Online Appendix: Trafficking Networks and the
Mexican Drug War
Melissa Dell
February 6, 2015
Contents
A-1 Estimation appendix A–3
A-1.1 The shortest paths problem . . . . . . . . . . . . . . . . . . . . . . . . A–3
A-1.2 Solving for the congested trafficking equilibrium . . . . . . . . . . . . . A–3
A-1.3 Moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A–4
A-1.4 Maximizing the simulated method of moments objective function . . . A–5
A-1.5 Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A–6
A-1.6 The government’s resource allocation problem . . . . . . . . . . . . . . A–7
A-2 Additional Results A–10
A-2.1 Robustness of Balance Checks . . . . . . . . . . . . . . . . . . . . . . A–10
A-2.2 Robustness of Regressions Discontinuity Analysis . . . . . . . . . . . . A–20
A-2.3 Robustness to Using Differences-in-Differences . . . . . . . . . . . . . . A–26
A-2.4 Police-Criminal Confrontations . . . . . . . . . . . . . . . . . . . . . . A–37
A-2.5 Robustness of Heterogeneity Results . . . . . . . . . . . . . . . . . . . A–39
A-2.6 Robustness of Results on Local Politics and Violence . . . . . . . . . . A–50
A-2.7 Corruption and Other Results . . . . . . . . . . . . . . . . . . . . . . . A–61
A-2.8 Robustness of Spillover Results . . . . . . . . . . . . . . . . . . . . . . A–65
A-2.9 Law Enforcement Allocation Table . . . . . . . . . . . . . . . . . . . . A–78
A-2.10 Map of Close PAN Elections . . . . . . . . . . . . . . . . . . . . . . . A–80
A-2.11 Balance Figures for Pre-Characteristics . . . . . . . . . . . . . . . . . A–82
A-2.12 Balance Figures for the Predicted Homicide Rate . . . . . . . . . . . . A–87
A–1
A-2.13 McCrary Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A–89
A-2.14 Homicide RD Figures - Robustness . . . . . . . . . . . . . . . . . . . . A–92
A-2.15 Homicide RD Figures - Neighbors’ Homicide Rates . . . . . . . . . . .A–108
A-2.16 Robustness to Varying the Length of the Analysis Period . . . . . . .A–110
A-2.17 Spillovers Model Placeo Check . . . . . . . . . . . . . . . . . . . . . .A–113
A-2.18 Law Enforcement Allocation Figure . . . . . . . . . . . . . . . . . . .A–115
A–2
A-1 Estimation appendix
A-1.1 The shortest paths problem
The model setup is as follows: let N = (V , E) be an undirected graph representing the
Mexican road network, which consists of sets V of vertices and E of edges. Traffickers
transport drugs across the network from a set of origins to a set of destinations. Trafficking
paths connect origins to destinations. Formally, a trafficking path is an ordered set of nodes
such that an edge exists between two successive nodes. Each edge e ∈ E has a cost function
ce(le), where le is the length of the edge in kilometers. The total cost to traverse path p is
w(p) =∑
e∈p ce(le), which equals the length of the path. Close PAN victories remove edges
from the network. Let Pi denote the set of all possible paths between producing municipality
i and the United States. Each trafficker solves:
minp∈Pi
w(p) (A-1)
This problem, which amounts to choosing the shortest path between each producing munici-
pality and the nearest U.S. point of entry, can be solved using Dijkstra’s algorithm (Dijkstra,
1959).
A-1.2 Solving for the congested trafficking equilibrium
An equilibrium routing pattern must satisfy the following conditions (Wardrop, 1952):
1. For all p, p′ ∈ Pi with xp, xp′ > 0,∑
e∈p′ ce(xe, le) =∑
e∈p ce(xe, le).
2. For all p, p′ ∈ Pi with xp > 0 xp′ = 0,∑
e∈p′ ce(xe, le) ≥∑
e∈p ce(xe, le).
where xp is total flows on path p, xe is total flows on edge e, and ce(·) is the cost to traverse
edge e. The equilibrium routing pattern satisfying these conditions is the Nash equilibrium
of the game.
Beckmann, McGuire, and Winsten (1956) proved that the equilibrium can be character-
ized by a straightforward optimization problem. Specifically, the routing pattern x∗ is an
equilibrium if and only if it is a solution to:
min∑e∈E
∫ xe
0
ce(z)dz (A-2)
s.t.∑
p∈P|e∈p
xp = xe ∀e ∈ E (A-3)
A–3
∑p∈Pi
xp = 1 ∀i = 1, 2, . . . , ∀p ∈ P (A-4)
xp ≥ 0 ∀p ∈ P (A-5)
The first constraint requires that the flow of traffic on the paths traversing an edge sum
to the total flow of traffic on that edge, the second constraint requires that supply (equal
to 1 for each producer i) be conserved, and the third constraint requires flows to be non-
negative. By Weierstrass’s Theorem, a solution to the above problem exists, and thus a
trafficking equilibrium always exists.
While this problem does not have a closed-form solution, for a given network and spec-
ification of the congestion costs ce(·) it can be solved using numerical methods. I use the
Frank-Wolfe algorithm (1956), which generalizes Dantzig’s simplex algorithm to non-linear
programming problems. The Frank-Wolfe algorithm alternates between solving a linear pro-
gram defined by a tangential approximation of the objective function in (A-2) and a line
search that minimizes the objective over the line segment connecting the current iterate
and the solution to the linear programming problem. The linear subproblem determines the
direction of movement, and the line search selects the optimal step length in that direction.
At the end of each iteration, the current iterate is updated to the xe selected by the line
search problem. The linear subproblem defines a lower bound on the optimal value, which
is used in the termination criterion.
The tangential approximation to the objective given in (A-2) is a simple shortest paths
problem in which the costs to traverse each edge ce(xe, le) are evaluated at the current
iterate’s flows xke . In other words, the linear subproblem finds the shortest path between
each producing municipality and the nearest U.S. point of entry given edge costs of ce(xke , le)
at iteration k. The linear subproblem is solved using Dijkstra’s algorithm (Dijkstra, 1959).
The line search problem is solved using the golden section method (Kiefer, 1953).
A-1.3 Moments
In the baseline congestion model, the moments match the mean model predicted and observed
confiscations at ports, at terrestrial bordering crossings, and on interior edges. They also
match the interactions between port confiscations and the port’s container capacity, between
terrestrial crossing confiscations and the crossing’s number of commercial lanes, between
interior confiscations and the length of the interior edge, and between interior confiscations
and the length of the detour required to circumvent the edge. Finally, the moment conditions
match the model predicted and observed variance of confiscations across U.S. points of entry
A–4
and across interior edges. For the congestion models reported in the appendix that estimate
six separate crossing congestion parameters, the moment conditions match mean model
predicted and observed confiscations for each of the six separate groups of crossings, instead
of matching mean confiscations for all ports and for all terrestrial border crossings.
The model with DTO territorial costs and no congestion matches mean confiscations,
mean confiscations interacted with DTO presence, and mean confiscations interacted with
the share of Mexico’s territory (if any) that the municipality’s DTO controls. The model
that includes congestion matches the same moments as in the baseline congestion model as
well as the two moments that interact confiscations and DTO presence/share.
The model with a PAN cost parameter and no congestion matches the mean monthly
change in confiscations in municipalities that do not have a PAN mayor elected during
the Calderon period. The sample is limited to these municipalities because it is plausible
that enforcement remains constant. The model also matches the mean monthly change in
confiscations in municipalities bordering a municipality with a PAN mayor elected during
the sample period. These municipalities are useful for estimating the PAN cost parameter
because drug traffic is often diverted to them. The model that includes both a PAN cost
parameter and congestion matches the same moments as in the baseline congestion model,
as well as the two moments that summarize changes in confiscations.
A-1.4 Maximizing the simulated method of moments objective
function
The simulated method of moments (SMM) estimator θ minimizes a weighted quadratic form:
θ = argminθ∈Θ
1
M
[M∑m=1
g(Xm, θ)
]′Σ
[M∑m=1
g(Xm, θ)
](A-6)
where g(·) is an estimate of the true moment function, M is the number of municipalities in
the sample, and Σ is an L x L positive semi-definite weighting matrix.
The SMM objective function is not globally convex, and thus standard gradient methods
may perform poorly. Instead, I use simulated annealing (Kirkpatrick, Gelatt, and Vecchi,
1983), which is more suitable for problems that lack a globally convex objective.1 Simulated
annealing is a non-gradient iterative method that differs from gradient methods in permitting
movements that increase the objective function being minimized.
Given a value of θs for the congestion parameters at the sth iteration, the algo-
1See Goffe, Ferrier, and Rogers (1994) for a comprehensive review and Cameron and Trivedi (2005, p.347) for a textbook treatment.
A–5
rithm perturbs the jth component of θs so as to obtain a new trial value of θ∗s =
θs + [0 . . . 0 (λsrs) 0 . . . 0]′, where λs is a pre-specified step length and rs is a draw from
a uniform distribution on (−1, 1). The method sets θs+1 = θ∗s if the perturbation decreases
the objective function. If θ∗s does not decrease the objective, it is accepted with probability1
1+exp( ∆Ts
), where ∆ is the change in value of the objective and Ts is a positive scaling pa-
rameter called the temperature. Uphill moves are accepted with a probability that declines
with the change in the objective function and increases with the temperature.2 The temper-
ature is set to T0 at the initial iteration and updated according to the temperature schedule
Tk = T0/k. The annealing parameter k is initially set equal to the iteration number. If after
a given number of iterations convergence has not been achieved, k is set to some value less
than the iteration number so that the temperature increases and the algorithm can move
to a potentially more promising region of the parameter space. The dependency between
the temperature and acceptance probability is such that the current solution changes almost
randomly when T is large and increasingly downhill as T goes to zero.
The algorithm runs until the average change in value of the objective function over a
given number of iterations is less than some small number ε. I choose the starting value
using a grid search over the parameter space. Results (available upon request) are robust to
the use of different starting values and annealing parameters, with these choices primarily
affecting the speed with which the algorithm converges.
A-1.5 Inference
Predicted confiscations on a given edge are not independent of predicted confiscations else-
where in the network, introducing spatial dependence. Conley (1999) explores method of
moments estimators for data exhibiting spatial dependence, showing that the sufficient condi-
tions for consistency and normality require the dependence amongst observations to die away
as the distance between the observations increases. This condition appears likely to hold
in the current application, since drugs are typically trafficked to relatively close crossings.
With the presence of spatial dependence, the asymptotic covariance matrix Λ is replaced by
a weighted average of spatial autocovariance terms with zero weights for observations farther
than a certain distance (Conley, 1999):
λ =1
M
∑m
∑s∈Munm
[g(Xm, θ)g(Xm, θ)′] (A-7)
where Munm is the set of all municipalities within 250 kilometers of municipality m, in-
2Since both ∆ and Ts are positive, the probably of acceptance is between zero and one half.
A–6
cluding municipality m. The implicit assumption is that the correlation between observations
is negligible for municipalities beyond 250 kilometers.
A-1.6 The government’s resource allocation problem
To apply the trafficking framework to policy analysis, I embed the trafficking model in
a Stackelberg network game (Bas and Srikant, 2002). In the first stage, the government
(a single player) decides how to allocate law enforcement resources to edges in the road
network, subject to a budget constraint. The edges selected by the government are referred
to as vital edges. Traffickers’ costs of traversing an edge increase when law enforcement
resources are placed on it. The network model best predicts the diversion of drug traffic
following PAN victories when I assume that they increase trafficking costs by a factor of
three. Thus, I assume that each police checkpoint increases the effective length of selected
edges by 3×9 = 27 kilometers, where 9 kilometers is the average edge length in the network.3
With more information on the resources deployed in PAN crackdowns, it would be possible
to construct more precise estimates of the costs that law enforcement resources impose on
traffickers.
In the second stage, traffickers simultaneously select least cost routes to the U.S. The
government’s objective is to maximize the total costs that traffickers incur, and each trafficker
minimizes his own costs. The scenario in which traffickers respond to the government’s action
by choosing the shortest path to the U.S. is a special case in which congestion costs are zero.
Ball, Golden, and Vohra (1989) showed that this special case is NP hard, and thus it follows
that the more general problem is also NP-hard. That is, the time required to solve for the
optimum increases quickly as the size of the problem grows. Even if we focused on the
simpler model with no congestion costs, solving for the optimum using an exhaustive search
would have an order of complexity of O(V !), where V (the number of vertices) equals 13,969,
and thus would take trillions of years to run.
Developing algorithms for problems similar to the one described here is an active area
of operations research and computer science. For example, researchers have examined the
problem of identifying vital edges in critical infrastructure networks, such as oil pipelines
and electricity grids, so that these edges can be better defended against terrorist attacks
and the systems made more robust (see, for example, Brown, Carlyle, Salmeron and Wood,
2005). To the best of my knowledge there are currently no known algorithms for solving the
3An alternative assumption is that police checkpoints multiply the effective length of edges by a givenfactor. However, this would imply that checkpoints increase the costs of longer edges by more than theyincrease the costs of shorter edges. The multiplicative costs assumption appears reasonable for PAN crack-downs, as larger municipalities have more police and are likely to receive larger federal police and militarycontingents, but the assumption appears less appropriate for police checkpoints.
A–7
government’s resource allocation problem that are both exact (guaranteed to converge to
optimality) and feasible given the size of the network, either for the network with congestion
or for the simpler problem in which congestion costs are zero.4 Developing a fast, exact
algorithm for this problem is a challenging endeavor that is significantly beyond the scope
of the current study. Thus, I instead use the following approximate heuristic to solve for the
k vital edges:
1. For each of k iterations, calculate how total trafficking costs respond to individuallyincreasing the edge lengths of each of the N most trafficked edges in the network.
2. Assign each element of this set of N edges a rank, m = 1 . . . N , such that the removalof edge m = 1 would increase trafficking costs the most, the removal of edge m = 2would increase trafficking costs the second most . . . and the removal of edge m = Nwould increase trafficking costs the least.
3. Increase the effective length of the edge with m = 1 by a pre-specified amount.
4. Terminate if k iterations have been completed and return to step 1 otherwise.
Appendix Figure A-28 plots the results of this exercise with k = 25 and N = 250,
highlighting municipalities that contain a vital edge in yellow. The average monthly drug
trade-related homicide rate between 2007 and 2009 is plotted in the background. Allocating
police checkpoints to these 25 edges increases the total length of the network by 0.043 percent
and increases total trafficking costs by 17 percent. Appendix Table A-61 documents that
results are similar when I instead: a) choose values of N ranging from 100 to 500, b) alternate
in step 3 between selecting the edges with m = 1 and m = 2, c) alternate in step 3 between
selecting the edges with m = 1, m = 2, and m = 3, and d) remove the edge with m = 2,
m = 3, m = 4, or m = 5 when k = 1 and remove the edge with m = 1 when k = 2 . . . 25.
4Malik, Mittal, and Gupta (1989) suggest an algorithm for finding k vital edges in the shortest pathproblem, but unfortunately it is theoretically flawed (see Israeli and Wood (2002) for a discussion). Themost closely related work is by Israeli and Wood (2002), who develop an efficient algorithm for solvingfor k vital edges in the context of a shortest path problem on a directed graph with a single origin anddestination. Even if the algorithm, which involves considerable mathematical machinery, could be extendedto this paper’s undirected graph with multiple origins, it is unlikely to be feasible on a network of the sizeexamined here and does not accommodate congestion costs. Existing vital edge algorithms focus on shortestpath or max flow problems (i.e. Lim and Smith, 2007) , and to the best of my knowledge researchers havenot examined the vital edge problem in a congested network.
A–8
References
Acemoglu, D. and M. Dell (2009): “Beyond Neoclassical Growth: Technology, Hu-man Capital, Institutions and Within-Country Differences,” American Economic Journal:Macroeconomics, 2, 169–188.
Ball, M., B. Golden, and R. Vohra (1989): “Finding the most vital arcs in a network,”Operations Research Letters, 8, 73–76.
Beckmann, M., C. McGuire, and C. Winston (1956): Studies in the Economics ofTransportation, Yale University Press.
Brown, G., M. Carlyle, J. Salmeron, and K. Wood (2005): “Analyzing the vul-nerability of critical infrastructure to attack and planning defenses,” INFORMS tutorialsin operations research, 102–123.
Cameron, A. (2005): Microeconometrics: methods and applications, Cambridge universitypress.
Dijkstra, E. (1959): “A note on two problems in connection with graphs,” Numerischemathematik, 1, 269–271.
Goffe, W., G. Ferrier, and J. Rogers (1994): “Global optimization of statisticalfunctions with simulated annealing,” Journal of Econometrics, 60, 65–99.
Israeli, E. and R. Wood (2002): “Shortest-path network interdiction,” Networks, 40,97–111.
Kiefer, J. (1953): “Sequential minimax search for a maximum,” in Proceedings of theAmerican Mathematical Society, vol. 4, 502–506.
Kirkpatrick, S., C. Gelatt, and M. Vecchi (1983): “Optimization by simulatedannealing,” Science, 220, 671–680.
Lim, C. and J. Smith (2007): “Algorithms for discrete and continuous multicommodityflow network interdiction problems,” IIE Transactions, 39, 15–26.
Malik, K., A. Mittal, and S. Gupta (1989): “The k most vital arcs in the shortestpath problem,” Operations Research Letters, 8, 223–227.
Wardrop, J. (1952): “Some Theoretical Aspects of Road Traffic Research,” Proceedings ofthe Institution of Civil Engineers, 1, 325–378.
