The University of New Haven Department of Economics...tactics outfitted with modern weapons. The...

27
The University of New Haven Department of Economics Special studies series no. 1604 Department Special Studies Series are preliminary materials circulated to stimulate discussion and critical comment. The analyses and conclusions set forth are those of the authors and do not necessarily reflect the views of other members of the Department, the College of Business, the University of New Haven or its Board of Governors. Upon request, single copies of the paper will be provided. References in publications to Department Special Studies Series should be cleared with the Individual author to protect the tentative character of these papers. An examination of possible unintended effects of drone activity in Pakistan Kevin Lauber May, 2016

Transcript of The University of New Haven Department of Economics...tactics outfitted with modern weapons. The...

Page 1: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

The University of New Haven

Department of

Economics

Special studies series no. 1604

Department Special Studies Series are preliminary materials circulated to stimulate discussion and critical comment. The analyses and conclusions set forth are those of the authors and do not necessarily reflect the views of other members of the Department, the College of Business, the University of New Haven or its Board of Governors. Upon request, single copies of the paper will be provided. References in publications to Department Special Studies Series should be cleared with the Individual author to protect the tentative character of these papers.

An examination of possible unintended

effects of drone activity in Pakistan

Kevin Lauber

May, 2016

Page 2: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

AN EXAMINATION OF POSSIBLE

UNINTENDED EFFECTS OF DRONE

ACTIVITY IN PAKISTAN

Kevin Lauber*

Abstract

The US drone program in Pakistan started in 2004 in order to counter

Pakistani insurgent militant activity in Pakistan. The program has been

both criticized and hailed extensively in the years since it began. To fully

appraise the pros and cons of the program, it is important to establish

whether the drone strike activity has had any secondary impacts, especially

unintended ones. It is distinctly possible that the use of unmanned aerial

vehicles depress regional economic activity and thereby contribute to an

environment of deprivation and uncertainty. The resulting bleak

circumstances could perhaps foster more insurgent activity. In this paper

we examine whether drone strike activity has had an impact on the economic

circumstances in Pakistan. Specifically, we construct an Economic Misery

Index to account for economic circumstances. We specifically test the

hypothesis of a relationship between drone-strike activity and the Economic

Misery Index. We carefully account for possible endogeneity between

drone-strikes activity and economic circumstances.

Based on available applicable data collected for the region we find

statistically significant support for the hypothesis. The results should be

interpreted with caution; still, they remain suggestive of a productive line of

inquiry ahead.

May 2016

* Department of Economics, University of New Haven; Email: [email protected]

Page 3: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 2 of 25

Introduction

In 2004, Pervez Musharraf, Pakistan’s President, gave the approval of

the use of drones in Waziristan, Pakistan as terrorism and violence

spread uncontrollably throughout the region (Coll, 2014). A drone, or

Unmanned Aerial Vehicle (UAV), is a remoted piloted aircraft used

by the U.S. military and intelligence agencies to conduct intelligence,

surveillance, and reconnaissance purposes (ISR); and for combat

operations (Callam, 2010). Its primary objective in Pakistan’s

Federally Administered Tribal Areas (FATA) is to counter terrorist

militant activity in the region with surveillance and airstrikes (Callam,

2010).

There is a level of discretion by certain authorities over what

constitutes terrorism. Since the mid-19th century, it seems to have

been a fluid concept as one seeks a standard for accurate classification

(Whitaker, 2001). Even the U.N has had problems coming to a

consensus. (Schmid, 2004).

The Central Intelligence Agency and the United States Department of

State share a consensus on the definition of terrorism as it is defined

in the Code of Laws of the United States of America. For the purpose

of continuity as we examine drone strikes conducted by the Central

Intelligence Agency, this paper recognizes these parameters for

terrorism as well. Taken from Title 22 of the US Code, Section

2656f(d):

Page 4: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 3 of 25

The term "terrorism" means premeditated, politically motivated

violence perpetrated against noncombatant targets by subnational

groups or clandestine agents (Congress)

The program has been both criticized and hailed by terrorism and

intelligence experts alike. Critics believe it has unintended

consequences prompting terrorism to continue to exist and perhaps

even thrive. Studies on the matter have shown a positive change in

terrorist attacks immediately following manned air strikes and shows

of force in areas like Afghanistan (Lyall, 2014), while other evidence

suggests unmanned drones have a mitigating effect on terrorist

activities in Pakistan, Afghanistan’s neighboring country (Johnston

and Sarbahi, 2014). The results vary from circumstance to

circumstance.

