Security, Trade, and Political Violence - dse.univr.it

Security, Trade, and Political Violence

Francesco Amodio1 Leonardo Baccini1

Michele Di Maio2

1McGill University2University of Naples “Parthenope”

SSDEV 2017Prato

June 19, 2017

Introduction

I Issues of trade and security dominate the current policy debate

I Interlinkages exist between the two

I Governments often implement security-motivated restrictions on themovement of goods and people:

I U.S. bans exports of arms to and imports of charcoal from SomaliaI China bans export of products and technologies to North Korea

I Can security-motivated trade restrictions increase violence?I Trade barriers can have negative economic consequencesI Decrease in opportunity cost of violence, higher grievance

I Answer will depend onI Nature of restrictions × structural composition.

This Paper

I We study the case of the Israeli dual-use list

I Introduced by the Israeli government on December 31st, 2007

I Severe restrictions on imports to the West Bank of a detailed set ofmaterials that can be used for military application

I We compare economic and political outcomes across industrial sectorsand localities over time

I We find that:I Industrial output and wages decrease differentially more in

dual-use input intensive sectors after 2008I Labor market outcomes worsen differentially in localities where

employment is more concentrated in these sectorsI Incidence of political violence is differentially higher in these

same localities.

This Paper

I We identify these effects using a difference-in-differences strategy

I We use data on firms, labor market outcomes, and political violence inthe Occupied Palestinian Territories (OPT) from 1999 to 2014

I We derive baseline measures of dual-use input intensity andemployment concentration using

I US 2002 Input-Output matrixI 1997 Palestinian Census

I The Gaza Strip, under full embargo from 2007 to 2010, as placebo.

This Paper

I We provide causal evidence of one specific mechanism through whichthe list increases political violence

I Negative effect on the OPT economy

I This mechanism accounts forI 4.5% loss in the total value of industrial outputI 17.6% of events of political violence in the West Bank in 2008-2014

I Although ours is not a complete policy evaluation, these resultshighlight important interactions between security and trade policy

I Need for an integrated policy approach.

Literature and Contribution

I Unsettled debate on poverty and terrorism and conflictI Krueger and Maleckova (2003), Krueger and Laitin (2008), Berman

et al. (2011), Benmelech et al. (2012)I Blair et al. (2013) and Fair et al. (2016): poverty not correlated

with support for militant organizations

I Trade and conflictI Mansfield (1994), Mansfield and Pevehouse (2000), Martin et al.

(2008a,b), Schneider (2007)I We provide an empirical microfoundation of the link between

barriers to trade and conflict

I Economic shocks and conflictI We focus on a policy over which governments have direct control.

Literature: Economic shocks and conflict

I Do income shocks affect conflict outbreak and/or incidence?I Positive shock: higher gains from appropriationI Negative shocks: scarcity and higher incentives to fightI Labor markets and opportunity cost of fighting

I Rainfall as source of income shock(Miguel, Satyanath, and Sergenti 2004, Ciccone 2011)

Literature: Economic shocks and conflict

I More recent contributions exploit variation in international commodityprices

I Specificationyct = δt + γc + β sc × pt + uct (1)

I yct is conflict in geographical unit c at time t

I sc is source of cross-sectional variationI Export commodity of country c

(Bruckner and Ciccone 2010, Bazzi and Blattman 2014)I Agricultural specialization of region c (Berman and Couttenier 2015)I Agricultural suitability of region c (Dube and Vargas 2013)I Presence of mines in region c (Berman et al. 2014)

I pt is price index at time tI δt and γc are time and unit fixed effectsI Main identifying assumption: fluctuation in international prices is

exogenous to local conflict.

Outline

1. Introduction

2. The setting

3. Data

4. Empirical strategy

5. Results

6. Conclusion

The Setting

I The economy of the OPT is strictly dependent on the Israeli one

I 70% of imports coming from Israel in 2006I 15% of Palestinian workers commuting daily to Israel

I Security measures imposed by the Israeli Defense Forces (IDF) have anegative impact on the economy

(Calı and Miaari 2013, Amodio and Di Maio 2017)

I Restrictions on trade of dual-use items are particularly important

I Goods, services, or technologies that are intended for civilian use,but also have military applications

I Trade regulated by several international treaties.

