lg LIBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

ESSAYS ON THE ECONOMICS OF COCAINE
PRODUCTION IN COLOMBIA

presented by
LUlS CARLOS REYES-HERNANDEZ,

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Economics

U Qaior Professor’s Signature
08 / 20 / 20l0

Date

 

MSU is an Affirmative Action/Equal Opportunity Employer

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 KIProj/AafiPres/ClRC/DateDueJndd

 

ABSTRACT
ESSAYS ON THE ECONOMICS OF COCAINE PRODUCTION IN COLOMBIA
By

Luis Carlos Reyes-Hernandez

The United Nations estimate that there are 16 to 21 million cocaine users in the
world, and the majority of the production of this illegal drug takes place in Colombia.
This dissertation explores different drug control policies in the Andean country, in two
chapters.

In the first one I tackle the question of how drug control policies affect the
production of coca at the municipality level. I use a five year panel that includes
observations on aggregate coca production for every Colombian municipality. I consider
the effects of two drug control policies: eradication of coca crops, and alternative
development programs. I take an Instrumental Variable (IV) approach to estimate the
change in land allocated to coca by farmers as a consequence of eradication and
alternative development programs.

I use instruments that enter the expected cost function of the government as it
implements eradication and alternative development programs, yet affect farmers’ land
allocation decisions through no channel other than their effect on drug policy. The
instruments are derived from increases in the expected cost of coca eradication as crews
get far from the zone where Antinarcotics Police helicopters can protect them from illegal
armed groups that try to shoot them down. I find that there is a strong negative

. relationship between distance from this zone and the level of eradication that takes place.

Moreover, eradication decreases at a higher rate where the presence of illegal armed
group combatants is greater. Because of analogous safety limitations in getting alternative
development program administrators to their destination in coca growing regions, the
same correlation exists between distance from the safety zone and government spending
on alternative development programs. IV estimation of the effect of eradication shows
that a one percent increase in eradication results in a 0.3 percent increase in coca
cultivation. A one percent increase and alternative development program spending results
in coca cultivation increases of up to 0.5 percent.

In the second chapter I use a cross section of household level survey data
collected by the United Nations Office on Drug and Crime to analyze the impact of crop
eradication through fumigation in coca farms. I use once again IV estimation to deal with
policy endogeneity. The estimates suggest that, conditional on being a coca producer, a
one percent increase in eradication can result in a decrease of up to 0.2 percent in the

share of land that is devoted to coca.

To the memory of my grandfather, Alvaro Vicente Hernandez-Roj as

iv

ACKNOWLEDGEMENTS

Doing research on cocaine production is at times a peculiar task. I am grateful to
the people and the institutions who kindly shared their data with me, and in particular to
Martha Paredes, the United Nations Office on Drug and Crime, and the Colombian
Antinarcotics Police. I am also thankful to my advisor, Stacy Dickert-Conlin, my parents,
my friends from the Graduate Intervarsity Christian Fellowship at Michigan State
University, and Christopher and Priscilla Smith for their support and their encouragement
throughout my time in East Lansing; and to the customs Officers at Miami International
Airport who inquired about my doctoral research on my way back from Colombia, for
eventually coming around to believing that my interest in cocaine was merely academic

and not personal.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................ vii
LIST OF FIGURES ......................................................................................................... viii
CHAPTER 1: DO ERADICATION AND ALTERNATIVE CROP PROGRAMS
REDUCE COCA CULTIVATION IN COLOMBIA? ....................................................... 1
I. Introduction ................................................................................................................. 1
11. Literature Review ....................................................................................................... 3
III. Data ........................................................................................................................... 9
IV. Distance to a Fumigation Base and Antinarcotics Policy ....................................... 13
V. The Effect of Eradication and Alternative Development on Coca Cultivation ....... 17
VI. Conclusion .............................................................................................................. 22
CHAPTER 2: THE EFFECT OF ERADICATION ON COCA PRODUCTION AT THE
FARM LEVEL .................................................................................................................. 44
I. Introduction ............................................................................................................... 44
11. Institutional Framework ........................................................................................... 47
III. Data ......................................................................................................................... 50
IV. Empirical Strategy .................................................................................................. 53
V. Results ...................................................................................................................... 56
VI. Conclusion .............................................................................................................. 59
APPENDIX A: UNODC Estimates of Coca Cultivation ................................................. 66
REFERENCES ................................................................................................................. 67

vi

LIST OF TABLES

Table 1: Coca Cultivation by Year, by Region ................................................................. 26
Table 2: Coca Eradication by Year, by Region ................................................................ 27
Table 3: Spending in alternative development programs by year by region in millions of
1998 pesos ......................................................................................................................... 28
Table 4: Mean annual cultivation, fumigation, eradication, and alternative development
spending by location within 80 miles of a fumigation base .............................................. 29
Table 5: Average Coca Cultivation, Eradication, and Alternative Development Spending
per Municipality ................................................................................................................ 30
Table 6: First Stage Estimates with Eradication as a Dependent Variable ....................... 31

Table 7: First Stage Estimates with Alternative Development Spending as a Dependent
Variable ............................................................................................................................. 32

Table 8: First Stage Estimates with Eradication as a Dependent Variable (Replacing
Fumigation Bases with Largest Urban Areas) .................................................................. 33

Table 9: First Stage Estimates with Alternative Development Spending as a Dependent
Variable (Replacing Fumigation Bases with Largest Urban Areas) ................................. 34

Table 10: Estimates of the Effect of Eradication and Alternative Development Spending
on Coca Production ........................................................................................................... 35

Table 1]: Descriptive statistics of the Survey of Agro-Cultural Characteristics of Coca

Crops in Colombia ............................................................................................................ 61
Table 12: Descriptive statistics by farm ............................................................................ 62
Table 13: Descriptive statistics of the municipalities in the sample ................................. 63
Table 14: First stage regression ........................................................................................ 64

Table 15: OLS and IV regressions with share of cultivated land devoted to coca as a
dependent variable ...... - ........... 65

vii

LIST OF FIGURES

Figure 1: Map of Mean Annual Cultivation of Coca by Municipality, 2001-2006 .......... 36
Figure 2: Map of Change in Coca Cultivation, 2001—2006 .............................................. 37
Figure 3: Map of Kilograms of Coca Leaf per Hectare per Year by Region .................... 38

Figure 4: Map of Mean Annual Eradication of Coca by Municipality, 2001-2006 ......... 39

Figure 5: Map of Mean Annual Spending in Alternative Development Programs, 2001-

2006 (in Millions of Pesos) ............................................................................................... 40
Figure 6: Map of Distance from a F umigation Base (Except Larandia) ........................... 41
Figure 7: Coca eradication and cultivation by distance to a fumigation base .................. 42

Figure 8: Alternative development spending in millions of pesos by distance to a
fumigation base ................................................................................................................. 43

viii

CHAPTER 1: DO ERADICATION AND ALTERNATIVE CROP PROGRAMS
REDUCE COCA CULTIVATION IN COLOMBIA?

I. INTRODUCTION

Colombia is the main supplier of cocaine to the United States. Because of the
social costs resulting from production and consumption of the drug, both countries have
made it a priority to put an end to illegal drug manufacturing and trafficking. In order to
disrupt the flow of drugs, the governments of the two nations have targeted every step of
this process, allocating billions of dollars to antinarcotics military units, border controls,
interdiction of drug shipments, and the destruction of laboratories and the precursor
chemicals used in the production of cocaine.

However, little applied work has been done to assess the efficacy of drug control
policies. This is particularly true in the case of coca eradication and alternative
development programs, which target the farmers that produce coca leaf, the primary input
of cocaine. Only Moreno et al. (2003) and Dion and Russler (2008) have attempted to
estimate the effectiveness of coca eradication, and to date there are no econometric
evaluations of whether increased spending on alternative development programs leads to
a reduction in coca cultivation.

It is widely acknowledged in the economic literature on crime that empirical
estimation of the effects of law enforcement and other government programs on illegal
behavior is a task complicated by severe endogeneity of the policy variable (Johnson, R.
S., S. Kantor and P. V. F ishback 2007, Levitt, S. D. 1997). Yet the available studies of

the effect of eradication on coca cultivation do not address this issue. This is surprising,

because endogeneity in this setting is a serious concern. The Colombian government has
a clear incentive to concentrate its antinarcotics efforts on those regions of the country
where the production Of coca is most pervasive.

In this paper I use an Instrumental Variable (IV) approach to account for drug
policy endogeneity. I choose instruments that enter the expected cost function of the
government as it implements eradication and alternative development programs, yet
affect farmers’ land allocation decisions through no channel other than their effect on
drug policy. The instruments are derived from variations in the expected cost of coca
eradication as crews get far from the zone where Antinarcotics Police helicopters can
protect them from illegal armed groups on the ground that try to shoot them down. I find
that there is a strong negative relationship between distance from this zone and the level
Of eradication that takes place. Because of similar safety limitations in getting alternative
development program administrators to their destination in coca growing regions, the
same correlation exists between distance from the safety zone and government spending
on alternative development. The location of coca eradication bases is not determined by
coca production levels in the vicinity, but by the presence of pre-existing commercial and
military airports, which makes instruments derived from the distance to the safety zone
surrounding the bases ideal to estimate the causal effect of eradication and alternative
development programs on coca production.

By evaluating the effectiveness of eradication and alternative development
programs in reducing the cultivation of coca, I contribute to answering the wider question
of whether the current approaches to drug policy are succeeding. Eradication and

alternative development are a particularly good place to start because the outcome of

these programs - coca cultivation per municipality - can be observed directly. Six years of
municipality level data on coca cultivation for the entire country are available from
Project SIMCI II, a satellite survey of coca crops conducted by the United Nations Office
on Drug and Crime (UNODC). I construct a panel that combines SIMCI II data with
municipality level measures of eradication, alternative development, and the presence of
illegal armed groups, as collected by entities such as the Antinarcotics Police and the
Presidency Of the Republic.

This paper is divided in six sections. In Section II, I provide a comprehensive
review of the relevant literature on crop choice, coca farming, and crime. In the third
section I describe the data. Section IV describes the instrument and its correlation with
Colombian antinarcotics policies. The estimates of the causal effects of eradication and
alternative development on coca cultivation are presented in section V. Section VI

concludes.

II. LITERATURE REVIEW

Coca eradication and alternative development programs seek to reduce illegal
crop cultivation by modifying the economic incentives faced by farmers, and there is
evidence that coca farming households respond to such changes. In particular, studies of
coca farmers demonstrate that they respond to fluctuations in the profitability of their
crops. Angrist and Kugler (2008), for example, analyze the impact Of an exogenous rise
in the price of coca resulting from the sudden interruption of coca imports from-Bolivia
and Peru, a consequence of heightened enforcement at the border in 1994. Using data

from an annual survey of rural households, they find that after 1994 there was higher self

employment income in regions where coca was traditionally grown. They attribute the
rise in self employment to an increase in illegal crop farming. Furthermore, they were
able to link higher coca prices to an increase in child labor. Additionally, they found a
rise in violent deaths per capita in coca growing areas. This suggests that the traditional
assumption that illegal crop farming results in conflict over coca profits is true. Using the
same shift in coca production from Peru and Bolivia to Colombia, but focusing on its
consequences for Peruvian farmers, Dammert (2008) finds that reduced coca earnings led.
to an increase in child labor. Her findings further demonstrate that coca farmers respond
to economic incentives and that altering them is a viable approach to drug policy.
Whether eradication and alternative development change incentives in a way that
will lead to a reduction in coca farming is a complex question. At issue is the mechanism
through which policy intervention can reduce or eliminate production of illegal crops.
One way eradication can make farmers switch from coca to other crops is by increasing
the variability of coca yields. Agricultural households in the developing world have been
observed to forego profitable economic activities when they have highly variable returns
(Morduch, J. 1995). This stems from the households’ inability to buy insurance against
annual fluctuations in harvest yields and the associated difficulty in smoothing
consumption across years. Rosenzweig and Binswanger (1993) conclude that Indian
farmers in environments with more weather variability choose less profitable but less
risky crop portfolios. Kurosaki and F afchamps’ (2002) study of crop choices in Pakistan
finds that agricultural production choices are similarly affected by price and yield risk.
Yield variations in this literature are due to weather shocks, but the threat of government

sponsored eradication in Colombia acts in the same way, making the returns from coca

farming more uncertain. It is reasonable to assume that, just like their counterparts in
other developing countries, coca farmers in Colombia cannot fully insure against harvest
losses, thereby smoothing consumption across years. Consequently, they may diversify
their crop portfolio and replace some coca with other crops.