A–9
A-2 Additional Results
A-2.1 Robustness of Balance Checks
A–10
Table A-1: Baseline Characteristics (4% vote spread, 2007-2008)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 64.52 50.13 (0.96) 53.89 (0.89) 8.76 (0.28)Turnout 0.61 0.59 (0.58) 0.04 (0.56) 0.01 (0.13)PAN incumbent 0.26 0.29 (-0.34) 0.07 (0.27) 0.07 (0.63)PRD incumbent 0.16 0.13 (0.54) -0.12 (-0.65) -0.13 (-0.95)% alternations (1976-2006) 0.31 0.31 (0.08) 0.06 (0.64) 0.00 (0.07)PRI never lost (1976-2006) 0.08 0.08 (0.03) -0.21 (-1.52) -0.10 (-0.96)Demographic characteristicsPopulation (2005) 6.48 5.12 (0.45) 2.47 (0.24) -3.54 (-0.93)Population density (2005) 197.63 210.49 (-0.16) -615.53** (-1.99) -376.16** (-2.02)Migrants per capita (2005) 0.02 0.02 (-0.80) 0.00 (-0.54) -0.01 (-1.42)Economic characteristicsIncome per capita (2005) 4.37 4.40 (-0.06) -0.56 (-0.37) 0.61 (0.82)Malnutrition (2005) 32.18 31.52 (0.20) 2.06 (0.23) -7.76 (-1.20)Mean years schooling (2005) 6.23 6.17 (0.22) -1.24 (-1.44) -0.23 (-0.42)Infant mortality (2005) 22.35 21.97 (0.30) 1.80 (0.43) -1.59 (-0.69)HH w/o access to sewage (2005) 8.05 8.23 (-0.13) -2.43 (-0.73) -5.46* (-1.74)HH w/o access to water (2005) 17.15 15.93 (0.33) -15.69* (-1.84) -11.22 (-1.48)Marginality index (2005) -0.16 -0.12 (-0.26) -0.06 (-0.13) -0.44 (-1.25)Road network characteristicsDetour length (km) 29.40 24.16 (0.23) -76.85* (-1.90) -33.17* (-1.71)Road density 0.15 0.13 (0.80) -0.10 (-1.61) -0.11** (-2.01)Distance U.S. (km) 708.09 765.78 (-1.05) -104.77 (-0.59) -120.37 (-0.68)Geographic characteristicsElevation (m) 1365.84 1398.81 (-0.22) 426.08 (0.84) 392.43 (0.84)Slope (degrees) 3.65 3.38 (0.57) 0.13 (0.10) -0.24 (-0.23)Surface area (km2) 1951.44 535.23 (1.59) 1048.62 (0.68) 53.41 (0.05)Average min. temperature, C 7.29 7.76 (-0.46) -4.20 (-1.20) -3.79 (-1.15)Average max. temperature, C 22.52 23.22 (-0.95) -3.82 (-1.46) -3.66 (-1.56)Average precipitation, cm 1160.13 1056.88 (0.78) 21.76 (0.07) 11.08 (0.03)Observations 61 62 123 123
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the author’s own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections in 2007-2008. Column (6) and (7) examine these characteristics for municipalities that border amunicipality with a close election in 2007-2008. Column (3) reports the t-statistic on the difference in means betweenmunicipalities where the PAN barely won and where they barely lost. Columns (4) and (6) report the coefficient onPAN win from a standard RD specification where the respective characteristic is used as the dependent variable, andcolumns (5) and (7) report the respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-2: Baseline Characteristics (3% vote spread, 2007-2008)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 67.85 48.18 (1.13) 41.17 (0.54) -31.11 (-0.70)Turnout 0.60 0.60 (0.27) 0.06 (0.70) 0.00 (0.04)PAN incumbent 0.23 0.30 (-0.82) -0.10 (-0.33) 0.00 (0.02)PRD incumbent 0.21 0.11 (1.32) 0.08 (0.37) -0.04 (-0.27)% alternations (1976-2006) 0.32 0.30 (0.51) 0.11 (1.00) -0.03 (-0.43)PRI never lost (1976-2006) 0.08 0.11 (-0.41) -0.24 (-1.64) -0.03 (-0.28)Demographic characteristicsPopulation (2005) 7.87 5.64 (0.58) -2.39 (-0.18) -8.50 (-1.55)Population density (2005) 218.11 249.33 (-0.30) -836.07** (-2.23) -505.72** (-2.31)Migrants per capita (2005) 0.02 0.02 (-0.50) -0.01 (-0.99) -0.01** (-2.08)Economic characteristicsIncome per capita (2005) 4.46 4.43 (0.05) -0.95 (-0.50) -0.18 (-0.19)Malnutrition (2005) 30.77 31.24 (-0.12) 1.87 (0.18) -3.61 (-0.46)Mean years schooling (2005) 6.39 6.24 (0.49) -1.59 (-1.54) -0.82 (-1.26)Infant mortality (2005) 22.00 22.03 (-0.02) 2.05 (0.42) -1.04 (-0.39)HH w/o access to sewage (2005) 8.33 8.06 (0.16) -3.42 (-0.84) -5.21 (-1.42)HH w/o access to water (2005) 18.75 16.24 (0.56) -11.93 (-1.28) -10.01 (-1.14)Marginality index (2005) -0.20 -0.12 (-0.41) 0.21 (0.38) -0.11 (-0.26)Road network characteristicsDetour length (km) 36.72 32.31 (0.15) -86.28 (-1.63) -58.93** (-2.17)Road density 0.15 0.15 (0.27) -0.19** (-2.50) -0.16*** (-2.65)Distance U.S. (km) 680.18 770.18 (-1.57) -67.33 (-0.32) -79.38 (-0.38)Geographic characteristicsElevation (m) 1439.71 1380.10 (0.34) 432.09 (0.76) 256.36 (0.49)Slope (degrees) 3.57 3.46 (0.20) -0.24 (-0.17) -0.43 (-0.37)Surface area (km2) 2246.80 448.90 (1.60) 284.59 (0.11) -393.16 (-0.23)Average min. temperature, C 6.62 8.00 (-1.22) -4.89 (-1.20) -3.64 (-0.95)Average max. temperature, C 22.03 23.24 (-1.49) -3.86 (-1.25) -3.06 (-1.11)Average precipitation, cm 1106.39 1071.97 (0.24) 61.04 (0.16) 86.84 (0.23)Observations 48 46 94 94
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the authors own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections in 2007-2008. Column (6) and (7) examine these characteristics for municipalities that border amunicipality with a close election in 2007-2008. Column (3) reports the t-statistic on the difference in means betweenmunicipalities where the PAN barely won and where they barely lost. Columns (4) and (6) report the coefficient onPAN win from a standard RD specification where the respective characteristic is used as the dependent variable, andcolumns (5) and (7) report the respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-3: Baseline Characteristics (2% vote spread, 2007-2008)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 82.31 50.58 (1.32) 125.10 (1.24) 6.69 (0.12)Turnout 0.60 0.58 (0.48) 0.06 (0.62) 0.05 (0.73)PAN incumbent 0.27 0.28 (-0.03) 0.06 (0.14) 0.03 (0.19)PRD incumbent 0.15 0.14 (0.15) -0.13 (-0.51) 0.00 (-0.01)% alternations (1976-2006) 0.32 0.30 (0.28) 0.07 (0.51) -0.03 (-0.33)PRI never lost (1976-2006) 0.09 0.14 (-0.57) -0.26 (-1.42) 0.04 (0.23)Demographic characteristicsPopulation (2005) 10.06 5.43 (0.89) 5.54 (0.30) 3.71 (0.38)Population density (2005) 257.93 313.69 (-0.36) -813.01 (-1.52) -374.10 (-1.26)Migrants per capita (2005) 0.02 0.02 (-0.19) 0.00 (0.07) 0.00 (-0.53)Economic characteristicsIncome per capita (2005) 4.65 4.88 (-0.32) 0.61 (0.23) 0.38 (0.32)Malnutrition (2005) 30.11 26.33 (0.83) -5.51 (-0.39) -13.74 (-1.48)Mean years schooling (2005) 6.49 6.58 (-0.24) -0.93 (-0.66) -0.07 (-0.08)Infant mortality (2005) 22.25 20.90 (0.74) 4.88 (0.71) -0.10 (-0.03)HH w/o access to sewage (2005) 8.17 7.22 (0.49) -1.15 (-0.21) -8.01 (-1.54)HH w/o access to water (2005) 17.49 14.63 (0.53) 4.77 (0.42) -7.35 (-0.61)Marginality index (2005) -0.27 -0.28 (0.03) 0.28 (0.37) -0.37 (-0.70)Road network characteristicsDetour length (km) 45.17 49.45 (-0.10) 8.21 (0.16) -3.35 (-0.07)Road density 0.17 0.15 (0.48) -0.16 (-1.49) -0.06 (-0.74)Distance U.S. (km) 654.64 740.40 (-1.13) -262.07 (-0.99) -263.20 (-1.01)Geographic characteristicsElevation (m) 1473.84 1299.94 (0.81) 62.24 (0.08) -132.31 (-0.20)Slope (degrees) 3.55 3.14 (0.58) -1.22 (-0.63) -1.17 (-0.70)Surface area (km2) 2788.02 528.14 (1.39) 3712.91* (1.69) 4011.75 (1.56)Average min. temperature, C 6.26 7.94 (-1.21) -5.95 (-1.12) -4.24 (-0.84)Average max. temperature, C 21.58 23.32 (-1.72*) -5.68 (-1.43) -4.67 (-1.30)Average precipitation, cm 1065.19 1029.42 (0.21) 3.28 (0.01) 96.62 (0.20)Observations 33 29 62 62
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the authors own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections in 2007-2008. Column (6) and (7) examine these characteristics for municipalities that border amunicipality with a close election in 2007-2008. Column (3) reports the t-statistic on the difference in means betweenmunicipalities where the PAN barely won and where they barely lost. Columns (4) and (6) report the coefficient onPAN win from a standard RD specification where the respective characteristic is used as the dependent variable, andcolumns (5) and (7) report the respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-4: Baseline Characteristics (13.3% vote spread, 2007-2008)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 66.59 58.11 (0.23) 38.85 (1.28) 42.14* (1.65)Turnout 0.61 0.59 (0.99) 0.01 (0.15) 0.01 (0.20)PAN incumbent 0.25 0.32 (-0.61) 0.00 (0.04) 0.02 (0.24)PRD incumbent 0.14 0.13 (0.63) 0.06 (0.58) -0.06 (-0.91)% alternations (1976-2006) 0.32 0.30 (-0.01) 0.02 (0.41) -0.04 (-1.20)PRI never lost (1976-2006) 0.12 0.10 (-0.04) -0.12 (-1.17) -0.03 (-0.44)Demographic characteristicsPopulation (2005) 5.28 4.34 (0.35) 4.13 (0.70) 1.88 (0.73)Population density (2005) 207.02 195.96 (0.42) -197.74 (-1.06) -69.11 (-0.63)Migrants per capita (2005) 0.02 0.02 (-0.69) 0.00 (-0.27) 0.00 (0.64)Economic characteristicsIncome per capita (2005) 4.28 4.36 (-0.53) -0.12 (-0.15) 0.47 (0.89)Malnutrition (2005) 33.08 31.79 (0.53) 0.24 (0.04) -4.22 (-0.97)Mean years schooling (2005) 6.22 6.11 (0.32) -0.17 (-0.36) 0.13 (0.35)Infant mortality (2005) 22.56 22.54 (0.22) 1.14 (0.50) 0.61 (0.38)HH w/o access to sewage (2005) 8.29 9.04 (0.05) 0.72 (0.32) -0.63 (-0.33)HH w/o access to water (2005) 17.52 18.48 (-0.62) 1.80 (0.30) -2.28 (-0.52)Marginality index (2005) -0.10 -0.05 (-0.23) -0.08 (-0.27) -0.22 (-0.95)Road network characteristicsDetour length (km) 22.65 22.29 (0.19) -14.57 (-0.35) 6.21 (0.36)Road density 0.16 0.14 (0.98) -0.02 (-0.42) -0.03 (-0.76)Distance U.S. (km) 732.16 759.47 (-0.55) -127.59 (-1.37) -131.39 (-1.41)Geographic characteristicsElevation (m) 1363.85 1367.75 (0.26) 327.64 (1.19) 273.58 (1.08)Slope (degrees) 3.60 3.32 (1.02) 0.25 (0.29) -0.02 (-0.02)Surface area (km2) 1613.60 748.56 (1.36) 2422.76* (1.73) 1463.56* (1.76)Average min. temperature, C 7.61 7.79 (-0.46) -3.41* (-1.92) -3.04* (-1.82)Average max. temperature, C 22.64 23.19 (-0.53) -2.54* (-1.91) -2.38** (-1.99)Average precipitation, cm 1217.80 1112.02 (0.65) -55.13 (-0.28) -62.91 (-0.32)Observations 168 212 380 380
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the authors own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections in 2007-2008. Column (6) and (7) examine these characteristics for municipalities that border amunicipality with a close election in 2007-2008. Column (3) reports the t-statistic on the difference in means betweenmunicipalities where the PAN barely won and where they barely lost. Columns (4) and (6) report the coefficient onPAN win from a standard RD specification where the respective characteristic is used as the dependent variable, andcolumns (5) and (7) report the respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-5: Baseline Characteristics (5% vote spread, 2007-2010)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 76.24 85.61 (-0.69) -27.04 (-0.38) 7.28 (0.17)Turnout 0.67 0.64 (1.54) -0.03 (-0.50) -0.04 (-0.90)PAN incumbent 0.37 0.37 (0.12) 0.08 (0.51) 0.10 (1.13)PRD incumbent 0.11 0.10 (0.37) 0.09 (0.82) -0.05 (-0.67)% alternations (1976-2006) 0.27 0.29 (-1.04) 0.04 (0.85) 0.04 (1.34)PRI never lost (1976-2006) 0.13 0.10 (0.90) -0.09 (-0.78) -0.16** (-2.42)Demographic characteristicsPopulation (2005) 3.74 6.11 (-1.43) 1.83 (0.36) -3.00 (-1.53)Population density (2005) 136.31 226.96 (-1.32) -242.43 (-1.46) -192.40 (-1.57)Migrants per capita (2005) 0.02 0.02 (-0.15) 0.01 (1.41) 0.00 (0.53)Economic characteristicsIncome per capita (2005) 4.57 4.94 (-1.58) -0.58 (-0.69) -0.09 (-0.17)Malnutrition (2005) 27.45 26.52 (0.50) 2.30 (0.41) -3.57 (-0.83)Mean years schooling (2005) 6.27 6.41 (-0.90) -0.49 (-0.97) -0.18 (-0.51)Infant mortality (2005) 22.76 22.08 (0.79) -0.30 (-0.11) 0.05 (0.03)HH w/o access to sewage (2005) 11.11 10.55 (0.41) -1.47 (-0.31) -1.27 (-0.41)HH w/o access to water (2005) 13.98 14.43 (-0.22) -5.09 (-0.97) -2.77 (-0.62)Marginality index (2005) -0.29 -0.30 (0.17) -0.12 (-0.41) -0.17 (-0.71)Road network characteristicsDetour length (km) 17.67 17.46 (0.02) -21.32 (-0.94) -3.57 (-0.36)Road density 0.13 0.13 (0.01) -0.04 (-1.02) -0.02 (-0.63)Distance U.S. (km) 776.52 781.72 (-0.10) -111.11 (-0.72) -113.45 (-0.74)Geographic characteristicsElevation (m) 1276.19 1264.79 (0.12) 401.26 (1.46) 406.91 (1.57)Slope (degrees) 3.10 2.84 (0.92) 0.29 (0.31) 0.15 (0.21)Surface area (km2) 1372.19 1084.88 (0.73) 911.82 (1.14) 422.22 (0.60)Average min. temperature, C 7.66 7.83 (-0.28) -3.01 (-1.56) -2.86 (-1.54)Average max. temperature, C 23.22 23.22 0.00 -2.42 (-1.59) -2.47* (-1.78)Average precipitation, cm 948.41 941.35 (0.11) -72.78 (-0.39) -61.14 (-0.34)Observations 155 155 310 310
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the authors own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections. Column (6) and (7) examine these characteristics for municipalities that border a municipality with a closeelection. Column (3) reports the t-statistic on the difference in means between municipalities where the PAN barelywon and where they barely lost. Columns (4) and (6) report the coefficient on PAN win from a standard RDspecification where the respective characteristic is used as the dependent variable, and columns (5) and (7) reportthe respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-6: Baseline Characteristics (4% vote spread, 2007-2010)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 79.03 78.40 (0.04) -41.42 (-0.47) 38.88 (0.80)Turnout 0.67 0.65 (0.53) -0.06 (-0.75) -0.04 (-0.70)PAN incumbent 0.36 0.36 (0.06) -0.01 (-0.07) 0.08 (0.77)PRD incumbent 0.11 0.08 (0.72) 0.01 (0.13) -0.07 (-0.84)% alternations (1976-2006) 0.27 0.29 (-0.73) 0.06 (1.09) 0.04 (1.21)PRI never lost (1976-2006) 0.14 0.10 (1.01) -0.04 (-0.34) -0.16** (-2.20)Demographic characteristicsPopulation (2005) 4.14 4.79 (-0.39) 1.65 (0.28) -2.04 (-0.86)Population density (2005) 118.32 157.30 (-0.88) -318.49* (-1.82) -225.44* (-1.96)Migrants per capita (2005) 0.02 0.02 (-0.08) 0.00 (0.71) 0.00 (0.21)Economic characteristicsIncome per capita (2005) 4.59 4.82 (-0.85) -0.40 (-0.40) 0.42 (0.65)Malnutrition (2005) 27.22 26.91 (0.15) 0.88 (0.13) -6.53 (-1.38)Mean years schooling (2005) 6.29 6.34 (-0.32) -0.58 (-0.98) 0.00 0.00Infant mortality (2005) 22.74 22.27 (0.51) -0.14 (-0.05) -1.76 (-0.98)HH w/o access to sewage (2005) 10.06 11.48 (-0.95) -3.13 (-0.55) -3.72 (-1.03)HH w/o access to water (2005) 14.34 13.24 (0.49) -8.96 (-1.54) -5.65 (-1.13)Marginality index (2005) -0.31 -0.26 (-0.42) -0.21 (-0.58) -0.38 (-1.40)Road network characteristicsDetour length (km) 20.58 16.63 (0.35) -35.07* (-1.70) -8.45 (-0.84)Road density 0.13 0.13 (0.40) -0.04 (-1.13) -0.04 (-1.16)Distance U.S. (km) 763.38 816.01 (-0.89) -63.01 (-0.35) -69.21 (-0.38)Geographic characteristicsElevation (m) 1249.72 1212.67 (0.34) 472.14 (1.49) 476.90 (1.60)Slope (degrees) 3.22 2.87 (1.09) 0.21 (0.20) 0.08 (0.11)Surface area (km2) 1513.15 1028.66 (1.04) 818.71 (0.98) 454.31 (0.55)Average min. temperature, C 7.71 8.26 (-0.78) -3.18 (-1.39) -3.08 (-1.41)Average max. temperature, C 23.22 23.49 (-0.54) -2.68 (-1.49) -2.73* (-1.67)Average precipitation, cm 966.66 925.62 (0.57) -21.92 (-0.11) -10.31 (-0.05)Observations 129 122 251 251
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the authors own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections. Column (6) and (7) examine these characteristics for municipalities that border a municipality with a closeelection. Column (3) reports the t-statistic on the difference in means between municipalities where the PAN barelywon and where they barely lost. Columns (4) and (6) report the coefficient on PAN win from a standard RDspecification where the respective characteristic is used as the dependent variable, and columns (5) and (7) reportthe respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-7: Baseline Characteristics (3% vote spread, 2007-2010)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 85.90 80.76 (0.26) -72.69 (-0.65) 10.60 (0.19)Turnout 0.64 0.64 (0.06) -0.05 (-0.53) -0.03 (-0.56)PAN incumbent 0.39 0.35 (0.53) -0.02 (-0.10) 0.03 (0.29)PRD incumbent 0.14 0.07 (1.61) 0.09 (0.73) -0.03 (-0.36)% alternations (1976-2006) 0.27 0.28 (-0.57) 0.04 (0.63) 0.02 (0.40)PRI never lost (1976-2006) 0.15 0.11 (0.76) 0.07 (0.51) -0.10 (-1.19)Demographic characteristicsPopulation (2005) 5.11 4.28 (0.40) 0.12 (0.02) -3.91 (-1.28)Population density (2005) 134.39 153.45 (-0.35) -384.00* (-1.88) -259.43** (-2.17)Migrants per capita (2005) 0.02 0.02 (0.66) 0.00 (0.61) 0.00 (-0.48)Economic characteristicsIncome per capita (2005) 4.61 4.80 (-0.56) -0.50 (-0.41) 0.19 (0.25)Malnutrition (2005) 26.86 27.05 (-0.08) -0.93 (-0.12) -5.56 (-1.04)Mean years schooling (2005) 6.38 6.39 (-0.05) -0.57 (-0.82) -0.12 (-0.26)Infant mortality (2005) 22.73 22.56 (0.15) -0.60 (-0.17) -2.05 (-0.99)HH w/o access to sewage (2005) 9.79 11.62 (-1.00) -4.10 (-0.59) -3.25 (-0.78)HH w/o access to water (2005) 15.94 13.88 (0.73) -8.13 (-1.26) -6.57 (-1.13)Marginality index (2005) -0.34 -0.26 (-0.62) -0.17 (-0.40) -0.30 (-0.99)Road network characteristicsDetour length (km) 26.90 19.51 (0.49) -41.76* (-1.73) -20.51 (-1.56)Road density 0.13 0.13 (0.13) -0.07* (-1.66) -0.06* (-1.89)Distance U.S. (km) 679.02 766.78 (-1.42) -11.13 (-0.05) -14.72 (-0.07)Geographic characteristicsElevation (m) 1325.30 1247.59 (0.63) 438.35 (1.22) 379.92 (1.13)Slope (degrees) 3.39 3.02 (0.95) -0.58 (-0.49) -0.35 (-0.41)Surface area (km2) 1729.22 1060.67 (1.08) 674.43 (0.54) 818.35 (0.81)Average min. temperature, C 6.92 7.95 (-1.34) -3.21 (-1.21) -2.73 (-1.08)Average max. temperature, C 22.66 23.24 (-1.01) -2.29 (-1.08) -2.01 (-1.04)Average precipitation, cm 934.10 916.70 (0.21) -11.08 (-0.05) 17.62 (0.08)Observations 95 91 186 186
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the authors own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections. Column (6) and (7) examine these characteristics for municipalities that border a municipality with a closeelection. Column (3) reports the t-statistic on the difference in means between municipalities where the PAN barelywon and where they barely lost. Columns (4) and (6) report the coefficient on PAN win from a standard RDspecification where the respective characteristic is used as the dependent variable, and columns (5) and (7) reportthe respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-8: Baseline Characteristics (2% vote spread, 2007-2010)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 98.68 89.03 (0.36) -87.77 (-0.58) -31.19 (-0.38)Turnout 0.63 0.65 (-0.39) -0.04 (-0.35) 0.00 (0.05)PAN incumbent 0.42 0.37 (0.54) 0.16 (0.59) 0.07 (0.54)PRD incumbent 0.12 0.08 (0.87) -0.08 (-0.66) -0.03 (-0.30)% alternations (1976-2006) 0.27 0.27 (-0.18) -0.02 (-0.30) -0.02 (-0.49)PRI never lost (1976-2006) 0.15 0.14 (0.25) 0.10 (0.53) -0.02 (-0.19)Demographic characteristicsPopulation (2005) 5.82 4.15 (0.62) 0.55 (0.06) -4.98 (-1.31)Population density (2005) 149.18 170.68 (-0.29) -474.11* (-1.81) -289.46* (-1.96)Migrants per capita (2005) 0.02 0.02 (0.80) 0.00 (-0.52) 0.00 (-0.66)Economic characteristicsIncome per capita (2005) 4.72 5.07 (-0.84) -0.34 (-0.22) -0.25 (-0.26)Malnutrition (2005) 26.37 23.99 (0.83) 1.48 (0.15) -4.70 (-0.73)Mean years schooling (2005) 6.42 6.60 (-0.81) -0.50 (-0.56) -0.23 (-0.40)Infant mortality (2005) 23.03 21.86 (0.86) 3.88 (0.80) 1.02 (0.38)HH w/o access to sewage (2005) 9.79 10.95 (-0.52) -0.05 (-0.01) 0.00 (-0.00)HH w/o access to water (2005) 14.48 12.52 (0.61) 2.64 (0.34) -0.33 (-0.05)Marginality index (2005) -0.40 -0.37 (-0.19) 0.07 (0.13) -0.12 (-0.30)Road network characteristicsDetour length (km) 33.06 25.45 (0.36) -10.91 (-0.41) -9.92 (-0.54)Road density 0.14 0.13 (0.59) -0.11** (-2.07) -0.08** (-2.12)Distance U.S. (km) 656.29 752.45 (-1.24) -125.61 (-0.46) -130.07 (-0.48)Geographic characteristicsElevation (m) 1360.00 1170.44 (1.32) 298.37 (0.69) 279.43 (0.69)Slope (degrees) 3.53 2.88 (1.37) -0.09 (-0.06) 0.01 0.00Surface area (km2) 2020.80 1218.50 (0.90) 554.11 (0.37) 1515.10 (1.23)Average min. temperature, C 6.54 8.00 (-1.63) -3.36 (-1.01) -3.13 (-0.98)Average max. temperature, C 22.27 23.35 (-1.58) -2.75 (-1.04) -2.65 (-1.09)Average precipitation, cm 898.98 870.95 (0.29) -38.56 (-0.15) 3.53 (0.01)Observations 65 65 130 130
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the author’s own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections. Column (6) and (7) examine these characteristics for municipalities that border a municipality with a closeelection. Column (3) reports the t-statistic on the difference in means between municipalities where the PAN barelywon and where they barely lost. Columns (4) and (6) report the coefficient on PAN win from a standard RDspecification where the respective characteristic is used as the dependent variable, and columns (5) and (7) reportthe respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-9: Baseline Characteristics (13.3% vote spread, 2007-2010)
(1) (2) (3) (4) (5) (6) (7)
Own municipality Neighboring muns.
5% vote spread t-stat on t-stat on t-stat onPAN PAN means RD RD RD RDwon lost difference estimate estimate estimate estimate
Political characteristicsMun. taxes per capita (2005) 85.93 88.95 (-0.69) 6.18 (0.15) 16.05 (0.50)Turnout 0.66 0.64 (1.54) -0.04 (-0.83) -0.04 (-1.25)PAN incumbent 0.38 0.38 (0.12) 0.06 (0.55) 0.04 (0.65)PRD incumbent 0.11 0.10 (0.37) 0.07 (1.04) -0.04 (-0.96)% alternations (1976-2006) 0.28 0.28 (-1.03) -0.01 (-0.19) 0.01 (0.48)PRI never lost (1976-2006) 0.13 0.11 (0.90) 0.03 (0.36) -0.07 (-1.62)Demographic characteristicsPopulation (2005) 4.18 5.39 (-1.43) 2.16 (0.64) -0.77 (-0.43)Population density (2005) 160.46 236.82 (-1.32) -84.14 (-0.76) -48.70 (-0.62)Migrants per capita (2005) 0.02 0.02 (-0.15) 0.00 (1.04) 0.00 (1.45)Economic characteristicsIncome per capita (2005) 4.67 4.86 (-1.58) -0.39 (-0.72) 0.16 (0.39)Malnutrition (2005) 27.54 26.80 (0.49) 0.52 (0.14) -2.73 (-0.91)Mean years schooling (2005) 6.34 6.38 (-0.90) -0.20 (-0.65) -0.01 (-0.03)Infant mortality (2005) 22.57 22.26 (0.78) 0.57 (0.34) 0.39 (0.34)HH w/o access to sewage (2005) 10.52 11.12 (0.41) -3.40 (-1.17) -2.96 (-1.42)HH w/o access to water (2005) 14.08 14.64 (-0.22) 1.68 (0.44) 0.24 (0.08)Marginality index (2005) -0.29 -0.28 (0.17) -0.13 (-0.65) -0.17 (-1.05)Road network characteristicsDetour length (km) 15.85 15.96 (0.02) 2.08 (0.09) 8.38 (0.87)Road density 0.14 0.14 (0.01) -0.01 (-0.40) -0.01 (-0.45)Distance U.S. (km) 777.21 793.53 (-0.10) -130.20 (-1.35) -133.43 (-1.38)Geographic characteristicsElevation (m) 1302.49 1265.37 (0.12) 221.93 (1.22) 239.94 (1.40)Slope (degrees) 3.09 2.91 (0.91) 0.62 (1.01) 0.38 (0.83)Surface area (km2) 1272.77 1003.83 (0.73) 1098.05 (1.30) 681.57 (1.20)Average min. temperature, C 7.68 8.03 (-0.28) -2.10* (-1.78) -2.11* (-1.86)Average max. temperature, C 23.20 23.48 0.00 -1.39 (-1.50) -1.53* (-1.82)Average precipitation, cm 988.68 962.91 (0.11) 3.48 (0.03) -0.68 (-0.01)Observations 366 398 764 764
Notes: Data on population, population density, mean years of schooling, and migrants per capita are from II Conteode Poblacion y Vivienda, INEGI (National Institute of Statistics and Geography, 2005). Data on municipal taxcollection are from Sistema de Cuentas Municipales, INEGI. Data on housecold access to sewage and water are fromCONAPO (National Population Council) (2005). Data on malnutrition are from CONEVAL (National Council forEvaluating Social Development Policy), Indice de Reazgo Social (2005). Data on infant mortality are from PNUDMexico (UN Development Program, 2005). The marginality index is from CONAPO (2005). Data on distance to theU.S. and other road network characteristics are from the author’s own calculations. Electoral data are from MexicoElectoral-Banamex and electoral results published by the Electoral Tribunals of each state. For 11 states, data onthe total number of eligible voters, required to calculate turnout, are not reported. The geographic characteristics arefrom Acemoglu and Dell (2009). Columns (1) through (5) examine these variables for municipalities with closeelections. Column (6) and (7) examine these characteristics for municipalities that border a municipality with a closeelection. Column (3) reports the t-statistic on the difference in means between municipalities where the PAN barelywon and where they barely lost. Columns (4) and (6) report the coefficient on PAN win from a standard RDspecification where the respective characteristic is used as the dependent variable, and columns (5) and (7) reportthe respective t-statistic. * significant at 10%, ** significant at 5%, *** significant at 1%.
A-2.2 Robustness of Regressions Discontinuity Analysis
Tab
leA
-10:
PA
NE
lect
ion
s(2
007-2
008)
an
dD
rug
Tra
de-R
ela
ted
Hom
icid
es
5%b
and
wid
th4%
3%2%
13.3
%P
ost
Lam
eP
reP
ost
Pre
Pos
tP
reP
ost
Pre
Post
Pre
inau
g.d
uck
elec
.in
aug.
elec
.in
aug.
elec
.in
au
g.
elec
.in
au
g.
elec
.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Lin
ear
32.9
81**
*4.