In World War I, trench warfare retained remnants of dated Napoleonic

tactics outfitted with modern weapons. The result was extremely

bloody and sometimes reckless disregard for casualties. Today, we

face the same challenges as advances in technology allow for us to

conduct warfare in ways never previously thought possible. It is

important to ensure that history does not repeat itself and to account

for all consequences of such a tool of warfare which even today is still

in its infancy. This paper attempts to examine whether drone strikes

have an active impact on the economic environment and speculate as

to the likely consequences. The costs of war and the effectiveness of

Page 5: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 4 of 25

our tactics cannot be quantified in the same ways as they used to.

Studies have concluded terrorism to be related to economic instability

(Choi, 2014), and perhaps we can identify the effects of US Drone

strikes on economic stability and thusly on terrorism in the area.

Given the relevance and sensitivity around the issue of drone strikes,

it is important to clarify the scope of this study. This paper

hypothesizes drone strikes to contribute to economic stagnation in

Pakistan. This may be an indicator that the use of unmanned aerial

vehicles does have unintended consequences, fostering an

environment in which terrorism continues to occur. With so many

variables and with the classified nature of these operations, conclusive

analysis with accurate and unbiased data is all but impossible. Years

of study is needed to come to consensus on whether drone strikes are

being employed properly today, so that we can be a more effective

global force for the future.

The paper does not address the moral questions surrounding the drone

program. Nor does it address the broader questions regarding the

soundness of American military involvement in a foreign country.

These are important questions for further work. Still, an important

objective of this study is to address the effectiveness of the drone

program in a narrow field – which has ready implications for broader

morality of the initiative.

Page 6: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 5 of 25

Review of the Literature

With every new study analyzing airstrike data comes more questions

than answers. Jason Lyall’s (2015) analysis on bombings in

Afghanistan concludes that, within 90 days of an airstrike, instances

of terrorism in an area increase in relation to before the airstrike, and

in relation to areas where an airstrike did not occur. Furthermore, he

concludes that civilian casualties are not the major explanatory factor.

This is in contrast to Johnston and Sarbahi’s (2014) Pakistan

conclusion that unmanned drone strikes play a role in mitigating

terrorism. The obvious differences here are the locations (Pakistan

versus Afghanistan), and manned versus unmanned methods of aerial

bombings. Jaeger and Siddique’s (2011) study analyzed drone strikes

in Pakistan and monitored the weeks following a drone strike for

terrorist activity. They found a decrease in terrorist activity following

unsuccessful drone strikes (strikes that did not kill a high value

leadership target) but a strong positive impact on terrorist activity

following successful drone strikes. This indicates what they call a

“vengeance effect,” and that drone strikes have a strong deterrent

effect, but when a strike is thought to incapacitate through elimination

of leadership, it has quite the opposite effect (Jaeger and Siddique,

2011).

This paper seeks to determine whether there are any economic

impacts as a result of drone strikes in the area. Choi (2014) finds that

positive economic indicators are “are less disposed to domestic and

Page 7: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 6 of 25

international terrorist events, but are more likely to experience suicide

attacks.” Here, we see one type of attack decreasing while another

increases. The variables multiply.

The drone strike data used in this paper will be from the gatherings of

the Bureau of Investigative Journalism. There are other sources which

offer conflicting data, however. The Long War Journal, and The New

America Foundation have their own compiled drone strike casualty

data (Data Team, 2015). They seem to be closely correlated in the

total number of drone strikes and total casualties, but their reports on

civilian casualties vary greatly. This is testament to the elusive nature

of reliable data in one of the world’s most dangerous and elusive

frontiers.

Empirical Model

To test the proposed hypothesis we rely on a linear multiple regression

model:

Y = β0 + β1X1 + β2X2 + μ (1)

Where Y is the dependent variable – the Economic Misery Index, X1

is the explanatory variable, drone strikes, and X2 is another control

variable, in this case, a Corruption indicator; the last term is an error

term (μ).