While eradication can increase income variability, a serious problem with
government attempts to force a change in the farmers’ crop portfolio is that coca offers
the highest expected profit, approximately double that of the next best legal crop
(Peterson, S. 2002, Thoumi, F. E. 2002). Some authors have explored the mechanisms
through which eradication and alternative development can overcome this obstacle. In a
study of opium production in Afghanistan, Clemens (2008) sketches a theoretical model
of illegal crop farming in which farmers face the possibility of eradication and/or receive
some kind of alternative development support. The farming household maximizes
expected utility from consumption, which depends positively on the returns of a crop
portfolio, and negatively on moral aversion to poppy cultivation. Each crop has returns
that vary across different states of nature. By increasing the returns to legal crOps through
price support, cost reduction, or higher yields, or by using eradication to raise the
probability that returns from illegal crop cultivation will be zero, and perhaps by finding
ways to increase moral aversion to coca farming, resources would be reallocated from
illegal crops to legal ones.

It is clear that illegal crop eradication lowers yields and may cause a reallocation
of land away from coca. But it is possible that an inelastic demand causes prices to
increase with eradication, as Vargas-Manrique (2004) notes. This would keep profits high

and minimize the effect of eradication. Inelastic demand would result in farmers

responding to eradication programs which reduce expected yield by increasing land used.
The incentive to do this would be provided by higher prices, resulting in minimal changes
in the quantity of coca harvested.

Because land allocated to coca can conceivably increase or decrease as a result of
eradication, the question of whether drug control policies lead to a reduction in coca
cultivation is ultimately empirical. Yet few empirical analyses of the effect of eradication
and alternative development spending on coca cultivation exist.

An early study of the effect of eradication is Moreno-Sanchez et al. (2003). It uses
national level data. Regressing hectares of coca cultivated in Colombia on hectares of
coca eradicated, they find that eradication does not effectively control the supply of coca.
Rather, it is associated with greater levels of cultivation. They attribute this outcome to
farmers compensating for the destruction of their crops by cultivating greater extensions
of land. Their specification accounts for the effects of coca cultivation in Bolivia and
Peru and the prices of coca and plantain. Coca cultivation in Colombia is found to be
unitary elastic to production in Bolivia and Peru, inelastic to its own price, and elastic to
plantain price. Moreno-Sanchez et al. conclude that unilateral enforcement by one
country does not work because the unitary elasticity of coca production in Colombia to
production in Bolivia and Peru suggests that production will simply move from country
to country as levels of enforcement vary. Because they find that the elasticity of coca
cultivation to the price of other crops seems to be higher than that to the price of coca,
they take it as evidence that alternative development efforts may be more successful than

eradication.

More recently, Dion and Russler (2008) have used panel data on Colombian
departments (administratively equivalent to US states) to examine the effect of
fumigation and other variables on coca cultivation. The dependent variable in their study
is hectares cultivated per capita. The main independent variable is the percentage of
department land area fumigated by the government. They find that aerial eradication, the
size of the incoming population of displaced persons, department Gross Domestic
Product, and corruption indicators do not have an effect on the level of coca cultivation.
Government spending per capita, the number of outgoing persons displaced by violence
and market access are associated with lower levels of coca cultivation. Poverty and legal
agricultural output are associated with increased coca production. While the results are
interesting, a weakness of these studies is that they do not try to account for the
endogeneity of eradication policy.

Yet there is a substantial literature on crime that has better addressed the
endogeneity of enforcement, whose approach I incorporate to the analysis of the
relationship between drug policy and coca cultivation. By finding exogenous shocks to
enforcement, the economic literature on crime has been able to address issues of
causality. A seminal paper is Levitt (1997), which measures the effect of the size of
police departments on crime in the United States. It reviews several previous studies
which found a counterintuitive positive relation between police and crime. It also
documents a previously unknown relationship between mayoral and gubernatorial
elections and changes in the size of police departments. Levitt constructs a panel of US

city level data and regresses changes in violent crime on changes in police size, using

election cycles as an instrument for police officers. He finds a negative relation between
the number of police officers and crime.

In the same vein, Klick and Tabarrok (2005) use terror alert levels as a source of
exogenous variation in police levels in Washington DC. They claim that more police are
deployed when terror alert levels — which are considered exogenous to local crime - are
high, therefore having a deterrent effect on crime. Regressing the daily number of crimes
on terror alert levels, they find that a high terror alert level has a negative effect on crime,
which they attribute to the additional numbers of police officers on the streets. A final
example is Johnson et al. (2007), which evaluates the effect of welfare spending per
capita on crime rates during the Great Depression, using panel data on major US cities.
Because welfare spending was allocated in part with the goal to reduce the criminality
that could spring from widespread unemployment, the government targeted federal aid to
cities that were considered more likely to harbor crime. As a result of the endogeneity of
the allocation, OLS estimates are likely to be biased. The authors therefore use an IV
approach, instrumenting spending with variables based on state size and federal land
ownership.

I apply the insights of the crime literature to the problem of estimating the effect
of drug policy on coca cultivation. An IV approach that accounts for the endogeneity of
coca eradication and alternative development spending allows me to estimate the causal

relationship between enforcement and coca cultivation.

III. DATA

The data I use are a six year panel of the 257 Colombian municipalities that grew
coca at some point between 2001 and 2006. The United Nations Office on Drug and

Crime in Bogota conducts satellite surveys of coca crops in every municipality of the

country Since 2001]. The surveys use satellite photography and are designed to measure

. . . . . st
the number of hectares of coca In a given munrcrpalrty on December 31 of each year.

These measures are reasonably accurate, and the methodology behind them is discussed
in the data appendix.

Figure 1 shows a map of the average annual level of coca cultivation per
municipality for the period 2001-2006. 57 percent of cultivation took place in the
Colombian Amazon rainforest, 17 percent in the Pacific coast, 15 percent in the Andes
Mountains, 8 percent in the eastern plains bordering Venezuela, and the remaining 3
percent in the Caribbean regionz. Entire municipalities entered and left the market: in
2001, there were 164 coca producing municipalities, and by 2006 they are 199.
Moreover, 257 municipalities cultivated coca at least once, implying that some entered

and left the market over this period. This suggests that coca production can shift across

 

l Cultivation data per municipality are publicly available at
http://www.biesimci.org/llicitos/cultivosilicitos/cocampios.html

Colombia is divided into five natural regions: Amazonian, Andean, Caribbean, Eastern Plains, and
Pacific. These regions vary in socioeconomic conditions, climate, soil, flora, and fauna, and they provide
different conditions for the cultivation of coca. When analyzing regional differences in coca cultivation,
UNODC further subdivides the Amazonian region into Meta-Guaviare and Putumayo-Caqueta, and the
Andean region into Catatumbo and Sur de Bolivar.

municipalities, in a local version of the “balloon effect” sometimes mentioned in the
literature on drug production (Rouse, S. M. and M. Arce 2006).

National and regional levels of cultivation by year can be found in Table l.
Cultivation across the country fell by 46 percent, from 144,808 hectares in 2001 to
77,870 in 2006, and the reduction came mainly from lower cultivation in the Amazonian
and Andean regions. But in every region there were municipalities where coca cultivation
went up, increasing or staying constant in 157 municipalities and falling in 100. A map
showing these changes in coca production is given in Figure 2.

A reduction in coca cultivation need not result in lower cocaine production.

Recent estimates of productivity per hectare3 are higher than those obtained by previous
studies, and indicate that cocaine production in Colombia fell by only 1 percent, from 617
to 610 metric tons (UNODC 2006). Furthermore, because coca leaf yields vary across
regions as shown by the map in Figure 3, production could have migrated from low to
high yield locations. However, this does not seem to be the case, as cultivation has fallen
in the regions with the highest yields.

The Colombian Antinarcotics Police maintain a national database of eradication
and provided their data for this study. The data are automatically recorded by GPS units
on board of the fumigation planes, or by the agents conducting manual eradication. Table
2 gives national and regional annual eradication levels, and the map in Figure 4 shows
average annual eradication per municipality. Eradication increased dramatically between
2001 and 2006: it more than doubled, going from roughly 94,000 hectares to 210,000.

The number of hectares eradicated at the national level exceeded the number of hectares

 

3 . . . . . .
Cocaine production per coca hectare Is not estimated by UNODC on an annual basrs, and methodologies
vary across studies, which limits the usefulness of comparing total cocaine production estimates over time.

10

at the end of every year. The reason is that the coca bush can regenerate in 6 to 8 months,
so that the same area of coca may be eradicated more than once per year.

The police locate coca crops using data from SIMCI II as well as less thorough
aerial surveys conducted every six months. Early in the period under consideration there
was a marked emphasis on eradication in the areas with the highest levels of production.
From 2001 to 2003, for example, not a single hectare of coca was eradicated in the
Caribbean region. Yet between 2001 and 2006 there was a uniform trend toward higher
levels of eradication in every region of the country. Moreover, eradication increased or
stayed at 2001 levels in 239 out of 257 municipalities.

The methods of eradication also changed slightly over time. The government has
used fumigation with glyphosate throughout the period covered by the data, and this
continues to be the most widely used technique. Glyphosate kills coca leafs and keeps the
bush from producing another harvest for around six months. From 2005 onwards, manual
destruction of coca bushes replaced fumigation in areas near the border with Ecuador and
in national parks. The primary reason was political pressure resulting from environmental
concerns. As I describe below, the pattern of application of both methods is essentially
the same, so I have added the number of hectares aerially sprayed to the number of
hectares eradicated manually to obtain a single eradication variable.

Countrywide data on spending in alternative development programs are collected
by the Presidency of the Republic, and were provided by UNODC. Table 3 shows
spending data at the national and regional level and the map in Figure 5 shows average

annual spending per municipality.

ll

The funding for alternative development programs fell by 81 percent between
2001 and 2006, increasing only in 6 out of 257 municipalities. The secular decline in
alternative development spending probably resulted from a change in government. 93
percent of the programs were started before or during 2001, roughly coinciding with the
presidency of Andres Pastrana. There was a sharp increase in spending toward the end of
his government, and a gradual decline after he was succeeded by Alvaro Uribe in 2002.

The diversity of alternative development programs is remarkable, and it limits
somewhat our ability to measure their effectiveness. Over the period of interest, there
were 1379 different types of crop substitution programs, many of which provided training
or subsidized inputs for the production of alternative crops. But they also included the
introduction of new agricultural techniques, help to commercialize farm products,
technical assistance, and construction of processing facilities and various types of
infrastructure. By considering spending on alternative development as a single policy
variable I am able to evaluate the effects of the average program, but I cannot determine
the performance of individual ones.

Aside item the data on coca cultivation, eradication, and alternative development
spending, I obtained data on other variables that are likely to affect coca cultivation
levels. Population data come from the 2005 census and are available from the National
Administrative Department of Statistics (DANE). To control for the varying availability
of land across municipalities, I used data on municipality areas from the Agustin Codazzi
Geographical Institute in Bogota. Finally, data on the presence of illegal armed groups

per municipality for the year 2005 were provided by UNODC.

12

IV. DISTANCE TO A FUMIGATION BASE AND ANTTNARCOTICS POLICY

According to the Colombian Antinarcotics Police, there is a strong link between
distance from a fumigation base and the amount of coca eradication that goes on in a
municipality. Fumigation of coca crops with glyphosate, the primary method of
eradication, is conducted by police personnel flying various types of unarmed aircraft.
They include OV-lO Bronco military planes, modified so that they have no weapons.
Instead, they carry modern fumigation equipment. There are also Air Tractor AT-802 and
Turbo Thrush spraying planes, which are small aircraft widely used in agriculture that
have no artillery (Luna, A. O. 2007). While the planes can fly to every region of the
country, security as they perform their task is often at stake.