967
-2.0
3836
.423
***
-2.4
6638
.064
***
-2.7
1047.1
11***
-5.7
61
25.6
21***
-0.2
26
(9.3
46)
(4.1
22)
(3.7
76)
(8.9
69)
(4.0
54)
(8.5
87)
(4.0
43)
(10.8
17)
(5.8
30)
(8.4
84)
(3.0
20)
Lin
ear
FE
15.8
99*
0.50
4-1
.211
18.4
60*
-0.6
7517
.545
-1.0
8737.4
45**
1.0
01
14.9
01**
-0.0
07
(8.7
36)
(2.8
15)
(2.7
09)
(9.9
23)
(2.8
30)
(12.
540)
(3.0
83)
(15.2
09)
(3.9
84)
(6.4
43)
(1.9
70)
Lin
ear
FE
contr
ols
16.7
86**
0.50
5-0
.989
19.4
88**
0.22
620
.637
*0.
445
40.2
25***
1.7
82
13.0
94**
-0.5
67
(7.7
62)
(3.1
19)
(2.6
43)
(8.5
12)
(2.8
83)
(10.
511)
(3.0
88)
(12.4
72)
(3.9
62)
(5.9
73)
(1.9
09)
Qu
adra
tic
41.6
58**
*3.
875
-4.5
3741
.436
***
-6.1
3542
.559
***
-7.9
31
29.4
69*
-9.7
13
34.9
24***
-3.2
61
(8.1
94)
(3.8
88)
(4.2
06)
(9.6
63)
(5.4
96)
(11.
390)
(6.8
76)
(15.4
31)
(9.1
52)
(9.5
34)
(3.7
17)
Qu
adra
tic
FE
29.6
06**
*6.
049*
**-1
.923
29.6
18**
-3.4
4735
.337
**-1
.426
22.8
67
-3.5
10
18.0
52**
-2.6
35
(9.5
38)
(2.2
26)
(3.6
61)
(14.
169)
(5.0
45)
(17.
279)
(4.4
28)
(21.1
84)
(5.1
55)
(8.5
31)
(2.9
93)
Qu
adra
tic
FE
Con
trol
s33
.271
***
6.95
8**
-0.7
8639
.390
***
0.51
443
.177
***
1.02
633.6
05*
-0.0
98
18.3
31**
-2.7
05
(8.2
62)
(2.8
72)
(3.3
36)
(11.
498)
(4.0
71)
(14.
609)
(4.1
85)
(16.9
85)
(5.3
63)
(7.4
54)
(2.7
06)
Cu
bic
40.9
96**
*-1
.483
-7.4
3740
.655
***
-9.1
5441
.769
***
-9.2
2864.2
31**
-8.1
59
39.1
30***
-4.8
12
(11.
568)
(3.3
23)
(7.0
88)
(13.
353)
(8.6
88)
(15.
468)
(9.9
19)
(31.0
15)
(14.2
08)
(8.5
13)
(4.0
40)
Cu
bic
FE
33.7
55**
4.05
4*-4
.527
34.8
65*
-2.3
8527
.018
-4.1
83
38.6
26
-5.1
44
19.3
02*
-4.0
82
(14.
914)
(2.2
00)
(6.4
22)
(18.
264)
(6.0
58)
(23.
226)
(6.4
06)
(32.7
52)
(10.0
63)
(10.2
65)
(3.6
62)
Cu
bic
FE
contr
ols
46.5
36**
*6.
879
-0.1
7347
.227
***
2.51
743
.331
**2.
464
90.7
85***
13.3
95
21.7
72**
-3.5
66
(11.
706)
(4.8
54)
(4.8
27)
(14.
629)
(4.8
33)
(18.
836)
(5.6
33)
(29.3
37)
(9.7
79)
(8.9
54)
(3.1
37)
Qu
arti
c41
.610
***
-2.3
93-1
0.52
246
.432
**-1
0.30
641
.921
-11.
420
211.0
72***
-11.6
09
43.6
79***
-4.7
69
(14.
486)
(3.8
23)
(9.4
01)
(19.
849)
(11.
231)
(31.
551)
(14.
458)
(40.1
56)
(25.8
41)
(8.7
16)
(4.8
74)
Qu
arti
cF
E38
.295
**2.
163
-4.6
1825
.043
-6.3
5815
.129
-9.2
08144.0
96***
-10.4
22
27.5
73**
-3.3
10
(17.
054)
(2.9
05)
(7.1
36)
(23.
980)
(8.1
32)
(32.
741)
(10.
622)
(48.6
17)
(21.0
34)
(11.6
66)
(4.6
61)
Qu
arti
cF
Eco
ntr
ols
53.3
62**
*5.
450
0.48
750
.675
***
2.58
579
.041
**9.
955
191.8
58***
9.0
95
32.8
52***
-1.9
26
(12.
543)
(4.7
41)
(5.1
99)
(18.
686)
(6.3
74)
(33.
582)
(11.
251)
(33.1
38)
(18.2
89)
(9.6
66)
(3.6
38)
Ob
serv
atio
ns
152
152
152
123
123
9494
62
62
380
380
Notes:
Inco
lum
ns
(1),
(4),
(6),
(8),
and
(10)
the
dep
end
ent
vari
ab
leis
the
dru
gtr
ad
eh
om
icid
era
ted
uri
ng
the
may
or’
ste
rm;
inco
lum
n(2
)it
isth
ed
rug
hom
icid
era
ted
uri
ng
the
lam
ed
uck
per
iod
,an
din
colu
mn
s(3
),(5
),(7
),(9
),an
d(1
1)
itis
the
dru
gh
om
icid
era
ted
uri
ng
the
pre
-ele
ctio
np
erio
d.
All
row
san
dco
lum
ns
rep
ort
the
coeffi
cien
ton
the
PA
Nw
inin
dic
ator.
Th
ero
ws
corr
esp
on
dto
diff
eren
tsp
ecifi
cati
on
sof
the
RD
poly
nom
ial,
fixed
effec
ts,
an
dco
ntr
ols
.T
he
colu
mn
sco
rres
pon
dto
diff
eren
tsp
ecifi
cati
ons
ofth
eb
an
dw
idth
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-11:
PA
NE
lect
ion
s(2
007-2
010)
an
dD
rug
Tra
de-R
ela
ted
Hom
icid
es
5%b
and
wid
th4%
3%2%
13.3
%P
ost
Lam
eP
reP
ost
Pre
Pos
tP
reP
ost
Pre
Post
Pre
inau
g.d
uck
elec
.in
aug.
elec
.in
aug.
elec
.in
au
g.
elec
.in
au
g.
elec
.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Lin
ear
26.7
35**
*6.
331
3.83
022
.540
***
0.39
124
.392
***
1.00
317.5
22***
-3.7
68
15.5
80**
3.6
44
(8.5
60)
(5.2
12)
(4.6
88)
(8.0
09)
(4.5
95)
(7.8
51)
(5.0
70)
(6.0
15)
(4.3
38)
(7.1
00)
(3.4
81)
Lin
ear
FE
14.0
05**
*0.
918
-0.6
1715
.259
***
-1.5
1316
.153
***
-0.6
32
10.5
85
-5.5
22
8.5
31*
1.2
89
(5.3
55)
(5.1
81)
(4.7
74)
(5.0
87)
(4.4
89)
(5.6
51)
(4.8
17)
(8.2
75)
(7.7
76)
(5.0
98)
(3.8
35)
Lin
ear
FE
contr
ols
15.8
07**
*-0
.375
-1.2
7015
.881
***
-3.9
5017
.261
***
-2.4
54
15.3
52*
-9.0
01
8.3
69*
1.0
49
(4.3
94)
(4.2
68)
(3.7
18)
(4.6
72)
(3.8
09)
(4.9
21)
(3.8
64)
(8.1
44)
(8.2
85)
(4.2
98)
(3.3
50)
Qu
adra
tic
17.6
35**
*1.
132
-4.5
7419
.315
***
-4.2
4716
.306
**-6
.675
15.6
94
-9.3
75
23.4
53***
3.2
23
(6.4
89)
(4.8
95)
(4.7
74)
(4.9
95)
(5.0
93)
(7.6
80)
(5.3
68)
(10.3
40)
(7.5
11)
(7.7
11)
(4.3
71)
Qu
adra
tic
FE
11.5
67-2
.129
-6.9
5315
.197
**-5
.925
10.9
59-9
.178
6.9
91
-14.0
21
15.1
90***
0.2
87
(7.9
79)
(5.9
67)
(6.6
22)
(6.0
83)
(7.1
00)
(9.1
54)
(8.4
71)
(12.9
60)
(11.1
87)
(5.3
66)
(4.9
33)
Qu
adra
tic
FE
Con
trol
s16
.940
***
-2.0
07-6
.367
17.9
64**
*-8
.265
16.1
97-1
3.01
812.1
72
-21.2
90
15.2
37***
0.0
69
(6.2
85)
(6.1
63)
(5.0
49)
(6.2
22)
(6.6
41)
(9.9
02)
(9.7
88)
(12.6
45)
(13.6
16)
(4.5
53)
(4.1
08)
Cu
bic
18.2
31**
-5.5
30-5
.005
13.2
06-8
.174
14.8
17-8
.618
22.1
39
-12.9
79
25.3
50***
-1.7
05
(7.8
72)
(6.7
05)
(5.8
30)
(9.4
75)
(6.5
46)
(11.
067)
(8.9
69)
(18.2
59)
(14.6
29)
(7.3
13)
(4.9
80)
Cu
bic
FE
9.42
9-1
3.56
7-1
0.91
89.
384
-12.
567
4.34
1-1
4.08
6-3
.774
-17.7
89
15.4
61***
-5.8
14
(9.5
27)
(12.
302)
(9.1
44)
(9.8
14)
(9.3
85)
(12.
967)
(11.
801)
(18.6
43)
(18.0
66)
(5.2
61)
(5.9
43)
Cu
bic
FE
contr
ols
20.2
47**
-16.
038
-10.
364
14.6
00-1
8.10
9*14
.672
-18.
646
23.3
89
-21.1
22
16.7
38***
-5.6
31
(9.7
82)
(15.
974)
(8.4
26)
(10.
930)
(10.
576)
(13.
691)
(14.
287)
(19.1
26)
(28.6
63)
(4.8
25)
(4.7
56)
Qu
arti
c12
.756
-13.
076
-10.
507
19.4
49-1
0.38
820
.465
-15.
134
65.4
56
3.0
98
18.4
35***
-4.3
38
(11.
024)
(9.2
25)
(7.8
63)
(12.
099)
(10.
598)
(18.
263)
(15.
251)
(41.3
76)
(20.8
61)
(5.2
28)
(4.5
91)
Qu
arti
cF
E4.
364
-21.
661
-17.
814
8.96
5-1
5.17
22.
913
-16.
848
44.8
66
3.3
16
11.7
22*
-8.1
14
(12.
795)
(14.
819)
(11.
313)
(12.
962)
(11.
998)
(17.
952)
(16.
351)
(37.1
67)
(20.0
58)
(6.5
31)
(6.5
93)
Qu
arti
cF
Eco
ntr
ols
21.0
00-2
1.75
1-1
5.17
118
.177
-21.
380
21.8
41-2
1.71
258.7
63
6.1
17
16.4
70***
-6.6
44
(13.
113)
(17.
760)
(10.
061)
(12.
693)
(13.
262)
(16.
409)
(20.
772)
(38.9
80)
(21.1
41)
(5.6
89)
(5.1
42)
Clu
ster
s30
730
730
724
924
918
618
6130
130
746
746
Ob
serv
atio
ns
310
310
310
251
251
186
186
130
130
764
764
Notes:
Inco
lum
ns
(1),
(4),
(6),
(8),
and
(10)
the
dep
end
ent
vari
ab
leis
the
dru
gtr
ad
eh
om
icid
era
ted
uri
ng
the
post
-in
au
gu
rati
on
per
iod
;in
colu
mn
(2)
itis
the
dru
gh
omic
ide
rate
du
rin
gth
ela
me
du
ckp
erio
d,
and
inco
lum
ns
(3),
(5),
(7),
(9),
an
d(1
1)
itis
the
dru
gh
om
icid
era
ted
uri
ng
the
pre
-ele
ctio
np
erio
d.
All
row
san
dco
lum
ns
rep
ort
the
coeffi
cien
ton
the
PA
Nw
inin
dic
ato
r.T
he
row
sco
rres
pon
dto
diff
eren
tsp
ecifi
cati
on
sof
the
RD
poly
nom
ial,
fixed
effec
ts,
an
dco
ntr
ols
.T
he
colu
mn
sco
rres
pon
dto
diff
eren
tsp
ecifi
cati
ons
ofth
eb
an
dw
idth
.*
sign
ifica
nt
at
10%
,**
signifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-12:
PA
NE
lect
ion
s(2
007-2
008)
an
dO
vera
llH
om
icid
es
5%b
and
wid
th4%
3%2%
13.3
%P
ost
Lam
eP
reP
ost
Pre
Pos
tP
reP
ost
Pre
Post
Pre
inau
g.d
uck
elec
.in
aug.
elec
.in
aug.
elec
.in
au
g.
elec
.in
au
g.
elec
.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Lin
ear
56.6
30**
*2.
457
3.08
862
.219
***
1.62
663
.787
***
0.18
575.7
71***
-3.6
54
44.5
51***
3.6
53
(12.
768)
(2.9
22)
(4.3
61)
(11.
444)
(4.8
46)
(10.
791)
(5.0
00)
(14.4
34)
(7.5
32)
(11.9
67)
(3.2
45)
Lin
ear
FE
32.7
29**
-0.6
110.
719
37.4
78**
*1.
039
36.4
49**
-0.8
5364.7
81***
1.3
62
29.1
87***
3.3
85
(12.
579)
(3.5
41)
(3.8
46)
(13.
862)
(4.1
26)
(17.
457)
(4.6
67)
(20.2
62)
(5.1
91)
(9.6
69)
(2.3
02)
Lin
ear
FE
contr
ols
34.4
30**
*-1
.764
1.43
238
.963
***
2.96
240
.079
**2.
493
68.6
96***
4.0
53
26.1
89***
3.6
86*
(11.
205)
(3.2
11)
(4.1
16)
(12.
165)
(4.0
82)
(15.
365)
(4.6
29)
(19.2
28)
(5.3
87)
(8.8
72)
(2.1
62)
Qu
adra
tic
68.5
50**
*4.
442
-1.9
5168
.130
***
-5.9
1972
.084
***
-6.2
9961.4
34***
-11.8
15
59.6
02***
1.6
53
(10.
105)
(3.1
42)
(5.5
71)
(12.
772)
(7.1
33)
(15.
029)
(8.8
91)
(19.8
91)
(11.6
25)
(12.6
68)
(4.3
82)
Qu
adra
tic
FE
55.3
48**
*4.
903
-1.0
5156
.208
***
-5.1
6763
.300
***
-1.5
7851.9
54*
-3.7
71
37.5
71***
1.3
16
(13.
698)
(3.8
77)
(5.6
71)
(18.
629)
(7.9
59)
(22.
861)
(7.1
56)
(27.1
93)
(6.8
80)
(11.9
85)
(3.6
99)
Qu
adra
tic
FE
contr
ols
61.4
15**
*4.
737
1.09
168
.388
***
1.39
973
.884
***
3.58
168.5
88**
3.9
09
37.9
70***
2.1
85
(12.
687)
(3.3
38)
(5.0
38)
(17.
461)
(6.4
69)
(21.
715)
(7.2
16)
(26.0
84)
(6.1
00)
(10.5
44)
(3.5
71)
Cu
bic
71.1
97**
*-0
.063
-7.8
6971
.760
***
-9.5
3077
.883
***
-12.
089
113.2
70***
-23.2
80
66.1
31***
-0.5
22
(15.
385)
(4.0
71)
(9.1
84)
(17.
171)
(11.
132)
(19.
531)
(12.
627)
(41.0
32)
(17.7
77)
(10.6
63)
(5.2
20)
Cu
bic
FE
66.4
41**
*0.
464
-6.5
9265
.402
***
-5.4
9257
.932
*-8
.386
69.1
67
-19.4
62*
41.9
77***
-1.0
43
(18.
634)
(5.4
76)
(9.7
70)
(23.
193)
(9.4
92)
(30.
227)
(9.7
88)
(45.0
54)
(10.8
46)
(13.8
75)
(4.9
11)
Cu
bic
FE
contr
ols
87.5
33**
*2.
666
1.46
781
.458
***
2.63
880
.244
***
5.70
1139.9
02***
0.2
09
45.9
12***
0.4
95
(16.
743)
(4.5
82)
(8.0
17)
(21.
835)
(8.2
80)
(27.
709)
(8.8
35)
(46.8
36)
(14.3
78)
(12.4
94)
(4.5
53)
Qu
arti
c70
.782
***
1.63
8-1
1.84
689
.305
***
-14.
355
88.6
94**
-30.
021
314.6
81***
-20.9
10
70.8
07***
-3.8
20
(18.
881)
(5.8
98)
(12.
066)
(25.
679)
(14.
014)
(40.
734)
(18.
335)
(52.2
71)
(28.2
29)
(11.1
08)
(6.3
33)
Qu
arti
cF
E66
.807
***
2.31
3-8
.116
57.8
57*
-13.
788
39.9
03-3
4.19
0**
197.5
22***
-23.5
17
54.1
24***
-3.0
66
(22.
414)
(6.8
69)
(11.
003)
(31.
878)
(11.
834)
(44.
767)
(13.
661)
(63.6
47)
(21.8
13)
(15.6
38)
(6.5
30)
Qu
arti
cF
Eco
ntr
ols
91.7
60**
*5.
284
1.81
094
.476
***
3.43
512
7.39
7**
-8.2
09254.6
27***
-8.1
43
62.0
25***
0.1
10
(18.
590)
(6.0
54)
(8.8
38)
(27.
883)
(10.
118)
(51.
547)
(14.
102)
(51.1
94)
(22.9
45)
(14.0
23)
(5.5
64)
Ob
serv
atio
ns
152
152
152
123
123
9494
62
62
380
380
Notes:
Inco
lum
ns
(1),
(4),
(6),
(8),
and
(10)
the
dep
end
ent
vari
ab
leis
the
dru
gtr
ad
eh
om
icid
era
ted
uri
ng
the
may
or’
ste
rm;
inco
lum
n(2
)it
isth
ed
rug
hom
icid
era
ted
uri
ng
the
lam
ed
uck
per
iod
,an
din
colu
mn
s(3
),(5
),(7
),(9
),an
d(1
1)
itis
the
dru
gh
om
icid
era
ted
uri
ng
the
pre
-ele
ctio
np
erio
d.
All
row
san
dco
lum
ns
rep
ort
the
coeffi
cien
ton
the
PA
Nw
inin
dic
ator.
Th
ero
ws
corr
esp
on
dto
diff
eren
tsp
ecifi
cati
on
sof
the
RD
poly
nom
ial,
fixed
effec
ts,
an
dco
ntr
ols
.T
he
colu
mn
sco
rres
pon
dto
diff
eren
tsp
ecifi
cati
ons
ofth
eb
an
dw
idth
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-13:
PA
NE
lect
ion
s(2
007-2
010)
an
dO
vera
llH
om
icid
es
5%b
and
wid
th4%
3%2%
13.3
%P
ost
Lam
eP
reP
ost
Pre
Pos
tP
reP
ost
Pre
Post
Pre
inau
g.d
uck
elec
.in
aug.
elec
.in
aug.
elec
.in
au
g.
elec
.in
au
g.
elec
.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Lin
ear
44.8
20**
*6.
622
4.74
041
.830
***
3.02
342
.184
***
2.17
232.2
47***
-0.3
52
29.6
42***
3.6
97
(12.
289)
(6.6
65)
(3.4
98)
(12.
027)
(3.9
28)
(11.
541)
(4.4
08)
(11.6
85)
(4.6
81)
(10.9
13)
(2.4
48)
Lin
ear
FE
28.4
64**
*-0
.026
1.83
230
.120
***
2.03
330
.030
***
1.12
225.6
38*
0.6
92
19.3
16***
1.7
73
(8.1
78)
(7.2
83)
(2.9
95)
(8.7
58)
(3.0
23)
(9.5
47)
(3.7
76)
(13.4
78)
(4.4
21)
(7.0
85)
(1.9
90)
Lin
ear
FE
contr
ols
29.3
85**
*-2
.218
2.53
630
.469
***
2.82
331
.591
***
2.57
231.5
89**
2.5
42
18.3
99***
1.8
81
(7.6
57)
(5.5
95)
(2.8
40)
(8.0
45)
(2.9
28)
(8.6
00)
(3.2
67)
(13.0
13)
(4.6
20)
(6.2
78)
(1.9
29)
Qu
adra
tic
35.9
70**
*-2
.048
-0.6
6932
.601
***
-2.0
2129
.678
*-1
.824
27.2
29
-5.6
89
40.7
99***
3.4
82
(9.3
51)
(5.6
65)
(4.7
67)
(10.
070)
(5.1
40)
(15.
186)
(6.5
66)
(20.0
89)
(8.8
82)
(11.8
27)
(3.4
99)
Qu
adra
tic
FE
30.0
73**
*-7
.716
-0.3
0428
.374
**-0
.950
24.3
09-0
.476
21.5
69
-2.8
81
29.4
70***
1.9
11
(10.
626)
(8.9
69)
(4.7
56)
(11.
969)
(5.5
38)
(16.
664)
(6.2
81)
(22.5
81)
(7.4
04)
(8.5
57)
(3.1
89)
Qu
adra
tic
FE
Con
trol
s35
.409
***
-7.5
970.
864
32.1
56**
*1.
261
31.7
35*
3.87
530.6
80
2.6
55
28.5
97***
2.2
78
(10.
135)
(8.3
62)
(4.4
11)
(11.
135)
(4.9
74)
(16.
147)
(5.7
19)
(21.0
10)
(6.2
73)
(7.5
78)
(3.0
21)
Cu
bic
29.0
08*
-6.6
48-2
.118
25.1
65-3
.813
25.8
58-7
.825
51.1
01*
-12.3
97
43.4
34***
2.2
14
(15.
412)
(8.8
30)
(6.9
28)
(18.
519)
(7.9
34)
(21.
975)
(9.9
77)
(29.0
94)
(12.2
82)
(10.6
07)
(4.4
09)
Cu
bic
FE
22.3
61-2
2.51
0-1
.630
21.2
05-2
.884
14.9
00-7
.392
10.6
52
-16.7
40*
30.7
12***
1.1
05
(16.
504)
(17.
344)
(7.8
18)
(19.
012)
(7.9
28)
(24.
761)
(9.4
79)
(32.3
58)
(9.6
76)
(9.0
36)
(3.9
95)
Cu
bic
FE
contr
ols
33.2
65**
-25.
009
2.46
828
.943
2.46
731
.132
1.55
253.6
59
-4.1
70
31.2
67***
1.8
20
(15.
873)
(21.
577)
(6.7
64)
(18.
109)
(6.8
64)
(23.
605)
(7.6
14)
(32.6
62)
(8.7
75)
(8.7
61)
(3.8
11)
Qu
arti
c23
.244
-12.
152
-7.3
1336
.261
-6.7
2750
.659
*-1
4.26
3106.4
28*
-13.2
29
36.0
97***
-1.4
00
(21.
227)
(12.
413)
(9.3
23)
(23.
206)
(10.
635)
(30.
272)
(12.
929)
(63.4
40)
(17.9
26)
(9.5
27)
(5.2
95)
Qu
arti
cF
E17
.154
-29.
959
-7.1
0421
.076
-7.2
7621
.133
-17.
062
68.7
71
-19.2
63
28.9
60***
-0.5
96
(21.
824)
(20.
973)
(10.
080)
(25.
462)
(10.
220)
(33.
436)
(11.
330)
(59.4
45)
(15.0
78)
(10.9
66)
(5.6
12)
Qu
arti
cF
Eco
ntr
ols
36.2
16*
-27.
800
-1.8
0535
.657
0.91
452
.288
*-2
.224
95.2
17
-6.8
43
34.2
87***
1.1
82
(20.
522)
(24.
123)
(8.6
21)
(22.
803)
(8.3
62)
(31.
