Politicians, Criminality and Mining Booms in India Sam Asher and - - PowerPoint PPT Presentation

SLIDE 1

Politicians, Criminality and Mining Booms in India

Sam Asher and Paul Novosad March 24, 2015

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 1 / 20

SLIDE 2

Overview

Question: Does mineral wealth lead to bad political outcomes? Context: India, 1980-present Empirical strategy: Instrument local mineral wealth with geological deposits, local production, and global prices Outcomes: Election results, politician criminality Results

Mining booms lead to criminal politicians Electoral competition falls:

Bigger win margins Incumbency advantages increase

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 2 / 20

SLIDE 3

Motivation

Could resource wealth be bad for growth?

Economic Factors

Dutch Disease Volatility

Political factors

Rent-seeking Conflict

Why is resource extraction special?

Spatial concentration → ownership concentration High fixed cost, low variable cost: rents Highly regulated

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 3 / 20

SLIDE 4

Our approach

Focus on testing one mechanism of the political resource curse

Resource studies often measure the sum of multiple positive and negative effects Results do not rule out mechanisms in other direction Different mechanisms imply different solutions

Outcomes

Politician behavior Voter behavior

Context

Democracy Institutionally weak: significant corruption

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 4 / 20

SLIDE 5

Mechanisms for a political resource curse

Resource wealth shocks increase potential rents

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 5 / 20

SLIDE 6

Mechanisms for a political resource curse

Resource wealth shocks increase potential rents

Adverse Selection

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 6 / 20

SLIDE 7

Mechanisms for a political resource curse

Resource wealth shocks increase potential rents

Adverse Selection Election Success

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 7 / 20

SLIDE 8

Mechanisms for a political resource curse

Resource wealth shocks increase potential rents

Adverse Selection Election Success Moral Hazard

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 8 / 20

SLIDE 9

Mining in India

Half of districts have at least one large mineral deposit 2.5% of GDP Mix of private / public Taxes and royalties to state and federal government only

No local profit-sharing

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 9 / 20

SLIDE 10

Empirical strategy

Challenge: mineral wealth is largely static

Compensating differentials suggest mineral-rich places will lack

• ther natural advantages

Our approach:

Sample: places with production at baseline Define point source of wealth with geological deposits Predict change in value over time using district production and global price changes

Natural experiment:

Value of subsurface minerals increases in treatment regions, remains unchanged in control

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 10 / 20

SLIDE 11

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 11 / 20

SLIDE 12

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 12 / 20

SLIDE 13

Cross section: Mining places vs non-mining places

Criminal Serious Criminal Criminal Count Assets Graduate Deposit

• 0.004

0.007

• 0.018
• 0.152
• 0.003

(0.009) (0.006) (0.036) (0.038)*** (0.014) Log population

• 0.001
• 0.000
• 0.008
• 0.070
• 0.002

(0.005) (0.003) (0.023) (0.029)** (0.006) Rural pop share 0.003

• 0.002
• 0.006

0.016

• 0.001

(0.010) (0.006) (0.049) (0.044) (0.011) Employment share 0.003

• 0.065

0.480 1.452 0.046 (0.046) (0.034)* (0.540) (0.489)*** (0.147) Firm size

• 0.005

0.001 0.014 0.059

• 0.012

(0.005) (0.003) (0.040) (0.053) (0.009) Rural electrification

• 0.003

0.001 0.007 0.404 0.099 (0.023) (0.021) (0.076) (0.110)*** (0.037)** Primary schools per capita

• 8.284
• 2.750
• 43.325
• 81.313
• 17.385

(6.120) (3.800) (26.072) (40.783)* (6.398)** Government employment share

• 0.067
• 0.021
• 0.125
• 0.578

0.164 (0.034)* (0.029) (0.169) (0.219)** (0.073)** N 4983 4983 4983 4980 4943 r2 0.21 0.13 0.12 0.34 0.14

∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 13 / 20

SLIDE 14

Dependent variable: MLA has at least one criminal case

1 2 3 4 5 6 Price shock 0.094 0.099 0.101 (0.037)** (0.036)** (0.035)*** Price shock (large deposits) 0.079 0.085 0.085 (0.029)** (0.028)*** (0.028)*** Deposit count

• 0.001
• 0.001
• 0.000

(0.002) (0.002) (0.002) Large deposit count

• 0.002
• 0.001
• 0.000

(0.002) (0.002) (0.002) Log population

• 0.021
• 0.023
• 0.032
• 0.030

(0.019) (0.019) (0.021) (0.021) Rural pop share

• 0.003
• 0.002

0.030 0.029 (0.033) (0.032) (0.038) (0.038) Employment share 0.195 0.109 0.250 0.159 (0.272) (0.276) (0.349) (0.382) Firm size

• 0.029
• 0.029
• 0.034
• 0.033

(0.019) (0.023) (0.023) (0.028) Rural electrification

• 0.027
• 0.002

(0.098) (0.134) Primary schools per capita

• 50.581
• 52.274

(21.805)** (26.166)* Government employment share

• 0.119
• 0.119

(0.187) (0.217) N 1812 1755 1755 1453 1408 1408 r2 0.12 0.12 0.12 0.11 0.11 0.12

∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 14 / 20

SLIDE 15

Dependent var: share of candidates facing criminal cases

1 2 3 4 5 6 Price shock 0.026 0.026 0.028 (0.021) (0.021) (0.020) Price shock (large deposits) 0.023 0.025 0.025 (0.019) (0.019) (0.019) Deposit count

• 0.000
• 0.000
• 0.000

(0.001) (0.001) (0.001) Large deposit count

• 0.001
• 0.000
• 0.000

(0.001) (0.001) (0.001) Log population

• 0.015
• 0.017
• 0.018
• 0.017

(0.009)* (0.009)* (0.010)* (0.010)* Rural pop share 0.003 0.004 0.014 0.014 (0.018) (0.017) (0.020) (0.019) Employment share 0.087 0.044 0.121 0.056 (0.106) (0.118) (0.144) (0.166) Firm size

