) (0.022) (0.017) (0.002) (0.024)
Mean o dep. a . 0.929 0.030 0.939 0.081 0.005 0.143
Obse a ions 69,008 57,596 59,569 59,569 59,569 57,176
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.002 -0.165*** -0.116*** -0.018*** -0.124*** -0.500***
(0.006) (0.046) (0.033) (0.006) (0.033) (0.185)
Mean o dep. a . 0.028 0.402 0.682 0.950 0.316 2.980
Obse a ions 56,496 57,445 57,568 57,784 57,420 59,569
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.026*** -0.103*** -0.040* -0.032** -0.057*** -0.055***
(0.007) (0.028) (0.022) (0.012) (0.018) (0.019)
Mean o dep. a . 0.924 0.421 0.169 0.949 0.359 0.498
Obse a ions 199,162 193,681 193,681 57,596 55,693 193,681
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) a e excluding 10% o he obse -
a ions wi h he highes le els o he i s p incipal componen om nongang a eas. Fo he dwelling cha ac e is ics, we use he i s p incipal
componen o he dwelling cha ac e is ics; o he household cha ac e is ics— he i s p incipal componen o he household cha ac e is ics;
o he indi idual cha ac e is ics— he i s p incipal componen o he indi idual cha ac e is ics. When mo e han 10% o obse a ions had he
i s p incipal componen less han o equal o he alue o he 10 h pe cen ile, we exclude a andom subse o obse a ions o which he i s
p incipal componen is exac ly equal o he 10 h pe cen ile. The es ima es do no depend on which subsample o obse a ions a e excluded.
In pa icula , we pe o m 1,000 i e a ions o his p ocedu e, and o each a iable epo he mos conse a i e esul s, i.e., when hey a e leas
signi ican . All he a iables come om he 2007 census. The uni o obse a ion is a dwelling, household, o indi idual, depending on which
cha ac e is ics a e being conside ed. In he indi idual-le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols include
a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en-
heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).

GANGS, LABOR MOBILITY, AND DEVELOPMENT 85
TABLE S.VII
HOUSING RENT
Log o housing en Housing en
(1) (2)
Gang e i o y -0.191*** -203.20***
(0.052) (56.33)
Numbe o ooms in he apa men :
2 ooms 0.210*** 19.93
(0.053) (30.79)
3 ooms 0.296*** 87.65**
(0.059) (42.09)
4 ooms 0.189** 33.14
(0.070) (73.21)
5 ooms 0.134 2.46
(0.107) (124.27)
6 ooms 0.383*** 330.19**
(0.089) (148.86)
7+ ooms 0.365*** 378.31*
(0.124) (194.71)
Numbe o ba h ooms in he apa men :
2 ba h ooms 0.507*** 209.67***
(0.073) (49.22)
3 ba h ooms 0.718*** 350.97***
(0.062) (46.61)
4 ba h ooms 0.836*** 473.41***
(0.066) (82.91)
5 ba h ooms 0.992*** 650.37***
(0.080) (130.00)
6 ba h ooms 1.095*** 1,028.51***
(0.113) (213.85)
7+ ba h ooms 0.979*** 786.86***
(0.160) (233.44)
Squa e me e s 0.140*** 190.59***
(0.018) (22.68)
Squa e me e s squa ed -0.003*** -4.29***
(0.000) (0.61)
Ren ed ou by an agency 0.269*** 242.29***
(0.034) (15.55)
Mean dep. a 6.731 1,008.81
Obse a ions 1,537 1,537
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) o housing en eques ed by
landlo ds, con olling o he cha ac e is ics o he apa men s ha a e being en ed ou . The uni o obse a ion is an apa men . Omi ed
con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d
e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he
bounda ies).