497)
(8.8
57)
(63.0
04)
(12.1
19)
(10.0
80)
(5.3
54)
Clu
ster
s30
730
730
724
924
918
618
6130
130
746
746
Ob
serv
atio
ns
310
310
310
251
251
186
186
130
130
764
764
Notes:
Inco
lum
ns
(1),
(4),
(6),
(8),
and
(10)
the
dep
end
ent
vari
ab
leis
the
dru
gtr
ad
eh
om
icid
era
ted
uri
ng
the
post
-in
au
gu
rati
on
per
iod
;in
colu
mn
(2)
itis
the
dru
gh
omic
ide
rate
du
rin
gth
ela
me
du
ckp
erio
d,
and
inco
lum
ns
(3),
(5),
(7),
(9),
an
d(1
1)
itis
the
dru
gh
om
icid
era
ted
uri
ng
the
pre
-ele
ctio
np
erio
d.
All
row
san
dco
lum
ns
rep
ort
the
coeffi
cien
ton
the
PA
Nw
inin
dic
ato
r.T
he
row
sco
rres
pon
dto
diff
eren
tsp
ecifi
cati
on
sof
the
RD
poly
nom
ial,
regio
nfi
xed
effec
ts,
an
dco
ntr
ols.
Th
eco
lum
ns
corr
esp
ond
tod
iffer
ent
spec
ifica
tion
sof
the
ban
dw
idth
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Table A-14: PAN Elections and Violence (All Municipalities)
Drug-Related Hom. Overall Hom.07-08 07-10 07-08 07-10
(1) (2) (3) (4)
PAN win 13.576 9.418 30.908** 17.775*(9.382) (6.962) (13.109) (10.608)
Clusters 621 1166 621 1166Observations 621 1,205 621 1,205R-squared 0.032 0.014 0.044 0.019
Notes: The sample includes all elections where the PAN was the winner or runner-up. Columns (1) and(2) examine the drug trade-related death rate and columns (3) and (4) examine the overall homicide rate.Columns (1) and (3) utilize elections that occurred in 2007-2008. Columns (2) and (4) utilize electionsoccurring in 2007-2010. Standard errors are clustered by municipality. * significant at 10%, ** significantat 5%, *** significant at 1%.
A-2.3 Robustness to Using Differences-in-Differences
Tab
leA
-15:
Clo
seP
AN
Ele
ctio
ns
an
dD
rug
Tra
de-R
ela
ted
Hom
icid
es
(DD
stra
tegy;
5%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
Pan
elA:20
07-200
8election
sP
AN
win
x1.
238
5.90
20.
490
1.34
56.
205
0.59
8la
me
duck
(5.3
55)
(5.5
15)
(3.9
78)
(5.6
65)
(5.5
00)
(4.2
60)
PA
Nw
inx
30.5
39**
*30
.539
***
26.3
29**
28.6
68**
*29
.448
***
24.4
75*
pos
t-in
aug.
(9.5
17)
(10.
197)
(12.
443)
(9.9
04)
(10.
182)
(12.
571)
R-s
quar
ed0.
085
0.16
50.
085
0.08
30.
165
0.08
3C
lust
ers
152
152
152
152
152
152
Obse
rvat
ions
8,81
68,
816
8,81
68,
816
8,81
68,
816
Pan
elB:20
07-201
0election
sP
AN
win
x-3
.212
2.83
0-3
.191
-3.3
213.
456
-3.2
24la
me
duck
(6.1
73)
(4.6
43)
(5.0
06)
(7.0
11)
(4.4
68)
(5.6
95)
PA
Nw
inx
22.2
99**
24.8
22**
22.3
91**
18.5
36*
23.3
07**
18.9
63*
pos
t-in
aug.
(9.5
01)
(10.
020)
(11.
175)
(9.9
02)
(9.8
57)
(10.
610)
R-s
quar
ed0.
038
0.10
30.
038
0.03
60.
103
0.03
6C
lust
ers
307
307
307
307
307
307
Obse
rvat
ions
17,9
8017
,980
17,9
8017
,980
17,9
8017
,980
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-16:
Clo
seP
AN
Ele
ctio
ns
an
dD
rug
Tra
de-R
ela
ted
Hom
icid
es
(DD
stra
tegy;
4%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
Pan
elA:20
07-200
8election
sP
AN
win
x3.
030
8.37
52.
304
3.75
58.
340
3.03
0la
me
duck
(5.8
09)
(6.4
94)
(4.1
76)
(6.0
83)
(6.2
90)
(4.4
37)
PA
Nw
inx
34.1
68**
*31
.252
***
30.2
21**
30.8
20**
30.2
24**
*26
.897
*p
ost-
inau
g.(1
0.76
3)(1
0.40
7)(1
3.03
4)(1
2.41
9)(1
0.45
2)(1
5.21
5)R
-squar
ed0.
105
0.18
60.
105
0.10
00.
186
0.09
9C
lust
ers
123
123
123
123
123
123
Obse
rvat
ions
7,13
47,
134
7,13
47,
134
7,13
47,
134
Pan
elB:20
07-201
0election
sP
AN
win
x-5
.467
4.67
3-5
.051
-8.2
844.
245
-7.7
18la
me
duck
(6.7
04)
(5.7
90)
(5.4
78)
(7.8
80)
(5.6
27)
(6.3
68)
PA
Nw
inx
20.1
04**
26.1
14**
21.9
88*
13.6
1023
.606
**16
.115
pos
t-in
aug.
(9.8
93)
(10.
446)
(11.
811)
(11.
134)
(10.
342)
(12.
877)
R-s
quar
ed0.
053
0.12
20.
053
0.04
60.
121
0.04
6C
lust
ers
249
249
249
249
249
249
Obse
rvat
ions
14,5
5814
,558
14,5
5814
,558
14,5
5814
,558
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-17:
Clo
seP
AN
Ele
ctio
ns
an
dD
rug
Tra
de-R
ela
ted
Hom
icid
es
(DD
stra
tegy;
3%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
Pan
elA:20
07-200
8election
sP
AN
win
x2.
235
7.96
01.
593
3.58
77.
926
2.95
2la
me
duck
(6.2
67)
(7.1
74)
(4.5
51)
(6.6
82)
(6.7
13)
(4.9
47)
PA
Nw
inx
36.9
40**
*31
.023
***
33.4
14**
31.5
73**
30.1
96**
*28
.102
*p
ost-
inau
g.(1
1.61
9)(9
.744
)(1
3.36
1)(1
4.21
9)(1
0.13
9)(1
6.78
5)R
-squar
ed0.
120
0.19
80.
119
0.11
10.
198
0.11
1C
lust
ers
9494
9494
9494
Obse
rvat
ions
5,45
25,
452
5,45
25,
452
5,45
25,
452
Pan
elB:20
07-201
0election
sP
AN
win
x-3
.969
4.63
2-4
.030
-3.4
214.
809
-3.2
98la
me
duck
(7.0
76)
(5.9
23)
(5.5
95)
(8.2
85)
(6.1
64)
(6.4
05)
PA
Nw
inx
24.6
42**
25.1
71**
24.4
18*
21.3
6724
.005
**21
.805
pos
t-in
aug.
(10.
865)
(9.8
87)
(12.
733)
(13.
329)
(9.8
46)
(15.
405)
R-s
quar
ed0.
064
0.13
30.
064
0.05
70.
132
0.05
7C
lust
ers
186
186
186
186
186
186
Obse
rvat
ions
10,7
8810
,788
10,7
8810
,788
10,7
8810
,788
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-18:
Clo
seP
AN
Ele
ctio
ns
an
dD
rug
Tra
de-R
ela
ted
Hom
icid
es
(DD
stra
tegy;
2%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
Pan
elA:20
07-200
8election
sP
AN
win
x-1
.080
-1.8
39-2
.870
-1.7
37-3
.311
-3.4
89la
me
duck
(6.1
38)
(6.1
36)
(6.0
34)
(5.2
98)
(4.4
31)
(5.5
91)
PA
Nw
inx
62.3
72**
*26
.180
***
52.2
88**
*65
.278
***
21.3
47**
55.2
00**
*p
ost-
inau
g.(1
1.93
2)(9
.778
)(1
1.46
8)(1
3.01
8)(8
.117
)(1
3.99
4)R
-squar
ed0.
182
0.25
40.
180
0.18
20.
253
0.17
9C
lust
ers
6262
6262
6262
Obse
rvat
ions
3,59
63,
596
3,59
63,
596
3,59
63,
596
Pan
elB:20
07-201
0election
sP
AN
win
x-1
4.46
7**
-4.0
72-1
5.20
6**
-19.
703*
**-3
.825
-20.
396*
**la
me
duck
(6.6
15)
(3.9
98)
(6.9
50)
(6.2
39)
(5.3
99)
(6.0
26)
PA
Nw
inx
39.5
83**
*20
.491
**35
.769
**35
.248
**21
.181
***
31.5
88*
pos
t-in
aug.
(13.
640)
(8.6
82)
(15.
615)
(15.
864)
(7.2
56)
(18.
729)
R-s
quar
ed0.
096
0.16
50.
096
0.09
20.
165
0.09
2C
lust
ers
130
130
130
130
130
130
Obse
rvat
ions
7,54
07,
540
7,54
07,
540
7,54
07,
540
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-19:
Clo
seP
AN
Ele
ctio
ns
an
dD
rug-R
ela
ted
Hom
icid
es
(DD
stra
tegy;
13.3
%vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
Pan
elA:20
07-200
8election
sP
AN
win
x2.
626
3.36
11.
120
2.40
53.
247
0.89
9la
me
duck
(4.3
21)
(3.8
41)
(3.0
72)
(4.4
50)
(3.9
07)
(3.1
70)
PA
Nw
inx
29.2
28**
*24
.241
***
20.9
68**
29.1
95**
*24
.390
***
20.9
38**
pos
t-in
aug.
(7.8
79)
(8.8
01)
(9.3
84)
(8.1
76)
(9.3
37)
(9.5
39)
R-s
quar
ed0.
046
0.14
00.
045
0.04
60.
139
0.04
5C
lust
ers
380
380
380
380
380
380
Obse
rvat
ions
22,0
4022
,040
22,0
4022
,040
22,0
4022
,040
Pan
elB:20
07-201
0election
sP
AN
win
x-0
.437
3.02
8-1
.737
-1.3
332.
731
-2.6
05la
me
duck
(5.5
92)
(3.1
17)
(4.1
80)
(5.8
08)
(3.0
80)
(4.3
42)
PA
Nw
inx
16.2
31**
13.7
91*
12.0
0715
.104
**13
.118
10.9
76p
ost-
inau
g.(7
.048
)(8
.338
)(7
.728
)(7
.305
)(8
.640
)(7
.691
)R
-squar
ed0.
026
0.10
80.
026
0.02
60.
108
0.02
5C
lust
ers
746
746
746
746
746
746
Obse
rvat
ions
44,3
1244
,312
44,3
1244
,312
44,3
1244
,312
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-20:
Clo
seP
AN
Ele
ctio
ns
an
dO
vera
llH
om
icid
es
(DD
stra
tegy;
5%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
PA
Nw
inx
-9.4
37**
-14.
058*
**-4
.807
-8.8
80**
-13.
610*
**-4
.272
lam
educk
(3.9
92)
(4.2
13)
(3.5
75)
(4.1
15)
(4.3
59)
(3.6
99)
PA
Nw
inx
48.4
63**
*41
.445
***
53.9
83**
*45
.426
**38
.318
**50
.945
***
pos
t-in
aug.
(15.
376)
(15.
796)
(15.
830)
(18.
009)
(18.
406)
(18.
490)
R-s
quar
ed0.
178
0.23
50.
176
0.16
70.
229
0.16
5C
lust
ers
152
152
152
152
152
152
Obse
rvat
ions
39,2
6939
,269
39,2
6939
,269
39,2
6939
,269
Pan
elB:20
07-201
0election
sP
AN
win
x-5
.948
-8.6
47-2
.276
-5.9
14-8
.398
-2.2
27la
me
duck
(5.8
12)
(5.8
36)
(5.2
58)
(5.9
08)
(5.7
75)
(5.2
89)
PA
Nw
inx
37.2
94**
*33
.527
***
41.5
76**
*32
.368
***
28.9
19**
36.6
48**
*p
ost-
inau
g.(1
1.41
9)(1
1.25
9)(1
2.32
4)(1
2.11
2)(1
1.83
5)(1
2.88
1)R
-squar
ed0.
072
0.11
40.
071
0.06
80.
112
0.06
7C
lust
ers
307
307
307
307
307
307
Obse
rvat
ions
73,8
7573
,875
73,8
7573
,875
73,8
7573
,875
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-21:
Clo
seP
AN
Ele
ctio
ns
an
dO
vera
llH
om
icid
es
(DD
stra
tegy;
4%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
PA
Nw
inx
-6.4
89**
-9.0
79**
*-0
.247
-6.0
02*
-7.5
78**
0.25
1la
me
duck
(3.2
34)
(2.6
85)
(3.1
09)
(3.3
73)
(3.0
41)
(3.2
72)
PA
Nw
inx
53.8
30**
*48
.647
***
61.0
20**
*49
.333
**44
.441
**56
.573
***
pos
t-in
aug.
(14.
621)
(15.
404)
(15.
041)
(20.
426)
(20.
793)
(20.
817)
R-s
quar
ed0.
222
0.27
30.
220
0.20
10.
262
0.19
8C
lust
ers
123
123
123
123
123
123
Obse
rvat
ions
31,7
7331
,773
31,7
7331
,773
31,7
7331
,773
Pan
elB:20
07-201
0election
sP
AN
win
x-5
.738
-8.0
13-1
.428
-5.9
90-8
.388
*-1
.680
lam
educk
(5.2
93)
(5.1
75)
(5.0
37)
(5.1
78)
(5.0
11)
(4.9
14)
PA
Nw
inx
35.4
09**
*32
.132
***
40.4
31**
*30
.298
**27
.320
**35
.316
***
pos
t-in
aug.
(10.
123)
(10.
595)
(11.
218)
(11.
747)
(12.
309)
(12.
691)
R-s
quar
ed0.
089
0.13
00.
087
0.08
00.
124
0.07
9C
lust
ers
249
249
249
249
249
249
Obse
rvat
ions
59,8
0959
,809
59,8
0959
,809
59,8
0959
,809
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-22:
Clo
seP
AN
Ele
ctio
ns
an
dO
vera
llH
om
icid
es
(DD
stra
tegy;
3%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
PA
Nw
inx
-6.8
35**
-8.6
50**
*0.
010
-6.3
77**
-6.7
93**
0.49
8la
me
duck
(3.1
74)
(2.5
92)
(3.4
89)
(3.1
67)
(2.8
55)
(3.3
83)
PA
Nw
inx
57.3
04**
*53
.231
***
65.2
74**
*50
.182
**46
.527
**58
.230
***
pos
t-in
aug.
(14.
212)
(15.
044)
(14.
540)
(21.
524)
(22.
163)
(21.
965)
R-s
quar
ed0.
251
0.30
00.
248
0.22
50.
287
0.22
2C
lust
ers
9494
9494
9494
Obse
rvat
ions
24,2
8724
,287
24,2
8724
,287
24,2
8724
,287
Pan
elB:20
07-201
0election
sP
AN
win
x-3
.055
-5.0
791.
863
-3.2
08-4
.922
1.73
1la
me
duck
(5.3
54)
(4.7
30)
(5.3
09)
(5.3
44)
(4.6
95)
(5.2
93)
PA
Nw
inx
37.0
70**
*33
.725
***
42.5
97**
*33
.486
***
30.5
12**
39.0
35**
*p
ost-
inau
g.(9
.057
)(1
0.04
3)(1
0.14
2)(1
2.04
6)(1
3.14
7)(1
2.88
4)R
-squar
ed0.
110
0.15
30.
108
0.10
00.
147
0.09
7C
lust
ers
186
186
186
186
186
186
Obse
rvat
ions
44,3
3744
,337
44,3
3744
,337
44,3
3744
,337
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-23:
Clo
seP
AN
Ele
ctio
ns
an
dO
vera
llH
om
icid
es
(DD
stra
tegy;
2%
vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
PA
Nw
inx
-8.3
61*
-7.8
06-1
.053
-9.8
27**
-8.5
74-2
.597
lam
educk
(4.8
09)
(6.0
89)
(5.4
92)
(3.8
04)
(6.1
82)
(4.4
28)
PA
Nw
inx
73.9
58**
*72
.483
***
83.5
04**
*75
.102
***
73.9
56**
*84
.659
***
pos
t-in
aug.
(15.
766)
(15.
425)
(15.
613)
(17.
163)
(17.
102)
(16.
919)
R-s
quar
ed0.
342
0.39
40.
338
0.34
20.
394
0.33
8C
lust
ers
6262
6262
6262
Obse
rvat
ions
16,0
2216
,022
16,0
2216
,022
16,0
2216
,022
Pan
elB:20
07-201
0election
sP
AN
win
x-1
9.60
3***
-17.
354*
**-1
3.40
8**
-19.
280*
*-1
7.23
9**
-13.
088*
lam
educk
(6.8
79)
(6.2
17)
(6.5
99)
(7.9
85)
(8.0
41)
(7.7
21)
PA
Nw
inx
27.9
07**
29.1
23**
35.3
56**
*27
.354
*28
.422
*34
.802
**p
ost-
inau
g.(1
3.19
4)(1
3.41
4)(1
2.94
8)(1
4.92
0)(1
5.79
0)(1
4.71
8)R
-squar
ed0.
157
0.20
30.
153
0.15
60.
202
0.15
2C
lust
ers
130
130
130
130
130
130
Obse
rvat
ions
30,9
9730
,997
30,9
9730
,997
30,9
9730
,997
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-24:
Clo
seP
AN
Ele
ctio
ns
an
dO
vera
llH
om
icid
es
(DD
stra
tegy;
13.3
%vote
spre
ad
)
Quad
rati
cvo
tesp
read
pol
ynom
ial
Lin
ear
vote
spre
adp
olynom
ial
Cal
endar
Munic
ipal
ity
No
Cal
endar
Munic
ipal
ity
No
tim
etr
end(s
)ti
me
tren
d(s
)
(1)
(2)
(3)
(4)
(5)
(6)
PA
Nw
inx
-6.6
33*
-8.9
04**
-3.0
21-6
.556
*-9
.019
**-2
.944
lam
educk
(3.7
88)
(3.8
57)
(3.0
95)
(3.7
29)
(3.8
75)
(3.0
60)
PA
Nw
inx
38.0
91**
*33
.937
***
42.5
30**
*38
.525
***
34.2
62**
42.9
61**
*p
ost-
inau
g.(1
2.82
1)(1
3.02
8)(1
3.47
7)(1
4.08
0)(1
4.02
6)(1
4.79
0)R
-squar
ed0.
128
0.18
70.
127
0.12
40.
185
0.12
3C
lust
ers
380
380
380
380
380
380
Obse
rvat
ions
98,1
7998
,179
98,1
7998
,179
98,1
7998
,179
Pan
elB:20
07-201
0election
sP
AN
win
x-1
.024
-3.5
551.
429
-1.2
98-3
.859
1.15
2la
me
duck
(4.2
32)
(4.6
34)
(3.8
74)
(4.1
63)
(4.6
19)
(3.8
48)
PA
Nw
inx
24.5
37**
21.5
08**
27.3
57**
*23
.302
**20
.458
**26
.117
**p
ost-
inau
g.(9
.642
)(9
.488
)(1
0.56
6)(1
0.31
6)(1
0.08
3)(1
1.23
9)R
-squar
ed0.
047
0.08
90.
046
0.04
40.
087
0.04
3C
lust
ers
746
746
746
746
746
746
Obse
rvat
ions
182,
104
182,
104
182,
104
182,
104
182,
104
182,
104
Notes:
Th
ed
epen
den
tva
riab
leis
the
hom
icid
era
tein
agiv
em
un
icip
ali
ty-m
onth
.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,la
me
du
ckis
anin
dic
ator
equ
alto
one
ifth
eob
serv
atio
nocc
urr
edd
uri
ng
the
lam
ed
uck
per
iod
,an
dp
ost
-in
au
g.
isan
ind
icato
req
ual
toon
eif
the
ob
serv
ati
on
occ
urr
edd
uri
ng
the
pos
t-in
augu
rati
onp
erio
d.
Col
um
ns
(1)
thro
ugh
(3)
incl
ud
ea
qu
ad
rati
cvo
tesp
read
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-in
au
gu
rati
on
ind
icat
ors,
and
Col
um
ns
(4)
thro
ugh
(6)
incl
ud
ea
lin
ear
vote
spre
ad
poly
nom
ial
inte
ract
edw
ith
the
lam
ed
uck
an
dp
ost
-inau
gu
rati
on
ind
icato
rs.
All
colu
mn
sin
clu
de
mu
nic
ipal
ity
and
mon
thfi
xed
effec
ts,
and
stan
dard
erro
rsare
clu
ster
edby
mu
nic
ipali
ty.
*si
gn
ifica
nt
at10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
A-2.4 Police-Criminal Confrontations
Tab
leA
-25:
Clo
seP
AN
Ele
ctio
ns
an
dD
eath
sin
Poli
ce-C
rim
inal
Con
fronta
tion
s
Con
fron
tati
onP
robab
ilit
yC
onfr
onta
tion
Dea
ths
Quad
rati
cR
DP
olynom
ial
Lin
ear
RD
Pol
ynom
ial
Quad
rati
cR
DP
olynom
ial
Lin
ear
RD
Pol
ynom
ial
Pos
tL
ame
Pre
Pos
tL
ame
Pre
Pos
tL
ame
Pre
Pos
tL
ame
Pre
inau
g.duck
elec
tion
inau
g.duck
elec
tion
inau
g.duck
elec
tion
inau
g.duck
elec
tion
Pan
elA:20
07-200
8Elections
PA
Nw
in0.
031
0.01
4-0
.002
0.03
5-0
.002
0.01
58.
658*
*0.
805
0.18
123
.454
***
3.63
64.
040
(0.0
58)
(0.0
10)
(0.0
29)
(0.0
37)
(0.0
12)
(0.0
26)
(3.6
97)
(1.3
88)
(4.3
32)
(8.5
59)
(2.9
83)
(6.6
13)
Obs.
152
152
152
152
152
152
152
152
152
152
152
152
R2
0.03
70.
051
0.02
00.
021
0.03
10.
016
0.36
00.
331
0.20
90.
222
0.21
20.
124
Pan
elB:20
07-201
0Elections
PA
Nw
in0.
002
-0.0
070.
001
0.02
50.
017
0.01
515
.261
**-0
.104
0.00
726
.770
*0.
453*
0.64
9(0
.033
)(0
.019
)(0
.018
)(0
.021
)(0
.016
)(0
.016
)(7
.537
)(0
.160
)(0
.406
)(1
4.12
0)(0
.263
)(0
.549
)
Clu
ster
s30
730
730
730
730
730
730
730
730
730
730
730
7O
bs.
310
310
310
310
310
310
310
310
310
310
310
310
R2
0.03
60.
010
0.01
10.
016
0.00
50.
008
0.23
10.
100
0.07
20.
200
0.04
70.
038
Notes:
Th
ed
epen
den
tva
riab
leis
dea
ths
inp
olic
e-cr
imin
al
con
fronta
tion
s.P
AN
win
isan
ind
icato
req
ual
toon
eif
aP
AN
can
did
ate
won
the
elec
tion
,an
dth
esa
mp
lein
clu
des
elec
tion
sin
wh
ich
the
PA
Nw
asfi
rst
or
seco
nd
by
a5
per
centa
ge
poin
tor
less
vote
spre
ad
marg
in.
Colu
mn
s(1
)th
rou
gh
(3)
an
d(7
)th
rou
gh
(9)
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
les
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.C
olu
mn
s(4
)th
rou
gh
(6)
an
d(1
0)
thro
ugh
(12)
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
thre
shold
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
A-2.5 Robustness of Heterogeneity Results
A–39
Tab
leA
-26:
Hete
rogen
eit
y(5
%b
an
dw
idth
)
Dep
end
ent
vari
able
:d
rug-
rela
ted
hom
icid
era
te
2007
-200
8el
ecti
ons
2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
PA
Nw
in32
.981
***
37.6
99**
*37
.333
***
-2.3
4426
.735
***
32.5
94***
30.3
33***
-0.7
81
(9.3
46)
(9.0
00)
(9.7
52)
(4.0
20)
(8.5
60)
(10.2
61)
(8.7
25)
(8.1
64)
PA
Nw
inx
-49.