• 0.007
• 0.007
• 0.008
• 0.006

(0.008) (0.008) (0.008) (0.010) Rural electrification

• 0.035
• 0.026

(0.052) (0.064) Primary schools per capita

• 24.293
• 25.889

(12.400)* (16.305) Government employment share

• 0.064
• 0.099

(0.073) (0.084) N 1812 1755 1755 1453 1408 1408 r2 0.24 0.24 0.24 0.25 0.24 0.25

∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 15 / 20

SLIDE 16

Mineral price shocks and election outcomes

Candidates ENOP Margin Turnout Price shock

• 0.019

0.006 0.017 0.006 (0.304) (0.029) (0.005)*** (0.006) Deposit count

• 0.003

0.006

• 0.000
• 0.000

(0.009) (0.002)** (0.000) (0.000) Log population

• 0.298

0.030

• 0.002

0.004 (0.177) (0.028) (0.004) (0.005) Rural pop share

• 0.462

0.042 0.005 0.014 (0.248)* (0.044) (0.005) (0.005)** Employment share 2.300

• 0.468
• 0.095

0.085 (1.944) (0.363) (0.048)* (0.058) Firm size 0.835 0.023 0.007

• 0.026

(0.313)** (0.031) (0.004)* (0.007)*** Rural electrification 2.424

• 0.121
• 0.014

0.059 (0.886)** (0.104) (0.014) (0.019)*** Primary schools per capita

• 343.153

22.384 9.847 3.021 (235.055) (23.307) (3.232)*** (7.943) Government employment share

• 5.490
• 0.752

0.029

• 0.184

(1.869)*** (0.329)** (0.038) (0.058)*** N 7421 8315 8326 7421 r2 0.51 0.40 0.20 0.66

∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 16 / 20

SLIDE 17

Mineral price shocks and incumbent performance

Local Inc Margin Local Inc Win State Inc Margin State Inc Win Price shock 0.034 0.084 0.026 0.042 (0.013)** (0.041)* (0.012)** (0.026) Deposit count

• 0.001
• 0.002
• 0.000
• 0.001

(0.000) (0.001) (0.000) (0.001) Log population

• 0.013
• 0.030
• 0.000
• 0.008

(0.010) (0.019) (0.006) (0.011) Rural pop share

• 0.003

0.004 0.004 0.004 (0.008) (0.029) (0.007) (0.018) Employment share

• 0.152
• 0.185
• 0.010
• 0.005

(0.102) (0.260) (0.050) (0.127) Firm size 0.013 0.018

• 0.002
• 0.009

(0.007)* (0.016) (0.008) (0.014) Rural electrification 0.018 0.072 0.006

• 0.032

(0.034) (0.060) (0.027) (0.052) Primary schools per capita 17.134 32.981

• 1.980
• 1.566

(9.275)* (22.589) (6.633) (16.299) Government employment share

• 0.011

0.096 0.021 0.040 (0.071) (0.152) (0.049) (0.100) N 4312 4312 4635 4635 r2 0.21 0.15 0.38 0.26

∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 17 / 20

SLIDE 18

Discussion

Main results

More criminal MLA after a mining boom Incumbents are more entrenched and elections less competitive Voter turnout unaffected

Mechanisms

Selection into candidate pool: doesn’t explain full effect Moral Hazard: Timing of criminal cases is wrong

Non-incumbents haven’t been exposed to mining rents yet Can test with candidate time series

Criminals win more elections during mining booms

Candidate effort Voter indifference Voter preference

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 18 / 20

SLIDE 19

Conclusions

Testing microfoundations of the political resource curse Mining booms reduce electoral competition and put criminal politicians in power

Mechanism appears to be criminal success in elections

Next steps

Test moral hazard with candidate panel Observe impact of criminals on mine operation / local public services Examine impact of mining on local residents

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 19 / 20

SLIDE 20

THANK YOU!

Sam Asher and Paul Novosad Politicians, Criminality and Mining Booms in India 20 / 20

SLIDE 21

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

Nemera Mamo

(with A. Moradi and S. Bhattacharyya) Economics Department, University of Sussex

CSAE Conference: Economic Development in Africa 22nd - 24th March 2015 Oxford University

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 22

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Preview of the result:

◮ causal effect of mineral discoveries and extraction in sub-Saharan

Africa between 1992 - 2012

◮ mineral discovery and extraction improves local living stan-

dards in a panel of 3635 districts

◮ no evidence of resource curse at regional and national levels ◮ mining investments are not driven by preexisting levels of de-

velopment (as measured by nighttime lights and infrastructure)

◮ main results hold after a battery of robustness checks

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 23

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Contents

• 1. Motivation
• 2. Data and Variables
• 3. Results and Discussion
• 4. Reverse Causality
• 5. General Equilibrium Effects
• 6. Conclusions

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 24

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Motivation

Macro literature documented negative correlation between growth rates of GDP per capita and resource dependence (van der Ploeg, 2011 and Ross , 2011)

◮ resource wealth adversely affects development ◮ oil wealth in Nigeria and mineral wealth in D.R. Congo

Yet establishing causality has remained somewhat illusive

◮ absence of credible exogenous variation in the data

Explore the causal effect of mineral discovery and mineral extraction

◮ quasi-natural experiment to establish causality ◮ use nighttime lights to create measure of living standards

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 25

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

◮ Ihosy District in Madagascar: discovery of sapphire in 1998

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 26

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Data and Variables

Subnational administrative units of 2000 (FAO GeoNetwork)

◮ 3635 districts, 519 regions and 42 countries over 1992 - 2012

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 27

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Data and Variables

◮ production and status of mining (IntierraRMG): 548 industrial

size mines

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 28

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Data and Variables

◮ mineral discovery (MinEx): 263 giant and major deposits

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 29

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Data and Variables

Measuring local economic development from outer space:

◮ nightlights as a proxy measure of development at the local level

(National Oceanic and Atmospheric Administration)

◮ credibility of lights as a proxy measure of living standards

◮ Henderson, Storeygard and Weil (2012) in AER ◮ Chen and Nordhaus (2010) as NBER Working Paper ◮ Michalopoulos and Papaioannou (2013) in Econometrica ◮ Hodler and Raschky (2014) in QJE

◮ the measure comes on scale from 0 to 63 (digital number) ◮ we construct lights density and lights per capita

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 30

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Data and Variables

Instrument:

◮ exogenous international commodity price

Controls:

◮ population density ◮ geographical variables: altitude, ruggedness, soil fertility, coastal

proximity and land surface area

◮ climate variables: average annual rainfall, tropical climate,

arid climate and temperate climate

◮ political economy variables: capital city indicator, distance

to the capital city and ethnic fractionalization

◮ infrastructure variables: paved road density, railway density

and electric grid density

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 31

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Econometric Model

Estimate local effects of mineral production on development: LDdrt = αr + βt + γMPdrt + θPDdrt + X

′

drtΦ + ǫdrt

where:

