BAD APPLES: THE PEER EFFECTS OF VIOLENCE Sidra Ahmad Q: Does - PowerPoint PPT Presentation

BAD APPLES: THE PEER EFFECTS OF VIOLENCE Sidra Ahmad

Q: Does violence spread within groups of diverse individuals in a common institutional setting? A: Yes.* *Violent behavior spreads from the group to each of its members but not between members. Violent culture doesn’t spread at all.

Q: Why does this matter? A: Membership in institutions alters states’ behavior— positively 1 and negatively 2 — through the diffusion of norms of behavior. What happens to individual political actors? 1 (Mansfield and Pevehouse 2008) 2 (Hafner-Burton and Montgomery 2008)

LITERATURE AND THEORY

Peer Effects • Gould and Kaplan (2008): A baseball player’s performance, especially with regard to power hitting, significantly improves after being on the same team with Canseco • Guryan et al. ( 2009): No relationship between a golfer’s score and the ability, relative or absolute, of his “groupmates” Lasting Impact of Civil War • Civil war disturbs the status quo and legitimizes violence; individuals are socialized into aggression • Miguel et al. (2011): Higher incidence of conflict in a player’s home country → more yellow/red cards he receives Learning Mechanisms • Social Learning Theory (Bandura 1971): People learn through direct experience and observation • Mirror Neurons: Fire during observation and execution; believed to be responsible for imitation

HYPOTHESES

H 5 H 1 : Team I ’s aggregate exposure to civil war exerts a peer effect on player i ’s behavior. Team I Player j Civil War Exposure H 2 : Team I ’s aggregate violent behavior exerts a peer effect on player i ’s behavior. H 1 H 3 H 3 : Player j ’s exposure to civil war exerts a peer effect on player i ’s behavior. H 4 : Player j ’s violent behavior exerts a peer Violent Behavior effect on player i ’s behavior. H 2 H 4 H 5 : Peer effects exerted by player j will be stronger than peer effects exerted by team I .

DATA, VARIABLES, AND METHODOLOGY

• “Civil War Exposure and Violence” (Miguel et al. 2011) – Five national leagues: England, France, Germany, Italy, and Spain – One supra-national league: Champions League – Two seasons: 2004-05, 2005-06 – Compiled from ESPN Soccernet , UCDP/PRIO Armed Conflict Data, Worldwide Governance Indicators (World Bank), Football Manager, 2005 and World Soccer Manager, 2006 • Modifications – Dyads composed of players i and j – Spatial effect variables representing the aggregate influence of team I ’s civil war exposure and behavior – Team I control variables: mean metrics among players on the team

• Team I ’s Aggregate Civil War Exposure Spatial Effect – Weighted average of the spatially lagged years of civil war in each teammate’s home country Variables – Weight: proportion of games played together 𝑌 𝑗𝑡 = 𝑥 𝑗𝑘𝑡 𝑦 𝑗𝑘𝑡 𝑘≠𝑗 • Team I ’s Aggregate Violent Behavior – Weighted average of the yellow/red/total cards received by each teammate – Weight: proportion of games played together 𝑍 𝑗𝑡 = 𝑥 𝑗𝑘𝑡 𝑧 𝑗𝑘𝑡 𝑘≠𝑗 𝑆 𝑗𝑡 = 𝑥 𝑗𝑘𝑡 𝑠 𝑗𝑘𝑡 𝑘≠𝑗 𝑈 𝑗𝑡 = 𝑥 𝑗𝑘𝑡 𝑢 𝑗𝑘𝑡 𝑘≠𝑗 i = player of analysis; j = i’s teammate; s = season; w = weight; x = years of civil war in player’s home country; y = yellow cards; r = red cards, t = total cards

• Dependent (player i ’s behavior) • Controls – Yellow, Red, and Total Cards Received – Field Position • Independent – Number of Games as a Starter – Number of Games as a Substitute – Team I ’s Aggregate Civil War Exposure – Age – Team I ’s Aggregate Behavior – Player Quality – Player j ’s Civil War Exposure • Goals Scored – Player j ’s Aggregate Behavior • Ln(Transfer Fee) – Rule of Law in Player i ’s Home Country • Fixed Effects: League, World Region

Methodology • Team-to-Player Peer Effects – Statistical analysis using monadic spatial effect variables • Player-to-Player Peer Effects – Statistical analysis within dyads composed of players i and j • Negative binomial regression • Standard errors clustered by player i ’s home country • Robustness Check: Zero-inflated negative binomial regression