Armed attacks from the ground against fumigation planes are common. Police
protect from the attacks by escorting the planes with armed helicopters like the Huey II
and the UH-60 Black Hawk. According to Antinarcotics Officers, the range of the
helicopters is 80 miles from the base, a point after which they must return to the base they
departed from. The map in Figure 6 shows which coca municipalities are within reach of
a base. Beyond the 80 mile range, fumigation planes must go unprotected, greatly
increasing the expected cost of eradication missions. Being vulnerable to attack by
disgruntled coca producers (or by the armed groups acting on their behalf), the planes are
liable to being shot down, resulting in losses of lives and equipment.

The Observed patterns of coca cultivation, eradication, and alternative
development efforts across Colombian municipalities suggest that the lack of protection
of fumigation planes results in major shifts in drug policy. A comparison Of the mean

number of hectares of coca grown per municipality for the years 2001 through 2006

13

shows that cultivation is roughly equal across locations within 80 miles of a base and
outside of this range. The mean for the first group is 397.7 hectares, and 328 for the
second group. Instead, fumigation within 80 miles of a base is significantly higher than
beyond this distance. Mean fumigation per municipality within the safety zone is 681.7
hectares, but only 208.2 hectares outside of it. The trend of more intensive eradication in
areas closer to the fumigation bases remains unaltered when manual eradication is
included into the eradication calculations (see Table 4). Figure 7 Shows the relationship
between distance from a fumigation base, coca cultivation, and eradication per
municipality. The number of hectares of coca fumigated is greater than the number of
hectares allocated to coca cultivation for the first 80 miles. At that point the relationship
is turned around, with cultivation surpassing fumigation thereafter.

Another noticeable difference between the two zones emerges when mean
alternative development spending per municipality is calculated: average funding for
these programs is far greater in municipalities within the 80 mile range of a base than in
the rest, by a ratio of 5 to 1. This results from the same security issues as in the case of
fumigation, and from the government’s preference to concentrate alternative development
efforts in areas where a greater rate of eradication makes alternative crops more attractive
to farmers. Figure 8 shows the relationship between alternative development spending
and distance from a fumigation base.

Naturally, the relationships I have presented do not take into consideration other
factors that may affect eradication and alternative development levels, such as the
presence of illegal armed groups. For example, the police may want to pursue different

levels of eradication in municipalities with left wing militias that seek to overthrow the

14

government, such as the Revolutionary Armed Forces of Colombia (FARC) and the
National Liberation Army (ELN), or where the right wing United Self—Defense Forces of
Colombia (AUC) are present. A more thorough model of the effect of distance fiom a
fumigation base on drug policy is given in Equation 1.1, which is the first stage equation

for IV estimation of the effects of drug policy on coca cultivation:

(1-1) Fit: 0 + Di’13+ Xi’t + 8r + er + Uit

Fit is the policy in question, which can be either eradication or alternative development
spending in municipality i in period t. Di is a vector of variables that reflect cost
variations in drug policy, Xi is a vector of control variables, and ar and et stand for

regional and year fixed effects. Finally, “it is an error term with zero mean.

Table 6 shows OLS estimates of three different specifications of Equation 1.1.
The dependent variable is hectares of coca eradicated per municipality per year. All
regressions control for the number of illegal armed combatants belonging to FARC, ELN,
and AUC, as well as for municipality area and population. Indicators for year and
regional fixed effects are also included.

Specification (1) includes two variables that proxy for the expected cost of drug
policy. The first is an indicator for whether a municipality is within the safety zone
surrounding a fumigation base. As a result of the lower cost of drug policy
implementation in this area, the coefficient on the indicator should take a positive value.

This would indicate that drug policies are implemented more intensively where they are

15

cheaper. The second variable is distance from the safety zone. The longer an eradication
plane stays away from the safety zone, the more likely it is to get shot. Because
unprotected exposure to potential attacks will last longer as distance to a destination
increases, 1 take the probability of an attack on a fumigation plane to be increasing on
distance from the safety zone. Therefore, the coefficient on distance to the safety zone
should take a negative value, reflecting lower levels of drug policy implementation as
costs increase. The estimated coefficients on the drug policy cost proxy variables are
statistically significant and take on the predicted signs.

Specification (2) adds interactions between distance from the safety zone and the
number of combatants in illegal left wing (FARC and ELN) and right wing (AUC) armed
organizations. These interactions exploit variation in the expected cost of drug policy
implementation across municipalities that are equidistant from a base. Because the
likelihood of an attack (and therefore the expected cost of drug policy) is increasing on
the number of illegal group members at the destination, the Sign of the coefficients on the
interactions should be negative. The coefficients are indeed negative, and the one on the
distance and left wing combatant interaction statistically significant.

Finally, specification (3) includes three more regressors, intended to capture
variation in expected drug policy costs within the safety zone. One is the interaction of
the safety zone indicator with distance from the fumigation base, and the other two are
interactions of this variable with the number of left and right wing combatants. For
reasons analogous to those given above, the coefficients on these interactions should be
negative. The estimates of two of these additional interaction terms are, indeed, negative

and statistically significant.

16

Table 7 presents estimates of Equation 1.1 where the dependent variable is
changed from coca eradication to alternative development spending. Otherwise, the
specifications are the same as in Table 6. The estimated coefficients take values that are
generally consistent with the predicted variation in the cost of drug policy
implementation, and several of them are statistically significant. However, the variation
in the number of illegal armed combatants across municipalities located at the same
distance from a base seems to have less influence on alternative development spending
than on coca eradication.

To verify that these correlations do not arise randomly or because of unaccounted
factors related to a municipality’s distance from a major urban area, I estimate Equation
1.1 again, replacing the location of the ten fumigation bases with the location of the ten
largest Colombian cities without a fumigation base. Predicted eradication is shown in
Table 8. The coefficients on the instruments are generally statistically insignificant, and
in the few instances where they are significant they have the “wrong” Sign. Table 9 shows
predicted alternative development spending. All but one of the coefficients on the
instruments turns out to be statistically insignificant. The weakness of these alternative
specifications is reassuring, demonstrating that the measures of drug policy

implementation that I use are actually picking up the effect of expected cost variations.

V. THE EFFECT OF ERADICATION AND ALTERNATIVE DEVELOPMENT ON
COCA CULTIVATION

The previous section documents the negative relationship between Colombian
drug policy and the ability of the Antinarcotics Police to protect its coca eradication

crews. I now take advantage of the exogenous variation in drug policy costs demonstrated

17

by this correlation to estimate the causal effect of eradication and alternative development
on coca production in Colombian municipalities.

If distance from the zone where eradication aircraft can fly safely is to be used as
an instrument for drug policy, distance from this zone cannot itself be related to a
municipality’s propensity tO produce coca. The exclusion restriction is not satisfied if the
second stage regression does not account for variables that are correlated with distance to
this zone and with coca production levels. The most obvious violation of this condition
would occur if the bases had been intentionally located in areas with high levels of
production of coca. However, the bases were not built with this criterion. Rather, they

made use of preexisting structures such as airports in large cities that made them suitable

to host a fleet of airplanes and helicopters.4

The location Of the bases being exogenous, the issue becomes whether there are
systematic differences between the more urban areas where the bases are located and the
rural regions out of their reach. To account for this I control for a number of variables.
More rural areas have lower population densities, so I control for population and
municipality area. I also include in the regressions the number of FARC, ELN, and AUC
combatants present in each municipality, because they concentrate in rural areas and
benefit from the drug trade. Furthermore, regional fixed effects control for the impact of
variations in climate, local institutions, and history, on coca cultivation decisions. For
example, regional variations in growing conditions result in significant differences in

annual yields per hectare that could affect land allocation decisions. Moreover, the

 

4 Because it is possible that unobserved characteristics of rural areas may have prompted the construction
of military bases and created a predisposition to grow coca, 1 have excluded from the calculations the only
fumigation base not located in an urban area, the Larandia military fort in rural Caqueta. The results do not
change significantly when it is included.

18

existence of trading routes, markets, and experience in coca cultivation and processing
may also be better developed in parts of the country like the Amazonian region, where
illegal crop farming has taken place for decades, than in others such as the Pacific region,
where coca production is of more recent date (UN ODC 2006). Finally, year fixed effects
account for shocks affecting all municipalities in a given year, such as changes in the
world market for cocaine, shocks to US funding for the war on drugs, and changes in the
central government resulting from the transition between the Pastrana and the Uribe
administrations.

With these controls, distance from the safety zone of fumigation planes should be
an appropriate instrument for Colombian anti-coca policies.

I estimate the effect of coca eradication and alternative development programs
using an Instrumental Variables (IV) regression. The empirical specification is the

following:

(12) Hit = 5 + Fit’q) + xi,6 + br + ft + Vit

where Hit is the number of coca hectares per municipality at year end, Fit is a vector of
drug policies, and Xi is a vector of exogenous covariates. br and ft represent regional and

year effects, while Vit is a zero mean error term. While OLS estimates may be biased

because of a correlation between the policy variables and the error term, if distance from

the safety zone is uncorrelated with Vit the IV estimation should result in accurate

l9

estimates Of the causal effect of eradication and alternative development on coca
cultivation.

In all regressions, I use standard errors that are robust to heteroskedasticity and
clustered by municipality. Clustering is essential because some regressors, as well as the
instrument, are fixed at the municipality level. Moulton (1986) shows that when a
regressor does not vary within groups of observations, the conventional standard errors
underestimate the variance of the coefficients, and Shore—Sheppard (1996) demonstrates
that an analogous downward bias exists when an instrument takes the same value for
clusters of observations. Fortunately, as long as the number of clusters is large, this
concern can be put aside by clustering at the level of aggregation of the regressors
(Angrist, J. D. and J .-S. Pischke 2009). With 257 municipalities in the sample, there are
enough clusters to ensure the asymptotic validity of the estimates of the standard errors.

IV and OLS estimates of Equation 1.2 are presented in Table 10. The policy
variables are hectares of coca eradicated and government spending on alternative
development per municipality. All specifications control for regional and year fixed
effects, municipality area, and population.

OLS estimates in column (1) Show a statistically significant, positive effect of
both coca eradication and alternative development spending on coca production. They
suggest that a one percent increase in coca eradication will lead to an increase of around
0.24 percent in cultivation, and that a one percent increase in alternative development
results in a 0.15 percent increase in hectares of coca cultivated.

Similar results are given by the just identified IV regression in column (2), where

the instruments are distance beyond the 80 mile radius of a base and an indicator for

20

whether a municipality is less than 80 miles from a fumigation base. A one percent
increase in eradication is estimated to result in a 0.18 percent increase in cultivation, and
a one percent increase in alternative development spending leads to a 0.51 percent rise in
coca cultivation. However, the coefficients are not statistically significant. In column (3),
where interactions between distance beyond 80 miles of a base and the number of right
wing/left wing combatants are added to the instrument set, a one percent increase in
eradication is estimated to increase coca cultivation by 0.34 percent, but the coefficient is
not statistically Significant. A one percent increase in alternative development spending
results in a 0.39 percent increase in coca cultivation, and the coefficient is significant at
the 10 percent level. Finally, specification (4) further adds to the instrument set an
interaction between the indicator for distance from a base being less than 80 miles and
distance from the base, plus interactions between this variable and the number of right
and left wing combatants. The coefficients on eradication and alternative development
are of the same order of magnitude as in previous specifications: 3 one percent increase in
eradication results in a 0.30 increase in cultivation, an a one percent increase in
alternative development spending leads to a 0.28 percent increase in cultivation.
Moreover, the coefficients are significant at the 1 and the 5 percent levels, respectively.
For specifications (3) and (4), the Sargan-Hansen test Of the overidentifying
restrictions cannot reject the null that the instruments are exogenous, even at the 10
percent level. I estimate (3) and (4) by the LIML method (not reported) to verify the IV
results, because while having the same asymptotic distribution of IV, LIML has lower
finite-sample bias when the model is overidentified (Angrist, J. D. and J .-S. Pischke

2009). The point estimates remain practically unchanged.