The Israeli Dual-use List

I First restrictions to trade of chemicals introduced in 1976

I List expanded during the Second Intifada (2000-2006)

I Defense Export Control Law of 2007

I Regulatory framework: restrictions become systematicI Bill enacted by the Israeli Parliament on December 31, 2007

I 56 items: chemicals, fertilizers, raw materials for industry, steel pipes,milling machines, optical equipment

List

I Entry monitored by the Trade and Industry Department of the CivilAdministration (TIDCA)

I Palestinian importers need to obtain a licenseI Application to be repeated for every truck loadI Average application time varies from 4 to 8 weeksI Each license lasts 21 days (TIDCA 2012).

Anecdotal Evidence

I National Aluminum and Profile Company (NAPCO)

I Industrial aluminum firmI The list includes inputs essential for aluminum anodizing

(oxidizations) and nitrationI Forced to complete the required processing steps in IsraelI Lower margins: cut wages in response.

I Pal Karm Company

I Leading industrial cosmetics firmI The list includes glycerine, essential production inputI Israeli Health Authorities require glycerine to be part of final

productI Between 2008 and 2010, the company estimates a 30% drop in

exports to Israel.

Data

I Palestinian Industry Survey 1999-2012 (PCBS)I Repeated cross-section of Palestinian establishments in the

manufacturing sector (33,000+ observations)I Information on sector of activity (ISIC-4)

I US Input-Output Matrix 2002 (BEA)

I Palestinian Labor Force Survey 1999-2012 (PCBS)I Rotating panel, data aggregated at the locality level (570 localities)

I Palestinian Census 1997 (PCBS)I Employment figures per locality per sector (ISIC-2)

I ICEWS 1999-2014 (LMATL)I Integrated Crisis Early Warning SystemI Records any interaction between socio-political actorsI We identified hostile and violent interactions

I UN-OCHA 2004-2012I Checkpoints, observation towers and road blocks.

Dual-use Input Intensity

I 2002 US economy as benchmark

I We identify the 10-digit HS code of each item in the list, and linked it toBEA commodity codes

More

I For each ISIC 4-digit sector s we compute

ms =1ns

∑i∈s

dis (2)

I ns is the number of commodities i in sector sI di is value of dual-use inputs needed to deliver a dollar of i

I Captures dual-use input requirements of each sector.

Employment Concentration in Dual-use Input Intensive Sectors

I Palestinian Census 1997 as benchmark

I For each locality l we compute

ml =∑

s

Lls ms

Ll (3)

I Ll is the total number of workers in locality l in 1997I Ll

s is number of workers in sector sI ms is sector-level measure of dual-use input intensity

I Captures employment concentration in dual-use input intensiveindustries at baseline.

Difference-in-differences

I We compare economic and political outcomes across industrial sectorsand localities over time

I More vs. less dual-use input intensive sectors (ms)I More vs. less employment in dual-use input intensive sectors (ml)I Before and after 2008

I Identifying assumption

I Had the list not been issued, economic and political outcomeswould have not evolved differentially according to ms or ml

I We derive these measures using 2002 US and 1997 OPT as benchmark

I Measures themselves are not informed by the list.

ms Across Sectors

ISIC 4 ms DescriptionLeast Intensive Sectors

1600 0.0001 Manufacture of tobacco products1532 0.0001 Manufacture of starches and starch products1543 0.0002 Manufacture of cocoa, chocolate and sugar confectionery1542 0.0003 Manufacture of sugar1554 0.0010 Manufacture of soft drinks; production of mineral waters1549 0.0013 Manufacture of other food products n.e.c.1553 0.0014 Manufacture of malt liquors and malt

Most Intensive Sectors

2922 0.4142 Manufacture of machine tools2732 0.4343 Casting of non-ferrous metals2731 0.4343 Casting of iron and steel2696 0.4687 Cutting, shaping and finishing of stone3592 0.4911 Manufacture of bicycles and invalid carriages2411 0.4930 Manufacture of basic chemicals2421 0.5637 Manufacture of pesticides and other agrochemical products

ml Across Localities

Notes. The Figure shows the location of each locality in both the West Bank and the Gaza Strip. Colorscorrespond to the degree of intensity in dual-use inputs in each location according to their quintile of thedistribution of the ml variable, from yellow to red (Sources: BEA, PCBS).