364*
**-3
4.487***
far
from
U.S
.(1
2.67
4)(1
1.2
31)
PA
Nw
inx
-51.
267*
**-3
8.2
62***
low
vio
len
ce(1
1.52
4)(9
.709)
PA
Nw
inx
0.33
11.0
00
loca
lga
ng
(14.
867)
(9.9
78)
PA
Nw
inx
33.7
47**
*33.6
80**
riva
l(1
0.82
7)(1
6.7
47)
PA
Nw
inx
11.5
224.9
00
ally
(10.
992)
(9.4
05)
R-s
qu
ared
0.32
60.
433
0.44
30.
504
0.10
20.2
01
0.2
16
0.2
20
Clu
ster
s15
215
215
215
230
7307
307
307
Ob
serv
atio
ns
152
152
152
152
310
310
310
310
PA
Nw
ineff
ect
-11.
670
-1.8
92
(far
from
US
)(8
.924
)(4
.566)
PA
Nw
ineff
ect
-13.
930*
*-7
.928*
(low
vio
len
ce)
(6.1
40)
(4.2
59)
PA
Nw
ineff
ect
-2.0
130.2
19
(loca
lga
ng)
(14.
310)
(5.7
36)
PA
Nw
ineff
ect
31.4
00**
*32.9
00**
(riv
al)
(10.
050)
(14.6
20)
PA
Nw
ineff
ect
9.17
84.1
19
(all
y)
(10.
230)
(4.6
69)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sli
mit
the
sam
ple
tom
un
icip
alit
ies
wh
ere
aP
AN
can
did
ate
was
the
win
ner
or
run
ner
-up
by
less
than
afi
vep
erce
nta
ge
poin
tvo
tesp
read
marg
inan
din
clu
de
ali
nea
rR
Dp
oly
nom
ial
esti
mat
edse
par
atel
yon
eith
ersi
de
ofth
eP
AN
win
-los
sth
resh
old
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
aren
thes
es.
*si
gnifi
cant
at10
%,
**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-27:
Hete
rogen
eit
y(4
%b
an
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Drug-relatedhomiciderate
2007-2008sample
2007-2010sample
PA
Nw
in36
.423
***
43.6
31**
*43
.031
***
-5.6
6622
.540
***
26.1
06***
23.8
36***
-1.9
74
(8.9
69)
(7.7
10)
(9.1
87)
(5.1
76)
(8.0
09)
(9.0
68)
(7.5
02)
(9.5
17)
PA
Nw
inx
-53.
324*
**-3
0.7
72***
far
from
U.S
.(1
2.09
0)(1
0.6
73)
PA
Nw
inx
-59.
589*
**-3
4.9
39***
low
vio
len
ce(1
1.33
5)(8
.642)
PA
Nw
inx
9.40
52.9
72
loca
lga
ng
(14.
139)
(11.0
99)
PA
Nw
inx
38.5
11**
*19.9
79
riva
l(1
0.41
8)(1
5.3
44)
PA
Nw
inx
21.8
61**
7.0
98
ally
(8.3
41)
(10.8
15)
R-s
qu
ared
0.39
20.
504
0.49
90.
587
0.20
30.2
69
0.3
08
0.3
22
Ob
serv
atio
ns
123
123
123
123
251
251
251
251
PA
Nw
ineff
ect
-9.6
93-4
.666
(far
from
US
)(9
.312
)(5
.629)
PA
Nw
ineff
ect
-16.
560*
*-1
1.1
0**
(low
vio
len
ce)
(6.6
39)
(4.2
90)
PA
Nw
ineff
ect
3.73
80.9
97
(loca
lga
ng)
(13.
160)
(5.7
13)
PA
Nw
ineff
ect
32.8
40**
*18.0
0(r
ival
)(9
.041
)(1
2.0
4)
PA
Nw
ineff
ect
16.1
90**
5.1
23
(all
y)
(6.5
41)
(5.1
39)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-28:
Hete
rogen
eit
y(3
%b
an
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Drug-relatedhomiciderate
2007-2008sample
2007-2010sample
PA
Nw
in38
.064
***
43.3
33**
*45
.302
***
-6.2
8724
.392
***
26.6
75***
24.3
02***
-2.0
64
(8.5
87)
(7.6
03)
(9.4
29)
(6.0
72)
(7.8
51)
(9.6
34)
(6.9
78)
(10.5
29)
PA
Nw
inx
-52.
538*
**-2
7.9
99**
far
from
U.S
.(1
2.85
4)(1
0.9
92)
PA
Nw
inx
-62.
145*
**-3
4.1
04***
low
vio
lence
(11.
631)
(8.2
30)
PA
Nw
inx
11.4
892.9
32
loca
lga
ng
(13.
437)
(12.0
05)
PA
Nw
inx
43.2
27**
*22.2
53
riva
l(1
0.78
0)(1
6.6
95)
PA
Nw
inx
22.0
37**
14.1
41
ally
(9.0
44)
(11.8
10)
R-s
qu
ared
0.41
40.
524
0.52
30.
612
0.16
20.2
38
0.2
77
0.2
91
Ob
serv
atio
ns
9494
9494
186
186
186
186
PA
Nw
ineff
ect
-9.2
06-1
.324
(far
from
US)
(10.
360)
(5.2
92)
PA
Nw
ineff
ect
-16.
840*
*-9
.802**
(low
vio
len
ce)
(6.8
10)
(4.3
65)
PA
Nw
ineff
ect
5.20
30.8
68
(loca
lga
ng)
(11.
990)
(5.7
68)
PA
Nw
ineff
ect
36.9
40**
*20.1
9(r
ival
)(8
.907
)(1
2.9
6)
PA
Nw
ineff
ect
15.7
5**
12.0
8**
(all
y)
(6.7
02)
(5.3
51)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-29:
Hete
rogen
eit
y(2
%b
an
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Drug-relatedhomiciderate
2007-2008sample
2007-2010sample
PA
Nw
in47
.111
***
62.9
14**
*36
.860
**-0
.535
17.5
22**
*18
.520***
16.7
88*
-16.8
49
(10.
817)
(9.8
89)
(15.
046)
(5.4
35)
(6.0
15)
(6.4
53)
(9.4
11)
(10.4
99)
PA
Nw
inx
-70.
701*
**-1
4.4
75
far
from
U.S
.(1
9.08
5)(9
.007)
PA
Nw
inx
-46.
786*
**-2
3.4
94**
low
vio
len
ce(1
6.45
4)(1
0.5
67)
PA
Nw
inx
26.9
029.3
62
loca
lga
ng
(25.
339)
(13.1
92)
PA
Nw
inx
49.3
99**
*45.1
04***
riva
l(1
6.37
8)(1
4.8
24)
PA
Nw
inx
25.9
80**
*31.9
71**
ally
(7.3
90)
(13.6
10)
R-s
qu
ared
0.34
90.
542
0.55
80.
663
0.12
50.2
21
0.2
32
0.3
11
Ob
serv
atio
ns
6262
6262
130
130
130
130
PA
Nw
ineff
ect
-7.7
884.0
45
(far
from
US
)(1
6.32
0)(6
.284)
PA
Nw
ineff
ect
-9.9
26-6
.705
(low
vio
len
ce)
(6.6
61)
(4.8
06)
PA
Nw
ineff
ect
26.3
70-7
.487
(loca
lga
ng)
(24.
750)
(7.9
87)
PA
Nw
ineff
ect
48.8
60**
*28.2
5***
(riv
al)
(15.
450)
(10.4
7)
PA
Nw
ineff
ect
25.4
50**
*15.1
2*
(all
y)
(5.0
08)
(8.6
61)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-30:
Hete
rogen
eit
y(1
3%
ban
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Drug-relatedhomiciderate
2007-2008sample
2007-2010sample
PA
Nw
in25
.621
***
30.7
01**
*30
.306
***
-0.1
9615
.580
**16
.849**
18.4
95**
-1.9
14
(8.4
84)
(9.1
08)
(8.8
86)
(2.2
51)
(7.1
00)
(8.1
30)
(8.0
38)
(4.8
17)
PA
Nw
inx
-36.
108*
**-1
7.739**
far
from
U.S
.(1
0.79
2)(8
.673)
PA
Nw
inx
-36.
721*
**-2
2.6
62***
low
vio
len
ce(1
0.47
8)(8
.601)
PA
Nw
inx
-6.5
593.1
13
loca
lga
ng
(9.6
54)
(6.0
83)
PA
Nw
inx
29.0
27**
20.6
75*
riva
l(1
1.31
1)(1
0.9
54)
PA
Nw
inx
7.81
75.5
65
ally
(6.8
54)
(5.9
05)
Ob
serv
atio
ns
380
380
380
380
764
764
764
764
R-s
qu
ared
0.18
40.
289
0.30
20.
365
0.08
00.1
45
0.1
66
0.1
91
PA
Nw
ineff
ect
-5.4
08-0
.890
(far
from
US
)(5
.789
)(3
.020)
PA
Nw
ineff
ect
-6.4
15-4
.167
(low
vio
lence
)(5
.552
)(3
.062)
PA
Nw
ineff
ect
-6.7
551.1
99
(loca
lga
ng)
(9.3
88)
(3.7
15)
PA
Nw
ineff
ect
28.8
3***
18.7
6**
(riv
al)
(11.
08)
(9.8
38)
PA
Nw
ineff
ect
7.62
23.6
52
(all
y)
(6.4
74)
(3.4
16)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-31:
Hete
rogen
eit
y(o
vera
llh
om
icid
es,
5%
ban
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Overallhomiciderate
2007-2008elections
2007-2010elections
PA
Nw
in56
.630
***
66.2
35**
*59
.683
***
-8.7
9844
.820
***
54.5
53***
46.3
71***
-3.6
56
(12.
768)
(11.
669)
(12.
309)
(5.5
11)
(12.
289)
(12.7
27)
(11.6
20)
(5.0
31)
PA
Nw
inx
-84.
627*
**-5
8.3
41***
far
from
U.S
.(1
6.10
6)(1
4.7
32)
PA
Nw
inx
-71.
613*
**-4
9.2
56***
low
vio
lence
(14.
238)
(12.9
85)
PA
Nw
inx
8.13
71.8
38
bor
der
slo
cal
gan
g(1
7.53
9)(7
.406)
PA
Nw
inx
73.1
91**
*56.7
24***
bor
der
sri
val
(14.
488)
(17.7
64)
PA
Nw
inx
16.0
4518.6
12**
bor
der
sal
ly(1
8.96
0)(8
.874)
R-s
qu
ared
0.39
60.
521
0.53
60.
593
0.23
70.3
60
0.4
19
0.4
12
Ob
serv
atio
ns
152
152
152
152
310
310
310
310
PA
Nw
ineff
ect
-18.
39-3
.787
(far
from
US)
(11.
10)
(7.4
21)
PA
Nw
ineff
ect
-11.
930*
-2.8
85
(low
vio
lence
)(7
.156
)(5
.796)
PA
Nw
ineff
ect
-0.6
61-1
.818
(bor
der
slo
cal
gan
g)(1
6.65
)(5
.435)
PA
Nw
ineff
ect
64.3
9***
53.0
7***
(bor
der
sri
val)
(13.
40)
(17.0
4)
PA
Nw
ineff
ect
7.24
714.9
6**
(bor
der
sal
ly)
(18.
140)
(7.3
11)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-32:
Hete
rogen
eit
y(o
vera
llh
om
icid
es,
4%
ban
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Overallhomiciderate
2007-2008elections
2007-2010elections
PA
Nw
in62
.219
***
73.7
06**
*68
.397
***
-10.
449
41.8
30**
*50
.865***
42.1
32***
-5.9
63
(11.
444)
(8.8
38)
(10.
357)
(6.9
78)
(12.
027)
(12.3
44)
(10.9
54)
(6.1
35)
PA
Nw
inx
-86.
308*
**-5
8.5
65***
far
from
U.S
.(1
3.09
8)(1
5.1
81)
PA
Nw
inx
-84.
534*
**-5
2.6
56***
low
vio
lence
(12.
926)
(12.4
58)
PA
Nw
inx
14.8
984.6
29
loca
lga
ng
(17.
122)
(8.1
35)
PA
Nw
inx
70.7
33**
*39.3
02**
bor
der
sri
val
(11.
601)
(17.7
28)
PA
Nw
inx
35.4
34**
*27.6
89***
bor
der
sal
ly(1
2.70
8)(8
.589)
Ob
serv
atio
ns
123
123
123
123
251
251
251
251
R-s
qu
ared
0.49
10.
625
0.62
50.
708
0.34
10.4
48
0.5
15
0.5
73
PA
Nw
ineff
ect
-12.
60-7
.700
(far
from
US)
(9.6
66)
(8.8
36)
PA
Nw
ineff
ect
-16.
14**
-10.5
2*
(low
vio
lence
)(7
.735
)(5
.936)
PA
Nw
ineff
ect
4.44
9-1
.334
(loca
lga
ng)
(15.
64)
(5.3
42)
PA
Nw
ineff
ect
60.2
8***
33.3
4**
(bor
der
sri
val)
(9.2
68)
(16.6
3)
PA
Nw
ineff
ect
24.9
9**
21.7
3***
(bor
der
sal
ly)
(10.
62)
(6.0
11)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-33:
Hete
rogen
eit
y(o
vera
llh
om
icid
es,
3%
ban
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Overallhomiciderate
2007-2008elections
2007-2010elections
PA
Nw
in63
.787
***
73.0
82**
*72
.177
***
-9.7
2742
.184
***
49.7
43***
38.5
28***
-4.3
02
(10.
791)
(8.6
94)
(10.
543)
(8.2
95)
(11.
541)
(12.9
27)
(11.3
85)
(6.1
95)
PA
Nw
inx
-84.
231*
**-5
7.5
44***
far
from
U.S
.(1
3.80
1)(1
6.2
51)
PA
Nw
inx
-88.
859*
**-5
0.0
16***
low
vio
lence
(13.
198)
(12.9
54)
PA
Nw
inx
13.8
86-0
.710
loca
lga
ng
(16.
519)
(8.7
05)
PA
Nw
inx
74.4
66**
*39.9
61**
(bor
der
sri
val)
(12.
398)
(18.9
60)
PA
Nw
inx
36.9
57**
*31.5
99***
bor
der
sal
ly(1
3.38
2)(8
.882)
Ob
serv
atio
ns
9494
9494
186
186
186
186
R-s
qu
ared
0.50
50.
640
0.63
80.
721
0.32
50.4
49
0.5
06
0.5
67
PA
Nw
ineff
ect
-11.
15-7
.801
(far
from
US)
(10.
72)
(9.8
48)
PA
Nw
ineff
ect
-16.
68**
-11.4
9*
(low
vio
lence
)(7
.939
)(6
.179)
PA
Nw
ineff
ect
4.15
9-5
.011
(loca
lga
ng)
(14.
28)
(6.1
16)
PA
Nw
ineff
ect
64.7
4***
35.6
6**
(bor
der
sri
val)
(9.2
13)
(17.9
2)
PA
Nw
ineff
ect
27.2
3***
27.3
0***
(bor
der
sal
ly)
(10.
50)
(6.3
65)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-34:
Hete
rogen
eit
y(o
vera
llh
om
icid
es,
2%
ban
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Overallhomiciderate
2007-2008elections
2007-2010elections
PA
Nw
in75
.771
***
96.2
69**
*72
.528
***
8.12
132
.247
***
39.7
75***
25.7
45
-16.4
21
(14.
434)
(12.
337)
(21.
121)
(10.
202)
(11.
685)
(10.9
70)
(15.8
91)
(12.7
04)
PA
Nw
inx
-98.
543*
**-4
5.1
18***
far
from
U.S
.(2
1.96
4)(1
6.7
66)
PA
Nw
inx
-80.
222*
**-3
0.3
23*
low
vio
lence
(22.
263)
(17.4
50)
PA
Nw
inx
31.6
8821.2
40
loca
lga
ng
(32.
210)
(16.3
91)
PA
Nw
inx
71.1
76**
*50.4
82**
bor
der
sri
val
(18.
826)
(21.5
73)
PA
Nw
inx
36.7
08**
39.9
46***
bor
der
sal
ly(1
4.58
8)(1
5.0
55)
Ob
serv
atio
ns
6262
6262
130
130
130
130
R-s
qu
ared
0.42
80.
641
0.64
40.
757
0.26
90.4
30
0.4
65
0.5
82
PA
Nw
ineff
ect
-2.2
73-5
.344
(far
from
US)
(18.
17)
(12.6
8)
PA
Nw
ineff
ect
-7.6
94-4
.577
(low
vio
lence
)(7
.039
)(7
.209)
PA
Nw
ineff
ect
39.8
14.8
20
(loca
lga
ng)
(30.
55)
(10.3
6)
PA
Nw
ineff
ect
79.3
0***
34.0
6**
(bor
der
sri
val)
(15.
82)
(17.4
4)
PA
Nw
ineff
ect
44.8
3***
23.5
3***
(bor
der
sal
ly)
(10.
43)
(8.0
79)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-35:
Hete
rogen
eit
y(o
vera
llh
om
icid
es,
13.3
%b
an
dw
idth
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
var:
Overallhomiciderate
2007-2008elections
2007-2010elections
Dep
var:
Overallhomiciderate
Dep
var:
Overallhomiciderate
PA
Nw
in44
.551
***
53.3
98**
*50
.809
***
-7.3
32*
29.6
42**
*35
.266***
33.9
76***
-4.2
48
(11.
967)
(12.
223)
(11.
327)
(4.1
39)
(10.
913)
(11.4
89)
(10.6
43)
(3.4
01)
PA
Nw
inx
-59.
568*
**-3
6.5
90***
far
from
U.S
.(1
3.56
5)(1
2.4
64)
PA
Nw
inx
-56.
665*
**-3
6.5
36***
low
vio
lence
(12.
876)
(11.5
63)
PA
Nw
inx
-1.6
773.8
59
loca
lga
ng
(12.
021)
(5.5
96)
PA
Nw
inx
60.6
79**
*42.5
09***
bor
der
sri
val
(14.
663)
(13.8
02)
PA
Nw
inx
21.9
12**
19.6
23***
bor
der
sal
ly(1
1.02
9)(6
.379)
Ob
serv
atio
ns
380
380
380
380
764
764
764
764
R-s
qu
ared
0.27
00.
389
0.43
50.
481
0.16
50.2
56
0.3
28
0.3
26
PA
Nw
ineff
ect
-6.1
70-1
.324
(far
from
US)
(5.8
83)
(4.8
33)
PA
Nw
ineff
ect
-5.8
56-2
.561
(low
vio
lence
)(6
.124
)(4
.520)
PA
Nw
ineff
ect
-9.0
10-0
.389
(loca
lga
ng)
(11.
29)
(4.4
43)
PA
Nw
ineff
ect
53.3
5***
38.2
6***
(bor
der
sri
val)
(14.
07)
(13.3
8)
PA
Nw
ineff
ect
14.5
815.3
8***
(bor
der
sal
ly)
(10.
22)
(5.3
97)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,fa
rfr
om
U.S
.is
an
ind
icato
req
ual
to1
ifth
em
un
icip
ali
tyis
ab
ove
med
ian
dis
tan
cefr
omth
eU
.S.,
low
vio
len
ceis
anin
dic
ator
equ
al
to1
ifth
em
un
icip
ali
tyh
ad
ab
elow
med
ian
hom
icid
era
ted
uri
ng
2004-2
006,
loca
lgan
gis
an
ind
icato
req
ual
toon
eif
the
mu
nic
ipal
ity
conta
ins
only
alo
cal
gan
g,
riva
lis
an
ind
icato
req
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
db
ord
ers
terr
itory
contr
oll
edby
ari
val
DT
O,
and
ally
isan
ind
icat
oreq
ual
toon
eif
itco
nta
ins
am
ajo
rD
TO
an
dd
oes
not
bord
erte
rrit
ory
contr
oll
edby
ari
val
DT
O.
All
colu
mn
sin
clu
de
ali
nea
rR
Dp
olyn
omia
les
tim
ated
sep
arat
ely
onei
ther
sid
eof
the
PA
Nw
in-l
oss
thre
shold
.In
ad
dit
ion
toth
ein
tera
ctio
ns,
main
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
ity.
*si
gnifi
cant
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
A-2.6 Robustness of Results on Local Politics and Violence
Tab
leA
-36:
Loca
lP
oli
tics
an
dD
rug-R
ela
ted
Hom
icid
es
(5%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
ons
2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in32
.981
***
30.1
34**
*34.0
38***
26.7
35***
33.3
36***
24.1
23***
(9.3
46)
(8.0
82)
(11.1
73)
(8.5
60)
(9.8
01)
(6.5
31)
PA
Nw
inx
-32.
965*
**-3
2.8
12***
PA
Nin
cum
b.
(9.7
04)
(10.9
99)
Alt
er(P
AN
)8.
147
2.9
96
(6.3
13)
(6.0
41)
PR
Iw
in11.5
23
1.6
93
(10.5
50)
(13.0
92)
Alt
er(P
RI/
PR
D)
4.4
19
-2.7
95
(3.7
28)
(6.0
78)
PA
Nw
inx
3.6
17
0.4
15
PA
Ngo
v.
(15.4
94)
(17.3
42)
Clu
ster
s15
215
215
2142
142
152
307
307
307
181
181
307
Ob
serv
atio
ns
152
152
152
142
142
152
310
310
310
183
183
310
R-s
qu
ared
0.32
60.
470
0.10
40.0
38
0.0
39
0.3
42
0.1
02
0.255
0.1
87
0.0
89
0.0
96
0.1
19
PA
Nw
ineff
ect
-2.8
310.
524
(PA
Nin
cum
b.)
(5.3
70)
(4.9
93)
PA
Nw
ineff
ect
37.6
60***
24.5
40
(PA
Ngo
v.)
(10.7
30)
(16.0
70)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
pby
less
than
afi
ve
per
centa
ge
poin
tvo
tesp
read
marg
in;
an
dco
lum
ns
(4),
(5),
(10),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipal
itie
sw
ith
acl
ose
elec
tion
bet
wee
nP
RI
and
PR
Dca
nd
idate
s.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
poly
nom
ial
esti
mate
dse
para
tely
on
eith
ersi
de
of
the
thre
shol
d.
Inco
lum
ns
(2),
(6),
(8),
and
(12)
,m
ain
effec
tsare
als
oin
clu
ded
.S
tan
dard
erro
rs,
clu
ster
edby
mu
nic
ipali
ty,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at5%
,**
*si
gnifi
cant
at1%
.
Tab
leA
-37:
Loca
lP
oli
tics
an
dD
rug-R
ela
ted
Hom
icid
es
(4%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
on
s2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in36
.423
***
32.7
89**
*36.2
08***
22.5
40***
22.3
82***
19.4
03***
(8.9
69)
(7.8
86)
(10.3
25)
(8.0
09)
(7.4
29)
(4.8
42)
PA
Nw
inx
-34.
597*
**-2
1.2
79**
PA
Nin
cum
b.
(8.3
80)
(9.1
81)
Alt
er(P
AN
)8.
687
10.1
57**
(6.3
70)
(4.6
25)
PR
Iw
in15.7
29
18.9
13
(10.3
47)
(11.6
87)
Alt
er(P
RI/
PR
D)
2.3
96
1.0
49
(3.7
47)
(4.8
52)
PA
Nw
inx
31.3
77*
-0.5
38
PA
Ngo
v.
(18.7
41)
(20.7
65)
Ob
serv
atio
ns
123
123
123
116
116
123
251
251
251
147
147
251
R-s
qu
ared
0.39
20.
523
0.15
50.0
87
0.0
17
0.4
10
0.2
03
0.2
82
0.1
70
0.0
53
0.0
03
0.2
27
PA
Nw
ineff
ect
-1.8
081.1
03
(PA
Nin
cum
b.)
(2.8
32)
(5.3
94)
PA
Nw
ineff
ect
67.5
80***
18.8
70
(PA
Ngo
v.)