Page 8: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 7 of 25

Under appropriate conditions, we anticipate an inverse relationship

between drone-strike activity and the economic misery index (created

for the purpose of this study). However, it is conceivable that there

exists endogeneity in the formulation of the above model. While there

may be an impact of drone strike activity on the regional economic

environment, there may simultaneously arise a reverse effect;

specifically, an effect proceeding from declining economic activity to

drone strikes. Again, to the extent that the thesis advanced here is

correct, it would follow that a depressed economic environment may

foster the conditions leading or compelling individuals or groups

towards enhanced terrorist activity, thereby increasing retaliatory

drone strikes.

In the presence of endogeneity, Ordinary Least Squares is known to

result in biased and inconsistent parameter estimates. The problem

may be addressed with suitable instrumental variables.

If there a set of variables Z correlated with X but not correlated with

the error term μ so that E(Z,μ) ≠ 0 and E(Z,X) = 0 then the variables

are called instrumental variables. A class of instrumental variable

methods can then be used to consistently estimate the impact of X on

Y. Preferably, the Z variables should be as highly linearly correlated

as possible with the X variable. Modest correlations between Z and X

result in weak instruments; we will have more to say on these below

(Murray, 2006).

Page 9: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 8 of 25

Constructed Instruments

The fundamental idea in constructing an instrument is to modify the

available endogenous variable in a manner that retains the signal or

pattern present in the data but leaves out some portion of noise so as

to reduce the endogeneity with the error term. A Wald-type grouping

variable is used here as a constructed estimator.

Wald described a method that did not make an assumption about the

error structure (Wald, 1940). Theil suggests two variants (Theil,

1971). Both approaches first order the observed pair (x,y) sorted on

the values of x, then divide the observations into a specific number of

groups. We examine two variants of Wald’s grouping method.

Wald’s grouping method is equivalent to using the following

instrumental variable:

ZWald 1 = {

1 𝑖𝑓 𝑥 > 𝑚𝑒𝑑𝑖𝑎𝑛(𝑥1, 𝑥2, . . . , 𝑥𝑛)

−1 𝑖𝑓 𝑥 < 𝑚𝑒𝑑𝑖𝑎𝑛(𝑥1, 𝑥2, … , 𝑥𝑛)

0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

Bartlett’s variant of Wald’s grouping method is equivalent to the

following (Bartlett, 1949):

Page 10: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 9 of 25

ZWald 2 = {

1 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑙𝑎𝑟𝑔𝑒𝑠𝑡𝑁

3 𝑜𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛𝑠

−1 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑠𝑚𝑎𝑙𝑙𝑒𝑠𝑡𝑁

3 𝑜𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛𝑠

0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

Data

The statistics regarding US drone strikes was extracted from the

Bureau of Investigative Journalism, a non-profit, philanthropically

funded news agency based in London. The covert drone war is one of

their major investigations. The data reveals the beginning of the drone

strike strategy in 2004 with strikes totaling less than ten in each of the

years 2004, 2005, 2006, and 2007 respectively. Strikes increased

dramatically from five in 2007 to 38 in 2008. The data shows a

continued increase to 128 total strikes in 2010 before a steady

decrease to 2016 where the observable data ends. (Bureau of

Investigative Journalism, 2016)

Pakistan’s economic data comes from the World Bank, a multilateral

organization based in Washington DC dedicated to ending extreme

world poverty and promoting shared prosperity. The data shows a

positive and steady increase in drone activity from 2004 until 2008. It

also displays a point of inflection from 2008 to 2009.

To represent the “economic environment” we construct an economic

misery index. The index is a sum of the normalized value of each of

three economic variables. The variables were normalized using the

following method, for each Xi:

Page 11: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 10 of 25

Xi_norm = (X – Xmin)/(Xmax – Xmin)

Where i = 1, 2, 3

ECON Index = Σ (X1 + X2 + X3)*100

The index is a equally-weighted average of the following three

economic variables: Unemployment Rate, GDP per capita, the

change in the volume of exports of goods and services.

We use Gross Domestic Product per capita as a measure of economic

performance. Pakistan’s GDP and GDP per capita experienced

negative growth before correcting in 2010 and returning to its original

rate of growth from 2011 onward (World Bank, 2016). A graphical

representation shows an apparent correlation between drone strikes

and GDP per capita. Correlation does not constitute causality,

however. Other factors must be observed.