◮ LDdrt is the natural log of lights density ◮ MPdrt is the natural log of mineral production value ◮ PDdrt is the natural log of population density ◮ Xdrc is a vector of district level control variables ◮ αr is a region dummy variable, indicating the use of region FE ◮ βt is a year dummy variable controlling for time varying shocks

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 32

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Results

◮ annual, decadal and growth effect of mineral production

Dependent Variable: Natural Logarithm of Nighttime Lights Density Annual Data Decadal Data Growth OLS OLS 2SLS OLS 2SLS Arellano-Bond (1) (2) (3) (4) (5) (6) Ln(mineral production) 0.04*** 0.02*** 0.06*** 0.02*** 0.06*** 0.01*** (0.01) (0.01) (0.02) (0.01) (0.01) (0.003) Ln(population density) 0.50*** 0.58*** 0.60*** 0.58*** 0.62*** 0.15* (0.06) (0.10) (0.10) (0.11) (0.10) (0.09) F test of excluded instruments 119.59 82.60 Weak identification F statistic 303.53 49.37 Adj R-squared 0.76 0.94 0.93 0.92 0.90 Observations 4599 4599 4599 657 657 4161 Control variables Yes Yes Yes Yes Yes Yes Year and Region FE No Yes Yes Yes Yes Yes

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 33

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

◮ non-parametric: effect of mineral production start year

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 34

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Treatment and control group comparison:

◮ no evidence of divergent trend prior to the production

treatment

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 35

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Econometric Model

Identify effect of discovery shocks on local development: LDdrt = αr + βt + γMDdrt + θPDdrt + φYDdrt + X

′

drtΦ + ǫdrt

where:

◮ LDdrt is the natural log of lights density ◮ MDdrt is the indicator for giant and major mineral discovery ◮ YDdrt is the number of years with discovery from t-10 to t-1 ◮ PDdrt is the natural log of population density ◮ Xdrc is a vector of district level control variables ◮ αr is a region dummy variable, indicating the use of region FE ◮ βt is a year dummy variable controlling for time varying shocks

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 36

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

◮ effect of discovery shocks on nighttime lights density

Dependent Variable: Natural Logarithm of Night-Time Lights Density Outcome in year: t t+2 t+4 t+6 t+8 t+10 (1) (2) (3) (4) (5) (6) Discovery

• 0.02
• 0.001

0.07 0.19* 0.34** 0.31** (0.11) (0.11) (0.11) (0.11) (0.15) (0.16) Years with discovery from t-10 to t-1 0.24** 0.29*** 0.37*** 0.46*** 0.52*** 0.57*** (0.07) (0.08) (0.09) (0.09) (0.10) (0.10) Ln(population density) 0.67*** 0.68*** 0.69*** 0.70*** 0.52*** 0.57*** (0.05) (0.05) (0.05) (0.05) (0.05) (0.04) Adj R-squared 0.81 0.81 0.82 0.82 0.82 0.82 Observations 76335 69065 61795 54525 47255 39985 Control variables Yes Yes Yes Yes Yes Yes Year and Region FE No Yes Yes Yes Yes Yes

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 37

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Reverse Causality

Do mining investments only go to desirable locations?

◮ relate the stages of investment in 2012 to district level indicators

• f development in the year 2000

◮ no evidence of systematic relationship between preexisting

development and future mining investments

Grassroots Exploration Advanced Exploration Pre-Feasibility Feasibility Construction (1) (2) (3) (4) (5) (6) Ln(nighttime lights density) 0.003

• 0.002
• 0.001
• 0.001

0.002 0.001 (0.005) (0.004) (0.004) (0.002) (0.002) (0.001) Ln(population density)

• 0.007
• 0.001

0.006

• 0.002

0.007 0.003* (0.008) (0.006) (0.006) (0.003) (0.004) (0.002) Ln(paved road density) 0.09 0.06

• 0.04
• 0.07

0.06

• 0.02

(0.09) (0.08) (0.08) (0.05) (0.05) (0.04) Ln(railway density)

• 0.005

0.002

• 0.007*
• 0.001
• 0.002

0.0005 (0.005) (0.005) (0.004) (0.003) (0.003) (0.001) Adj R-squared 0.17 0.17 0.10 0.10 0.04 0.003 Observations 3635 3635 3635 3635 3635 3635 Control Variables Yes Yes Yes Yes Yes Yes Region Fixed Effects Yes Yes Yes Yes Yes Yes Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 38

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

General Equilibrium Effects

Resource curse as a general equilibrium phenomenon:

◮ positive effect in a mining district could be coupled with greater

negative effects in neighbouring non-mining districts

◮ the net effect of a resource boom in a country is negative? ◮ no evidence of resource curse at the regional and country levels

Region Level Effects Country Level Effects (1) (2) (3) (4) (5) (6) (7) (8) Ln(mineral production) 0.06*** 0.01*** 0.02 0.02 0.003 0.02 (0.01) (0.003) (0.02) (0.02) (0.002) (0.03) Discovery 0.02 0.01 (0.03) (0.04) Adj R-squared 0.42 0.97 0.97 0.68 0.18 0.98 0.96 0.45 Observations 1932 1932 1932 10899 672 672 672 528 Year Fixed Effects No Yes Yes Yes No Yes Yes Yes Region Fixed Effects No Yes Yes Yes No No No No Country Fixed Effects No No No No No Yes Yes Yes Years with discovery from t-10 to t-1 No No No Yes No No No Yes Population Density Yes Yes Yes Yes Yes Yes Yes Yes Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 39

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Robustness Tests

◮ zero luminosity and potential noise ◮ lights intensity adjusted to population ◮ sparsely populated area - lights may be dominated by noise ◮ assigning reasonably equal weights to each country ◮ grid level analysis: administrative boundaries are endogenous

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 40

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Summary

◮ mineral discovery and extraction improves local living stan-

dards in a panel of 3635 districts from 42 sub-Saharan African countries observed over the period 1992 - 2012

◮ policy implications:

◮ high commodity prices and new natural resource discoveries can

provide a major new source of development finance

◮ opportunity to improve local living standards via investments in

• ther sectors

◮ investment in infrastructure and economies of scale Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 41

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Thank You!