86
TABLE S.VIII
ESTIMATING THE EFFECTS SEPARATELY FOR MS-13 AND 18TH STREET
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
MS-13 -0.051*** 0.024** -0.058** -0.079*** 0.006*** -0.141***
(0.017) (0.010) (0.025) (0.021) (0.001) (0.031)
18 h S ee -0.044** 0.027** -0.045** -0.078*** 0.005* -0.126***
(0.017) (0.011) (0.021) (0.022) (0.003) (0.031)
Mean o dep. a . 0.932 0.028 0.941 0.108 0.005 0.181
Obse a ions 72,087 60,675 62,169 62,169 62,169 59,776
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
MS-13 -0.015** -0.242*** -0.163*** -0.025*** -0.198*** -0.829***
(0.006) (0.050) (0.034) (0.006) (0.039) (0.194)
18 h S ee -0.012* -0.187*** -0.119*** -0.019*** -0.159*** -0.615***
(0.006) (0.049) (0.036) (0.006) (0.037) (0.212)
Mean o dep. a . 0.033 0.429 0.697 0.952 0.346 3.093
Obse a ions 59,096 60,045 60,168 60,384 60,020 62,169
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
MS-13 -0.036*** -0.179*** -0.145*** -0.036*** -0.102*** -0.119***
(0.007) (0.030) (0.027) (0.012) (0.021) (0.020)
18 h S ee -0.029*** -0.138*** -0.108*** -0.036** -0.082*** -0.091***
(0.008) (0.031) (0.027) (0.013) (0.021) (0.021)
Mean o dep. a . 0.928 0.449 0.208 0.952 0.378 0.522
Obse a ions 208,416 202,935 202,935 60,675 58,293 202,935
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) wi h he dummy o gang e i o y
eplaced wi h wo dummies o a eas con olled by MS-13 and a eas con olled by 18 h S ee . All he a iables come om he 2007 census.
The uni o obse a ion is a dwelling, household, o indi idual, depending on which cha ac e is ics a e being conside ed. In he indi idual-le el
eg essions, he sample consis s o he en i e popula ion. Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y,
sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o
he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 87
TABLE S.IX
EXCLUDING AREAS WITHIN 150 METERS OF THE RIVAL GANG
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Gang e i o y -0.041*** 0.025** -0.060*** -0.076*** 0.004*** -0.123***
(0.015) (0.010) (0.020) (0.020) (0.001) (0.027)
Mean o dep. a . 0.942 0.027 0.939 0.122 0.003 0.206
Obse a ions 60,187 50,742 51,933 51,933 51,933 49,948
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Gang e i o y -0.012** -0.191*** -0.122*** -0.021*** -0.161*** -0.612***
(0.006) (0.044) (0.031) (0.006) (0.032) (0.192)
Mean o dep. a . 0.035 0.475 0.734 0.958 0.383 3.249
Obse a ions 49,271 50,178 50,306 50,480 50,144 51,933
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Gang e i o y -0.030*** -0.151*** -0.117*** -0.034*** -0.083*** -0.098***
(0.007) (0.028) (0.024) (0.011) (0.018) (0.019)
Mean o dep. a . 0.932 0.475 0.231 0.957 0.397 0.540
Obse a ions 174,465 169,910 169,910 50,742 48,619 169,910
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) a e excluding gang-con olled
neighbo hoods ha a e loca ed wi hin 150 me e s o he i al gang. The uni o obse a ion is a dwelling, household, o indi idual, depending
on which cha ac e is ics a e being conside ed. All he a iable come om he 2007 census. In he indi idual-le el eg essions, he sample
consis s o he en i e popula ion. Omi ed con ols include a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions
on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing he dis ance o he bounda ies o gang
e i o y (sepa a ely o each side o he bounda ies).