An annual variable representing the degree of corruption is obtained

from Transparency International Global Corruption Barometer

Survey available online at the World Bank’s Governance Indicators

Database. Data encompass years: 2004:2014. Estimates of

corruption for 2015 and 2016 were obtained from a linear projection

of existing data (World Bank, 2016)

Page 12: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 11 of 25

The period examined here witnessed the global recession. A dummy

variable is included in the regression to account for any independent

impact of the recession on the effects measured, whereas the years in

which the global recession was present is represented with a

numerical value of 1, and the years in which the recession was not

present, 0.

The formal model is an instrumental variables regression of the

Economic Misery Index on Drone Strikes and Corruption Index over

the period 2004-2016. As appropriate instruments for the Drone

Strike Activity variable we rely on the constructed instrument (Wald

or Bartlett) as well as the Corruption Index and the Recession

Dummy.

GRAPHS

Page 13: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 12 of 25

Results

Page 14: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 13 of 25

The results of an ordinary least squares of the multiple regression

model as well as the results of the instrumental variable procedure are

reported here.

Multiple Regression Results

==========================================

Dependent variable:

---------------------------

econindex

-----------------------------------------------

strikes -0.162

(0.179)

corruption 0.453

(0.999)

recession -0.260*

(0.136)

Constant 0.301

(0.507)

-----------------------------------------------

Observations 13

R2 0.519

Adjusted R2 0.359

Residual Std. Error 0.125 (df = 9)

F Statistic 3.243* (df = 3; 9)

==========================================

Note: *p<0.1; **p<0.05; ***p<0.01

Page 15: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 14 of 25

As expected, the results provided by ordinary least squared are biased

an unreliable.

IV Regression Results

================================================

Dependent variable:

----------------------------

econindex

(Wald) (Bartlett)

----------------------------------------------------------

strikes -0.408* -0.535**

(0.200) (0.217)

corruption -1.004 -1.337

(0.863) (0.919)

Constant 1.034** 1.228**

(0.455) (0.486)

----------------------------------------------------------

Observations 13 13

R2 0.318 0.247

Adjusted R2 0.182 0.096

Res Std. Error (df = 10)

0.142 0.149

================================================

Note: *p<0.1; **p<0.05; ***p<0.01

The results of the instrumental variable algorithm are reported above.

The column labeled Wald reflects the use of the Wald version of the

constructed estimator. The column labeled Bartlett reflects the

Bartlett version of the constructed estimator.

Page 16: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 15 of 25

The results are startling, revealing a highly significant relationship

between drone activity and the Economic Misery Index.

Concluding Comments and Further Work

The empirical results suggest a statistically significant relationship

between drone-strike activity and the economic misery index once

we account for the possible endogeneity between drone-strikes and

the economic misery index. However, while supportive of the

general thesis advanced here, the results should be interpreted with

caution. There are multiple variables left unexamined and numerous

avenues of causality. Importantly, the thesis might be incorrect and

the results reported here are picking up spurious results.

A more complete model would acknowledge other variables.

Variables, ranging from the largesse of foreign sympathizers to the

intensity of domestic military efforts, to the incidence of corruption,

to the effectiveness of government programs acting as bulwarks to

terrorist activity, to education variables, and others.

An obvious limitation with the data is that of the small numbers bias,

whereas the unavailability of a large number of observations can

give inaccurate results. Unfortunately, we are limited only to the

data available to us. While there does exist monthly data for drone

Page 17: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 16 of 25

strikes, Pakistan’s economic data is only available yearly. This gives

us only 12 observations between 2004 and 2016. For this reason, the

results of this thesis should be approached with caution, as

previously mentioned. Nevertheless, the model shows positive

results, and thusly an implication that further research into this study

is needed as time allows for more observation.

When approaching an analysis of the data, it is also important to

understand that the instances of drone strikes are restricted to a

particular region, and the economic data is that of the entire nation

as a whole. However, the results suggest an economic spillover

effect of the regional drone activity into the rest of the nation.

Speculatively, this is perhaps due to apprehension of potential

foreign investors, or the emigration of wealth from Pakistan itself.

Political instability and surreptitious warfare may dramatically

heighten risk for individuals and businesses seeking a community in

which to invest. Again, this is speculative, but the intention of this

study is to shed light on the possibilities of these effects occurring.

Policy implications

There are significant policy implications in this field of study. The

United States drone policy in Pakistan and in other areas is in effect

specifically to deter the insurgency of militant groups who intend to

inflict harm upon surrounding communities and around the world.