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 42

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

Non-parametric: effect of mineral production start date on nighttime lights density

◮ nighttime lights improving two years prior to the actual

production start date

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 43

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

◮ non-parametric: effect of discovery shocks

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 44

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

◮ treat discovery as a permanent shock rather than temporary

Dependent Variable: Natural Logarithm of Night-Time Lights Density (1) (2) (3) (4) (5) (6) Discovery 0.13 0.25 0.25 (0.13) (0.16) (0.16) 5 year pre discovery period (1=yes)

• 0.01

0.09

• 0.14

(0.11) (0.14) (0.15) 0 to 5 years post discovery (1=yes)

• 0.11
• 0.05
• 0.22

(0.12) (0.14) (0.14) 6 Year onwards post discovery (1=yes) 0.32*** 0.26*** 0.36*** (0.12) (0.13) (0.17) Years with discovery from t-10 to t-1 0.18*** 0.18*** 0.15* 0.21*** 0.24*** 0.21*** (0.08) (0.08) (0.08) (0.08) (0.08) (0.08) Ln(population density) 0.67*** 0.67*** 0.67*** 0.67*** 0.67*** 0.67*** (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) Discovery size Giant + Major Discovery Major Discovery Giant Discovery Adj R-squared 0.81 0.81 0.81 0.81 0.81 0.81 Observations 76335 76335 76335 76335 76335 76335 Geographical Controls Yes Yes Yes Yes Yes Yes Climate Controls Yes Yes Yes Yes Yes Yes Political Economy Controls Yes Yes Yes Yes Yes Yes Infrastructure Controls Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Region Fixed Effects Yes Yes Yes Yes Yes Yes Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 45

Motivation Data and Variables Results and Discussion Reverse Causality General Equilibrium Effects Conclusions

◮ First-Stage Regression of Production on Price Index

Mineral Production (Annual) Mineral Production (Decadal) Ln (commodity price index) 3.63*** 3.93*** 0.33 (0.80) F test of excluded instruments 119.59 82.60 (Prob F) (0.00) (0.00) Underidentification LM statistic 2.83 4.11 (Chi-sq(1) P-val) (0.11) (0.04) Weak idenitification F statistic 303.53 49.37 Stock-Yogo critical values 16.38/5.53 16.38/5.53 Observations 4599 657 Control Variables Yes Yes Year and Region FE Yes Yes

Nemera Mamo Economic Consequences of Mineral Discovery and Extraction in Sub-Saharan Africa: Is There a Curse?

SLIDE 46

March 24, 2015 CSAE Conference Oxford

Weapon of Choice

Axel Dreher (Heidelberg University) Merle Kreibaum (University of Goettingen)

SLIDE 47

Motivation

• Broad literature on `resource curse‘ and civil war

– Close to consensus that `greed’ dominates over `grievances’ (Collier and Hoeffler 2004, Collier et al. 2009, Fearon and Laitin 2005, Soysa and Neumayer 2007)

• Country level

– Recent exception is Hunziker and Cedermann (2012), finding evidence in favour of `grievances‘ by interacting oil with group participation in power

• Ethnic groups
• No serious attempt to test the resource curse with respect to

terrorism

• Few studies including both at country level

March 24, 2015 CSAE Conference Oxford

SLIDE 48

Research Question

• What determines if an ethno-political group remains peaceful,

reverts to terrorism or starts an insurgency?

• Look at geo-coded fossil fuel resources and political groups over

time

– i.e., sub-national, groups, panel data

• Political groups as `rational actors‘ who weigh risks and

potential benefits with regards to their strategy

– Accounting for own characteristics and state‘s behaviour

March 24, 2015 CSAE Conference Oxford

SLIDE 49

Hypothesis

• Oil on a group‘s territory plays a role, but also interacted with

– Power sharing/ political discrimination – Regional autonomy – Economic discrimination – Support by a foreign state

March 24, 2015 CSAE Conference Oxford

SLIDE 50

Data and Method

• Minorities at Risk Organizational Behavior (MAROB)

– 118 ethno-political organisations (maximum of 105 in regressions) – 13 MENA countries – 1980-2004

• Multilevel Random Intercept Model (Multinomial Logit)

WEAPONi,t = α+βRESOURCESi,t-1+γXi,t-1+δRESOURCESi,t-1*Xi,t-

1+ζZi,t+εi,t

– Terror and insurgencies as distinct choices – Estimate choices in one model – Choices clustered by organisation

March 24, 2015 CSAE Conference Oxford

SLIDE 51

Terror and Insurgencies

• Apply action- and actor-based characteristics for distinguishing

forms of violence

• Narrow definition of terrorism: violent attacks on civilians and

non-security state personnel, excluding those on state security institutions

• Large-scale violent events include

– those where violence is a regular strategy of a group – those targeting security personnel and state institutions in anything from a local rebellion, guerilla activity, up to a civil war – a group that has already gained control of a specific area

March 24, 2015 CSAE Conference Oxford

SLIDE 52

Regional Oil Production

• Geocoded data on (giant) oil reserves (Horn 2010)

– >500 billion recoverable barrels – Varies over time as new fields are discovered and fields `grow‘

• Assume that field contribution is in proportion to its size
• Multiply share with yearly value of oil production (at the country

level)

• Use log(oil regional production in millions constant US$)
• Test for robustness using a dummy for oil fields in a region

(PRIO 2007)

– Varies over time only when new fields are discovered

March 24, 2015 CSAE Conference Oxford

SLIDE 53

Control Variables

• Group-level

– Goals: eliminate political discrimination, eliminate economic discrimination, autonomy, eliminate cultural discrimination, establish Islamic state – State uses violence against group – State negotiated with organisation

• Ethnicity-level

– Ethnic group has regional autonomy, Ethnic group shares political power with others

• Country-level

– log(national oil production), log(GDP p.c.), log(population), Democracy, Ethno-linguistic fractionalisation

March 24, 2015 CSAE Conference Oxford

SLIDE 54

Results: Main specification and policy

March 24, 2015 CSAE Conference Oxford (1) (2) (3) (4) (5) (6) Terror Insurgency Terror Insurgency Terror Insurgency Log(Group oil production) 0.9863 (0.863) 1.3739*** (0.000) 0.9353 (0.464) 1.2995*** (0.002) 0.9510 (0.565) 1.574*** (0.000) Log(National oil production) 1.0113 (0.897) 0.7995** (0.040) 1.1241 (0.268) 0.7694*** (0.022) 1.8782** (0.013) 0.7794** (0.050) Interaction term oil and political discrimination 1.0011 (0.960) 1.0445 (0.108) Political discrimination 1.8595** (0.038) 0.6683 (0.255) Interaction term oil and power sharing 1.0757 (0.440) 0.7439*** (0.001) Ethnic group shares power with others 3.2593 (0.375) 15.5840* (0.082) Number of groups 105 103 88 Number of