ml and Locality Characteristics

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)Variables ml ml ml ml ml ml ml ml ml ml ml

Population 0.0001(0.000)

Daily wage 0.000(0.000)

Working days (per month) -0.000(0.002)

Share of manufacturing 0.041(0.031)

Share of agriculture 0.046(0.048)

Share of workers in public sector -0.065(0.090)

Share of self-employed workers 0.036(0.082)

Unemployment -0.373(0.281)

Out of the labor force -0.186(0.129)

Non-schooling 0.056(0.089)

High education -0.021(0.046)

Observations 187 187 187 187 187 187 187 187 187 187 187R-squared 0.000 0.003 0.000 0.011 0.011 0.007 0.002 0.014 0.030 0.003 0.001

Notes. Unit of observation is a locality surveyed in 1999 (Sources: BEA, PCBS Labor Force Survey).Summary Stats

Sector-level Analysis

I Difference-in-differences specification:

yst = δt + γs + β ms × Post2008t + ust (4)

I yst is outcome of sector s in year tI ms is benchmark dual-use input intensityI Post2008t is equal to one after 2008I δt and γs are year and sector fixed effectsI Clustering at the sector level

I β captures differential changes in yst after 2008 according to dual-useinput intensity.

Sector-level Analysis: Results

Output Value Output Value Price Output Wages4-digit PPI

ms × Post2008t -0.704** -0.646** 0.044 -0.691*** -1.428***(0.303) (0.257) (0.110) (0.242) (0.325)

Year FE Yes Yes Yes Yes YesSector FE Yes Yes Yes Yes Yes

Observations 1039 607 619 607 946R2 0.893 0.884 0.789 0.872 0.924

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis. Unit of observationis a 4-digit sector in a year. ms is intensity of each sector in dual-use inputs as derived from US Input-Outputmatrix. All dependent variables are in log. Post2008 is a dummy equal to 1 for observations belonging to theyear 2008 or after. Observations are weighted by the number of establishments per sector. Standard errors areclustered at the 4-digit sector level (Sources: BEA, PCBS Industry Survey).

Wages Across Sectors

-6-4

-20

2C

oeffi

cien

t Est

imat

es

intd2

004

intd2

005

intd2

006

intd2

007

intd2

008

intd2

009

intd2

010

intd2

012

Effects with Respect to Intensity

West Bank

Notes. Dependent variable is the log of wages. The Figure plots the estimated coefficient of the interaction ofthe dual-use input intensity variable ms with the corresponding year dummy. The solid vertical lines showthe 95% confidence interval of each estimate, while the dash horizontal line indicates zero (Sources: BEA,Industry Survey).

Discussion

I Output and wages decrease differentially after 2008 in dual-use inputintensive sectors

I 25th to 75th percentile of ms: 11% differential loss in output value,22% differentially lower wages

I No effect after 2010: firm exit or change in technologyI No change in prices: dual-use input intensive sectors face lowly

elastic demand

I No evidence of pre-trends

I Support for the identifying assumptionI Rules out concerns that list composition was informed by

willingness to hurt rising (or declining) sectors

I Robustness

I Robust to linear or quadratic sector-specific trendsI Not driven by export or import intensityI No effect in the Gaza Strip. More

Local Labor Markets

I Difference-in-differences specification:

ylt = δt + γl + β ml × Post2008t + ult (5)

I ylt is outcome of locality l in year tI ml is baseline employment concentration in dual-use input

intensive industriesI Post2008t is equal to one after 2008I δt and γl are year and locality fixed effectsI Clustering at the locality level

I β captures differential changes in ylt after 2008 according toemployment concentration in dual-use input intensive industries.

Local Labor Markets: Results

Daily Wage(1) (2) (3) (4) Log

ml × Post2008t -15.988 -18.953** -33.501* -20.538* -0.198*(10.285) (9.546) (17.611) (11.162) (0.113)

Share of Manuf 18.985*** 13.723*** 13.906** 0.242***(4.495) (4.411) (6.772) (0.053)

Share of Agric -7.661 -5.475 -15.001*** -0.111(5.313) (5.184) (5.641) (0.075)

Trends No No Yes No NoObstacles No No No Yes No

Year FE Yes Yes Yes Yes YesLocality FE Yes Yes Yes Yes Yes

Observations 2769 2571 2571 1585 2571R2 0.723 0.730 0.854 0.772 0.732

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis, clustered at the localitylevel. Unit of observation is an OPT locality in which was surveyed in the Labor Force Survey a year. Dependentvariable is average daily wage among employed individuals surveyed in the locality. Observations are weightedaccording to estimated sampling probabilities and surveyed population in each location (Sources: BEA, PCBS LaborForce Survey).