88
TABLE S.X
ISLANDS OF GANG TERRITORY
Dwelling cha ac e is ics Household cha ac e is ics
Walls made Ba e loo Has sewe age Use elec ici y o No ba h oom Has in e ne
o conc e e in as uc u e ligh ing and cooking
(1) (2) (3) (4) (5) (6)
Island o gang e i o y -0.029** 0.023** -0.084** -0.065*** 0.006*** -0.103***
(0.013) (0.009) (0.038) (0.020) (0.001) (0.030)
Res o gang e i o y -0.057*** 0.027** -0.028 -0.087*** 0.006* -0.148***
(0.020) (0.010) (0.028) (0.022) (0.003) (0.030)
Mean o dep. a . 0.932 0.028 0.941 0.108 0.005 0.181
Obse a ions 72,087 60,675 62,169 62,169 62,169 59,776
Household cha ac e is ics
Has a mo o cycle Has a ca Has a phone Has a TV Has a compu e Numbe o ooms
(7) (8) (9) (10) (11) (12)
Island o gang e i o y -0.011* -0.216*** -0.130*** -0.018*** -0.167*** -0.709***
(0.006) (0.050) (0.029) (0.005) (0.038) (0.189)
Res o gang e i o y -0.014** -0.202*** -0.139*** -0.024*** -0.177*** -0.684***
(0.006) (0.048) (0.037) (0.007) (0.037) (0.203)
Mean o dep. a . 0.033 0.429 0.697 0.952 0.346 3.093
Obse a ions 59,096 60,045 60,168 60,384 60,020 62,169
Indi idual cha ac e is ics 1s p incipal componen o he:
Can ead Has a high Has a uni e si y Dwelling Household Indi idual
and w i e school deg ee deg ee cha ac e is ics cha ac e is ics cha ac e is ics
(13) (14) (15) (16) (17) (18)
Island o gang e i o y -0.040*** -0.194*** -0.148*** -0.026** -0.087*** -0.127***
(0.007) (0.028) (0.025) (0.010) (0.020) (0.019)
Res o gang e i o y -0.026*** -0.125*** -0.104*** -0.043*** -0.091*** -0.084***
(0.007) (0.033) (0.028) (0.014) (0.020) (0.022)
Mean o dep. a . 0.928 0.449 0.208 0.952 0.378 0.522
Obse a ions 208,416 202,935 202,935 60,675 58,293 202,935
No e: *** p<0.01, ** p<0.05, * p<0.1. The able p esen s he esul s o es ima ing Speci ica ion (1) wi h he dummy o gang e i o y
eplaced wi h dummies o he islands o gang e i o y and o he o he gang-con olled loca ions. All he a iables come om he 2007
census. The uni o obse a ion is a dwelling, household, o indi idual, depending on which cha ac e is ics a e being conside ed. In he
indi idual-le el eg essions, he sample consis s o he en i e popula ion. Omi ed con ols include a linea end in dis ance o he bounda ies
o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s in pa en heses a e clus e ed by 30 me e bins, deno ing
he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 95
FIGURE S.4.—Socioeconomic Condi ions Be o e he Gangs’ A i al: Indi idual Cha ac e is ics
No e: The igu e illus a es he esul s o he indi idual cha ac e is ics om Table II. All he a iables come om he 1992 census. The
uni o obse a ion is an indi idual. All he a iables ep esen he sha e o indi iduals ha ha e he ou come a iable (can ead and w i e, ha e
a high school deg ee, e c.). The e ical axis ep esen s he a e age alue o he ou comes a iable; he ho izon al axis—dis ance (in me e s)
o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side o gang e i o y; a eas o he igh a e
con olled by he gangs. The do s ep esen he a e age alue o he ou come a iable in ha 30 me e bin.

96
FIGURE S.5.—In-sample Mig a ion Is No D i ing he Resul s: Educa ion and Fi m Size
No e: The igu e illus a es he esul s om Appendix Table A.I. The le -hand side o he igu e p esen s he esul s o he ull sample
(Panel A o Appendix Table A.I), he igh -hand side— o he subsample o indi iduals who ha e li ed in he same loca ion all hei li e (Panel
B o Appendix Table A.I). The esul s a e e y simila . The e ical axis ep esen s he a e age alue o he ou come a iable; he ho izon al
axis—dis ance (in me e s) o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side o gang e i o y;
a eas o he igh a e con olled by he gangs. The do s ep esen he a e age alue o he ou come a iable in ha 30 me e bin.