2]

The coefficient on municipality area is always positive and statistically
significant, a natural result if the availability of more land results in greater production of
coca. The coefficients on all the other exogenous variables are statistically insignificant.
A one percent increase in the number of FARC combatants is associated with an increase
in coca cultivation Of less than one hundredth of a percent in all specifications. Because
the presence of illegal armed combatants is generally thought to be linked to greater
levels of production of coca, it is interesting that all specifications show that increases in
the number of ELN combatants are associated with a reduction in levels of coca
cultivation. This may be explained by these organizations having strong historical links
with certain areas of the country, tending to stay in those regions for reasons other than
coca production levels. At any rate, the effect is small, and in all specifications a one
percent increase in ELN combatants leads to reductions in coca cultivation of less than
one hundredth of a percent. In the case of the coefficient on the number of AUC
combatants, the Sign of the point estimates does not take a consistent value. Depending
on the specification, the effect of a one percent increase in AUC combatants on coca

cultivation ranges between a reduction of 0.03 percent and an increase of 0.04 percent.

VI. CONCLUSION

The data collected by the United Nations Office on Drug and Crime provide an
excellent opportunity to shed light on the efficiency Of the Colombian government’s
antinarcotics policies. I evaluate the effectiveness of two major drug control initiatives,

coca crop eradication and alternative development programs. The principal challenge to

22

accurately estimating the effectiveness of these programs is the endogeneity of the policy
variables. I address it using an IV strategy, exploiting exogenous shocks to the ability of
the Colombian government to implement drug control policies.

While the social cost of coca and cocaine production and consumption is difficult
to estimate precisely, it is clear that it is high and that the government has good reason to
fight drug production. Yet the appropriate course of action remains unclear because coca
eradication and alternative development have significant extemalities, and because the
estimates of their effect on cultivation presented in the previous section suggest they are
unsuccessful.

A substantial burden on the state results from the use of drug revenues to finance
illegal armed groups. Taxation of the drug trade is thought to account for around half the
revenue of F ARC, ELN, and AUC, organizations that attack and weaken the state; They
bomb the country’s oil pipelines, electrical, and telecommunications infrastructure;
forcefully displace the rural population; and are the cause of excess defense spending and
human capital losses from the combat deaths of members of the armed forces (Pinto
Borrego, M. E., A. Vergara Ballen and Y. Lahuerta Percipiano 2005). Furthermore, drug
consumption brings about human capital losses at home and abroad.

Addressing the problem of coca and cocaine production by offering farmers
alternative sources of income has several positive extemalities. The creation of new, legal
sources of income raises welfare in impoverished areas of the country, which is
particularly desirable in view of the common claim that coca farmers are so impoverished
that they “turn to illegal crops out of desperation” (Haugaard, L. 2001). The United

Nations recommend that coca farmers be seen as “candidates for development rather than

23

as criminals” (UNODC 2005), and alternative development is favored by farmer
advocates and sectors of the international community like the European Union, because it
is seen as superior to eradication on humanitarian grounds (Hagen, J. T. 2001).
Furthermore, alternative development has favorable political consequences in coca
farming regions, where illegal armed groups compete with the authority of the state. Coca
farmers tend to distrust a government they perceive as distant, corrupt, and uninterested
in their well being (Farthing, L. and B. Kohl 2005, Hagen, J. T. 2001, UNODC 2005). By
providing economic opportunities, alternative development can make strides against local
ambivalence towards the intentions of the state. If nothing else, alternative development
spending has the potential to increase farmer welfare while enhancing the legitimacy of
the Colombian state at the local level and internationally.

By contrast, illegal crop eradication has major negative extemalities. In particular,
aerial eradication may be detrimental to the environment. The area of the country targeted
by the spraying campaign is very biologically diverse and it is home to several unique
and endangered species, which makes the potential damage caused by herbicides severe
(Peterson, S. 2002). There are conflicting claims about the ecological impact of
glyphosate: Peterson contends that glyphosate has “well documented deleterious effects
on soil micro-organisms, mammalian life (including humans), invertebrates, and aquatic
organisms, especially fish.” On the other hand, the US. Department of State affirms that
the substance is safe. It points out that the US. Environmental Protection Agency
approved the general use of glyphosate in 1974, and that studies have shown it to be
easily biodegraded by soil microbes and to be essentially non-toxic to fish and farm

animals (USDS 2004). While the effects of glyphosate on the environment may remain

24

controversial, an undisputed environmental extemality of eradication has to do with the
displacement of illegal crops. Large amounts of rainforest have been cleared to make way
for new coca fields, in order to avoid eradication by moving production to plots that have
not yet been pinpointed as coca land (Bigwood, J. and P. Coffin 2005). Governmental
intervention may thus encourage the destruction of valuable environmental resources.

The bottom line, however, is that neither eradication nor alternative development
spending seem to be very effective in reducing coca cultivation. Eradication actually
seems to be increasing the area cultivated with coca slightly. Even if coca leaf production
is falling because of lower coca yields resulting from eradication, increased productivity
in cocaine manufacturing seems to be making up for it and maintaining supply of the
drug constant, as mentioned in section 11. With regard to alternative development
spending, the estimates obtained above Show that a one percent increase in spending in
alternative development programs is associated with an increase in coca cultivation of
between 0.18 and 0.51 percent. One explanation for this is that resources such as
agricultural inputs destined to alternative crops could be channeled toward coca
production by the farmers, resulting in increased illegal farming.

Coca production has large costs for the Colombian nation, and a high level of
investment on antinarcotics efforts seems to be warranted. Nevertheless, these resources
should be allocated to programs that reduce coca cultivation and the supply of drugs, and
neither eradication nor alternative development programs are furthering this goal. The
most cost effective course of action may be to explore new approaches to the problem of

coca production in Colombia.

25

Table 1: Coca Cultivation by Year, by Region

Hectares of Coca Detected
Region/Year

Catatumbo

(Andean Region)
Meta-Guaviare
(Amazonian Region)
Eastern Plains

Pacific Region
Putumayo-Caqueta
(Amazonian Region)

Caribbean Region

Sur de Bolivar
(Andean Region)

Total

2001

9145

36978

11915

11171

65404

865

9329

144807

2002

8042

36599

7124

17364

25155

998

6789

1 02071

2003

4471

28977

4356

19561

17297

759

10910

86331

2004

3056

28510

6244

15788

13474

1262

12014

80348

2005

844

25963

9709

17633

16271

542

14788

85750

2006

488

20540

6829

1 8809

19125

437

l 1642

77870

Total

26046

177567

46177

100326

156726

4863

65472

577177

 

Source: United Nations Office on Drug and Crime

26

Table 2: Coca Eradication by Year, by Region
Hectares of Coca Eradicated

Region/Year 2001 2002 2003 2004 2005
Catatumbo 10308 9186 13822 5686 3108
(Andean Region)

Meta-Guaviare 10729 8703 44467 34787 28826
(Amazonian
Region)
Eastern Plains 2820 O 1 1734 6782 2860
Pacific Region 8957 17962 38219 33366 69853
Putumayo-Caqueta 49758 90458 9402 34797 19418
(Amazonian
Region)
Caribbean Region 0 0 O 2224 3809
Sur de Bolivar 11581 4055 15173 211 19 40802
(Andean Region)
94153 130364 132817 138762 168677
Total

2006
3217

46906

7269

71548

37778

1151

42442

210311

Total
45327

174418

31465

239904

241612

7185

135172

875083

 

Source: Antinarcotics Police

27

Table 3: Spending in alternative develgmnent programs by year by region in millions of 1998 pesos

Region/Year
Catatumbo
(Andean Region)
Meta-Guaviare
(Amazonian
Region)

Eastern Plains

Pacific Region

Putumayo-Caqueta
(Amazonian
Region)

Caribbean Region

Sur de Bolivar
(Andean Region)

Total

Source: UNODC

Alternative Development Spending

2001
13160

2195

172

4231

51642

10

6315

77725

2002
7994

1 099

4668

45536

2003
6887

288

1651

15351

10

5098

29285

28

2004
1634

941

1056

10522

3626

17779

2005
0

141

1284

10371

1138

12932

2006
0

355

1057

10211

2005

896

14524

Total
29675

5018

172

11524

127617

2035

21740

197781

Table 4: Mean annual cultivation, fumigation, eradication, and alternative development spending by
location within 80 miles of a fumigation base.

 

Coca hectares Fumi gated coca Eradicated coca Alternative
per hectares per hectares per development
municipality municipality municipality spending per

municipality in
millions of pesos

Distance to a 397.7 681.7 736.7 174.9
firmigation base is (34.0) (71.2) (75.2) (30.9)
less than 80 miles

(N=1020)

Distance to a 328.6 208.2 236.9 37.1
fumigation base is 80 (49.6) (44.9) (48.3) (1 1.8)
miles or more

(N=522)

 

29

Table 5: Average Coca Cultivation, Eradication, and Alternative Development Spending per
Municipality

 

Year Mean number of Mean number of Mean number Mean spending
hectares cultivated hectares fumi gated of hectares on alternative
per municipality per municipality manually development
eradicated per programs per
municipality municipality
2001 563.45 366.35 0 302.43
(1709.19) (1572.04) (1577.4)
2002 397.16 507.25 0 177.18
(1273.80) (2496.13) (938.82)
2003 335.92 516.80 0 113.95
(880.41) (2020.36) (586.76)
2004 312.64 531.04 8.89 69.18
(762.41) (1598.477) (33.38) (346.98)
2005 333.66 539.14 117.19 50.32
(871.03) (2061.02) (476.63) (307.64)
2006 303 .00 667.87 150.46 56.51
(778.312) (1835.82) (493.79) (313.48)

 

Source: United Nations Office on Drug and Crime and Antinarcotics Police.

30

Table 6: First Stage Estimates with Eradication as a Dependent Variable

 

Dependent Variable is Eradicated Coca

ILectAes
(l) (2) (3)
Distance beyond 80 mile radius ofa base -5.142*** -4.470** -4.326***
(1.778) (1.810) (1.655)
Municipality is less than 80 miles from a fumigation base 5625*" 4112*" 9789*"
(172.4) (152.7) (331.9)
-0.1 12" -0.0562
Distance beyond 80 miles of a base * Number of left wing (0.0543) (0.0560)
combatants
Distance beyond 80 miles ofa base * Number of right wing -0.356 -l.08l*
combatants (0.252) (0.577)
Municipality is less than 80 miles from a fumigation base * -12.84**
Distance from base (5.025)
Municipality is less than 80 miles from a fumigation base * 0.132
Distance from base * Number of left wing combatants (0.0868)
Municipality is less than 80 miles from a fumigation base * -0.970*
Distance from base * Number of right wing combatants (0.551)
F ARC guerrilla combatants 1.720 2.391 2.681
(2.407) (2.779) (2.847)
AUC paramilitary combatants 18.08 20.15 55.39*
(1 1.20) (12.60) (28.60)
ELN guerrilla combatants 25.63" 28.54" 21.06*
(10.99) (11.20) (11.56)
Population 0.00330 0.00327 -0.000483
(0.00307) (0.00292) (0.00221)
Area in square kilometers 0.0310M 0.0326" 0.0195
(0.0143) (0.0145) (0.0153)
Observations 1536 1536 1536
R-squared 0.161 0.172 0.242

 

Robust standard errors clustered by municipality in parentheses. All regressions include a constant, year
indicators, and indicators for each of the 7 UNODC coca growing regions.

m p<0.01, ** p<0.05, * p<0.l

31

Table 7: First Stage Estimates with Alternative Development Spendingas a Dependent Variable

Dependent Variable is Spending in
Altemgtive Development Programs

Distance beyond 80 mile radius of a base
Municipality is less than 80 miles from a fumigation base

Distance beyond 80 miles of a base * Number of left wing
combatants

Distance beyond 80 miles of a base * Number of right wing
combatants

Municipality is less than 80 miles from a fumigation base *
Distance from base

Municipality is less than 80 miles from a fumigation base *
Distance from base * Number of left wing combatants

Municipality is less than 80 miles from a fumigation base *
Distance from base " Number of right wing combatants
FARC guerrilla combatants