Wages Across Localities

-60

-40

-20

020

40Ef

fect

s on

Lin

ear P

redi

ctio

n

intd2

004

intd2

005

intd2

006

intd2

007

intd2

008

intd2

009

intd2

010

intd2

011


West Bank

Notes. Dependent variable is the daily wage in the locality. The Figures plot the estimated coefficient of theinteraction of the dual-use input intensity variable ml with the corresponding year dummy. The solid verticallines show the 95% confidence interval of each estimate, while the dash horizontal line indicates zero (Sources:BEA, Labor Force Survey).

Discussion

I Wages decrease differentially after 2008 in localities where employmentis more concentrated in dual-use intensive industries

I 25th to 75th percentile of ml: 1% lower average wagesI Ignoring reallocation, this is consistent with 4.5% of labor force

experiencing 22% wage lossI Pattern of effect by year exactly mirrors the one for sectorsI Non-significant positive effect on unemployment and negative on

days worked in a month More

I As before, no evidence of pre-trends

I Further support for the identifying assumptionI Rules out concerns that list composition was informed by

willingness to target specific localities where wages were on therise (or declining)

I Robustness

I No effect in the Gaza Strip. More

Political Violence in ICEWS Dataset

I Hostile and violent eventsI OPT (non-government) as sourceI OPT or Israel as targetI 19,982 events between 1999 and 2014.

Political Violence: Results

Number of Violent Events(1) (2) (3) (4) (5)

Poisson Log

ml × Post2008t 1.671** 2.008** 2.575* 7.850*** 0.061*(0.759) (1.009) (1.538) (1.224) (0.036)

Trends No Yes No No NoObstacles No No Yes No No


Observations 7488 7488 3600 1728 7488R2 0.661 0.785 0.687 0.798

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis. Unit of observationis an OPT locality in a year. ml is intensity of each locality in dual-use inputs as derived from the US Input-Output matrix and employment in the 1997 Population Census. Post2008 is a dummy equal to 1 forobservations belonging to the year 2008 or after. Standard errors are clustered at the locality level (Sources:BEA, PCBS, ICEWS, UN-OCHA).

Political Violence Across Localities: Effect by Year

-20

24

Effe

cts

on L

inea

r Pre

dict

ion

intd2

004

intd2

005

intd2

006

intd2

007

intd2

008

intd2

009

intd2

010

intd2

011

intd2

012

intd2

013

intd2

014


West Bank

Notes. Dependent variable is the number of violent events in the locality. The Figures plot the estimatedcoefficient of the interaction of the dual-use input intensity variable ml with the corresponding year dummy.The solid vertical lines show the 95% confidence interval of each estimate, while the dash horizontal lineindicates zero (Sources: BEA, PCBS, ICEWS).

Discussion

I Political violence increases differentially after 2008 in localities whereemployment is more concentrated in dual-use intensive industries

I 25th to 75th percentile of ml: 8% increase over the meanI Effect already in 2008: rational expectations, grievancesI Outlasts impact on wages: lock-in effect, self-reinforcing cycle

I Some evidence of pre-trendsI Non-significant, also when restricting the sample to pre-2008I If list composition was informed by willingness to decrease

violence where it was rising, ours would be a lower bound

I Robustness and effect heterogeneityI No effect in the Gaza StripI Robust to spatial spillovers More

I Significant effect when focusing on violence from individuals withno affiliation to particular groups

I Significant effect when focusing on high intensity violenceI Significant effect when target is OPT: destabilizing effect. More

Conclusion

I Security-motivated trade restrictions can increase threats to security

I The Israeli dual-use list as a quasi-experimentI Negative impact on dual-use intensive industriesI Negative impact on specialized local labor marketsI Increase in political violence in these localities

I Based on our estimates, we calculate that this mechanism accounts forI 4.5% loss in industrial output in 2008-2012I 17.6% of episodes of political violence in 2008-2014

I External validityI Labor as main input for violenceI Not a policy evaluation: focus on one specific mechanismI Israeli context as a though test

I Important interactions between security and trade policy.