GANGS, LABOR MOBILITY, AND DEVELOPMENT 97
FIGURE S.6.—Al e na i e Bin Size
Panel A: 60-Me e Bins
Panel B: 20-Me e Bins
No e: The igu e illus a es he eg ession discon inui y plo s o he 1s p incipal componen s o he dwelling, household, and indi idual
cha ac e is ics om he 2007 census, using a di e en bandwid h han in he baseline speci ica ion. In Panel A, he do s ep esen he a e age
alue o he ou come a iable o 60 me e bins. In Panel B, he do s ep esen he a e age alue o he ou come a iable o 20 me e bins.
The uni o obse a ion is a dwelling, a household, and an indi idual, depending on he speci ica ion. All he a iables a e no malized o a y
be ween ze o and one wi h highe alues ep esen ing be e ou comes. The e ical axis ep esen s he a e age alue o he ou comes a iable;
he ho izon al axis—dis ance (in me e s) o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side
o gang e i o y; a eas o he igh a e con olled by he gangs.
98
FIGURE S.7.—Housing Ren
No e: The igu e illus a es he eg ession discon inui y plo s o he esidual o housing en and log housing en a e sub ac ing he
e ec s o all he con ols. The uni o obse a ion is an apa men lis ing. The e ical axis ep esen s he a e age alue o he ou comes a iable;
he ho izon al axis—dis ance (in me e s) o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side
o gang e i o y; a eas o he igh a e con olled by he gangs. Omi ed con ols include dummies o he numbe o ooms, dummies o he
numbe o ba h ooms, a quad a ic polynomial in squa e me e s, a dummy o whe he he apa men is being en ed ou by an agency a he
han an indi idual, and a linea end in dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies.
FIGURE S.8.—Gang Bo de Pe mu a ions
No e: The igu e p esen s he esul s o a pe mu a ion es , illus a ing how he eg ession discon inui y es ima es o he h ee main
ou come a iables change when he bo de s o gang e i o y a e shi ed in ei he di ec ion. Each es ima e comes om a sepa a e eg ession
speci ica ion. The x-axis cha ac e izes he shi in he bounda ies o gang e i o y (in me e s). Omi ed con ols include a linea end in
dis ance o he bounda ies o gang e i o y, sepa a ely o loca ions on each side o he bounda ies. S anda d e o s a e clus e ed by 30 me e
bins, deno ing he dis ance o he bounda ies o gang e i o y (sepa a ely o each side o he bounda ies).
GANGS, LABOR MOBILITY, AND DEVELOPMENT 99
FIGURE S.9.—A ailabili y and Quali y o Public Goods
No e: The op pa o he igu e p esen s he eg ession discon inui y plo s o he numbe o hospi als and schools pe squa e kilome e ;
he lowe pa o he igu e p esen s he eg ession discon inui y plo s o he ques ions abou sa is ac ion wi h he a ailabili y and quali y o
public goods om he 2019 su ey. In he op pa o he igu e, he uni o obse a ion is a 10 me e bin, deno ing dis ance o he bounda ies o
gang e i o y. In he lowe pa o he igu e, he uni o obse a ion is an indi idual. Fo he ques ions abou sa is ac ion wi h he a ailabili y
and quali y o public goods, he esponden s we e asked o a e he a ailabili y and quali y o public goods on a scale om 1 (ex emely
unsa is ied) o 7 (ex emely sa is ied). The e ical axis ep esen s he a e age alue o he ou comes a iable; he ho izon al axis—dis ance
(in me e s) o he bounda ies o gang e i o y. Neighbo hoods o he le o he dashed line a e loca ed ou side o gang e i o y; a eas o he
igh a e con olled by he gangs. The do s ep esen he a e age alue o he ou come a iable in ha 30 me e bin.