Their ability to operate being dependent on a stagnant local and

Page 18: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 17 of 25

national economy is highly suggestable. If drone strikes have a

significant enough impact on the economy such that it provides the

conditions in which militant activity can continue and even thrive,

then the whole purpose of the United States covert drone policy is

thereby defeated. A new or perhaps altered approach may be

required.

References

Bound, J., Jaeger, D. A., & Baker, R. M. (1995). Problems with

instrumental variables estimation when the correlation

between the instruments and the endogenous explanatory

variable is weak. Journal of the American statistical

association, 90(430), 443-450.

Bureau of Investigative Journalism (2016). CIA Drone Strikes in

Pakistan, 2004 to Present [Data File]. Retrieved from

https://www.thebureauinvestigates.com/category/projects/dro

nes/drones-graphs/

Page 19: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 18 of 25

Callam, A. (2010). Drone Wars: Armed Unmanned Aerial Vehicles.

International Affairs Review. Retrieved from http://www.iar-

gwu.org/node/144

Choi, S. W. (2015). Economic growth and terrorism: domestic,

international, and suicide. Oxford Economic Papers, 67(1),

157-181.

Coll, S. (2014). The Unblinking Stare. The New Yorker. Retrieved

from

http://www.newyorker.com/magazine/2014/11/24/unblinking

-stare

Congress (2013). 22 U.S. Code § 2656f - Annual Country Reports

on Terrorism. Ithaca, New York: Cornell Law School.

Data Team (2015). Drone Strikes: Cause or Effect. The Economist.

Retrieved from

http://www.economist.com/blogs/graphicdetail/2015/09/dail

y-chart-drone-attacks-and-terrorism-pakistan

Johnston, P. B., & Sarbahi, A. (2012). The impact of US Drone

Strikes on Terrorism in Pakistan. Santa Monica, California:

Rand Corporation.

Lyall, J. (2014) Bombing to Lose? Airpower and the Dynamics of

Violence in Counterinsurgency Wars. New Haven, CT: Yale

University.

Murray, M. P. (2006). Avoiding invalid instruments and coping with

weak instruments. The journal of economic perspectives,

20(4), 111-132

World Bank (2016). World Development Indicators [Data File].

Retrieved from http://data.worldbank.org/country/pakistan.

Page 20: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 19 of 25

Schmid, A., A. (2004) Terrorism – The Definitional Problem. Case

Western Reserve Journal of International Law. 36 (2)

Whitaker, B. (2001). The Definition of Terrorism. The Guardian.

Retrieved from

http://www.theguardian.com/world/2001/may/07/terrorism.

Page 21: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 20 of 25

Appendix

Descriptive Statistics

====================================================

Statistic N Mean St. Dev. Min Max

----------------------------------------------------

unemp 13 6.2 0.8 5.2 7.7

gdppercap 13 51,919.9 3,370.6 44,717 57,465

change.exports 13 4.2 4.9 -2.4 16.5

corruption 13 0.5 0.1 0.4 0.6

Wald 13 0.0 1.0 -1 1

Bartlett 13 0.0 0.8 -1 1

strikes 13 32.5 37.3 1 128

recession 13 0.0 1.0 0 1

----------------------------------------------------

Diagnostic Tests

Diagnostic tests:

df1 df2 statistic p-value

Weak instruments 2 9 7.852 0.0106 *

Wu-Hausman 1 9 2.250 0.1679

Sargan 1 NA 1.447 0.2290

---

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’

1

A weak instrument is one with a low correlation with the

endogenous explanatory variable. This could result in a coefficient

with a much larger variance, and thusly severe finite-sample bias.

"The cure can be worse than the disease" (Bound, Jaeger, Baker,

1993/1995). For our model the null is rejected, so we can move

Page 22: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 21 of 25

forward with the assumption that the instruments are sufficiently

strong.

A Wu-Hausman test is one which tests whether the Instrumental

Variable is as consistent as OLS, and since OLS is more efficient, it

would be preferable. The null hypothesis is that they are equally

consistent; in this output, Wu-Hausman is significant at slightly

more than the p =0.1 level.

The Sargan Test tests whether the model is overidentified, meaning

there is more than one instrument per endogenous variable, and

thusly some excess information. In order for the inferences to be

correct, all of the instruments must be valid. Simply put, the Sargan

Test tests whether exogenous instruments are in fact exogenous, and

uncorrelated with the model residuals. If it is significant, then we do

not have valid instruments (somewhere in there, as this is a global

test). In this case, this isn't a concern.