• bservations

3360 3336 2496 Log-Likelihood

• 424.210
• 414.307
• 308.901

SLIDE 55

Results: Autonomy, econ. discrimination, support

March 24, 2015 CSAE Conference Oxford (1) (2) (3) (4) (5) (6) Terror Insurgency Terror Insurgency Terror Insurgency Log(Group oil production) 1.0168 (0.801) 1.3070*** (0.000) 0.9763 (0.777) 1.3220*** (0.001) 0.9709 (0.716) 1.3192*** (0.004) Log(National oil production) 1.6052** (0.046) 0.8466 (0.101) 1.1082 (0.275) 0.8107** (0.036) 1.0092 (0.915) 0.7960** (0.039) Interaction term oil and autonomy 0.2985** (0.011) 1.2488** (0.020) Ethnic group has regional autonomy 0.4640 (0.486) 0.1562 (0.347) Interaction term oil and economic discrimination 0.9958 (0.865) 1.0137 (0.674) Economic discrimination 2.0703*** (0.005) 0.8947 (0.735) Interaction term oil and foreign support 1.0788 (0.192) 1.1511** (0.037) Group supported by foreign state 1.4092 (0.593) 2.4767* (0.068) 3.0868* (0.058) 3.1423** (0.017) 2.7315* (0.094) 1.6774 (0.225) Number of groups 88 103 105 Number of observations 2517 3336 3360 Log-Likelihood

• 315.477
• 415.383
• 421.664

SLIDE 56

Conclusions

• Organisations are more likely to rebel in countries rich in oil
• The `resource curse‘ does not extend to the realm of terrorism

– Highlights the importance of `greed‘ over `grievances‘

• Regional distribution of resources matters: concentration in an

area with ambitions for autonomy or receiving support from foreign states more likely to spark insurgency

• Policy implication: giving the population a say in the distribution
• f revenues can increase acceptance of extraction/ decrease

incentive for violence (see power sharing)

March 24, 2015 CSAE Conference Oxford

SLIDE 57

Thank you for your attention!

March 24, 2015 CSAE Conference Oxford

SLIDE 58

Appendix

March 24, 2015 CSAE Conference Oxford

SLIDE 59

Countries and Territories Included

• Algeria
• Bahrain
• Cyprus
• Iran
• Iraq
• Israel
• Jordan
• Lebanon
• Morocco
• Saudi Arabia
• Syria
• Turkey
• West Bank and Gaza

March 24, 2015 CSAE Conference Oxford

SLIDE 60

Model

• Clustered data, i.e., multiple observations per group over time
• Unobserved heterogeneity between groups – two observations

from the same cluster more similar than observations in different clusters

• Residual has two parts: random intercept per cluster (deviation

from overall mean) and unit-specific one (deviation from cluster mean)

March 24, 2015 CSAE Conference Oxford

SLIDE 61

Oil Production and Discovered Reserves

Source: Hunziker and Cedermann (2012) March 24, 2015 CSAE Conference Oxford

SLIDE 62

March 24, 2015 CSAE Conference Oxford

Alternative oil measure

Terror Insurgency Terror Insurgency Terror Insurgency Oil or gas field on group's territory 0.0468** 0.339 0.0471 0.122 0.0147** 0.440 (0.0454) (0.534) (0.0778) (0.248) (0.0205) (0.946) Log(national oil production) 1.081 1.091 1.172* 1.126 1.798* 1.107 (0.0551) (0.0888) (0.0847) (0.0941) (0.433) (0.0795) Ethnic group politically discriminated 2.085** 1.104 (0.489) (0.371) Interaction oil and political discrimination 0.809 1.379 (0.332) (0.789) Interaction oil and power sharing 10.74 0.0147* (19.72) (0.0305) Ethnic group shares political power with others 0.994 1.118 (1.220) (1.861) Group supported by foreign state 3.951* 3.640** 3.289* 3.581* 1.329 2.914 (2.235) (1.778) (1.817) (1.788) (0.800) (1.645) (1) (2) (3)

SLIDE 63

March 24, 2015 CSAE Conference Oxford

Terror Insurgency Terror Insurgency Terror Insurgency Oil or gas field on group's territory 0.0599 0.204 0.0957 0.452 0.0259*** 0.0701 (0.0993) (0.428) (0.141) (0.865) (0.0269) (0.121) Log(national oil production) 1.635* 1.035 1.162* 1.084 1.086 1.103 (0.360) (0.0932) (0.0871) (0.0885) (0.0540) (0.0896) Interaction oil and regional autonomy 0.00000886*** 234.6* (0.0000138) (517.7) Ethnic group has regional autonomy 0.484 0.208 (0.483) (0.391) Ethnic group economically discriminated 1.989** 1.027 (0.503) (0.340) Interaction oil and economic discrimination 0.792 0.831 (0.394) (0.616) Interaction oil and foreign state support 5.315 38.40* (5.990) (57.61) (1) (2) (3)

Alternative oil measure (cont’d.)

SLIDE 64

Descriptive Statistics

March 24, 2015 CSAE Conference Oxford

Mean SD Min Max N `Weapon of choice' 0.54 0.83 2 1120 Regional oil production (in Mio. US$) 1886.85 6610.29 76699 1120 National oil production (in Mio. US$) 8697.32 19409.70 162612 1120 GDP per capita 9780.94 6933.81 2162 28094 1120 Population (in 1000s) 13834.36 17306.95 374 69342 1120 Freedom House indicators 4.54 1.90 1 7 1120 Polity IV 0.71 8.13

• 10

10 761 Ethno-linguistic fractionaliation 0.50 0.20 1 1120 Ethno-linguistic polarisation 0.62 0.18 1 793 Goal: eliminate discrimination 0.60 0.49 1 1119 Goal: autonomy 0.25 0.44 1 1119 Goal: eliminate economic discrimination 0.19 0.39 1 1120 Goal: eliminate cultural discrimination 0.32 0.47 1 1120 Group supported by foreign state 0.35 0.48 1 1100 Goal: Islamic state 0.08 0.27 1 1120 State uses violence against group 0.13 0.33 1 1117 Group provides social services 0.21 0.41 1 1116 State negotiated with organization 0.11 0.32 1 1120 Group economically discriminated 2.05 1.61 4 1029 Group politically discriminated 2.25 1.73 4 1029 Ethnic group has regional autonomy 0.15 0.36 1 832 Ethnic group shares political power with