RESULTS

Violent Behavior Civil War Exposure H 2 H 1 Team I H 4 H 3 Player j H 5

Table 1: The Effect of Team I ’s (1) (2) (3) (4) (5) (6) Aggregate Violent Behavior Yellow Cards Yellow Cards Red Cards Red Cards Total Cards Total Cards on Player i ’s Behavior Individual Team Individual Team Individual Team Controls Controls Controls Controls Controls Controls VARIABLES • Negative binomial Independent regression 0.155*** 0.624*** Spatially Lagged Yellow Cards • Standard errors in (0.0145) (0.0203) Received by Player j, Weight = Games Played Together parentheses 0.427** 1.570*** Spatially Lagged Red Cards Received • Significance at 90% (*), (0.217) (0.183) by Player j, Weight = Games Played 95% (**), and 99% (***) Together confidence levels 0.154*** 0.586*** Spatially Lagged Total Cards Received by Player j, Weight = Games (0.0124) (0.0178) Played Together H 2 : Team I ’s aggregate violent behavior exerts a peer effect Controls on player i ’s behavior. Years of Civil War in Player i’s Home 0.00800*** 0.00827* 0.0150** 0.0173*** 0.00839*** 0.00861** Country (0.00284) (0.00426) (0.00664) (0.00629) (0.00290) (0.00425) Player i Characteristics Yes No Yes No Yes No Team I Characteristics No Yes No Yes No Yes Constant -3.268*** -1.225* -5.308*** -3.283** -3.250*** -1.347** (0.337) (0.697) (0.618) (1.621) (0.319) (0.636) Observations 5,064 5,386 5,064 5,386 5,064 5,386 League Fixed Effects Yes Yes Yes Yes Yes Yes World Region Fixed Effects Yes Yes Yes Yes Yes Yes

Table 2: The Effect of Team I ’s (1) (2) (3) (4) (5) (6) Aggregate Violent Behavior Yellow Cards Yellow Cards Red Cards Red Cards Total Cards Total Cards on Player i ’s Behavior Individual Team Individual Team Individual Team – Incidence Rate Ratios VARIABLES Controls Controls Controls Controls Controls Controls • Independent Negative binomial Spatially Lagged Yellow Cards 1.168 1.866 regression Received by Player j, Weight = Games (10.69)** (30.71)** • Z-value in parentheses Played Together • Significance at 90% (*), 1.533 4.805 Spatially Lagged Red Cards Received by Player j, Weight = Games Played (1.96)* (8.58)** 95% (**), and 99% (***) Together confidence levels Spatially Lagged Total Cards 1.167 1.797 Received by Player j, Weight = Games (12.44)** (33.01)** Played Together Controls Years of Civil War in Player i’s Home 1.008 1.008 1.015 1.017 1.008 1.009 Country (2.81)** (1.94) (2.26)* (2.74)** (2.90)** (2.03)* Player i Characteristics Yes No Yes No Yes No Team I Characteristics No Yes No Yes No Yes Observations 5,064 5,386 5,064 5,386 5,064 5,386 League Fixed Effects Yes Yes Yes Yes Yes Yes World Region Fixed Effects Yes Yes Yes Yes Yes Yes

Table 3: The Effect of Team I ’s Aggregate Violent Behavior (1) (2) (3) (4) on Player i ’s Behavior Yellow Cards Yellow Cards Total Cards Total Cards – Robustness Check* Individual Team Individual Team Controls Controls Controls Controls VARIABLES • Zero-inflated negative Independent binomial regression Spatially Lagged Yellow Cards 0.0738*** 0.268*** • Standard errors in Received by Player j, Weight = Games (0.0105) (0.0116) Played Together parentheses Spatially Lagged Total Cards 0.0765*** 0.257*** • Significance at 90% (*), (0.00876) (0.0118) Received by Player j, Weight = Games 95% (**), and 99% (***) Played Together confidence levels Controls Years of Civil War in Player i’s Home 0.00529** 0.00546* 0.00567** 0.00591* *The red card models did not Country (0.00226) (0.00304) (0.00222) (0.00312) converge. Player i Characteristics Yes No Yes No Team I Characteristics No Yes No Yes Constant -0.968*** 0.926** -1.028*** 0.866** (0.192) (0.398) (0.202) (0.442) Observations 5,064 5,386 5,064 5,386 League Fixed Effects Yes Yes Yes Yes World Region Fixed Effects Yes Yes Yes Yes

The Predicted Effect of Team I ’s Aggregate Behavior on Yellow Cards Received by an Average Player

The Predicted Effect of Team I ’s Aggregate Behavior on Red Cards Received by an Average Player

The Predicted Effect of Team I ’s Aggregate Behavior on Total Cards Received by an Average Player

Player Yellow Cards Team I ’s Aggregate Behavior DEPORTIVO LA CORUÑA 5.249 Aldo Duscher 16 SPANISH PRIMERA DIVISION 5.587 Pedro Alvarez Munitis 12 5.742 Juan Capdevila 10 2005-2006 4.523 Jorge Andrade 9 5.105 Delgado Manuel Juanma 8 5.00655 Julian De Guzmán 8 5.879 Manuel Pablo 7 5.958 Sergio 7 6.0433 Diego Tristán 6 Fabricio Coloccini 6 5.507 Sanchez Victor 5 5.0800 4.552 Berenguel Hector 3 3.613 César 3 6.259 Jose Francisco Molina 3 4.552 Lionel Scaloni 3 5.148 Enrique Romero 2 4.103 Iglesias Iago 2 4.861 Javier Arizmendi 2 5.576 Rubén 2 Francisco Jiménez Tejada 1 4.310 Iván Carril 1 4.0636 Jeronimo Cabrera Momo 1 3.301 3.970 Sebastián Taborda 1 5.277 Juan Carlos Valerón 0 2.849 Pablo Amo 0 3.230 Senel 0

The Effect of Deportivo La Coruña’s Aggregate Behavior on Yellow Cards Received by Each of Its Players

Conclusions • Violent culture does not spread among individuals. • Violent behavior spreads from groups to individuals. – Why not person-to-person? Incorrect empirical strategy? • Real world implication: European refugee crisis