AUC paramilitary combatants

ELN guerrilla combatants

Population

Area in square kilometers

Observations
R-squared

(1)

-2.670***
(0.984)
101.6*
(54.82)

0.559
(0.458)

2.111*
(1.197)

13.22
(8.520)

0.000640

(0.000551)

-0.00142
(0.00868)

1536
0.123

(2)

-2.546***
(0.955)
83.77
(59.39)
-o.0231
(0.0217)

-0.00692
(0.0354)

0.697
(0.550)

2.073
(1.349)

13.76
(8.841)

0.000702
(0.000572)

~0.00106
(0.00882)

1536
0.125

(3)

-2.492***
(0.863)
3420*"
(121.4)
0.000241
(0.0184)

-0.0466
(0.0433)

-6.578***
(2.027)

0.0766"
(0.0392)

-0.0677
(0.0433)

0.504
(0.339)

3.839“
(1.992)

7.865
(6.816)

0.000162
(0.000490)

-0.00559
(0.00772)

1536
0.148

Robust standard errors clustered by municipality in parentheses. All regressions include a constant, year
indicators, and indicators for each of the 7 UNODC coca growing regions.

m p<0.01, ** p<0.05, * p<0.1

32

Table 8: First Stage Estimates with Eradication as a Dependent Variable (Replacing Fumigation

Bases with Largest Urban Areas)

Dependent Variable is Eraflcated Coca

Elgar—es
(1) (2) (3)
Distance beyond 80 mile radius of 10 largest cities 3.297* —0.0873 -0.155
(1.964) (1.260) (1.287)
Municipality is less than 80 miles from one of the 10 largest -13.55 -54.08 -206.0
cities (155.7) (134.3) (280.4)
0103*" 0.102"
Distance beyond 80 miles of one of the 10 largest cities * (0.0395) (0.0404)
Number of left wing combatants
Distance beyond 80 miles of one of the 10 largest cities * 0.273” 0.331"
Number of right wing combatants (0.127) (0.156)
Municipality is less than 80 miles from one of 10 largest cities 0.741
* Distance from base (4.591)
Municipality is less than 80 miles from one of 10 largest cities 0.0306
* Number of left wing combatants (0.0916)
Municipality is less than 80 miles from one of 10 largest cities 0.339
* Distance from base * Number of right wing combatants (0.244)
FARC guerrilla combatants 2.003 -7.518** -7.575*
(2.657) (3.718) (3.854)
AUC paramilitary combatants 19.47* -19.66 -29.44*
(10.86) (12.14) (17.59)
ELN guerrilla combatants 22.32“ 30.57" 30.68"
(11.46) (13.36) (15.03)
Population 0.00365 0.00426* 0.00378“
(0.00330) (0.00228) (0.00221)
Area in square kilometers -0.00200 -0.00617 -0.00553
(0.0187) (0.0180) (0.0182)
Observations 1536 1536 1536
R-squared 0.104 0.109 0.124

 

Robust standard errors clustered by municipality in parentheses. All regressions include a constant, year
indicators, and indicators for each of the 7 UNODC coca growing regions.
*** p<0.01, ** p<0.05, * p<0.1

33

Table 9: First Stage Estimates with Alternative Development Spending as a Dependent Variable

(Replacing Fumigation Bases with Larggst Urban Areas)

Dependent Variable is Spending in
Alternative Development Promms

(l) (2) (3)
Distance beyond 80 mile radius of 10 largest cities 0.196 -0.225 -0.223
(0.417) (0.445) (0.467)
Municipality is less than 80 miles from one of the 10 largest cities 0.398 -8.157 129.3"
(36.78) (35.09) (62.50)
Distance beyond 80 miles of one of the 10 largest cities * Number 0.0209 0.0207
of left wing combatants (0.0127) (0.0137)
Distance beyond 80 miles of one of the 10 largest cities * Number 0.00924 000159
of right wing combatants (0.0181) (0.0305)
Municipality is less than 80 miles from one of 10 largest cities "‘ -1.915
Distance fiom base (1.460)
Municipality is less than 80 miles from one Of 10 largest cities * -0.0128
Number of left wing combatants (0.0437)
Municipality is less than 80 miles from one of 10 largest cities * -0.0603
Distance from base * Number of right wing combatants (0.0844)
FARC guerrilla combatants 0.691 -1.108 -l.063
(0.530) (0.993) (1.146)
AUC paramilitary combatants 2.708" 1.000 2.828
(1.253) (2.830) (5.107)
ELN guerrilla combatants 11.95 12.97 13.05
(8.495) (8.432) (9.068)
Population 0.000730 0.000783 0.000826
(0.000628) (0.000602) (0.000547)
Area in square kilometers -0.00954 -0.0115 -0.0117
(0.0104) (0.0105) (0.0106)
Observations 1536 1536 1536
R-squared 0.139 0.223 0.244

 

Robust standard errors clustered by municipality in parentheses. All regressions include a constant, year
indicators, and indicators for each of the 7 UNODC coca growing regions.

*** p<0.01, ** p<0.05, * p<0.l

34

Table 10: Estimates of the Effect of Eradication and Alternative Development Spending on Coca

Production

Dependent Variable is Hectares of Coca Detected at the End of the Year
in a Municipality

(1) (2) (3) (4)
OLS IV 1V IV
Eradicated Coca Hectares 0158*” 0.117 0.226 0198*"
(0.0365) (0.393) (0.160) (0.0749)
Spending on Alternative 0.437*** 1.491 1.146* 0.808"
Development Programs (0.0846) (1.250) (0.609) (0.319)
FARC guen’illa combatants 0.646 0.00307 0.0292 0.316
(0.814) (0.773) (0.680) (0.657)
AUC paramilitary combatants 1.952 -0.0816 -1 .342 0.168
(1.798) (5.029) (1.818) (1.324)
ELN guerrilla combatants -0.434 -12.09 -10.38 -5.719
(3.268) (1 1.30) (8.988) (5.335)
Population 0.000145 -0.000493 -0.000610 -0.000275
(0.000443) (0.000719) (0.000586) (0.000394)
Area in square kilometers 0.0596*** 0.0693*** 0.0651*** 0.0624***
(0.0152) (0.0229) (0.0154) (0.0140)
Observations 1542 1536 1536 1536
R-squared 0.518 -0.005 0.208 0.430

 

Instruments in (2) are distance beyond the 80 mile radius of a base and an indicator for whether a
municipality is less than 80 miles item a fumigation base. In (3), interactions between distance beyond 80
miles of a base and the number of right wing/left wing combatants are also included. Specification (4)
firrther adds an interaction between the indicator for distance from a base being less than 80 miles and
distance from the base, plus interactions between this variable and the number of right wing/left wing
combatants. All regressions include a constant, year indicators, and indicators for each of the 7 UNODC
coca growing regions. Robust standard errors clustered by municipality are shown in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

35

'nt .. ._
a“); [all .
. 4 . ,‘.-i‘."" ~"
~‘ ‘ ‘

Q

 

     

Mean Annual Cultivation in Hectares, 2001-200. . I ‘
5000-5889
4000-5000

3000-4000

‘ .1-1000 ’
- No Coca Cultivated

Figure 1: Map of Mean Annual Cultivation of Coca by Municipality, 2001-2006

  

 

 

   

 

       

Change in Coca Cultivation, 2001-2006

Increased or Stayed Constant

a Decreased

1:] No Coca Cultivated

Figure 2: Map of Change in Coca Cultivation, 2001-2006

37

4.

    
      
  

fiil‘:

I
.‘V'

o l

‘ .

\0"
‘ ‘\‘|b

a; '1‘
c. ‘\‘.1. '

.4 -

 
    

l

 
 

'7
I

a .N

 
 
 
  

1
d
It}
b '4
a

  
  
  
   
       

7;:
I
'4
t

l

- ‘3’!
e 3‘.
{Tr-t

in:
ibis

  

   
 
           

      

d‘

     

  

     

 

   
    
 

[51’
1:4

LII/h: ‘

  
   

Kgs. of Coca Leaf per Ha. per Ye
E] 2600 (Pacific Region)
4600 (Catatumbo)
E] 5400 (Caribbean Region)
1:] 5600 (Putumayo-Caqueta)
I 6600 (Sur de Bolivar)
I 7100 (Eastern Plains)

I 9900 (Meta-Guaviare) t.

1:] No Coca Cultivated

   

Figure 3: Map of Kilograms of Coca Leaf per Hectare per Year by Region.

38

’v

  
   

5
fl

 
  
  

5“"
0
I

II,

I
..t.«:

  
    
 
 

3‘

  
  
  
 

\
-\

“as

 
   
   
   
   
  

if.

    

rt.
fir.
it. ‘i"
.. Q. r
‘ira\
"In

     

1
s

r
.A ..

 
  

     
    
 
 
 

.IO '

 
      

    
   

 
  
      
   
 
 
   
 
  

‘ I.
I. V ' .‘I ‘1

4‘
i x
t"

1‘

     
 

   
  
    
 
  

I"
(is... :- 4
«‘1 '

 
  
 
 

 

  
 
  

 

4

.231
_ 1116i
5?!

 
     

 

   
 

Ir

. , a.
It?"
”VI

       
      
 

 
 
 

    

  
 

S I“;
’ll ‘
v.4
-1

Mean Annual Eradication in Hectares, 2001-200- I.

12500-15279
1000012500

750010000

5000-7500

2500-5000

.01-2500 "

No Coca Eradicated

 

Figure 4: Map of Mean Annual Eradication of Coca by Municipality, 2001-2006

39

 

 

       

‘\
ska

Mean Annual Spending in AD (Millions of Pesos)
4000-4680
3000-4000

2000-3000
1 000-2000

.1-1000
No AD Spending "

Figure 5: Map of Mean Annual Spending in Alternative Development Programs, 2001-2006 (in
Millions of Pesos)

40

      

  

Distance from a fumigation base

I Less than 80 miles

80 miles or more

D No Coca Cultivated

Figure 6: Map of Distance from a Fumigation Base (Except Larandia)

41

1 500

1000
1

500
l

”A- A

O-‘ “

 

 

—l

I

1 I
O 100 200 300
Distance from fumigation base, by 20 mile intervals

 

 

 

—O— Hec. eradicated per mun. —O— Hec. of coca per mun.

 

Figure 7: Coca eradication and cultivation by distance to a fumigation base

42

200 300
1 1

Government spending per municipality
100
l

 

 

I

T l
0 1 00 200 300
Distance from fumigation base. by 20 mile intervals

—4

Figure 8: Alternative development spending in millions of pesos by distance to a fumigation base

43

CHAPTER 2: THE EFFECT OF ERADICATION ON COCA PRODUCTION AT THE
FARM LEVEL

I. INTRODUCTION

Some 200 million people consumed illegal drugs at least once during the last year.
Cocaine accounted for 16-21 million users, ranking third most consumed after cannabis
(143-190 million) and amphetamine type drugs (16-51 million). The United Nations
estimates that the retail value of the overall drug market is around 321 billion dollars a
year. As a result of the health consequences of drug abuse, the international community
strongly condemned and agreed to fight the production and use of drugs in three major
treaties (UNODC 2005, 2009)

Many observers point out that the drug trade seems to flourish in spite of the
efforts of the governments of producer and consumer countries. The New York Times
(Kristof, N. 2009) reports that forty years after the United States declared a “war” on
drugs under President Nixon, they are “more readily available, at lower prices and higher
levels of potency” than before. The most recent United Nations World Drug Report
devotes its preface to countering what it acknowledges to be a “growing chorus” of
politicians, the press, and public opinion claiming that drug controls should be repealed
because they are not working (UNODC 2009). Yet the debate often misses the fact that
carefirl evaluations of the effects of enforcement efforts have rarely been undertaken,
Often because of the difficulties inherent in gathering data on illegal activity. Thus,
according to Reuter (2008), “the body of research and evaluation on drug policy

interventions, apart from drug treatment, is thin.”