• thers

0.18 0.38 1 832

SLIDE 65

March 24, 2015 CSAE Conference Oxford

"Weapon of Choice" 0 = no violent behavior; 1 = terrorism, i.e., attacks on civilians incl. non-security state personnel, no control of territory; 2 = insurgency, i.e., attacks targeting security personnel and state authorities, local rebellion, guerilla activity, civil war, control of territory Regional Oil Production Value of the share of the national hydrocarbon production in a year, located on an ethnic group's territory, in constant 2009 million US $ (for fields containing at least 500 million barrel oil or gas equivalents) National Oil Production Value of the national hydrocarbon production in a year in constant 2009 US $ (for fields containing at least 500 million barrel oil or gas equivalents) Oil Dummy Indicates that hydrocarbon reserves are located on an ethnic group's territory GDP p.c. GDP per capita, PPP, in constant 2005 US $ Population Measured in 1000s Freedom House Average of civil liberties and political rights indices; from 1 = free to 7 = not free; civil liberties: freedom of expression and belief, associational and organizational rights, rule of law, and personal autonomy and individual rights; political rights: electoral process, political pluralism and participation, and functioning of government Polity IV Polity scale ranges from +10 (strongly democratic) to -10 (strongly autocratic) Ethnolinguistic Fractionalization Probability that a randomly selected pair of individuals in a society will belong to different groups; from 0 = complete homogeneity to 1 = complete heterogeneity Ethnolinguistic Polarization Captures the distance of the distribution of ethnic groups from the bipolar distribution, ranging from 0 = least polarized to 1 = maximally polarized (bipolar) Definition

SLIDE 66

March 24, 2015 CSAE Conference Oxford

Goal: remedial policies Major organizational goals focused on eliminating discrimination and on creating increasing remedial policies Goal: autonomy, independence Major organizational goals focused on creating or strengthening autonomous status for group or on creating a separate state for the group or revanchist change in border of state Goal: eliminate economic discrimination Group expresses economic grievances focused on elimination of discrimination or on creating or strengthening economic remedial policies Goal: eliminate cultural discrimination Group expresses cultural grievances focused on elimination of discrimination or on strengthening economic remedial policies (i.e., establishing or increasing state funding for cultural protection and/or promotion) Group supported by foreign state Has org. received support from foreign state in year being coded - i.e. financial, humanitarian, political, or military support? Goal: Islamic state Has the organsiation expressed the goal of creating an islamic state/ an islamic government or of introducing islamic law? State uses violence against group Does the state use periodic or consistent lethal violence against the organization? Group provides social services The provision of social services is a minor or major strategy of the organization Ethnic group has regional autonomy Elite members of the group have no central power but have some influence at the subnational level (i.e., the provincial

• r district level, depending on the vertical organization of the state)

Ethnic group shares political power Any arrangement that divides executive power among leaders who claim to represent particular ethnic groups (formal or informal arrangements) State negotiated with group The state negotiated with the group in the year and the state might even have made concessions Political Discrimination 0 = no discrimination, 1=neglect but remedial policies; 2=neglect, no remedial policies; 3=social exclusion, neutral policies; 4 = exclusion/ repressive policy Economic Discrimination 0 = no discrimination, 1=neglect but remedial policies; 2=neglect, no remedial policies; 3=social exclusion, neutral policies; 4 = exclusion/ repressive policy Definition

SLIDE 67

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Extractive Industries, Production Shocks and Criminality: Evidence from a Middle-Income Country

Sebastian Axbard, Uppsala University Jonas Poulsen, Uppsala University Anja Tolonen, University of Gothenburg CSAE Conference, March 2015

SLIDE 68

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Mining and Development

Mineral resources are dominant in 81 countries, collectively accounting for a quarter of world GDP, half of world population and nearly 70% of those in extreme poverty (World Bank, 2014). 1% of the global workforce in mining (ILO, 2010) and could provide local labour market opportunities (Kotsadam & Tolonen, 2013; Aragon and Rud, 2013). However, large literature link mining to violence and conflict (recent: Berman, Couttenier, Rohner and Thoenig, 2014; Maystadt et al., 2014; Sanchez de la Sierra, 2014).

SLIDE 69

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Mining and Violence

Previous literature mostly focus on conflict in LIC countries. Know little about the link between extractive industries and violence/crime in MIC. Do we still need to worry about potential detrimental effects

• n criminal activity from this industry?

SLIDE 70

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

High Levels of Violence in Many Mineral Rich Countries

tj

Ç Ç Ç Ç Ç Ç Ç Ç Ç Ç Ç Ç Ç Ç tj ffi tj tu fi tj fi

Homicide rate

0.00

• 2.99

3.00

• 4.99

5.00

• 9.99

10.00

• 19.99

20.00

• 29.99

>=30.00 WHO estjmates No data available

Map 1.1: Homicide rates, by country or territory (2012 or latest year)

Note: The boundaries and names shown and the designations used on this map do not imply official endorsement or acceptance by the United Nations. Dashed lines represent undeter- mined boundaries. The dotted line represents approximately the Line of Control in Jammu and Kashmir agreed upon by India and Pakistan. The final status of Jammu and Kashmir has not yet been agreed upon by the parties. The final boundary between the Republic of Sudan and the Republic of South Sudan has not yet been determined. A dispute exists between the Governments of Argentina and the United Kingdom of Great Britain and Northern Ireland concerning sovereignty over the Falkland Islands (Malvinas).

Source: UNODC Homicide Statistics (2013).

SLIDE 71

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

This Paper

Question: What is the impact of mining activity on the South African local crime rate? Strategy: Use time and geographic variation in mining activity 2003-2012. Exploit fluctuations in international mineral prices. The effect on violent and property crime at the local level. Investigate potential mechanisms.

SLIDE 72

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

What to expect?

Two main competing mechanisms: Labour market

Direct or indirect effects on local economy. Alter the opportunity cost of engaging in criminal activity (Becker, 1968; Ehrlich, 1973).

Grabbing

More resources to steal (from the mine or a growing economy). Incentivize predatory behaviour (Ross, 2003; Collier and Hoeffler, 2004).

Outcome depends on which effect that dominates, related to relative factor intensities (Dal B´

• and Dal B´
• , 2011 & Dube

and Vargas, 2013).

SLIDE 73

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Preview of findings

Mining activity leads to lower levels of crime. Results largely hold for both property and violent crime. However, negative production shocks lead to increases in crime. Suggests that income opportunities provided by the mine are important (validated by using lights at night data & sub-analysis that focus on labour intensive mines).