(15.6
40)
(20.1
90)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-38:
Loca
lP
oli
tics
an
dD
rug-R
ela
ted
Hom
icid
es
(3%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
on
s2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in38
.064
***
33.0
36**
*36.9
67***
24.3
92***
22.5
90***
20.6
84***
(8.5
87)
(8.2
52)
(10.1
35)
(7.8
51)
(7.3
52)
(4.7
63)
PA
Nw
inx
-34.
499*
**-1
8.0
65*
PA
Nin
cum
b.
(8.8
17)
(9.3
01)
Alt
er(P
AN
)9.
565
10.1
93**
(6.8
95)
(4.6
39)
PR
Iw
in13.4
67
19.9
29
(13.2
11)
(14.9
89)
Alt
er(P
RI/
PR
D)
3.4
06
3.0
02
(3.9
48)
(4.8
30)
PA
Nw
inx
31.1
66
1.6
25
PA
Ngo
v.
(18.7
93)
(21.5
32)
Ob
serv
atio
ns
9494
9492
92
94
186
186
186
116
116
186
R-s
qu
ared
0.41
40.
537
0.14
90.0
86
0.0
28
0.4
24
0.1
62
0.2
44
0.1
47
0.0
53
0.0
14
0.1
96
PA
Nw
ineff
ect
-1.4
634.5
25
(PA
Nin
cum
b.)
(3.1
07)
(5.6
98)
PA
Nw
ineff
ect
68.1
30***
22.3
10
(PA
Ngo
v.)
(15.8
30)
(21.0
00)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-39:
Loca
lP
oli
tics
an
dD
rug-R
ela
ted
Hom
icid
es
(2%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
on
s2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in47
.111
***
45.5
37**
*49.8
28***
17.5
22***
11.8
39*
17.9
62**
(10.
817)
(13.
285)
(11.0
89)
(6.0
15)
(7.0
31)
(7.1
59)
PA
Nw
inx
-48.
438*
**-2
.714
PA
Nin
cum
b.
(15.
258)
(12.
234)
Alt
er(P
AN
)8.
199
10.1
04***
(11.
059)
(3.7
47)
PR
Iw
in20.4
07
37.9
53
(21.4
06)
(23.1
65)
Alt
er(P
RI/
PR
D)
5.7
82
5.2
72
(5.6
02)
(5.9
61)
PA
Nw
inx
58.1
36
-15.9
75
PA
Ngo
v.
(52.2
87)
(26.1
05)
Ob
serv
atio
ns
6262
6261
61
62
130
130
130
78
78
130
R-s
qu
ared
0.34
90.
529
0.21
80.1
16
0.0
40
0.4
01
0.1
25
0.263
0.2
25
0.0
89
0.0
09
0.1
34
PA
Nw
ineff
ect
-2.9
019.1
25
(PA
Nin
cum
b.)
(7.5
04)
(10.0
10)
PA
Nw
ineff
ect
108.0
00**
1.9
87
(PA
Ngo
v.)
(51.1
00)
(25.1
00)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-40:
Loca
lP
oli
tics
an
dD
rug-R
ela
ted
Hom
icid
es
(13.3
%B
an
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-2008
elec
tions
2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in25
.621
***
23.6
14**
*28.1
31**
15.5
80**
16.2
80**
14.1
32**
(8.4
84)
(7.8
72)
(11.5
73)
(7.1
00)
(7.9
99)
(5.9
45)
PA
Nw
inx
-20.
390*
*-1
5.9
39*
PA
Nin
cum
b.
(8.2
01)
(8.4
78)
Alt
er(P
AN
)6.
226
1.9
66
(5.8
97)
(4.6
49)
PR
Iw
in10.2
11
12.1
50
(6.4
61)
(8.6
80)
Alt
er(P
RI/
PR
D)
-0.1
03
-6.4
76
(3.5
14)
(6.5
58)
PA
Nw
inx
-2.4
49
0.2
60
PA
Ngo
v.
(14.7
20)
(13.2
37)
Ob
serv
atio
ns
380
380
380
308
308
380
764
764
764
423
423
764
R-s
qu
ared
0.18
40.
292
0.026
0.0
28
0.0
23
0.2
13
0.0
80
0.1
61
0.0
29
0.0
38
0.0
41
0.0
84
PA
Nw
ineff
ect
3.22
40.3
41
(PA
Nin
cum
b.)
(2.3
01)
(2.7
94)
PA
Nw
ineff
ect
25.6
80***
14.3
90
(PA
Ngo
v.)
(9.0
97)
(11.8
30)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-41:
Loca
lP
oli
tics
an
dO
vera
llH
om
icid
es
(5%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
ons
2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in56
.630
***
60.4
82**
*54.0
09***
44.8
20***
48.1
63***
35.7
23***
(12.
768)
(10.
886)
(16.8
49)
(12.2
89)
(12.0
96)
(11.0
52)
PA
Nw
inx
-66.
399*
**-3
5.1
06***
PA
Nin
cum
b.
(14.
555)
(13.3
35)
Alt
er(P
AN
)15
.459
8.1
51
(9.4
31)
(8.3
11)
PR
Iw
in11.5
50
-0.2
79
(11.0
69)
(12.1
98)
Alt
er(P
RI/
PR
D)
4.6
11
0.1
68
(4.6
26)
(5.8
45)
PA
Nw
inx
13.0
28
23.6
45
PA
Ngo
v.
(23.8
68)
(24.1
92)
Ob
serv
atio
ns
152
152
152
142
142
152
310
310
310
183
183
310
R-s
qu
ared
0.39
60.
535
0.16
70.0
33
0.0
16
0.4
07
0.2
37
0.4
01
0.2
02
0.0
32
0.0
43
0.2
62
PA
Nw
ineff
ect
-5.9
1813.0
60**
(PA
Nin
cum
b.)
(9.6
61)
(5.6
15)
PA
Nw
ineff
ect
67.0
40***
59.3
70***
(PA
Ngo
v.)
(16.9
10)
(21.5
20)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-42:
Loca
lP
oli
tics
an
dO
vera
llH
om
icid
es
(4%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
on
s2007-2
010
elec
tions
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in62
.219
***
62.1
76**
*58.8
42***
41.8
30***
39.
053***
33.2
53***
(11.
444)
(9.8
70)
(14.6
19)
(12.0
27)
(11.0
61)
(10.0
35)
PA
Nw
inx
-64.
698*
**-2
5.3
80**
PA
Nin
cum
b.
(11.
624)
(12.5
36)
Alt
er(P
AN
)18
.822
**
13.9
29*
(9.2
77)
(7.7
51)
PR
Iw
in14.7
21
16.3
24
(11.3
53)
(11.5
89)
Alt
er(P
RI/
PR
D)
2.1
95
2.9
42
(5.1
20)
(5.3
56)
PA
Nw
inx
51.3
09*
20.4
09
PA
Ngo
v.
(28.2
99)
(26.6
12)
Ob
serv
atio
ns
123
123
123
116
116
123
251
251
251
147
147
251
R-s
qu
ared
0.49
10.
612
0.21
70.0
80
0.0
21
0.5
14
0.3
41
0.4
56
0.2
42
0.0
56
0.0
20
0.3
70
PA
Nw
ineff
ect
-2.5
2213
.670**
(PA
Nin
cum
b.)
(6.1
41)
(5.9
00)
PA
Nw
ineff
ect
110.2
00***
53.6
60**
(PA
Ngo
v.)
(24.2
30)
(24.6
50)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-43:
Loca
lP
oli
tics
an
dO
vera
llH
om
icid
es
(3%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
on
s2007-2
010
elec
tions
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in63
.787
***
62.8
69**
*60.8
46***
42.1
84***
38.
620***
34.2
89***
(10.
791)
(10.
277)
(14.2
05)
(11.5
41)
(11.0
30)
(10.2
18)
PA
Nw
inx
-76.
492*
**-2
4.1
35*
PA
Nin
cum
b.
(13.
065)
(12.6
63)
Alt
er(P
AN
)20
.314
**
13.4
68*
(9.8
04)
(7.6
98)
PR
Iw
in10.6
39
16.5
70
(14.4
20)
(14.9
10)
Alt
er(P
RI/
PR
D)
4.1
47
5.5
25
(5.2
00)
(5.1
57)
PA
Nw
inx
49.6
64*
24.4
75
PA
Ngo
v.
(28.3
16)
(27.3
41)
Ob
serv
atio
ns
9494
9492
92
94
186
186
186
116
116
186
R-s
qu
ared
0.50
50.
617
0.22
20.0
82
0.0
29
0.5
25
0.3
25
0.4
40
0.2
50
0.0
54
0.0
31
0.3
70
PA
Nw
ineff
ect
-13.
620*
14.4
80**
(PA
Nin
cum
b.)
(8.0
66)
(6.2
20)
PA
Nw
ineff
ect
110.5
00***
58.7
60**
(PA
Ngo
v.)
(24.5
00)
(25.3
60)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-44:
Loca
lP
oli
tics
an
dO
vera
llH
om
icid
es
(2%
Ban
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-200
8el
ecti
on
s2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in75
.771
***
76.0
52**
*83.1
08***
32.2
47***
25.6
49*
38.1
82***
(14.
434)
(17.
870)
(13.7
76)
(11.6
85)
(14.4
57)
(12.3
92)
PA
Nw
inx
-107
.590
***
-10.1
49
PA
Nin
cum
b.
(21.
737)
(17.5
14)
Alt
er(P
AN
)20
.723
13.1
51**
(15.
258)
(6.3
76)
PR
Iw
in17.0
05
34.5
48
(23.1
66)
(22.2
97)
Alt
er(P
RI/
PR
D)
6.3
74
7.7
65
(6.5
86)
(6.3
05)
PA
Nw
inx
63.0
32
-17.3
89
PA
Ngo
v.
(94.4
14)
(30.3
07)
Ob
serv
atio
ns
6262
6261
61
62
130
130
130
78
78
130
R-s
qu
ared
0.42
80.
609
0.31
60.1
05
0.0
28
0.5
18
0.2
69
0.461
0.4
00
0.0
88
0.0
25
0.3
04
PA
Nw
ineff
ect
-31.
540*
*15.5
00
(PA
Nin
cum
b.)
(12.
380)
(9.8
85)
PA
Nw
ineff
ect
146.1
00
20.7
90
(PA
Ngo
v.)
(93.4
00)
(27.6
60)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-45:
Loca
lP
oli
tics
an
dO
vera
llH
om
icid
es
(13.3
%B
an
dw
dit
h)
Dep
end
ent
vari
ab
le:
dru
g-r
elate
dh
om
icid
era
te
2007
-2008
elec
tion
s2007-2
010
elec
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
PA
Nw
in44
.551
***
44.0
67**
*44.7
74***
29.6
42***
29.8
63***
24.3
42**
(11.
967)
(10.
326)
(17.0
80)
(10.9
13)
(10.6
96)
(10.8
18)
PA
Nw
inx
-35.
807*
**-2
1.8
65*
PA
Nin
cum
b.
(11.
183)
(11.3
68)
Alt
er(P
AN
)12
.353
4.9
50
(9.0
46)
(6.3
44)
PR
Iw
in6.8
37
7.0
93
(7.2
39)
(8.1
49)
Alt
er(P
RI/
PR
D)
-0.3
04
-3.7
52
(4.5
03)
(5.8
55)
PA
Nw
inx
3.2
74
11.4
55
PA
Ngo
v.
(22.4
02)
(20.1
87)
Ob
serv
atio
ns
380
380
380
308
308
380
764
764
764
423
423
764
R-s
qu
ared
0.27
00.
379
0.039
0.0
12
0.0
20
0.2
87
0.1
65
0.2
85
0.0
35
0.0
16
0.0
19
0.1
73
PA
Nw
ineff
ect
8.26
0*7.9
97**
(PA
Nin
cum
b.)
(4.2
94)
(3.8
27)
PA
Nw
ineff
ect
48.0
50***
35.8
00**
(PA
Ngo
v.)
(14.4
90)
(17.0
40)
Notes:
PA
Nw
inis
anin
dic
ator
equ
alto
one
ifa
PA
Nca
nd
idate
won
the
elec
tion
,P
AN
incu
mb
ent
isan
ind
icato
req
ual
to1
ifth
eP
AN
hel
dth
em
ayors
hip
du
rin
gth
ep
revio
us
term
,P
AN
gove
rnor
isan
ind
icat
or
equ
al
to1
ifth
est
ate
has
aP
AN
gov
ern
or,
PR
Iw
inis
an
ind
icato
req
ual
to1
ifth
eP
RI
won
the
elec
tion
,an
dal
ter
isa
du
mm
yeq
ual
one
ifth
ep
arty
contr
olli
ng
the
may
ors
hip
chan
ged
.C
olu
mn
s(1
)-
(3),
(6)
-(9
),an
d(1
2)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
wh
ere
aP
AN
can
did
ate
was
the
win
ner
orru
nn
er-u
p;
and
colu
mn
s(4
),(5
),(1
0),
an
d(1
1)
lim
itth
esa
mp
leto
mu
nic
ipaliti
esw
ith
acl
ose
elec
tion
bet
wee
nP
RI
an
dP
RD
can
did
ates
.A
llco
lum
ns
incl
ud
ea
lin
ear
RD
pol
yn
omia
les
tim
ate
dse
para
tely
on
eith
ersi
de
of
the
thre
shold
.In
colu
mn
s(2
),(6
),(8
),an
d(1
2),
main
effec
tsare
also
incl
ud
ed.
Sta
nd
ard
erro
rs,
clu
ster
edby
mu
nic
ipal
ity,
are
inp
are
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
A-2.7 Corruption and Other Results
Table A-46: Corruption
(1) (2) (3) (4) (5)
Bandwidth5% 4% 3% 2% 13.3%
Panel A: Means comparisonPAN win -0.022 -0.023 0.021 0.054 -0.007
(0.087) (0.097) (0.121) (0.152) (0.055)R-squared 0.001 0.001 0.000 0.003 0.000
Panel B: RD analysisPAN win 0.091 0.013 -0.034 -0.324 -0.005
(0.159) (0.174) (0.215) (0.295) (0.091)R-squared 0.124 0.164 0.133 0.109 0.027
Observations 102 84 62 44 237Mean dep. var. 0.245 0.262 0.323 0.409 0.231
Notes: PAN win is an indicator equal to one if a PAN candidate won the election, and the dependentvariable is an indicator equal to 1 if official government records document the mayor engaging in corruptionin 2008. Close elections from 2007 where the mayor had take office by the beginning of 2008 are included inthe sample. Panel B includes a linear RD polynomial estimated separately on either side of the PANwin-loss threshold. * significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-47: Violence and Corruption of the Losing Party
(1) (2) (3) (4)
Bandwidth5% 5% 13.3% 13.3%
PAN win 81.584* 43.017 37.418* 16.686(42.919) (37.565) (21.431) (13.875)
Loser corrupt 12.160 3.582(24.946) (8.288)
PAN win x 109.946** 83.278**Loser corrupt (50.657) (33.414)
Observations 61 61 165 165R-squared 0.200 0.303 0.099 0.204
Notes: The dependent variable is the homicide rate during the one year following the mayor’sinauguration. PAN win is an indicator equal to one if a PAN candidate won the election, and loser corruptis an indicator equal to 1 if official government records document that the losing party was engaged incorruption during the previous mayor’s term, in 2008. The only way to observe this is if the losing party isthe incumbent party, so in all municipalities with PAN win= 1, the PAN did not hold the mayorshippreviously. 2009-2010 close elections where the incumbent party lost form the sample. All columns includea linear RD polynomial estimated separately on either side of the PAN win-loss threshold. * significant at10%, ** significant at 5%, *** significant at 1%.
Table A-48: Political Competition and Violence
(1) (2) (3) (4) (5) (6) (7) (8)
Drug trade-related Overall Drug trade-related Overallhomicide rate homicide probability
07-08 07-10 07-08 07-10 07-08 07-10 07-08 07-10
5% bandwidthabs(spread) -1.165** -0.160 -0.604 -0.338 -0.021* -0.000 -0.057 0.025
(0.535) (1.152) (0.719) (0.627) (0.011) (0.009) (0.643) (0.428)4% bandwidthabs(spread) -1.234 -0.864 -1.128 -1.247 -0.036** -0.008 -0.186 -0.229
(0.809) (0.924) (1.188) (0.842) (0.016) (0.012) (0.988) (0.566)3% bandwidthabs(spread) -1.008 -0.913 -1.440 -1.216 -0.042* -0.021 1.472 1.413
(0.988) (1.106) (1.677) (1.285) (0.025) (0.016) (2.351) (1.405)2% bandwidthabs(spread) 0.621 3.290 3.037 2.859 -0.101* -0.020 -0.458 2.269
(3.194) (2.905) (2.811) (2.150) (0.058) (0.033) (2.778) (2.254)13.3% bandwidthabs(spread) -0.298* -0.265 -0.020 -0.158 -0.003 -0.003 -0.059 -0.083
(0.172) (0.251) (0.202) (0.155) (0.002) (0.002) (0.189) (0.110)
Notes: The table reports coefficients from regressing violence measures on the absolute value of the votespread. Each row considers a different vote spread bandwidth.
A-2.8 Robustness of Spillover Results
Tab
leA
-49:
Th
eD
ivers
ion
of
Dru
gT
raffi
c(2
007-2
010
Ele
ctio
ns)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Full
Sam
ple
Lim
ited
Sam
ple
Full
Sam
ple
Dom
esti
cIl
lici
tD
rug
Con
fisc
atio
ns
Coca
ine
Con
fisc
atio
ns
Dum
my
Val
ue
Val
ue
Dum
my
Val
ue
Val
ue
Dum
my
Val
ue
Val
ue
Pan
elA:Sho
rtestPaths
Pre
dic
ted
0.00
8*0.
080
0.00
70.
048
-0.0
010.
005
route
sdum
my
(0.0
05)
(0.0
60)
(0.0
08)
(0.0
93)
(0.0
05)
(0.0
25)
Pre
dic
ted
0.01
8***
0.01
60.
004
route
sco
unt
(0.0
06)
(0.0
10)
(0.0
03)
Pan
elB:Model
withCon
gestionCosts
Pre
dic
ted
0.00
6*0.
062
0.00
80.
093
0.00
40.
021
route
sdum
my
(0.0
04)
(0.0
41)
(0.0
06)
(0.0
61)
(0.0
04)
(0.0
20)
Pre
dic
ted
0.00
5*0.
007*
0.00
3ro
ute
sco
unt
(0.0
03)
(0.0
04)
(0.0
02)
Munic
ipal
itie
s1,
816
1,81
61,
816
937
937
937
1,81
61,
816
1,81
6O
bse
rvat
ions
88,9
8488
,984
88,9
8445
,913
45,9
1345
,913
88,9
8488
,984
88,9
84
Notes:
Th
edep
end
ent
vari
able
inco
lum
ns
(1),
and
(4)
isan
indic
ato
req
ual
to1
ifd
om
esti
cil
lici
td
rug
con
fisc
ati
on
sare
mad
ein
agiv
enm
un
icip
ali
ty-m
onth
;th
ed
epen
den
tva
riab
lein
colu
mn
s(2
),(3
),(5
),an
d(6
)is
the
log
valu
eof
dom
esti
cil
lici
td
rug
con
fisc
ati
on
s(o
r0
ifn
oco
nfi
scati
on
sare
mad
e);
the
dep
end
ent
vari
ab
lein
colu
mn
(7)
isan
ind
icat
oreq
ual
to1
ifco
cain
eco
nfi
scati
on
sare
mad
ein
agiv
enm
un
icip
ali
ty-m
onth
;an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(8
)an
d(9
)is
the
log
valu
eof
con
fisc
ated
coca
ine
(or
0if
no
con
fisc
atio
ns
are
mad
e).
Colu
mn
s(4
)th
rou
gh
(6)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
that
do
not
bord
era
mu
nic
ipali
tyth
ath
asex
per
ien
ced
acl
ose
PA
Nvic
tory
from
2007
to2010.
Pan
elA
pre
dic
tstr
affi
ckin
gro
ute
su
sin
gth
esh
ort
est
path
sm
od
el,
an
dP
an
elB
use
sth
em
od
elw
ith
con
gest
ion
cost
s.A
llco
lum
ns
incl
ud
em
onth
xst
ate
and
mu
nic
ipali
tyfi
xed
effec
ts.
Sta
nd
ard
erro
rscl
ust
ered
by
mu
nic
ipali
tyan
dm
onth
xst
ate
are
rep
ort
edin
par
enth
eses
.*
sign
ifica
nt
at10
%,
**si
gnifi
cant
at5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-50:
Vio
len
ceS
pil
lovers
(2007-2
010
Ele
ctio
ns)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Full
Sam
ple
Lim
ited
Sam
ple
Dep.var.:Drugtrad
e-relatedho
micide
Dep.Var.:
Drugtrad
e-relatedho
micide
dummy
rate
rate
dummy
rate
dummy
rate
rate
dummy
rate
Pan
elA:Sho
rtestPaths
Pre
dic
ted
0.00
31.
833
-0.0
111.
215
route
sdum
my
(0.0
05)
(1.3
68)
(0.0
09)
(2.0
58)
Pre
dic
ted
0.47
8**
0.40
6ro
ute
sco
unt
(0.2
22)
(0.2
63)
One
route
-0.0
01-3
.669
-0.0
180.
460
(0.0
06)
(3.2
86)
(0.0
12)
(1.3
45)
Mor
eth
an0.
007
6.02
2**
-0.0
071.
759
one
route
(0.0
07)
(2.5
53)
(0.0
11)
(2.9
02)
Pan
elB:Model
withCon
gestionCosts
Pre
dic
ted
0.00
31.
278
0.00
30.
601
route
sdum
my
(0.0
04)
(0.7
87)
(0.0
07)
(1.0
57)
Pre
dic
ted
0.03
60.
066
route
sco
unt
(0.0
45)
(0.0
76)
One
route
-0.0
040.
803
-0.0
060.
029
(0.0
06)
(1.2
93)
(0.0
09)
(0.8
93)
Mor
eth
an0.
006
1.43
00.
007
0.81
1on
ero
ute
(0.0
05)
(0.9
76)
(0.0
07)
(1.1
99)
Munic
ipal
itie
s1,
816
1,81
61,
816
1,81
61,
816
937
937
937
937
937
Obse
rvat
ions
88,9
8488
,984
88,9
8488
,984
88,9
8445
,913
45,9
1345
,913
45,9
1345
,913
Notes:
Th
ed
epen
den
tva
riab
lein
colu
mn
s(1
),(4
),(6
)an
d(9
)is
an
ind
icato
req
ual
to1
ifa
dru
gtr
ad
e-re
late
dh
om
icid
eocc
urr
edin
agiv
enm
un
icip
ali
ty-m
onth
,an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(2
),(3
),(5
),(7
),(8
),an
d(1
0)
isth
ed
rug
trad
e-re
late
dh
om
icid
era
tep
er100,0
00
mu
nic
ipal
inh
ab
itants
.C
olu
mn
s(6
)th
rou
gh(1
0)li
mit
the
sam
ple
tom
un
icip
alit
ies
that
do
not
bord
era
mu
nic
ipali
tyth
at
exp
erie
nce
da
close
PA
Nvic
tory
bet
wee
n2007
an
d2010.
All
colu
mn
sin
clu
de
mon
thx
stat
ean
dm
un
icip
alit
yfi
xed
effec
ts.
Sta
nd
ard
erro
rscl
ust
ered
by
mu
nic
ipali
tyan
dm
onth
xst
ate
are
rep
ort
edin
pare
nth
eses
.*
signifi
cant
at
10%
,**
sign
ifica
nt
at5%
,**
*si
gnifi
cant
at1%
.
Tab
leA
-51:
Th
eD
ivers
ion
of
Dru
gT
raffi
c(C
ontr
oll
ing
for
PA
Nm
ayors
)
Dep
.va
r.:
Dom
esti
cIl
lici
tD
rug
Con
fisc
atio
ns
Coca
ine
Con
fisc
atio
ns
Dum
my
Val
ue
Val
ue
Dum
my
Val
ue
Val
ue
Dum
my
Val
ue
Val
ue
Full
Sam
ple
Lim
ited
Sam
ple
Full
Sam
ple
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Pan
elA:Sho
rtestPaths
Pre
dic
ted
0.01
6***
0.17
0***
0.01
6**
0.17
0***
0.00
40.