Thank you!

Summary Statistics: West Bank

Variable Mean Std. Dev. Min. Max. N

ms 0.106 0.134 0 0.564 113

Output Value 18.851 1.455 8.034 20.899 1039Price 4.947 0.219 4.442 5.832 607Output 14.103 1.253 5.7 15.868 607Wages 8.315 1.666 1.675 11.676 924Import Intensity 0.387 0.296 0 1.635 74Export Intensity 0.611 1.563 0 13.378 113

Daily Wage 81.557 18.793 25 189.7 2865Monthly Days of Work 22.22 1.941 7 31 2865Unemployment Probability 0.083 0.04 0 0.302 2870Share of Manufacturing 0.307 0.141 0 1 2664Share of Agriculture 0.14 0.164 0 1 2664

Political Violence 0.92 9.44 0 255 8608Violence from New Fighters 0.742 7.67 0 220 8608High Intensity Violence 0.328 3.91 0 121 8608Low Intensity Violence 0.592 5.849 0 148 8608Violence against OPT 0.539 5.661 0 166 8608

Checkpoints 2.058 2.819 0 17 1841Observation Towers 1.032 1.505 0 8 1841Roadblocks 1.995 3.141 0 14 1841

ml 0.05 0.114 0 0.833 468

Notes. Output value, prices, output and wages per ISIC 4-digit sector in log. Values in NewIsraeli Shekel.

Summary Statistics: Gaza Strip

Variable Mean Std. Dev. Min. Max. N

Output Value 17.284 1.497 8.138 20.281 794Price 4.936 0.212 4.442 5.832 503Output 12.624 1.455 4.97 15.379 503Wages 7.952 1.566 2.308 11.226 601

Daily Wage 61.82 7.316 16.9 108.2 456Monthly Days of Work 23.597 0.882 12.9 29.3 456Unemployment Probability 0.111 0.03 0 0.28 456Share of Manufacturing 0.156 0.097 0 0.5 428Share of Agriculture 0.122 0.142 0 0.942 428

Political Violence 17.532 124.431 0 1632 688

ml 0.03 0.028 0 0.129 40

Notes. Output value, prices, output and wages per ISIC 4-digit sector in log. Values in NewIsraeli Shekel.

Back

HS Codes for Dual-use Inputs

HS Code Description Dual-use

2834210000 POTASSIUM NITRATES 12834290500 BISMUTH NITRATES 02834291000 CALCIUM NITRATES 12834292000 STRONTIUM NITRATES 0

3604109050 FIREWORKS & PYROTECHNICS, INCLUDING FLARES, IGNITERS, ETC 13604900000 OTHER PYROTECHNIC ARTICLES, NESOI 13605000000 MATCHES, EXC PYROTECHNIC ARTICLES OF HDG 3604 03605000030 MATCHES WITH NATURAL WOOD STEMS 0

3801100000 ARTIFICAL GRAPHITE 03801101000 ARTIFICIAL GRAPHITE PLATES ETC,FOR ELEC GENERATORS 1

8108903060 OTHER ARTICLES OF TITANIUM, NESOI 08108906030 TITANIUM BARS, RODS, PROFILES AND WIRE 08108906045 PLATES, SHEETS, STRIPS AND FOIL, TITANIUM 18108906060 TUBES AND PIPES, TITANIUM 1

Commodity Codes for Dual-use Inputs

Comm. Code Description Dual-use

331200 Steel product manufacturing from purchased steel 033131A Alumina refining and primary aluminum production 133131B Aluminum product manufacturing from purchased aluminum 1331411 Primary smelting and refining of copper 0

332913 Plumbing fixture fitting and trim manufacturing 033291A Valve and fittings other than plumbing 1

339920 Sporting and athletic goods manufacturing 1339930 Doll, toy, and game manufacturing 0339940 Office supplies (except paper) manufacturing 0