44

This paper’s goal is to contribute to the ongoing research on drug policy by
evaluating the effect of aerial eradication of illegal crops on the cultivation decisions
made by coca farmers in Colombia. Colombia is the source of the majority of the world’s
cocaine supply. At the same time, it has received over US$ 6 billion in aid to combat the
production of coca and cocaine through Plan Colombia, a joint initiative with the United
States government (GAO 2008). Unlike the two other major coca producing countries,
Peru and Bolivia, it has pursued aerial fumigation of coca in a vigorous attempt to
eliminate illegal crops (Mejia, D. and C. E. Posada 2007). Thus the importance of the
Andean country as a front in the war on drugs makes it an ideal place to conduct an
evaluation of the effectiveness of a major drug control policy.

I use a novel data set collected by the United Nations Office on Drug and Crime.

The Survey of Agro-Cultural Characteristics of Coca Crops in Colombia5 was conducted
by UNODC in 2005, by going directly into the sometimes dangerous areas where coca
leaf is grown and promising farmers that their cooperation would not in any way lead to
charges being brought against them. The survey resulted in a wealth of demographic and
socioeconomic data about coca producers, as well as a detailed measure of the plots of
legal and illegal crops owned by the farmers. I merge these data with another unique data
set containing information on aerial eradication, provided by the Colombian Anti-
Narcotics police. This allows me to estimate the effectiveness Of fumigation and the
effect of characteristics like level of education and mode of ownership of the land on
farmers’ cultivation decisions. This adds significantly to our empirical knowledge of the

economic behavior of coca farmers at this level of disaggregation, which at present

 

5 UNODC (2006): "Caracteristicas Agroculturales De Los Cultivos De Coca En Colombia." Bogota:
United Nations Office on Drug and Crime

45

consists of having documented an exogenous coca price increase that raised self
employment income and led to higher levels of child labor in Colombia (Angrist, J. D.
and A. D. Kugler 2008), and a fall in coca cultivation that seems to have resulted in an
increase in child labor in Peru (Dammert, A. C. 2008).

Estimating the effect of the fumigation of coca crops, like other optimally
distributed policy interventions, is econometrically challenging because of the
endogeneity of the policy variable (Rosenzweig, M. R. and K. I. Wolpin 1986). This
problem is commonly found in the broader economic literature on crime, which shows
that unless policy endogeneity is taken into account the effectiveness of anti-crime efforts
can be underestimated. A classical example is Levitt (1997), who remarks on the
seemingly paradoxical positive relationship between the size of police force and the
incidence of crime found by numerous studies. Using mayoral and gubernatorial election
cycles as an instrument for the potentially endogenous police force size, he estimates a
negative relation between police and crime. Klick and Tabarrok (2005) make a similar
point using terror alert levels as a source of exogenous variation in police deployments in
Washington DC, as do Johnson et al. (2007) in their study of the effect of government
spending on crime in the United States during the Great Depression.

To account for the endogeneity that arises as a consequence of targeting more
intensively the areas that have higher coca farming levels, 1 instrument fumigation with
variations in the expected cost faced by the Anti-Narcotics Police as it carries out its
eradication missions. According to the interviews I conducted with the police, their
ability to fumigate is restricted by the distance of a coca field from the safety zone

surrounding a fumigation base. The reason is that fumigation crews are sometimes

46

attacked by illegal armed groups when they approach their target. As a means of
protection, fumigation planes are escorted by military helicopters, which can travel up to
80 miles from a fumigation base without refueling. The fumigation planes can go beyond
that distance, but only unescorted and at a significant risk. As distance from the 80 mile
radius of a base increases, the expected cost of the mission goes up because Of the
increased probability of an airplane being shot down. Furthermore, expected costs
increase at a higher rate when the known presence of illegal armed groups is greater. The
resulting variation in firmigation allows me to use a municipality’s distance from a
fumigation base and the number of illegal armed combatants to generate instruments that
can be used when I regress coca cultivation on fumigation rates. I find that while the OLS
and IV estimates of the coefficient on fumigation are statistically insignificant, the sign of
the coefficient shifts from positive to negative when the instruments are used. This
suggests that enforcement affects the behavior of farmers in the direction expected by the
government, but that like in the relevant crime literature care must be taken to control for
the endogeneity of the policy.

The paper proceeds as follows. In section II I describe the institutional framework
of coca production in Colombia. Then I describe the data, and in Section IV I elaborate

on the empirical strategy. Section V gives the results, and section VI concludes.

II. TNSTITUTIONAL FRAMEWORK

Cocaine, like heroin, is different from drugs such as marijuana and amphetamines
in that its production requires large extensions of land. As a result, coca leaf farming has

prospered in remote areas of Colombia, Bolivia, and Peru, where the absence of the state

47

creates havens that allow large scale illegal activities to take place. Currently, Colombia
is the largest producer of cocaine in the world, with 53 percent of the world supply
(UNODC 2009)

There are four stages in the production of cocaine, which is chemically extracted
from coca leaves. First, farmers grow the leaves, which are then sold to intermediaries
who transform them into coca paste and then into coca base. Finally, coca base is bought
by cocaine producers, who use it to manufacture the drug in clandestine laboratories.
Farmers are often able to produce coca paste and coca base themselves, but not cocaine,
which requires the most sophisticated processes.

Even though growing coca is illegal, and despite the government’s ability to
detect illegal crops, coca farmers rarely face prosecution or lose their land as a result of
their activities. There are two main reasons for this. First, coca farmers are invariably
poor, which would make prosecution unpopular and thus politically challenging. Second,
it would be impractical to confiscate and oversee large tracts of land in the countryside,
where government presence is thin. As a result, the government fights coca production
primarily by aerially fumigating the coca bushes that produce the leaf, hoping to raise the
costs of production. Fumigation is not extensive or periodic enough so as to destroy every
coca harvest, but the government’s rationale is simple: if as a consequence of fumigation
coca bushes become a less productive crop in terms of the amount of coca leaf they
produce per year, legal crops will eventually become the more attractive alternative. A
rather similar situation, that of opium poppy growers facing eradication in Afghanistan,

has been modeled by Clemens (2008).

48

An annual census of coca crops is done through satellite imaging, and a
semiannual one through aerial surveys. The Colombian Antinarcotics Police use this
information to spray coca crops with glyphosate. When coca fields are sprayed the
harvest is lost, but the coca bush survives, and a new harvest can be produced within
three to six months. Multiple annual harvests result in fumigation sometimes occurring
more than once a year in a particular area. Unless the farmer decides to uproot the coca
bush and replace it with a new crop, the area of the land allocated to coca as recorded in
the Survey of Agro-Cultural Characteristics of Coca Crops in Colombia stays the same.
Because of this, and because new coca bushes can be planted year-round the area devoted
to coca can increase after eradication has occurred.

The government chooses the crops to be fumigated on the basis of the
concentration of coca in a region. Because the success of eradication campaigns is
measured by the number of coca hectares sprayed nationwide throughout the year,
fumigation concentrates in areas of the country where there is a high density of coca
fields. Once a region is chosen for fumigation, intelligence reports are used tO determine
the safety conditions for the fumigation planes, and plans to escort them with Black
Hawks are made accordingly. This is necessary because there are left and right wing
militias that roam the Colombian countryside that have a vested interest in protecting the
coca crops, whose production they tax. Occasionally, these militias attack the ftunigation
planes (Luna, A. O. 2007).

Legal crops are sometimes damaged in these operations, but only coca is targeted.
When legal crops are damaged, farmers can claim compensation from the government,

provided that they are not growing coca. As a firrther incentive to grow legal crops, there

49

are a variety Of alternative programs aimed at discouraging the production of coca and
encouraging that of cacao, coffee, rubber, palm trees, and others. Another variety of
alternative programs consists of training for the farmers, introduction of new agricultural
techniques, help to commercialize farm products, and various types of infrastructure such
as roads and electricity, all in the hope that farmers will abandon coca farming. However,
support for these programs had dwindled by 2005 from a peak in spending attained in
2001. Because the UNODC survey reports that only six percent of the farmers
interviewed had received some type of incentive to stop growing coca, here I focus

exclusively on the effects of coca eradication.

III. DATA

The data on coca cultivation comes from the 2005 survey of coca growers
conducted by UNODC and its associates. UNODC chose a random sample of coca fields
detected by satellite as fellows. First, the country was divided into seven regions. The
coca producing areas of each region were divided into plots of one square kilometer.
Then they chose a sample of coca plots for each region, each square having a probability
proportional to the size of the coca crops in it of being chosen. UNODC selected a sub-
plot within each of the initial squares using the same technique, and divided into parcels
that had identical probabilities of being included in the sample. The farmers in charge of
the chosen parcels were interviewed, as well as two neighboring coca farming ~

households. The interviews were conducted on-site.

50

Access to some of the farms was not possible because Of ongoing public order
issues involving illegal armed groups, so UNODC substituted 93 of the original 463
sampling sites with comparable ones. In the final sample, careful interaction with the
local authorities and clear assurances that the data collected would not be used to
prosecute the farmers ensured their cooperation. On each of the 463 sites, the
interviewers had to find one household where coca was in harvest so that they could
collect samples. For the other two households, the data reported was on the area covered
with coca bushes whether or not coca was in harvest.

There were 1389 surveys filled out. Descriptive statistics are shown in Table 11.
They offer a unique glimpse into the world of coca growers, who turn out to be farmers
of relatively modest means. The respondents are the heads of the households sampled,

and the fact that 91 percent of them are male (compared with 70 percent for the general

Colombian population6) is evidence of a more traditional family structure. The average
age of the heads of household is 38 years, below the national average of 46 years. As one
would expect, coca farmers are not highly educated: only 30 percent have studied beyond

elementary school. By contrast, 60 percent of all Colombians in the same age group have

surpassed this educational level7. By and large, coca is the farmers’ main source of
income: 82 percent report as much. Some of them control more than one coca farm, but
the prevalent tendency is toward having just one, as the average farmer controls 1.2
farms.

In contrast with countries like Bolivia, where coca farmers belong to indigenous

groups that culturally value the production of coca leaves, coca is not a traditional crop in

 

6 Data for the general Colombian population are taken from the 2005 census.

The comparison is between coca farmers from the UNODC survey and Colombians 14 years and older.

51

Colombia. Widespread cultivation started in the mid—1990s as a consequence of an abrupt
interruption in coca leaf imports from Bolivia and Peru resulting from heightened
enforcement at the border (Angrist, J. D. and A. D. Kugler 2008, Perafan, C. C. 1999).
The survey confirms that the motivations for growing coca are not primarily cultural, but
of an economic nature: 67 percent of farmers say they grow coca because of its good
market price, 40 percent say that they grow it because it is easy to market, and only 30
percent attribute their decision to a regional custom. 52 percent say they grow it because
there is no other option, and it is a bit unclear what they mean by this. There are several
other crops that could be grown instead of coca in these regions. In all likelihood, what is
meant is that the Opportunity cost of not growing coca is very high.

76 percent of farmers report hiring laborers for the harvest, and 9 percent of all
coca farms are rented. The majority (52 percent) of the land is legally titled to those who
occupy it. 32 percent is occupied without a title, and sharecropping and other modes of
ownership account for only 6 percent Of the total.

Fumigation data corresponding to 2004 come from the Colombian Antinarcotics
Police, who use GPS equipment in the fumigation planes used to spray coca mom and
record the number of hectares fumigated in each municipality they target. To obtain a
measure of the probability of fumigation faced by a farmer, I divide the number of coca
hectares fumigated in a municipality throughout the year by the area of the municipality.
1 match this information with the survey data at the municipality level. The Antinarcotics
Police also provided the location of their fumigation bases, which is key to the
development of the instrumental variable. Table 13 shows descriptive statistics for the 43

municipalities in the sample.