SLIDE 74

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Background & Data

SLIDE 75

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Mining and Crime in South Africa

Violent crime is a notorious problem in South Africa.

World Competitiveness Survey: worst out of 133 countries. Every day: 50 murders, 100 rapes & 330 armed robberies.

South Africa has the fifth largest mining industry in the world

8.8% (direct) & 18 % (indirect) to GDP (2011-2012). Employ 525,000, 0.7% of the workforce (2012).

Media attention: 2012 clashes at the Marikana Platinum Mine which lead to the death of 34 mine workers. Historically the mining industry in South Africa claimed to be

• ne of the contributors to the high crime rate (Kynoch, 2005

& 2007).

SLIDE 76

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Data on Mining

All large scale mining operations in South Africa 1975-2012 (InterraRMG). Know minerals produced and exact geographic location. Total 320 mines that produce 23 different minerals. Code mining activity as a dummy for each mineral-mine.

Figure: Mines in South Africa

SLIDE 77

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Mining in South Africa

Large fluctuations in mining production during sample period (industry is both expanding and contracting).

Graphs

Gold, copper, silver and zinc production decreased; whereas iron ore, cobalt and PGMs increased Important producer of platinum (and PGMs), chromite, vanadium, titanium, zirconium and manganese.

Graphs

SLIDE 78

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Data on Crime

Crimes reported from 1,083 police stations (SAPS). Reported for each financial year (April to March) from 2003-2012. 29 different categories condensed to property, violent and total crime.

Figure: Police Stations in South Africa

SLIDE 79

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Sample Construction

Follow literature and match mine to all police precincts within 20km to allow for spillovers.

Figure: Location of mines and police precincts

SLIDE 80

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Active Mines and Crime Rates in Mine and Non-mine Precincts

40 45 50 55 60 Crime per 1000 Inhabitants 210 220 230 240 250 Number of Active Mines 2002 2004 2006 2008 2010 2012 Year Active Mines No Mine Mine within 20 km

(a) Total Crime

20 25 30 35 Crime per 1000 Inhabitants 210 220 230 240 250 Number of Active Mines 2002 2004 2006 2008 2010 2012 Year Active Mines No Mine Mine within 20 km

(b) Property Crime

14 16 18 20 22 Crime per 1000 Inhabitants 210 220 230 240 250 Number of Active Mines 2002 2004 2006 2008 2010 2012 Year Active Mines No Mine Mine within 20 km

(c) Violent Crime

SLIDE 81

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Empirical Strategy & Results

SLIDE 82

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Empirical Strategies

Use two different strategies: fixed effect and IV. This presentation focus on IV - we instrument mining activity with international mineral price. Vast anecdotal evidence production decisions depend on prices. Demand elasticities typically low - minerals small share of production costs (Taurasi, 2014) - price driven by demand. Previously used: Sanchez de la Sierra (2014), Berman et al. (2014) & von der Goltz and Barnwal (2014).

SLIDE 83

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Instrumental Variable Approach

active minesijt = δpit + γij + λt + uijt (1) ln(yijt) = βactive minesijt + γij + λt + ǫijt (2) ln(yijt) is the log of the crime rate in police precinct j year t. active minesijt is the number of mines producing mineral i within 20 km from precinct j in year t. γij and λt represent mineral by precinct and time fixed effects. pit is the international mineral price of mineral i in year t.

SLIDE 84

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

International Price of Minerals

20 40 60 USD / Gram 2002 2004 2006 2008 2010 2012 Year Gold Palladium Platinum Vanadium Silver

(a) High Price

.02 .04 .06 .08 .1 Price (USD / Gram) 2002 2004 2006 2008 2010 2012 Year Antimony Copper Cobalt Lead Nickel Tin Zirconium Zinc

(b) Medium Price

.0001 .0002 .0003 .0004 Price (USD/ Gram) 2002 2004 2006 2008 2010 2012 Year Coal Chromite Iron Ore Manganese Ore Phosphate Rock Titanium

(c) Low Price

SLIDE 85

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Main Results

Outcome: Total Crime Property Crime Violent Crime 2SLS

• 0.073***
• 0.088***
• 0.066***

(0.020) (0.022) (0.024) [0.024] [0.027] [0.027] Reduced Form

• 0.00192***
• 0.00230***
• 0.00174***

(0.000504) (0.000551) (0.000615) First Stage 0.026*** (0.0018) [0.0024] F Statistic (one way cluster) 206.9 F Statistic (two way cluster) 121.1 Observations 5260 Mineral by Precinct FE Yes Yes Yes Year FE Yes Yes Yes Notes: Standard errors in parenthesis clustered at the precinct and standard errors in brackets are clustered at the precinct and the mineral-year level.

SLIDE 86

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Production Shocks

Start: ln(yijt) = πstartijt + γij + λt + ǫijt (3) startijt = κpit + γij + λt + uijt (4) Stop: ln(yijt) = πstopijt + γij + λt + ǫijt (5) stopijt = κpit + γij + λt + uijt (6) Purpose: Does production shocks in mining activity affect the crime rate in that particular year? startijt/stopijt count the net number of mineral i mines within 20km from precinct j that either start or stop producing in year t.

SLIDE 87

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Results on production starts

Outcome: Total Crime Property Crime Violent Crime A: Start Producing (incl. all mines) 2SLS 2.323 2.746 2.182 (1.890) (2.193) (1.865) First Stage

• 0.000893

(0.000675) F Statistic 1.748 Observations 4733 B: Start Producing (excl. new mines) 2SLS

• 2.086*
• 2.465*
• 1.959*

(1.077) (1.294) (1.059) First Stage 0.000994** (0.000464) F Statistic 4.594 Observations 4733 Mineral by Precinct FE Yes Yes Yes Year FE Yes Yes Yes

Notes: Standard errors in parenthesis clustered at the precinct.

SLIDE 88

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Results on production stops

Outcome: Total Property Violent C: Stop Producing 2SLS 0.86*** 1.02*** 0.81*** (0.27) (0.31) (0.31) [0.54] [0.64] [0.56] First Stage

• 0.0024***

(0.00048) [0.0015] F Statistic (one way cluster) 24.77 F Statistic (two way cluster) 2.689 Observations 4733 Mineral by Precinct FE Yes Yes Yes Year FE Yes Yes Yes

Notes: Standard errors in parenthesis clustered at the precinct and standard errors in brackets are clustered at the precinct and the mineral-year level.

SLIDE 89

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Why do we find these results?