028
route
sdum
my
(0.0
05)
(0.0
50)
(0.0
07)
(0.0
65)
(0.0
04)
(0.0
20)
Pre
dic
ted
0.02
2***
0.01
5*0.
006
route
sco
unt
(0.0
08)
(0.0
09)
(0.0
06)
Pan
elB:Model
withCon
gestionCosts
Pre
dic
ted
0.01
3**
0.14
9***
0.01
1*0.
129*
*0.
002
0.00
9ro
ute
sdum
my
(0.0
05)
(0.0
57)
(0.0
06)
(0.0
65)
(0.0
04)
(0.0
25)
Pre
dic
ted
0.00
40.
002
0.00
1ro
ute
sco
unt
(0.0
04)
(0.0
04)
(0.0
02)
Munic
ipal
itie
s18
6918
6918
6915
6215
6215
6218
6918
6918
69O
bse
rvat
ions
6915
369
153
6915
357
,794
57,7
9457
,794
6915
369
153
6915
3
Notes:
Th
edep
end
ent
vari
able
inco
lum
ns
(1),
and
(4)
isan
indic
ato
req
ual
to1
ifd
om
esti
cil
lici
td
rug
con
fisc
ati
on
sare
mad
ein
agiv
enm
un
icip
ali
ty-m
onth
;th
ed
epen
den
tva
riab
lein
colu
mn
s(2
),(3
),(5
),an
d(6
)is
the
log
valu
eof
dom
esti
cil
lici
td
rug
con
fisc
ati
on
s(o
r0
ifn
oco
nfi
scati
on
sare
mad
e);
the
dep
end
ent
vari
ab
lein
colu
mn
(7)
isan
ind
icat
oreq
ual
to1
ifco
cain
eco
nfi
scati
on
sare
mad
ein
agiv
enm
un
icip
ali
ty-m
onth
;an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(8
)an
d(9
)is
the
log
valu
eof
con
fisc
ated
coca
ine
(or
0if
no
con
fisc
atio
ns
are
mad
e).
Colu
mn
s(4
)th
rou
gh
(6)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
that
do
not
bord
era
mu
nic
ipali
tyth
ath
asex
per
ien
ced
acl
ose
PA
Nvic
tory
.P
anel
Ap
red
icts
traffi
ckin
gro
ute
su
sin
gth
esh
ort
est
path
sm
od
el,
and
Pan
elB
use
sth
em
od
elw
ith
con
ges
tion
cost
s.A
llco
lum
ns
incl
ud
em
onth
xst
ate
and
mu
nic
ipal
ity
fixed
effec
ts,
as
wel
las
an
ind
icato
req
ual
to1
ifth
eP
AN
curr
entl
yco
ntr
ols
the
may
ors
hip
inth
em
un
icip
ali
ty.
Sta
nd
ard
erro
rscl
ust
ered
by
mu
nic
ipal
ity
and
mon
thx
state
are
rep
ort
edin
pare
nth
eses
.*
sign
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-52:
Vio
len
ceS
pil
lovers
(Contr
oll
ing
for
PA
Nm
ayors
)
Dep
var:
dru
gtr
ade-
rela
ted
hom
icid
eD
epva
r:dru
gtr
ade-
rela
ted
hom
icid
edum
my
rate
rate
dum
my
rate
dum
my
rate
rate
dum
my
rate
Full
sam
ple
Lim
ited
sam
ple
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Pan
elA:Sho
rtestPaths
Pre
dic
ted
0.01
4***
1.17
50.
006
-0.5
14ro
ute
sdum
my
(0.0
05)
(1.2
00)
(0.0
06)
(1.1
64)
Pre
dic
ted
0.55
4*0.
460
route
sco
unt
(0.3
07)
(0.2
87)
One
route
0.01
7**
-6.0
640.
014
-5.2
78(0
.007
)(3
.758
)(0
.010
)(3
.421
)M
ore
than
0.01
110
.190
**-0
.000
6.17
9on
ero
ute
(0.0
08)
(4.7
02)
(0.0
10)
(4.4
93)
Pan
elB:Model
withCon
gestionCosts
Pre
dic
ted
0.01
7***
1.81
3**
0.01
9***
1.83
4**
route
sdum
my
(0.0
05)
(0.8
02)
(0.0
06)
(0.9
34)
Pre
dic
ted
-0.0
070.
001
route
sco
unt
(0.0
15)
(0.0
13)
One
route
0.01
02.
256
0.01
1.48
(0.0
06)
(1.6
38)
(0.0
07)
(0.9
56)
Mor
eth
an0.
020*
**1.
639
0.02
3***
1.98
8*on
ero
ute
(0.0
06)
(1.0
49)
(0.0
07)
(1.0
35)
Munic
ipal
itie
s18
6918
6918
6918
6918
6915
6215
6215
6215
6215
62O
bse
rvat
ions
69,1
5369
,153
69,1
5369
,153
69,1
5357
,794
57,7
9457
,794
57,7
9457
,794
Notes:
Th
ed
epen
den
tva
riab
lein
colu
mn
s(1
),(4
),(6
)an
d(9
)is
an
ind
icato
req
ual
to1
ifa
dru
gtr
ad
e-re
late
dh
om
icid
eocc
urr
edin
agiv
enm
un
icip
ali
ty-m
onth
,an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(2
),(3
),(5
),(7
),(8
),an
d(1
0)
isth
ed
rug
trad
e-re
late
dh
om
icid
era
tep
er100,0
00
mu
nic
ipal
inh
ab
itants
.C
olu
mn
s(6
)th
rou
gh(1
0)li
mit
the
sam
ple
tom
un
icip
alit
ies
that
do
not
bord
era
mu
nic
ipali
tyth
at
has
exp
erie
nce
da
close
PA
Nvic
tory
.A
llco
lum
ns
incl
ud
em
onth
xst
ate
and
mu
nic
ipal
ity
fixed
effec
ts,
asw
ell
asan
ind
icat
oreq
ual
to1
ifth
eP
AN
curr
entl
yco
ntr
ols
the
may
ors
hip
inth
em
un
icip
ali
ty.
Sta
nd
ard
erro
rscl
ust
ered
by
mu
nic
ipal
ity
and
mon
thx
stat
ear
ere
por
ted
inp
aren
thes
es.
*si
gn
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Table A-53: A Reduced Form Spillovers Model: Confiscations
(1) (2) (3)
Domestic ConfiscationsDummy Value Value
RF predicted 0.002 0.056routes dummy (0.006) (0.067)
RF predicted 0.029routes count (0.057)
R-squared 0.39 0.44 0.44Municipalities 1869 1869 1869Observations 69,153 69,153 69,153
Notes: The dependent variable in column (1) is an indicator equal to 1 if domestic illicit drug confiscations aremade in a given municipality-month, and the dependent variable in columns (2) and (3) is the log value of domesticillicit drug confiscations (or 0 if no confiscations are made). The RF predicted routes dummy is an indicator equalto 1 if the municipality borders a municipality that has inaugurated a closely elected PAN mayor during the sampleperiod. The RF predicted routes count is a count variable equal to the number of bordering municipalities thathave inaugurated a closely elected PAN mayor during the sample period. All columns include month x state andmunicipality fixed effects. Standard errors clustered by municipality and month x state are reported in parentheses.* significant at 10%, ** significant at 5%, *** significant at 1%.
Table A-54: A Reduced Form Spillovers Model: Violence
(1) (2) (3) (4) (5)
Dep. var.: Drug trade-related homicidedummy rate rate dummy rate
RF predicted -0.005 3.136routes dummy (0.007) (2.292)
RF predicted 2.204routes count (1.596)
One RF route -0.003 3.235(0.007) (2.443)
More than -0.017 2.522one RF route (0.014) (1.976)
R-squared 0.34 0.42 0.42 0.34 0.42Municipalities 1869 1869 1869 1869 1869Observations 69,153 69,153 69,153 69,153 69,153
Notes: The dependent variable in columns (1) and (4) is an indicator equal to 1 if a drug trade-related homicidesoccurred in a given municipality-month, and the dependent variable in columns (2), (3), and (5) is the drugtrade-related homicide rate per 100,000 municipal inhabitants. The RF predicted routes dummy is an indicatorequal to 1 if the municipality borders a municipality that has inaugurated a closely elected PAN mayor during thesample period. The RF predicted routes count is a count variable equal to the number of bordering municipalitiesthat have inaugurated a closely elected PAN mayor during the sample period, and analogously for the one RFroute and more than one RF route indicators. All columns include month x state and municipality fixed effects.Standard errors clustered by municipality and month x state are reported in parentheses. * significant at 10%, **significant at 5%, *** significant at 1%.
Table A-55: Trafficking Model Parameter Estimates
(1) (2) (3)
Crossing Costs Fullparsimonious flexible congestion
model model costs
φt 62.34***[2.72](1.41)
φp 36.48***[2.07](1.40)
φQ1t 3.24*** 13.00***
[0.30] [1.27](0.25) (1.19)
φQ2t 13.19*** 9.29***
[2.14] [0.34](1.89) (0.33)
φQ3t 13.86*** 21.26***
[4.37] [0.54](4.08) (0.52)
φQ4t 18.81*** 20.22***
[0.86] [0.62](0.83) (0.57)
φsmallp 64.47*** 70.990***
[9.76] [1.29](9.16) (1.28)
φlargep 55.34*** 43.50**[8.43] [21.73](7.46) (17.03)
φint 0.015***[0.004](0.003)
δ 1.88*** 1.57*** 1.86***[0.05] [0.15] [0.17](0.04) (0.12) (0.16)
γ 0.11**[0.06](0.05)
κ 0.763*** 0.91*** 0.79***[0.07] [0.08] [0.07](0.06) (0.07) (0.06)
Notes: Column 1 reports the simulated method of moments parameter estimates for the model with parsimoniouscongestion costs on U.S. points of entry, Column 2 reports the parameter estimates for the model with flexiblecongestion costs on U.S. points of entry, and Column 3 reports the parameter estimates for the model withcongestion costs on both U.S. points of entry and interior edges. Conley (1999) standard errors are in brackets, androbust standard errors are in parentheses.
Tab
leA
-56:
Th
eD
ivers
ion
of
Dru
gT
raffi
c(A
ltern
ati
ve
Con
gest
ion
Mod
els
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Full
Sam
ple
Lim
ited
Sam
ple
Full
Sam
ple
Dom
esti
cIl
lici
tD
rug
Con
fisc
atio
ns
Coca
ine
Con
fisc
atio
ns
Dum
my
Val
ue
Val
ue
Dum
my
Val
ue
Val
ue
Dum
my
Val
ue
Val
ue
Pan
elA:Con
gestionModel
(8Param
eters)
Pre
dic
ted
0.01
0***
0.10
6***
0.00
60.
063
0.00
30.
009
route
sdum
my
(0.0
04)
(0.0
41)
(0.0
04)
(0.0
48)
(0.0
03)
(0.0
27)
Pre
dic
ted
0.00
5-0
.002
-0.0
04ro
ute
sco
unt
(0.0
05)
(0.0
05)
(0.0
04)
Pan
elB:Con
gestionModel
(10Param
eters)
Pre
dic
ted
0.01
1***
0.12
8***
0.00
9**
0.10
5**
0.00
20.
014
route
sdum
my
(0.0
04)
(0.0
41)
(0.0
04)
(0.0
43)
(0.0
03)
(0.0
25)
Pre
dic
ted
0.00
1-0
.005
-0.0
05ro
ute
sco
unt
(0.0
04)
(0.0
04)
(0.0
04)
Munic
ipal
itie
s18
6918
6918
6915
6215
6215
6218
6918
6918
69O
bse
rvat
ions
69,1
5369
,153
69,1
5357
,794
57,7
9457
,794
69,1
5369
,153
69,1
53
Notes:
Th
edep
end
ent
vari
able
inco
lum
ns
(1),
and
(4)
isan
indic
ato
req
ual
to1
ifd
om
esti
cil
lici
td
rug
con
fisc
ati
on
sare
mad
ein
agiv
enm
un
icip
ali
ty-m
onth
;th
ed
epen
den
tva
riab
lein
colu
mn
s(2
),(3
),(5
),an
d(6
)is
the
log
valu
eof
dom
esti
cil
lici
td
rug
con
fisc
ati
on
s(o
r0
ifn
oco
nfi
scati
on
sare
mad
e);
the
dep
end
ent
vari
ab
lein
colu
mn
(7)
isan
ind
icat
oreq
ual
to1
ifco
cain
eco
nfi
scati
on
sare
mad
ein
agiv
enm
un
icip
ali
ty-m
onth
;an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(8
)an
d(9
)is
the
log
valu
eof
con
fisc
ated
coca
ine
(or
0if
no
con
fisc
atio
ns
are
mad
e).
Colu
mn
s(4
)th
rou
gh
(6)
lim
itth
esa
mp
leto
mu
nic
ipali
ties
that
do
not
bord
era
mu
nic
ipali
tyth
ath
asex
per
ien
ced
acl
ose
PA
Nvic
tory
.P
anel
Ap
red
icts
traffi
ckin
gro
ute
su
sin
gth
esh
ort
est
path
sm
od
el,
and
Pan
elB
use
sth
em
od
elw
ith
con
ges
tion
cost
s.A
llco
lum
ns
incl
ud
em
onth
xst
ate
and
mu
nic
ipal
ity
fixed
effec
ts.
Sta
nd
ard
erro
rscl
ust
ered
by
mun
icip
ali
tyan
dm
onth
xst
ate
are
rep
ort
edin
pare
nth
eses
.*
sign
ifica
nt
at10
%,
**si
gnifi
cant
at5%
,**
*si
gnifi
cant
at
1%
.
Tab
leA
-57:
Vio
len
ceS
pil
lovers
(Alt
ern
ati
ve
Con
gest
ion
Mod
els
)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Full
Sam
ple
Lim
ited
Sam
ple
Dep.var.:Drugtrad
e-relatedho
micide
Dep.Var.:
Drugtrad
e-relatedho
micide
dummy
rate
rate
dummy
rate
dummy
rate
rate
dummy
rate
Pan
elA:Con
gestionModel
(8Param
eters)
Pre
dic
ted
0.01
4***
0.56
80.
013*
**0.
035
route
sdum
my
(0.0
04)
(0.4
21)
(0.0
05)
(0.3
86)
Pre
dic
ted
0.00
60.
020
route
sco
unt
(0.0
22)
(0.0
19)
One
route
0.01
1*1.
094
0.01
00.
105
(0.0
06)
(1.3
05)
(0.0
08)
(0.9
70)
Mor
eth
an0.
015*
**0.
309
0.01
5***
0.00
1on
ero
ute
(0.0
05)
(0.7
27)
(0.0
05)
(0.5
40)
Pan
elB:Con
gestionModel
(10Param
eters
Pre
dic
ted
0.00
9**
0.76
50.
008*
0.32
0ro
ute
sdum
my
(0.0
04)
(0.8
40)
(0.0
04)
(0.9
16)
Pre
dic
ted
0.01
40.
024
route
sco
unt
(0.0
24)
(0.0
23)
One
route
0.00
71.
643
0.00
80.
806
(0.0
05)
(1.3
95)
(0.0
06)
(0.8
34)
Mor
eth
an0.
010*
*0.
360
0.00
80.
080
one
route
(0.0
05)
(1.1
22)
(0.0
05)
(1.1
00)
Munic
ipal
itie
s18
6918
6918
6918
6918
6915
6215
6215
6215
6215
62O
bse
rvat
ions
69,1
5369
,153
69,1
5369
,153
69,1
5357
,794
57,7
9457
,794
57,7
9457
,794
Notes:
Th
ed
epen
den
tva
riab
lein
colu
mn
s(1
),(4
),(6
)an
d(9
)is
an
ind
icato
req
ual
to1
ifa
dru
gtr
ad
e-re
late
dh
om
icid
eocc
urr
edin
agiv
enm
un
icip
ali
ty-m
onth
,an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(2
),(3
),(5
),(7
),(8
),an
d(1
0)
isth
ed
rug
trad
e-re
late
dh
om
icid
era
tep
er100,0
00
mu
nic
ipal
inh
ab
itants
.C
olu
mn
s(6
)th
rou
gh(1
0)li
mit
the
sam
ple
tom
un
icip
alit
ies
that
do
not
bord
era
mu
nic
ipali
tyth
at
exp
erie
nce
da
close
PA
Nvic
tory
bet
wee
n2007
an
d2008.
All
colu
mn
sin
clu
de
mon
thx
stat
ean
dm
un
icip
alit
yfi
xed
effec
ts.
Sta
nd
ard
erro
rscl
ust
ered
by
mu
nic
ipali
tyan
dm
onth
xst
ate
are
rep
ort
edin
pare
nth
eses
.*
signifi
cant
at
10%
,**
sign
ifica
nt
at5%
,**
*si
gnifi
cant
at1%
.
Tab
leA
-58:
Acc
ou
nti
ng
for
DT
OT
err
itory
wh
en
Pre
dic
tin
gR
ou
tes
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Con
fisc
atio
ns
Hom
icid
esdum
my
valu
eva
lue
dum
my
rate
rate
dum
my
rate
Pan
elA:Sho
rtestPathModel
Pre
dic
ted
0.00
8**
0.03
90.
009*
0.35
0ro
ute
sdum
my
(0.0
04)
(0.0
44)
(0.0
05)
(0.6
09)
Pre
dic
ted
0.01
2*0.
337*
route
sco
unt
(0.0
06)
(0.2
01)
One
route
0.01
4**
-2.2
51(0
.007
)(1
.891
)M
ore
than
0.00
33.
618
one
route
(0.0
06)
(2.4
95)
Pan
elB:Model
withCon
gestionCosts
Pre
dic
ted
0.00
7**
0.10
4***
0.00
7**
1.27
7ro
ute
sdum
my
(0.0
03)
(0.0
38)
(0.0
03)
(0.7
82)
Pre
dic
ted
0.00
40.
068*
route
sco
unt
(0.0
03)
(0.0
41)
One
route
0.00
8*1.
154*
(0.0
04)
(0.6
20)
Mor
eth
an0.
006
1.37
8on
ero
ute
(0.0
04)
(0.9
51)
Munic
ipal
itie
s18
6918
6918
6918
6918
6918
6918
6918
69O
bse
rvat
ions
69,2
6469
,264
69,2
6469
,264
69,2
6469
,264
69,2
6469
,264
Notes:
All
colu
mn
sin
clud
em
onth
xst
ate
and
mu
nic
ipali
tyfi
xed
effec
tsan
dom
itm
un
icip
ali
ties
that
exp
erie
nce
da
close
dP
AN
vic
tory
.S
tan
dard
erro
rscl
ust
ered
by
mu
nic
ipal
ity
and
mon
thx
stat
ear
ere
por
ted
inp
aren
thes
es.
*si
gn
ifica
nt
at
10%
,**
sign
ifica
nt
at
5%
,***
sign
ifica
nt
at
1%
.
Tab
leA
-59:
Vio
len
ceS
pil
lovers
ina
Mod
el
that
Est
imate
sP
oli
tica
lC
ost
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Full
Sam
ple
Lim
ited
Sam
ple
Dep.var.:Drugtrad
e-relatedho
micide
Dep.Var.:
Drugtrad
e-relatedho
micide
dummy
rate
rate
dummy
rate
dummy
rate
rate
dummy
rate
Pan
elA:20
07-200
8Elections
Pre
dic
ted
0.01
0***
0.81
4*0.
008*
0.98
3**
route
sdum
my
(0.0
04)
(0.4
58)
(0.0
04)
(0.4
94)
Pre
dic
ted
0.20
9*0.
171
route
sco
unt
(0.1
16)
(0.1
04)
One
route
0.01
3**
-1.9
470.
011
-0.5
95(0
.006
)(1
.834
)(0
.008
)(1
.608
)M
ore
than
0.00
9*2.
153*
*0.
006
1.84
3**
one
route
(0.0
05)
(1.0
71)
(0.0
05)
(0.9
21)
Obse
rvat
ions
69,1
5369
,153
69,1
5369
,153
69,1
5357
,794
57,7
9457
,794
57,7
9457
,794
Pan
elB:20
07-201
0Elections
Pre
dic
ted
0.01
1***
1.58
6**
0.01
0*0.
912
route
sdum
my
(0.0
04)
(0.6
43)
(0.0
06)
(0.7
13)
Pre
dic
ted
0.21
4**
0.12
9ro
ute
sco
unt
(0.1
04)
(0.1
00)
One
route
0.01
3**
-0.3
180.
009
1.54
1**
(0.0
06)
(1.7
27)
(0.0
09)
(0.6
84)
Mor
eth
an0.
009*
*2.
490*
*0.
011*
0.56
3on
ero
ute
(0.0
04)
(1.0
28)
(0.0
07)
(0.8
74)
Obse
rvat
ions
88,9
8488
,984
88,9
8488
,984
88,9
8445
,913
45,9
1345
,913
45,9
1345
,913
Notes:
Th
ed
epen
den
tva
riab
lein
colu
mn
s(1
),(4
),(6
)an
d(9
)is
an
ind
icato
req
ual
to1
ifa
dru
gtr
ad
e-re
late
dh
om
icid
eocc
urr
edin
agiv
enm
un
icip
ali
ty-m
onth
,an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(2
),(3
),(5
),(7
),(8
),an
d(1
0)
isth
ed
rug
trad
e-re
late
dh
om
icid
era
tep
er100,0
00
mu
nic
ipal
inh
ab
itants
.C
olu
mn
s(6
)th
rou
gh(1
0)li
mit
the
sam
ple
tom
un
icip
alit
ies
that
do
not
bord
era
mu
nic
ipali
tyth
at
has
exp
erie
nce
da
close
PA
Nvic
tory
.A
llco
lum
ns
incl
ud
em
onth
xst
ate
and
mu
nic
ipal
ity
fixed
effec
ts.
Sta
nd
ard
erro
rscl
ust
ered
by
mu
nic
ipali
tyand
month
xst
ate
are
rep
ort
edin
pare
nth
eses
.*
signifi
cant
at
10%
,**
sign
ifica
nt
at
5%
,**
*si
gnifi
cant
at1%
.
Tab
leA
-60:
Eco
nom
icS
pil
lovers
(1)
(2)
(3)
(4)
(5)
(6)
Full
sam
ple
Lim
ited
sam
ple
Mal
eF
emal
eF
orm
alIn
form
alF
emal
eIn
form
alpar
tici
pat
ion
sect
orlo
gw
ages
par
tici
pat
ion
wag
es
Pan
elA:Sho
rtestPaths
Pre
dic
ted
-0.1
24-0
.756
0.02
0-0
.023
-0.7
84-0
.030
route
sdum
my
(0.5
13)
(1.0
38)
(0.0
22)
(0.0
20)
(1.6
22)
(0.0
27)
Pan
elB:Model
withCon
gestionCosts
Pre
dic
ted
-0.2
42-1
.261
**0.
013
-0.0
22*
-1.5
58**
-0.0
28*
route
sdum
my
(0.3
02)
(0.5
70)
(0.0
12)
(0.0
13)
(0.6
73)
(0.0
17)
Sta
tex
quar
ter
FE
yes
yes
yes
yes
yes
yes
Munic
ipal
ity
FE
yes
yes
yes
yes
yes
yes
R2
0.52
0.79
0.18
0.09
0.79
0.09
Munic
ipal
itie
s88
088
087
987
170
970
3O
bse
rvat
ions
9,82
19,
821
407,
204
148,
302
7,88
711
4,63
3
Notes:
Th
ed
epen
den
tva
riab
lein
colu
mn
(1)
isav
erage
mu
nic
ipal
male
lab
or
forc
ep
art
icip
ati
on
,th
ed
epen
den
tva
riab
lein
colu
mn
s(2
)an
d(5
)is
aver
age
mu
nic
ipal
fem
ale
lab
orfo
rce
par
tici
pat
ion
,th
ed
epen
den
tva
riab
lein
colu
mn
(3)
islo
gw
ages
of
form
al
sect
or
work
ers,
an
dth
ed
epen
den
tva
riab
lein
colu
mn
s(4
)an
d(6
)is
log
wag
esof
info
rmal
sect
orw
orke
rs.
All
colu
mn
sin
clu
de
qu
art
erx
state
an
dm
un
icip
ali
tyfi
xed
effec
ts.