52

IV. EMPIRICAL STRATEGY

1 use a simple linear model, shown in Equation 2.1, to relate the share Of

cultivated land that farmer i of municipality m devotes to coca, Him, to the rate of
fumigation of the previous year in that municipality, fm, a vector of controls at the
individual level Xim, a vector of controls at the municipality level Zm, a regional fixed

8
effect bf, and an error term Viti

(2-1) I‘Iim = 5 + ‘me + xim,9 + Zmz + br + Vit

Vector Xim includes titling to the land, the farmer’s years of schooling, and whether the

head of the household is male. I control for titling to the land because untitled land cannot
be used as collateral in credit markets. Consequently, titling has the potential of affecting
farmers’ ability to make investments in the land, acquire capital goods and working
capital (Besley, T. 1995, Besley, T. and R. Burgess 2000, de Soto, H. 2000). Titling
could therefore widen the range of uses of the land to which the farmers have access,

acting as a disincentive to grow coca. Education and gender proxy for the value of the

farmers’ labor in alternative occupations. Zm includes the number of members of the

 

8 Colombia is divided into five natural regions: Amazonian, Andean, Caribbean, Eastern Plains, and
Pacific. These regions vary in socioeconomic conditions, climate, soil, flora, and fauna, and they provide
different conditions for the cultivation of coca. When analyzing regional differences in coca cultivation,
UNODC further subdivides the Amazonian region into Meta-Guaviare and Putumayo-Caqueta, and the
Andean region into Catatumbo and Sur de Bolivar. In all regressions, I use UNODC’s regional divisions.

53

Revolutionary Armed Forces Of Colombia (F ARC), the National Liberation Army (ELN),
and the United Self-Defense Forces of Colombia (AUC) in the farmer’s municipality. I
control for the presence of these illegal armed groups because they may change the
incentives for coca production. For example, they are known to protect coca farmers from
government sponsored eradication of their crops, as well as to tax the drug trade.

The parameter that measures the effect of the drug control policy is (p, the
coefficient on fumigation. If the government’s strategy of fumigating coca crops to
increase production costs and reduce cultivation is based on the amount of coca that
farmers grow, OLS estimates of (p may be biased. Because the government’s goal is to
maximize the number of hectares of coca eliminated, constrained by the amount of
resources available, it will choose to fumigate the areas where the marginal cost of
fumigation is lowest. There are economies of scale in fumigating areas with a high
density of coca crops, as several hectares can be fumigated in a single mission, so the
result is that the government targets more intensively those areas where more coca is
grown.

Therefore, the level of illegal crop fumigation is likely to be endogenous,
resulting in biased OLS estimates. When this is taken into account, the institutional
details learned from the Antinarcotics Police become crucial, as they contain the
exogenous source of variation necessary to identify (0. According to the police, for
security reasons the fumigation planes often have to be escorted by military helicopters
back and forth from their bases. The helicopters cannot travel beyond 80 miles from the
base without refueling, which limits their ability to escort planes that go past that

distance. Hence, the areas beyond the 80 mile radius of a fumigation base are less likely

54

to be fumi gated the greater their distance from a fumigation base. Because this variation
in fumigation results from a technological limitation that is not controlled by the police, it
is in no way a response to the land allocations made by the farmers. In other words,
distance to the 80 mile radius of a fumigation base is uncorrelated with the error term in
Equation 2.1. I use this as the exclusion restriction when instrumenting fumigation with
distance to the 80 mile radius of a fumigation base.

The exclusion restriction could break down if the location of the fumigation bases
were endogenously determined. The main scenario under which this could be true would
be one in which the bases were strategically situated near areas with high densities of
coca crops. Such a concern can be allayed by pointing out that the construction of airport
facilities is expensive and takes years. Therefore, the bases are usually located in urban
areas and near to previously existing airports, which were built to service the demands of
commercial flights before coca became a major cash crop in the mid 19905. The
exception to the rule is the Larandia military fort, located in a rural zone of the Caqueta
department. This base was excluded from the calculations of distance to the nearest
fumigation base.

The implementation Of the IV strategy furthermore requires correlation between

the instrument and the instrumented variable. The first stage is shown in Equation 2.2,

where fm - the share Of hectares of coca fumigated last year in the farmer’s municipality —
is a function of Dm, a vector of instruments, farmer characteristics Xim, municipality

level characteristics Zm, and regional characteristics a,:

55

(2-2) fm : 0. + BDm + Ximay + Zm’S + 3r + um

Table 14 shows OLS estimates of 2.2 using four different instrument sets. Column (1)
corresponds to the just identified case, with distance from the 80 mile radius of a base as
the instrument. Column (2) adds an indicator for whether the municipality is in the safety
zone. In column (3), interactions between distance from the safety zone and the number
of right wing and left wing combatants are included. Finally, column (4) adds the
interaction of the safety zone indicator and distance from the base, plus interactions of
this variable with the number of right and left wing combatants. The coefficient on (1) is
significant at the 5 percent level. While individual coefficients in (2) - (4) are not
statistically significant, a joint test of the instruments in (2) has a p-value of 0.11, and
joint tests of the instruments in (3) and (4) show them to be significant at the 1 percent

level.

V. RESULTS

OLS estimates of Equation 2.1 are reported in Table 15. The coefficient on
fumigation, while statistically insignificant, is positive. For the average farmer, it implies
that a one percent increase in fumigation would result in an increase of about 0.02 percent
in the share of cultivated land that a farmer devotes to coca. This implies that the policy is
likely not having its intended effect, a result that one should be careful to accept if the
endogeneity of fumigation has not been accounted for. To address this concern, Table 15

also shows the results of IV estimation of Equation 2.1. In contrast with the initial

56

attempt at OLS estimation, the coefficient on coca fumigation is negative, although still
statistically insignificant. It suggests that there may be an upward bias in the OLS
estimate. Depending on which IV point estimate is chosen, a one percent increase in
fumigation could result in a reduction of between 0.02 and 0.20 percent in the share of
cultivated land devoted to coca. The large standard errors, however, keep us fiom
reaching a conclusion about the effect of fumigation on coca cultivation, other than that it
does not seem to bring about any noticeable changes.

In view of the fact that eradication lowers coca yields and makes the crop less
attractive compared to others, the persistent levels of coca production may be the result of
an inelastic demand for coca. Higher production costs would thus be met with higher
prices Offered by coca leaf buyers, making the change in coca production as a result of
eradication negligible.

The coefficient on years of schooling is positive and statistically significant across
specifications. The OLS and IV estimates show that an additional year of schooling is
associated with an increase of about 2.5 percent in the share of land allocated to coca.
The fact that additional years of schooling are correlated with higher reliance on coca
production suggests that education provides human capital that is particularly useful for
the production and commercialization of illegal crops, more so than for other alternatives
in the labor market.

Interestingly, land titling has a negative and statistically significant coefficient,
lending support to the hypothesis that it may lead to improved access to credit markets
and thus widen the range of economic activities other than coca cultivation available to

the farmers. Acquiring a formal title to at least part of their land would reduce the share

57

of cultivated land devoted to coca by about 11% according to the OLS estimate, or
between 11 and 13 percent depending on the IV estimate chosen.

The coefficients on the number of illegal armed combatants are generally
statistically insignificant. In the OLS regression, all are associated with decreases in the
share of farmland that a farmer devotes to coca. An increase of one percent in the number
of FARC combatants in the farmer’s municipality is associated with a reduction in the
share of farmland that a farmer devotes to coca of less than a hundredth of a percent. For
both ELN and AUC, the reduction is close to 0.02 percent.

Depending on the IV specification, the coefficients imply that a one percent
increase in the number of FARC combatants leads to an increase of between less than a
hundredth of a percent and 0.04 percent in the Share of farmland with coca. A one percent
increase in the number of ELN combatants in the farmer’s municipality is associated with
a decrease in the share of farmland devoted to coca of between 0.02 and 0.03 percent. In
the case of AUC, a one percent increase in the number of combatants in the farmer’s
municipality results in somewhere between a reduction of less than a hundredth of a
percent and a 0.06 percent increase in the share of farmland devoted to coca. However,
because of the lack of statistical significance of most of the coefficients, it cannot be
established with certainty what influence if any these groups have on farmers’ coca

production decisions.

58

 

V1. CONCLUSION

This paper documents variations in the Colombian government’s expected cost of
fumigating coca crops. Because the distance between fumigation bases and their targets is
known and is uncorrelated with the unobserved determinants of the share of cultivated
land that a farmer devotes to coca, it is used to generate instrument sets that are used
when estimating the effect of municipality level fumigation on the share of land a farmer
devotes to coca. Neither OLS nor IV estimation find a statistically significant effect of
fumigation on coca cultivation. Using the IV results as a benchmark, it is estimated that a
one percent increase in fumigation may bring about a reduction of as much as 0.20
percent in the share of land devoted to coca. But while the IV estimates suggest a
reduction in the land allocated to coca crops is the outcome of enforcement policies, the
hypothesis that firmigation has no effect on coca cultivation cannot be rejected.

The results, then, do not find conclusive evidence that fumigation has its intended
results. One may conclude that the reduction in the expected yield of coca crops that
results from the threat of eradication is not dissuasive enough for farmers to significantly
reduce the area of land that they allocate to this illegal crop. This provides some
perspective to the current climate of public opinion on the war on drugs. At the very least
in the case of illegal crop eradication through fumigation, the ultimate goal of eliminating
the entire supply of coca remains elusive.

Beyond the main result, an interesting find is that a farmer’s level of education
has a statistically significant, positive impact on the share of land he or she devotes to
coca. Thus, increased education seems to enhance a farmer’s performance as a coca

grower. Another result worth highlighting is that titling to the land seems to be acting as a

59

disincentive to grow coca, presumably by improving access to credit markets as a result
of the ability to use the land as collateral.

Finally, it is worth emphasizing that the environmental and social costs of
fumigation, which have been documented elsewhere, must be taken into account in any
policy decision resulting from the estimates obtained here. In particular, the effects of
glyphosate on the local flora, and the low income of the coca farmers who often have few

sources of earnings aside from coca, must be incorporated into any policy decision.

60

Table 11: Descriptive statistics of the Survey of Agra-Cultural Characteristics of Coca Crops in

 

Colombia
Variable Mean
(s.d)
Male 0.910
Age 38.567
(1 1.643)
No education 0.101
Primary education 0.601
Secondary education 0.285
Some college 0.010
College 0.003
Coca is main source of income 0.823
F arms (“fincas”) per coca producer 1.252
(0.542)
Total area controlled by producer (in hectares) 22.042
(1 12.040)
Area with coca 1.641
(1.833)
Area with fallow land (0.405)
(1.831)
Share of farmland devoted to coca 0.628
(0.316)
Arrobas of coca sold per year 31 1.445
(284.398)
Price per arroba in Colombian pesos 25566.23
(10462.2)
A rrobas of coca per hectare 177.096
(47.540)
Coca producer pays for harvesting wages .768
Most productive age in years of coca bush 2.557
(.666)
Grows coca because of good prices 0.676
Grows coca because of regional custom 0.309
Grows coca because of easy marketing 0.403
Grows coca because there are no other Options 0.527
Grows coca because it is easy to finance 0.019
Has received support to replace coca crops 0.060

Sample size: 1389 coca producers

61

Table 12: Descriptive statistics by farm
Variable

Land is titled

Farm is occupied without a title
Farm is rented

Sharecropping

Other modes of ownership

Sample size: 1740 farms

Source: Author’s calculations from UNODC database

62

Mean

0.525

0.321

0.089

0.048

0.017

Table 13: Descriptive statistics of the municipalities in the sample

 

Variable Mean
(s.d)
Hectares of coca fumi gated in 2004 2134
(2946)
Hectares of coca manually eradicated in 2004 17.763
(42.348)
Municipality area in squared kilometers 3621
(10032)
Share of municipality fumigated in 2004 0.010
(0.0112)
Share of municipality manually eradicated in 2004 0.00033
(0.00139)
FARC combatants 52.047
(71.538)
ELN combatants 14.326
(27.682)
AUC combatants 20.628
(38.743)
Distance to the closest fumigation base in miles 52
(53)

 

Source: Author’s calculations from Anti-Narcotics Police database, cultivation data from Project SIMCI II,
lnstituto Geografico Agustin Codazzi.