Back

Trade Intensity


ms × Post2008t -1.752*** -1.704*** 0.309 -2.020*** -2.233***(0.470) (0.504) (0.234) (0.459) (0.774)

fs × Post2008t 0.442 0.550 -0.462** 1.014* 0.296(0.507) (0.660) (0.232) (0.584) (0.312)

es × Post2008t 0.055** 0.056* -0.018 0.074*** 0.041(0.023) (0.029) (0.012) (0.026) (0.031)


Observations 878 593 599 593 815R2 0.886 0.885 0.801 0.875 0.925

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis, clustered at the 4-digitsector level. Unit of observation is a 4-digit sector-year. fs is import intensity calculated by dividing the valueof imported materials by total output value in each sector in 2000. es is export intensity calculated by dividingthe value of external sales by total output value in each sector in 2000. All dependent variables are in log.Observations are weighted by the number of establishments per sector. Standard errors are (Sources: BEA,PCBS Industry Survey).

Industries in the Gaza Strip: Results


ms × Post2008t -0.456 -0.899 -0.013 -0.900 0.089(0.742) (0.659) (0.110) (0.573) (0.460)


Observations 794 503 569 503 636R2 0.853 0.851 0.803 0.849 0.898

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis. Unit of observationis a 4-digit sector in a year. ms is intensity of each sector in dual-use inputs as derived from US Input-Outputmatrix. All dependent variables are in log. Post2008 is a dummy equal to 1 for observations belonging to theyear 2008 or after. Observations are weighted by the number of establishments per sector. Standard errors areclustered at the 4-digit sector level (Sources: BEA, PCBS Industry Survey).

Back

Unemployment in the West Bank

Unemployment Probability(1) (2) (3) (4)

ml × Post2008t 0.069 0.072 0.152** 0.055(0.051) (0.053) (0.061) (0.042)

Share of Manuf -0.059*** -0.044*** -0.022(0.013) (0.015) (0.015)

Share of Agric -0.019 -0.049*** -0.023*(0.013) (0.014) (0.014)

Locality Trends No No Yes NoObstacles No No No Yes

Year FE Yes Yes Yes YesLocality FE Yes Yes Yes Yes

Observations 2774 2574 2574 1587R2 0.536 0.554 0.741 0.608

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis. De-pendent variable is average probability of unemployment among individuals surveyed in thelocality. Observations are weighted according to the locality population size in 1997. Standarderrors are clustered at the locality level (Sources: BEA, PCBS Labor Force Survey).

Monthly Days of Work in the West Bank

Monthly Days of Work(1) (2) (3) (4) (5)

Log

ml × Post2008t -0.299 -1.162 -2.916 -0.035 -0.059(0.680) (0.987) (2.640) (1.124) (0.048)

Share of Manuf -6.372*** -5.955*** -6.133*** -0.308***(0.603) (0.705) (0.815) (0.030)

Share of Agric -2.899*** -1.997** -2.820*** -0.147***(0.691) (0.801) (0.700) (0.034)

Locality Trends No No Yes No NoObstacles No No No Yes No


Observations 2754 2571 2571 1570 2571R2 0.544 0.593 0.720 0.668 0.580

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis. Dependent variable isaverage monthly days of work among employed individuals surveyed in the locality. Observations are weightedaccording to estimated sampling probabilities and surveyed population in each location. Standard errors are clusteredat the locality level (Sources: BEA, PCBS Labor Force Survey).

Back

Labor Markets in the Gaza Strip: Results

Daily Wage(1) (2) (3) (4) Log

ml × Post2008t 15.166 -15.318 20.252 37.366 -0.261(78.007) (85.732) (85.760) (64.547) (1.418)

Share of Manuf -12.118 1.926 -11.789 -0.231(13.943) (11.817) (13.873) (0.207)

Share of Agric 4.422 3.582 -2.558 0.086(5.812) (5.248) (5.606) (0.092)

Trends No No Yes No NoObstacles No No No Yes No


Observations 447 420 420 221 420R2 0.502 0.514 0.778 0.628 0.526

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis, clustered at the localitylevel. Unit of observation is an OPT locality in which was surveyed in the Labor Force Survey a year. Dependentvariable is average daily wage among employed individuals surveyed in the locality. Observations are weightedaccording to estimated sampling probabilities and surveyed population in each location. Standard errors are(Sources: BEA, PCBS Labor Force Survey).