Colu
mn
(1)
wei
ghts
by
the
squ
are
root
of
the
mu
nic
ipal
ity’s
mal
ep
opu
lati
onan
dco
lum
ns
(2)
and
(5)
wei
ght
by
the
squ
are
root
of
the
mu
nic
ipali
ty’s
fem
ale
pop
ula
tion
.T
he
sam
ple
inco
lum
ns
(5)
an
d(6
)ex
clu
des
mu
nic
ipaliti
esth
atb
ord
era
mu
nic
ipal
ity
that
has
exp
erie
nce
da
close
PA
Nvic
tory
.S
tan
dard
erro
rscl
ust
ered
by
mu
nic
ipali
tyan
dqu
art
erx
state
are
rep
orte
din
par
enth
eses
.*
sign
ifica
nt
at10
%,
**si
gnifi
cant
at
5%
,***
sign
ifica
nt
at
1%
.
A-2.9 Law Enforcement Allocation Table
A–78
Table A-61: Robustness of Policy Algorithm
(1)
Percentageincreasein totalcosts
Baseline (N = 250) 0.168N = 100 0.168N = 500 0.168Alternate between selecting edges with m = 1 and m = 2 0.105Alternate between selecting edges with m = 1, m = 2, and m = 3 0.106Select edge with m = 2 when k = 1 0.168Select edge with m = 3 when k = 1 0.168Select edge with m = 4 when k = 1 0.168Select edge with m = 5 when k = 1 0.168
Notes: The left column describes the variation in the policy algorithm (as described in theestimation appendix) and the right column gives the percentage increase in total trafficking costswhen the respective variant of the algorithm is used to select edges.
A-2.10 Map of Close PAN Elections
A–80
Fig
ure
A-1
:C
lose
Ele
ctio
ns
Notes:
Bla
ckci
rcle
sd
enot
eP
AN
vic
tori
esan
dgr
aysq
uare
sd
enote
PA
Nlo
sses
.T
he
sam
ple
isli
mit
edto
mu
nic
ipali
ties
wit
ha
vote
spre
ad
of
five
per
centa
ge
poi
nts
orle
ss.
A–81
A-2.11 Balance Figures for Pre-Characteristics
A–82
Figure A-2: Covariate Plots0
5010
015
020
0
−.05 0 .05PAN margin of victory
(a) Mun. taxes per capita (2005)
0.2
.4.6
.8
−.05 0 .05PAN margin of victory
(b) PAN incumbent
−.2
0.2
.4.6
−.05 0 .05PAN margin of victory
(c) PRD incumbent
.2.2
5.3
.35
.4.4
5
−.05 0 .05PAN margin of victory
(d) % alternations (1976-2006)
−.4
−.2
0.2
.4
−.05 0 .05PAN margin of victory
(e) PRI never lost (1976-2006)
−10
010
2030
−.05 0 .05PAN margin of victory
(f) Population (2005)
Figure A-3: Covariate Plots0
200
400
600
800
−.05 0 .05PAN margin of victory
(a) Population density (2005)
0.0
1.0
2.0
3.0
4
−.05 0 .05PAN margin of victory
(b) Migrants per capita (2005)
23
45
67
−.05 0 .05PAN margin of victory
(c) Income per capita (2005)
1020
3040
5060
−.05 0 .05PAN margin of victory
(d) Malnutrition (2005)
56
78
−.05 0 .05PAN margin of victory
(e) Mean years schooling (2005)
1520
2530
−.05 0 .05PAN margin of victory
(f) Infant mortality (2005)
Figure A-4: Covariate Plots
05
1015
20
−.05 0 .05PAN margin of victory
(a) Housecolds w/o access to sewage (2005)
−10
010
2030
40
−.05 0 .05PAN margin of victory
(b) Housecolds w/o access to water (2005)
−1
−.5
0.5
1
−.05 0 .05PAN margin of victory
(c) Marginality index (2005)
0.1
.2.3
−.05 0 .05PAN margin of victory
(d) Road density (km/km2)
400
600
800
1000
−.05 0 .05PAN margin of victory
(e) Distance U.S. (km)
500
1000
1500
2000
−.05 0 .05PAN margin of victory
(f) Elevation (m)
Figure A-5: Covariate Plots0
24
68
−.05 0 .05PAN margin of victory
(a) Slope (degrees)
−20
000
2000
4000
6000
8000
−.05 0 .05PAN margin of victory
(b) Surface area (km2)
24
68
1012
−.05 0 .05PAN margin of victory
(c) Average min. temperature, C (1950-2000)
1820
2224
2628
−.05 0 .05PAN margin of victory
(d) Average max. temperature, C (1950-2000)
500
1000
1500
2000
−.05 0 .05PAN margin of victory
(e) Average precipitation, cm (1950-2000)
A-2.12 Balance Figures for the Predicted Homicide Rate
A–87
Figure A-6: PAN victories and predicted homicides
0.2
.4.6
.81
Pre
dict
ed d
rug
hom
icid
e pr
obab
ility
−.05 0 .05PAN margin of victory
(a) Predicted drug-related homicide probability
010
2030
4050
6070
80P
redi
cted
dru
g ho
mic
ide
rate
−.05 0 .05PAN margin of victory
(b) Predicted drug-related homicide rate
0.2
.4.6
.81
Pre
dict
ed h
omic
ide
prob
abili
ty
−.05 0 .05PAN margin of victory
(c) Predicted overall homicide probability
010
2030
4050
6070
8090
100
Pre
dict
ed h
omic
ide
rate
−.05 0 .05PAN margin of victory
(d) Predicted overall homicide rate
Notes: This figure plots predicted homicide measures against the PAN margin of victory. The homicidemeasures are predicted using the characteristics in Table 1 and pre-period violence data. Each pointrepresents the average value of predicted homicides in vote spread bins of width one half of a percentagepoint. The solid line plots predicted values from an RD regression with separate vote spread polynomialsestimated on either side of the PAN win-loss threshold. The dashed lines show 95% confidence intervals.
A-2.13 McCrary Plots
A–89
Figure A-7: Vote Spread Density (2007-2008 Elections)
05
1015
20F
requ
ency
−.7 −.6 −.5 −.4 −.3 −.2 −.1 0 .1 .2 .3 .4 .5 .6 .7PAN margin of victory
PAN margin of victory in municipal elections (2007−2008)
Notes: This figure shows the frequency of mayoral elections (2007-2008) in one percentage point vote
spread bins. The solid line plots predicted values from a local linear regression of frequency on vote spread,
with separate vote spread trends estimated on either side of the PAN win-loss threshold. The dashed lines
show 95% confidence intervals. The bandwidth is chosen using the Imbens-Kalyanaraman bandwidth
selection rule (2009), and a rectangular kernel is used.
A–90
Figure A-8: Vote Spread Density (2007-2010 Elections)
010
2030
40F
requ
ency
−.7 −.6 −.5 −.4 −.3 −.2 −.1 0 .1 .2 .3 .4 .5 .6 .7PAN margin of victory
PAN margin of victory in municipal elections (2007−2010)
Notes: This figure shows the frequency of mayoral elections (2007-2010) in one percentage point vote
spread bins. The solid line plots predicted values from a local linear regression of frequency on vote spread,
with separate vote spread trends estimated on either side of the PAN win-loss threshold. The dashed lines
show 95% confidence intervals. The bandwidth is chosen using the Imbens-Kalyanaraman bandwidth
selection rule (2009), and a rectangular kernel is used.
A–91
A-2.14 Homicide RD Figures - Robustness
A–92
Figure A-9: Drug trade-related homicide RD figures (2007-2010 elections)0
.2.4
.6.8
1M
onth
ly p
roba
bilit
y of
dru
g−re
late
d ho
mic
ide
−.05 0 .05PAN margin of victory
(a) Post-inauguration (extensive margin)
−20
−10
010
2030
40D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(b) Post-inauguration (homicide rate)
0.2
.4.6
.81
Mon
thly
pro
babi
lity
of d
rug−
rela
ted
hom
icid
e
−.05 0 .05PAN margin of victory
(c) Lame duck (extensive margin)
−20
−10
010
2030
40D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(d) Lame duck (homicide rate)
0.2
.4.6
.81
Mon
thly
pro
babi
lity
of d
rug−
rela
ted
hom
icid
e
−.05 0 .05PAN margin of victory
(e) Pre-election (extensive margin)
−20
−10
010
2030
40D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(f) Pre-election (homicide rate)
Notes: This figure plots violence measures against the PAN margin of victory, with a negative margin indicating a PANloss. Each point represents the average value of the outcome in vote spread bins of width one half of a percentage point.The solid line plots predicted values, with separate quadratic vote spread trends estimated on either side of the PANwin-loss threshold. The dashed lines show 95% confidence intervals.
Figure A-10: All homicides RD figures (2007-2010 elections)0
.2.4
.6.8
1M
onth
ly p
roba
bilit
y of
hom
icid
e oc
curr
ing
−.05 0 .05PAN margin of victory
(a) Post-inauguration (extensive margin)
−20
020
4060
8010
012
0O
vera
ll ho
mic
ide
rate
−.05 0 .05PAN margin of victory
(b) Post-inauguration (homicide rate)
0.2
.4.6
.81
Mon
thly
pro
babi
lity
of h
omic
ide
occu
rrin
g
−.05 0 .05PAN margin of victory
(c) Lame duck (extensive margin)
−20
020
4060
8010
012
0O
vera
ll ho
mic
ide
rate
−.05 0 .05PAN margin of victory
(d) Lame duck (homicide rate)
0.2
.4.6
.81
Mon
thly
pro
babi
lity
of h
omic
ide
occu
rrin
g
−.05 0 .05PAN margin of victory
(e) Pre-election (extensive margin)
−20
020
4060
8010
012
0O
vera
ll ho
mic
ide
rate
−.05 0 .05PAN margin of victory
(f) Pre-election (homicide rate)
Notes: This figure plots violence measures against the PAN margin of victory, with a negative margin indicating a PANloss. Each point represents the average value of the outcome in vote spread bins of width one half of a percentage point.The solid line plots predicted values, with separate quadratic vote spread trends estimated on either side of the PANwin-loss threshold. The dashed lines show 95% confidence intervals.
Figure A-11: Drug trade-related homicide negative binomial RD figures−
200
2040
6080
Dru
g ho
mic
ide
rate
−.05 0 .05PAN margin of victory
(a) Post-inauguration (2007-2008 elections)
020
40D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(b) Post-inauguration (2007-2010 elections)
−20
020
4060
80D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(c) Lame duck (2007-2008 elections)
020
40D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(d) Lame duck (2007-2010 elections)
−20
020
4060
80D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(e) Pre-election (2007-2008 elections)
020
40D
rug
hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(f) Pre-election (2007-2010 elections)
Notes: This figure plots violence measures against the PAN margin of victory, with a negative margin indicating a PANloss. Each point represents the average value of the outcome in vote spread bins of width one half of a percentage point.The solid line plots predicted values from a negative binomial regression, with separate vote spread trends estimated oneither side of the PAN win-loss threshold. The dashed lines show 95% confidence intervals.
Figure A-12: All homicides negative binomial RD figures−
200
2040
6080
100
120
Hom
icid
e ra
te
−.05 0 .05PAN margin of victory
(a) Post-inauguration (2007-2008 elections)
020
4060
8010
0H
omic
ide
rate
−.05 0 .05PAN margin of victory
(b) Post-inauguration (2007-2010 elections)
−20
020
4060
8010
012
0H
omic
ide
rate
−.05 0 .05PAN margin of victory
(c) Lame duck (2007-2008 elections)
020
4060
8010
0H
omic
ide
rate
−.05 0 .05PAN margin of victory
(d) Lame duck (2007-2010 elections)
−20
020
4060
8010
012
0H
omic
ide
rate
−.05 0 .05PAN margin of victory
(e) Pre-election (2007-2008 elections)
020
4060
8010
0H
omic
ide
rate
−.05 0 .05PAN margin of victory
(f) Pre-election (2007-2010 elections)
Notes: This figure plots violence measures against the PAN margin of victory, with a negative margin indicating a PANloss. Each point represents the average value of the outcome in vote spread bins of width one half of a percentage point.The solid line plots predicted values from a negative binomial regression, with separate vote spread trends estimated oneither side of the PAN win-loss threshold. The dashed lines show 95% confidence intervals.
Fig
ure
A-1
3:
Month
lyh
om
icid
eR
Dfi
gu
res
●
●●
●
●●
●●
●●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
−5005010
0
−6
−3
LD3
69
1215
1821
2427
LDM
onth
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-r
elate
dh
om
icid
era
te
●●● ● ●
●● ● ●●● ●●●
●● ● ●● ●
●● ●●● ● ● ●● ●●●● ● ●● ●● ●
●●● ●
●●●● ●
● ● ● ●●●
●●●● ● ●●● ●●●
●●●● ●
●● ●● ● ●●● ●●● ●● ● ●● ●● ●●
● ●●●●
●●●● ●
● ●●● ●● ●● ●● ● ●
●●●●● ●● ●
●●● ●
●● ●● ●● ●● ●●● ● ●
●● ●● ●● ●● ● ●
●●
● ●●●
●● ●
●●●● ●● ● ●● ●
●● ●● ●
● ●● ●●●● ●
●● ● ●●●
●● ● ●●●● ● ●
●● ●●
● ●●● ●●
● ●●● ●
●●●● ●
●●●
●●●●
●●
●● ●● ● ● ●● ●●●●●● ● ● ● ●●●● ●
● ●●● ●●
● ●● ●● ● ●●●
0
100
200 −
205
−19
0−
175
−16
0−
145
−13
0−
115
−10
0−
85−
70−
55−
40−
25−
10LD
520
LD5
20Q
uart
er
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(b)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
acl
ose
PA
Nvic
tory
on
the
dru
g-r
elate
dh
om
icid
era
tein
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
acl
ose
PA
Nvic
tory
on
the
over
all
hom
icid
era
tein
am
un
icip
alit
y-m
onth
.T
he
lin
esp
lot
95%
con
fid
ence
inte
rvals
.
Fig
ure
A-1
4:
Tota
lh
om
icid
es
qu
art
erl
yR
Dest
imate
s(e
xte
nsi
ve
marg
in)
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●●
−0.
3
0.0
0.3
0.6
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Probability
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
Notes:
Eac
hp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
acl
ose
PA
Nvic
tory
on
wh
eth
era
hom
icid
eocc
ure
din
am
un
icip
ali
ty-q
uart
er.
Th
eli
nes
plo
t95
%co
nfi
den
cein
terv
als.
Fig
ure
A-1
5:
PA
NV
icto
ries
an
dH
om
icid
es
(4%
ban
dw
idth
)
●●
●●
●●
●
●
●
●
●●
●
−0.
6
−0.
3
0.0
0.3
0.6
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●
●
●●
●●
●
●
●
●
●
●
●
04080120
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
0
100
200
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-1
6:
PA
NV
icto
ries
an
dH
om
icid
es
(3%
ban
dw
idth
)
●●
●●
●
●●
●
●
●
●●
●
−0.
4
0.0
0.4
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●
●
●●
●●
●
●
●
●
●
●
●
050100
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
0
100
200
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-1
7:
PA
NV
icto
ries
an
dH
om
icid
es
(2%
ban
dw
idth
)
●●
●
●●
●●
●
●
●
●
●
●
−0.
5
0.0
0.5
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●●
●
●
●
●
●
●
●
●
●
●
●
050100
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●
●●
●●
●●
●
●●
●●
●●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
−10
00
100
200
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-1
8:
PA
NV
icto
ries
an
dH
om
icid
es
(13.3
%b
an
dw
idth
)
●●
●●
●●
●
●
●
●
●●
●
−0.
3
−0.
2
−0.
1
0.0
0.1
0.2
0.3
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●
●
●●
●
●
●
●
●
●
●
●
●
04080
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●
●
−5005010
0
150
200
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-1
9:
PA
NV
icto
ries
an
dH
om
icid
es
(5%
ban
dw
idth
,fi
xed
eff
ect
s)
●●
●
●
●●
●●
●
●
●
●●
−0.
25
0.00
0.25
0.50
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●
●
●●
●
●
●
●
●
●
●
●
●
0255075100
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
−5005010
0
150
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-2
0:
PA
NV
icto
ries
an
dH
om
icid
es
(4%
ban
dw
idth
,fi
xed
eff
ect
s)
●●
●
●
●
●●
●
●
●
●
●●
−0.
6
−0.
3
0.0
0.3
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●
●
●
●
●
●●
●
●
●
●
●
●
04080
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●●
●
●●
●●
●●
●
●
●
●●
●●
●●
●
●
●
●●
●
●
●
●
0
100
200
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-2
1:
PA
NV
icto
ries
an
dH
om
icid
es
(3%
ban
dw
idth
,fi
xed
eff
ect
s)
●
●
●●
●
●●
●
●
●
●
●
●
−0.
6
−0.
3
0.0
0.3
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●●
●
●
●
●
●
●
●
●
●
●
●
050100
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●
●
●
●●
●●
●
●●
●
●
●●
●
●
●
●
0
100
200
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-2
2:
PA
NV
icto
ries
an
dH
om
icid
es
(2%
ban
dw
idth
,fi
xed
eff
ect
s)
●●
●
●●
●
●
●●
●
●
●
●
−0.
5
0.0
0.5
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●●
●
●
●
●
●
●
●
●●
●
●
−5005010
0
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●
●●
●●
●●
●●
●●
●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●
●
●●
●●
●
●●
●
●●
●
●●
●
●
0
100
200
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
Fig
ure
A-2
3:
PA
NV
icto
ries
an
dH
om
icid
es
(13.3
%b
an
dw
idth
,fi
xed
eff
ect
s)
●
●
●●
●●
●
●
●
●
●
●
●
−0.
2
−0.
1
0.0
0.1
0.2
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●
●
●●
●
●
●
●●
●
●
●
●
0255075
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●
●●
●●
●●
●●
●●
●●
●●
●
●●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●
●
●
●●
●
●●
●●
●
●●
●
●
●
●
●
−5005010
0
150
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
aver
age
pro
bab
ilit
yth
at
ad
rug-r
elate
dh
om
icid
eocc
urr
edin
am
un
icip
alit
y-m
onth
.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
dru
g-r
elate
dh
om
icid
era
tein
agiv
enqu
arte
r.In
Pan
elC
,ea
chp
oint
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
close
PA
Nvic
tori
eson
the
over
all
hom
icid
era
tein
agiv
enqu
art
er.
All
regre
ssio
ns
incl
ud
ea
qu
adra
tic
RD
pol
yn
omia
l,es
tim
ated
sep
arat
ely
on
eith
ersi
de
of
the
PA
Nw
in-l
oss
thre
shold
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
A-2.15 Homicide RD Figures - Neighbors’ Homicide Rates
A–108
Fig
ure
A-2
4:
Neig
hb
or
Hom
icid
eR
DF
igu
res
●●
●●
●●
●●
●
●
●●
●
−0.
50
−0.
25
0.00
0.25
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Probability
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(a)
Dru
g-re
late
dh
omic
ides
(exte
nsi
ve
marg
in)
●●
●●
●
●
●
●●
●●
●
●
−40
−200204060
−2
−1
LD1
23
45
67
89
LDQ
uart
er
Homicide Rate
Per
iod
● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t−pe
riod
Lam
e D
uck
2
(b)
Dru
g-r
elate
dh
om
icid
era
te
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
●●
●
●●
●●
●●
●
●●
●
●
●●
●
●●
●●
●
●
●
−5005010
0
−69
−65
−60
−55
−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
LD1
48
LD1
47
Qua
rter
Homicide Rate
Per
iod
● ● ● ● ●
Pre
−pe
riod
Lam
e D
uck
1
Pos
t Ter
m 1
Lam
e D
uck
2
Pos
t Ter
m 2
(c)
All
hom
icid
es
Notes:
InP
anel
A,
each
poi
nt
plo
tsa
sep
arat
eR
Des
tim
ate
of
the
imp
act
of
acl
ose
PA
Nvic
tory
on
wh
eth
era
dru
g-r
elate
dh
om
icid
eocc
urr
edin
am
unic
ipali
ty’s
bor
der
ing
mu
nic
ipal
itie
s.In
Pan
elB
,ea
chp
oint
plo
tsa
sep
ara
teR
Des
tim
ate
of
the
imp
act
of
acl
ose
PA
Nvic
tory
on
the
dru
g-r
elate
dh
om
icid
era
tein
am
un
icip
alit
y’s
bor
der
ing
mu
nic
ipal
itie
s.In
Pan
elC
,ea
chp
oin
tp
lots
ase
para
teR
Des
tim
ate
of
the
imp
act
of
acl
ose
PA
Nvic
tory
on
the
over
all
hom
icid
era
tein
am
un
icip
alit
y’s
bor
der
ing
mu
nic
ipal
itie
s.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esplo
t90%
con
fid
ence
inte
rvals
.
A-2.16 Robustness to Varying the Length of the Analysis Period
Fig
ure
A-2
5:
Rob
ust
ness
top
eri
od
len
gth
:d
rug-r
ela
ted
hom
icid
es
●●
●●
●●
−200204060
12
34
56
Pre
−pe
riod
leng
th (
mon
ths)
Coefficient
(a)
Pre
-per
iod
●●
●●
●
−200204060
12
34
5La
me
duck
per
iod
leng
th (
mon
ths)
Coefficient
(b)
Lam
ed
uck
per
iod
●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
−200204060
15
1015
2025
3035
Pos
t−pe
riod
leng
th (
mon
ths)
Coefficient
(c)
Post
per
iod
Notes:
Pan
elA
rep
orts
RD
esti
mat
esof
the
imp
act
ofP
AN
vic
tori
eson
the
dru
gtr
ad
e-re
late
dh
om
icid
era
tefr
om
sep
ara
tere
gre
ssio
ns
that
vary
the
len
gth
of
the
pre
-per
iod
from
one
tosi
xm
onth
s.P
anel
Bva
ries
the
len
gth
of
the
lam
ed
uck
per
iod
,an
dP
an
elC
vari
esth
ele
ngth
of
the
post
-per
iod
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rval
s,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rvals
.
Fig
ure
A-2
6:
Rob
ust
ness
top
eri
od
len
gth
:overa
llh
om
icid
es
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
−25025507510
0
2040
6080
100
120
140
160
180
200
Pre
−pe
riod
leng
th (
mon
ths)
Coefficient
(a)
Pre
-per
iod
●●
●●
●
−25025507510
0
Lam
e du
ck p
erio
d le
ngth
(m
onth
s)
Coefficient
(b)
Lam
ed
uck
per
iod
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●●
●
−25025507510
0
510
1520
2530
3540
4550
55P
ost−
perio
d le
ngth
(m
onth
s)
Coefficient
(c)
Post
per
iod
Notes:
Pan
elA
rep
orts
RD
esti
mat
esof
the
imp
act
ofP
AN
vic
tori
eson
the
over
all
hom
icid
era
tefr
om
sep
ara
tere
gre
ssio
ns
that
vary
the
len
gth
of
the
pre
-per
iod
from
one
to20
5m
onth
s.P
anel
Bva
ries
the
len
gth
ofth
ela
me
du
ckp
erio
d,
an
dP
an
elC
vari
esth
ele
ngth
of
the
post
-per
iod
.T
he
thin
lin
esp
lot
95%
con
fid
ence
inte
rvals
,an
dth
eth
ick
lin
esp
lot
90%
con
fid
ence
inte
rval
s.
A-2.17 Spillovers Model Placeo Check
Figure A-27: Placebo Exercise
0.0
3.0
6.0
9.1
2D
ensi
ty
−3 s.d. −2 s.d. −1 s.d. mean +1 s.d. +2 s.d. +3 s.d. β∗
Coefficient on predicted routes dummy
Coefficients from Placebo Exercise
Notes: This figure plots the distribution of coefficients from the placebo exercise described in the text. β∗
is the baseline coefficient from Table 6, column (2). The mean of the distribution equals -0.005.
A-2.18 Law Enforcement Allocation Figure
Fig
ure
A-2
8:
Law
En
forc
em
ent
All
oca
tion
Notes:
Mu
nic
ipal
itie
sth
atco
nta
ina
sele
cted
edge
are
hig
hli
ghte
din
yell
ow.
Th
eav
erage
month
lyd
rug
trad
e-re
late
dh
om
icid
era
teb
etw
een
2007
an
d20
09is
plo
tted
inth
eb
ackgr
oun
d.