63

Table 14: First stage rsgression
(1)

(2)

(3)

(4)

Dependent variable is smre of the municipality
fumigated in 2004

Distance beyond 80 mile radius of a base -0.000181** -0.000113 -0.000332 -0.000209
(0.000086) (0.000182) (0.000277) (0.000371)
Municipality is less than 80 miles from a 0.00273 -0.000143 0.00646
fumigation base (0.00636) (0.00737) (0.00829)
Distance beyond 80 miles of a base * Number 0.000134 0.000139
of left wing combatants (0.000081) (0.000094)
Distance beyond 80 miles of a base * Number -0.000182 -0.000223
of right wing combatants (0.000138) (0.000184)
Municipality is less than 80 miles from 3 -0.000111
fumigation base * Distance from base (0.000124)
Municipality is less than 80 miles from a 0.000106
fumigation base * Distance fi'om base * (0.000085)
Number of left wing combatants
Municipality is less than 80 miles from a -0.000218*
fumigation base * Distance from base * (0.000121)
Number of right wing combatants
Male 0.000293 0.000317 0.000202 0.000003
(0.000712) (0.000683) (0.000683) (0.000592)
Years of schooling -0.000025 -0.000024 -0.000060 -0.000064
(0.0001 1 1) (0.0001 10) (0.0001 12) (0.0001 14)
Has title to at least part of the land -0.000247 -0.000274 -0.000433 -0.000353
(0.000601) (0.000609) (0.000557) (0.000504)
Hundreds of FARC combatants 0.00210* 0.00198 0.00166 0.000236
(0.00122) (0.00130) (0.00140) (0.00182)
Hundreds of ELN combatants 0.00493 0.00425 0.00504 -0.000419
(0.00463) (0.00452) (0.00472) (0.00885)
Hundreds of AUC combatants 0.00870" 0.00837" 0.00799" 0.00940“
(0.00335) (0.00372) (0.00391) (0.00505)
Constant 0.00302 0.000991 0.0041 1 0.001 17
(0.00388) (0.00556) (0.00620) - (0.00871)
Observations 1389 13 89 1389 1389
R-sguared 0.412 0.413 0.420 0.508

 

Robust standard errors clustered at the municipality level in parentheses. Regional fixed effects are

included in the regression.
*** p<0.01, ** p<0.05, * p<0.l

64

Table 15: OLS and IV regressions with share of cultivated land devoted to coca as a dependent
variable

Dependent variable is s_hare of cultivated land devoted to coca

( l) (2) (3) (4) (5)
OLS IV 1V [V IV
Share of municipality 1.098 -13.46 -1 3.53 -7.646 -1.415
fumigated in 2004 (1.362) (10.72) (10.93) (9.918) (3.768)
Male 0.00588 0.0120 0.0120 0.00955 0.00694

(0.0301) (0.0318) (0.0320) (0.0305) (0.0298)

School 0.0157*** 0.0153*** 0.0153*** 0.0155*** 0.0157***
(0.00354) (0.00401) (0.00402) (0.00381) (0.00356)

Has title to at least part ofthe land -0.0714*** -0.0805*** -0.0805*** -0.0769** -0.0730***
(0.0245) (0.0285) (0.0287) (0.0288) (0.0255)

Hundreds of FARC combatants -0.00175 0.0413 0.0415 0.0241 0.00569
(0.0179) (0.0322) (0.0322) (0.0267) (0.0140)
Hundreds of ELN combatants -0.0975 -0.125* -0.125* -0.114* -0.102
(0.0724) (0.0733) (0.0741) (0.0673) (0.0670)
Hundreds of AUC combatants -0.0434 0.114 0.114 0.0509 -0.0163
(0.0363) (0.132) (0.135) (0.1 18) (0.0563)
Constant 0552*” 0640*" 0.640*** 0605*" 0567*"
(0.0884) (0.0983) (0.0993) (0.109) (0.0907)
Observations I389 13 89 1389 1389 1389
R-squared 0.227 0.065 0.063 0.168 0.222

 

The instrument in (2) is distance beyond the 80 mile radius of a base. In (3), an indicator for whether a
municipality is less than 80 miles from a fumigation base is added. In (4), interactions between distance
beyond 80 miles of a base and the number of right wing/left wing combatants are also included.
Specification (5) further adds an interaction between the indicator for distance from a base being less than
80 miles and distance from the base, plus interactions between this variable and the number of right
wing/left wing combatants. All regressions include a constant, year indicators, and indicators for each of
the 7 UNODC coca growing regions. Robust standard errors clustered by municipality are shown in
parentheses.

*** p<0.01, ** p<0.05, * p<0.l

65

APPENDIX A: UNODC ESTIMATES OF COCA CULTIVATION

When the sky is clouded, satellite observations of coca fields cannot detect the
crops that lie below the clouds. Furthermore, sometimes technical difficulties create gaps
in the photographs. To account for clouds and gaps in the images, SIMCI II estimates the
number of coca hectares that are not observed, in the following way. First, coca hectares
that are up to one kilometer away from the perimeter of the cloud (300 meters for a gap in
the picture) are counted. Then, the number of coca hectares in the ring surrounding the
unobserved area is compared to the one from the previous year’s survey. Thus, a rate of
change is obtained, and it is applied to the number of coca hectares observed last year in
the area that is now below the cloud or gap.

A further difficulty is that satellite photographs are taken over a period that spans
five months (September through February). Because eradication of coca crops may take

place after an area is photographed, hectares that are manually eradicated after a picture

. st .
IS taken are subtracted from the Dec. 31 estimate. Sprayed coca bushes have an

estimated survival rate Of 12 percent. Thus, when fumigation takes place after a satellite
photograph is taken, the destroyed area is subtracted accordingly. Finally, SIMCI II

adjusts the data so that the photographs taken over the five month period reflect the

. . st . .
number of hectares of coca that ex13ted In December 31 . It Is unclear how this

adjustment is done. SIMCI II simply states that trends from previous surveys are used to

obtain an adjustment factor.

66

REFERENCES

ANGRIST, J. D., and A. D. KUGLER (2008): "Rural Windfall or a New Resource Curse?
Coca, Income, and Civil Conflict in Colombia," Review of Economics and
Statistics, 90, 1 91-21 5.

ANGRIST, J. D., and J.-S. PISCHKE (2009): Mostly Harmless Econometrics. Princeton:
Princeton University Press.

BESLEY, T. (1995): "Property Rights and Investment Incentives: Theory and Evidence
from Ghana," Journal of Political Economy, 103, 903-37.

BESLEY, T., and R. BURGESS (2000): "Land Reform, Poverty Reduction, and Growth:
Evidence from India," The Quarterly Journal of Economics, 115, 389-430.

BIGWOOD, J., and P. COFFIN (2005): "Coca Eradication" Foreign Policy in Focus,
http://www.fpif.org/reportS/coca_eradication.

CLEMENS, J. (2008): "Opium in Afghanistan: Prospects for the Success of Source
Country Drug Control Policies," Journal of Law & Economics, 51, 407-432.

DAMMERT, A. C. (2008): "Child Labor and Schooling Response to Changes in Coca
Production in Rural Peru," Journal of Development Economics, 86, 164-180.

DE SOTO, H. (2000): The Mystery of Capital: Why Capitalism Triumphs in the West and
Fails Everywhere Else. New York: Basic Books.

DION, M. L., and K. RUSSLER (2008): "Eradication Efforts, the State, Displacement and
Poverty: Explaining Coca Cultivation in Colombia During Plan Colombia,"
Journal of Latin American Studies, 40, 399-421.

FARTHING, L., and B. KOHL (2005): "Conflicting Agendas: The Politics of Development
Aid in Drug-Producing Areas," Development Policy Review, 23, 183-98.

GAO (2008): "Plan Colombia - Drug Reduction Goals Were Not Fully Met, but Security
Has Improved; U.S. Agencies Need More Detailed Plans for Reducing
Assistance," Government Accountability Office.

HAGEN, J. T. (2001): "Alternative Development Won't End Colombia's War" Foreign

Policy in Focus,
http://www.fpif.org/articles/altemative_development_wont_end_colombias_war.

67

HAUGAARD, L. (2001): "A Much Wider War" Foreign Policy in Focus,
http://www.fpif.org/articles/a_much_wider__war.

JOHNSON, R. S., S. KANTOR, and P. V. FISHBACK (2007): "Striking at the Roots of Crime:
The Impact of Social Welfare Spending on Crime During the Great Depression,"
National Bureau Of Economic Research, Inc.

KLICK, J ., and A. TABARROK (2005): "Using Terror Alert Levels to Estimate the Effect Of
Police on Crime," Journal of Law & Economics, 48, 267-79.

KRISTOF, N. (2009): "Drugs Won the War ".

KUROSAKI, T., and M. FAFCHAMPS (2002): "Insurance Market Efficiency and Crop
Choices in Pakistan," Journal of Development Economics, 67, 419-453.

LEVITT, S. D. (1997): "Using Electoral Cycles in Police Hiring to Estimate the Effect of
Police on Crime," American Economic Review, 87, 270-90.

LUNA, A. O. (2007): "Servicio Aéreo Policia Nacional," Bogota.

MEJIA, D., and C. E. POSADA (2007): "Cocaine Production and Trafficking: What DO We
Know?," BANCO DE LA REPUBLICA.

MORDUCH, J. (1995): "Income Smoothing and Consumption Smoothing," Journal of
Economic Perspectives, 9, 103-14.

MORENO-SANCHEZ, R., D. S. KRAYBILL, and S. R. THOMPSON (2003): "An Econometric
Analysis Of Coca Eradication Policy in Colombia," World Development, 31, 375-
383.

MOULTON, B. R. (1986): "Random Group Effects and the Precision Of Regression
Estimates," Journal of Econometrics, 32, 385-397.

PERAFAN, C. C. (1999): "Impacto De Cultivos Ilicitos En Pueblos Indigenas: El CaSO De
Colombia," Bogota: Proyecto de Desarrollo Altemativo (PLANTE).

PETERSON, S. (2002): "People and Ecosystems in Colombia: Casualties Of the Drug War,"
Independent Review, 6, 427-40.

PINTO BORREGO, M. E., A. VERGARA BALLEN, and Y. LAHUERTA PERCIPIANO (2005):
"Costos Generados Por La Violencia Armada En Colombia: 1993-2003,"
DEPARTAMENTO NACIONAL DE PLAN EACION

REUTER, P. (2008): "Can Production and Trafficking Of Illicit Drugs Be Reduced or
Merely Shifted ‘2," The World Bank Policy Research Working Paper Series: 4564.

68

ROSENZWEIG, M. R., and H. P. BINSWANGER (1993): "Wealth, Weather Risk and the
Composition and Profitability Of Agricultural Investments," Economic Journal,
103, 56-78.

ROSENZWEIG, M. R., and K. I. WOLPIN (1986): "Evaluating the Effects of Optimally
Distributed Public Programs: Child Health and Family Planning Interventions,"
American Economic Review, 76, 470-482.

ROUSE, S. M., and M. ARCE (2006): "The Drug-Laden Balloon: Us Military Assistance
and Coca Production in the Central Andes," Social Science Quarterly, 87, 540-
557.

SHORE-SHEPPARD, L. (1996): "The Precision Of Instrumental Variables Estimates with
Grouped Data," Princeton University, Department Of Economics, Industrial
Relations Section.

THOUMI, F. E. (2002): "Illegal Drugs in Colombia: From Illegal Economic Boom to
Social Crisis," The ANNALS of the American Academy of Political and Social
Science, 582, 102-116.

UNODC (2005): "Alternative Development: A Global Thematic Evaluation," New York:
United Nations Office on Drug and Crime.

— (2005): "World Drug Report 2005," New York: United Nations.

— (2006): "Caracteristicas Agroculturales De LOS Cultivos De Coca En Colombia,"
Bogota: United Nations Office on Drug and Crime

— (2009): "World Drug Report 2009," New York: United Nations.

USDS (2004): "International Narcotics Control Strategy Report," USDS, Bureau for
International Narcotics and Law Enforcement Affairs.

VARGAS MANRIQUE, C. E. (2004): "Cultivos Ilicitos Y Erradicacion Forzosa En
Colombia. (Growing Illicit Crops and Their Forced Eradication in Colombia.
With English Summary.)," Cuadernos de Economia Wational University of
Colombia), 23, 109-41.

69

MICHIGAN STATE UNIVERSITY LIBRARIES

ll ll llllllllll

l l

3 1293 03063 8708