Back

Political Violence in the Gaza Strip: Results

Number of Violent Events(1) (2) (3) (4) (5)

Poisson Log

ml × Post2008t -13.460 -43.666 231.514 -0.154 -2.598(57.289) (152.397) (261.167) (2.913) (2.826)

Trends No Yes No No NoObstacles No No Yes No No


Observations 640 640 252 272 640R2 0.647 0.848 0.797 0.840

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis. Unit of obser-vation is an OPT locality in a year. ml is intensity of each locality in dual-use inputs as derived from theUS Input-Output matrix and employment in the 1997 Population Census. Post2008 is a dummy equal to1 for observations belonging to the year 2008 or after. Standard errors are clustered at the locality level(Sources: BEA, PCBS, ICEWS, UN-OCHA).

Political Violence in the West Bank: Spatial Lags

Number of Violent Events(1) (2) (3) (4) (5) (6)

n = 5 n = 5 n = 5 n = 10 n = 10 n = 10

ml × Post2008t 2.334* 1.943* 2.199* 2.649* 2.771* 2.731*(1.205) (1.179) (1.206) (1.406) (1.676) (1.632)

m−l × Post2008t -0.723 4.729 2.385 -2.747 -0.331 -0.966(4.005) (3.121) (3.234) (3.224) (2.803) (2.582)

Locality Trends No Yes No No Yes NoObstacles No No Yes No No Yes

Year FE Yes Yes Yes Yes Yes YesLocality FE Yes Yes Yes Yes Yes Yes

Observations 6384 6384 3591 6400 6400 3600R2 0.661 0.785 0.687 0.661 0.785 0.687

Notes. (* p-value< 0.1; ** p-value<0.05; *** p-value<0.01) Standard errors in parenthesis. Unit of observation isan OPT locality in a year. ml is intensity of each locality in dual-use inputs as derived from the US Input-Outputmatrix and employment in the 1997 Population Census. m−l is average dual-use input intensity in the closestn localities, where the value of n is indicated on top of each column. Post2008 is a dummy equal to 1 forobservations belonging to the year 2008 or after. Standard errors are clustered at the locality level (Sources: BEA,PCBS, ICEWS, UN-OCHA).

Back

High Intensity Violence

-10

12

Effe

cts

on L

inea

r Pre

dict

ion

intd2

004

intd2

005

intd2

006

intd2

007

intd2

008

intd2

009

intd2

010

intd2

011

intd2

012

intd2

013

intd2

014


West Bank

Notes. Dependent variable is the number of violent events with high-intensity (i.e. intensity=-10) in thelocality. The Figures plot the estimated coefficient of the interaction of the dual-use input intensity variable mlwith the corresponding year dummy. The solid vertical lines show the 95% confidence interval of eachestimate, while the dash horizontal line indicates zero (Sources: BEA, PCBS, ICEWS).

Perpetrators with No Affiliation

-10

12

3Ef

fect

s on

Lin

ear P

redi

ctio

n

intd2

004

intd2

005

intd2

006

intd2

007

intd2

008

intd2

009

intd2

010

intd2

011

intd2

012

intd2

013

intd2

014


West Bank

Notes. Dependent variable is the number of violent events perpetrated by new fighters in the locality. TheFigures plot the estimated coefficient of the interaction of the dual-use input intensity variable ml with thecorresponding year dummy. The solid vertical lines show the 95% confidence interval of each estimate, whilethe dash horizontal line indicates zero (Sources: BEA, PCBS, ICEWS).

Violence Against OPT Targets

-2-1

01

2Ef

fect

s on

Lin

ear P

redi

ctio

n

intd2

004

intd2

005

intd2

006

intd2

007

intd2

008

intd2

009

intd2

010

intd2

011

intd2

012

intd2

013

intd2

014


West Bank

Notes. Dependent variable is the number of violent events in the locality having the OPT as target. TheFigures plot the estimated coefficient of the interaction of the dual-use input intensity variable ml with thecorresponding year dummy. The solid vertical lines show the 95% confidence interval of each estimate, whilethe dash horizontal line indicates zero (Sources: BEA, PCBS, ICEWS).

Back

Security, Trade, and Political Violence - dse.univr.it

Documents

Transcript of Security, Trade, and Political Violence - dse.univr.it