Results in line with prediction from economic theory (if the

• pportunity cost channel dominates grabbing).

Is this supported by the data? Hard question to answer:

Poor labour market data for local levels (only one census year during our sample period). Working on collecting labour survey data.

SLIDE 90

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Proposed solutions

Proxy economic activity using satellite data on lights at night.

Follow recent literature in Economics, such as Michalopoulos & Papaioannou (Econometrica, 2013) and Henderson et al. (AER, 2012)

Investigate effects in capital and labour intensive mines.

In South Africa open-pit mining often capital intensive, whereas underground mining is labor intensive.

SLIDE 91

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Mining and Economic Activity

Outcome Average Stable Lights at Night Active Mine 20 km 0.987*** (0.260) Start Producing (all mines)

• 29.51

(23.86) Start Producing (excl. new) 26.49** (13.07) Stop Producing

• 10.95***

(3.493) F Statistic 206.9 1.748 4.594 24.77 Observations 5260 4733 4733 4733 Precinct by Mineral FE Yes Yes Yes Yes Year FE Yes Yes Yes Yes Notes: Outcome is the average lights density in a precinct in a given year. Standard errors in parenthesis clustered at the precinct.

SLIDE 92

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Heterogeneous effects by mine type

Total Crime Property Crime Violent Crime A: Open-pit mining (capital intensive) 2SLS 0.132 0.0690 0.143 (0.313) (0.305) (0.385) First Stage 0.00993*** (0.00329) F Statistic 9.142 Observations 1105 B: Underground mining (labor intensive) 2SLS

• 0.0923***
• 0.116***
• 0.0548*

(0.0251) (0.0293) (0.0305) First Stage 0.0275*** (0.00216) F Statistic 162.9 Observations 3169 Mineral by Precinct FE Yes Yes Yes Year FE Yes Yes Yes

Notes: Standard errors clustered at the precinct level are reported in parenthesis.

SLIDE 93

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Robustness

Estimation & identification:

All results supported by alternative strategy with FE. Results robust to using log of prices as IV and count data as

• utcome.

Randomization test that preserves time structure of data suggest results not driven by mineral time trends. Results unaffected if exclude minerals for which South Africa is a main producer, suggest that prices are not driven by changes in domestic production.

Alternative mechanisms:

Mining activity is insignificantly associated with lower spending

• n crime prevention, suggesting reduction not driven by

increases in policing.

SLIDE 94

Motivation Background & Data Empirical Strategy & Main Results Mechanisms Robustness Conclusion

Conclusion

Mining activity leads to lower levels of crime. Results largely hold for both economic and violent crime. However, negative production shocks when mines stop producing lead to increases in crime. Income opportunities provided by the mine are arguably important for criminal outcomes. Results stand out compared to previous literature that have found that increases in mining activity increase conflict & violence. Mining activity may create conditions where crime occurs if an income shock hits. Important insight for policy, given high volatility in the mining industry.

SLIDE 95

Summary Statistics: Price Sample

MEAN SD MIN MAX OBS International Mineral Price (USD/gram) 16.6 20.1 0.0000024 60.6 5260 Lights at Night 15.4 20.6 63 5260 Population (1000’) 55.1 58.6 0.14 592.3 5260 Active Mine 20km 2.89 4.04 35 5260 # Start Producing 20km 0.12 0.42 5 4733 # Stop Producing 20km 0.061 0.28 3 4733 Total Crime per 1000 88.9 914.5 3.49 24794.3 5260 Property Crime per 1000 59.6 740.7 1.22 21531.7 5260 Violent Crime per 1000 19.9 56.2 1.39 1544.7 5260 Log(Total Crime per 1000) 3.71 0.70 1.25 10.1 5260 Log(Property Crime per 1000) 3.02 0.86 0.20 9.98 5260 Log(Violent Crime per 1000) 2.73 0.59 0.33 7.34 5260 Expenditure per capita (Rand) 7.11 9.08 0.062 43.4 5260

SLIDE 96

Mining Production by Mineral

3 4 5 6

Antimony (kt)

2000 2005 2010 Year

6 8 10 12

Chromite (Mt)

2000 2005 2010 Year

220 240 260

Coal (Mt)

2000 2005 2010 Year

0.5 1 1 2

Cobalt (kt)

2000 2005 2010 Year

100 120 140 160

Copper (kt)

2000 2005 2010 Year

8 12 16

Diamond carats (Mct)

2000 2005 2010 Year

250 350 450

Gold (t)

2000 2005 2010 Year

35 45 55

Iron ore (Mt)

2000 2005 2010 Year

40 60 80

Lead (kt)

2000 2005 2010 Year

2 6 10

Manganese (Mt)

2000 2005 2010 Year

30 35 40 45

Nickel (kt)

2000 2005 2010 Year

60 70 80 90

Palladium (t)

2000 2005 2010 Year

2.2 2.6 3

Phosphate (Mt)

2000 2005 2010 Year

220 260 300

PGMs (t)

2000 2005 2010 Year

60 100 140

Silver (t)

2000 2005 2010 Year

Tin (kt)

2000 2005 2010 Year

600 700 800

Titanium (kt)

2000 2005 2010 Year

15 20 25 30

Vanadium (kt)

2000 2005 2010 Year

30 40 50 60 70

Zinc (kt)

2000 2005 2010 Year

300 350 400 450

Zirconium (kt)

2000 2005 2010 Year

South Africa Total Annual Production Go Back

SLIDE 97

South Africa Share of World Production

Antimony

1990 2000 2010 Year 20 40 60 80 100

Chromite

1990 2000 2010 Year

Coal

1990 2000 2010 Year

Cobalt

1990 2000 2010 Year

Copper

1990 2000 2010 Year

Diamond carats

1990 2000 2010 Year 20 40 60 80 100

Gold

1990 2000 2010 Year

Iron ore

1990 2000 2010 Year

Lead

1990 2000 2010 Year

Manganese

1990 2000 2010 Year

Nickel

1990 2000 2010 Year 20 40 60 80 100

Phosphate

1990 2000 2010 Year

Platinum

1990 2000 2010 Year

PGMs

1990 2000 2010 Year

Silver

1990 2000 2010 Year

Tin

1990 2000 2010 Year 20 40 60 80 100

Titanium

1990 2000 2010 Year

Vanadium

1990 2000 2010 Year

Zinc

1990 2000 2010 Year

Zirconium

1990 2000 2010 Year

South Africa % Share of Annual World Production

20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100

Go Back