Page 23: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 22 of 25

This is a copy of the code used to obtain the result reported here.

# Kevin Lauber

#

# Thesis code/ Spring 2016

#

# UNH Economics

#

########## ##############################

#

remove(list =ls()) # tidy-up

#set working directory

dir()

# Input data

drone = read.csv("kevin.csv", header = TRUE, stringsAsFactors = FALSE)

str(drone)

View(drone)

names(drone)

drone$strikes = drone$CIA.Drone.Strikes

attach(drone)

# Initialize libraries

library(forecast) # for the Time Series

library(lmtest) # for Granger Causality

library(urca) # Dickey Fuller Tests

library(AER) # for instrumental variables regression

library(stargazer) # for printing output properly formatted

drone

names(drone)

## Plots

strikes = ts(strikes, c(2004))

unemp = ts(unemp, c(2004))

change.exports= ts(change.exports, c(2004))

gdppercap = ts(gdppercap, c(2004))

corruption = ts(corruption, c(2004))

Page 24: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 23 of 25

plot.ts(strikes)

plot.ts(unemp)

plot.ts(corruption)

# Plot two series

par(new = F)

plot(strikes, type='l', xlab='Year', ylab='strikes/unemp', col ="red")

par(new=T)

plot(unemp, type='o', xlab='', ylab='', axes=F)

par(new=F)

# Normalize the data and create Econ Index

normalize = function(x) {

temp = ((x -min(x))/(max(x) - min(x)))

return(temp)

}

names(drone)

drone_norm = lapply(drone[,2:6], normalize)

drone_norm = as.data.frame(drone_norm) #convert

strikes = drone_norm$CIA.Drone.Strikes

summary(strikes);

### Create Economic Index

# The index is the equally-weighted average of the normalized economic

variables.

names(drone)

attach(drone_norm)

econindex =

(drone_norm$unemp+drone_norm$gdppercap+drone_norm$change.export

s)/3

econindex = ts(econindex, c(2004))

Page 25: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 24 of 25

plot.ts(econindex)

plot.ts(strikes)

par(new = F)

plot(strikes, type='l', xlab='Year', ylab='Strikes/Econ Index', col

="red")

par(new=T)

plot(econindex, type='o', xlab='', ylab='', axes=F)

par(new=F)

################ Multivariate Regression

model_mvr_1 = lm(econindex ~ strikes + corruption+recession)

summary(model_mvr_1)

### Outputting Multivariate Regression Results

stargazer(model_mvr_1, type = "text", out="mvrreg.txt",

title="Multivariate Regression Results")

## Accounting for Endogeneity Using IV Reg package

Model_1_IVreg = ivreg(econindex~strikes + corruption| Wald+

corruption +recession)

summary(Model_1_IVreg, vcov = sandwich, diganostics =TRUE)

Model_2_IVreg = ivreg(econindex~strikes + corruption|

Bartlett+corruption+recession)

summary(Model_2_IVreg, diagnostics = TRUE)

### Outputting Instrumental Variables Results

Page 26: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

Page 25 of 25

stargazer(Model_1_IVreg, Model_2_IVreg, type = "text",

out="ivreg.txt", title="IV Regression Results")

### Printing Summary Stats

stargazer(drone, type = "text", title = "Descriptive Statistics",

digits = 1, out = "table1.txt")

Page 27: The University of New Haven Department of Economics...tactics outfitted with modern weapons. The result was extremely bloody and sometimes reckless disregard for casualties. Today,

University of New HavenDepartment of EconomicsSpecial Studies Series

No. Title Author Date1605 Predictive analytics in support of a retail

office site selection decision-making process Michael Parlato May 20161604 An examination of possible unintended

effects of drone activity in Pakistan Kevin Lauber May 20161603 Immigration and growth of GDP Howard Wayne May 2016

in the United States of America McGruder1602 Comparing the performance of classification &

regression trees and multiple regression in theappraisal of crime rates James Callan May 2016

1601 Macroeconomic uncertainty and the Olympics Henry Adegunle May 20161507 How low can you go? The performance of

a google-search enhanced forecast of employmentin a small region Dalton White May 2015

1506 Cigarettes: alcohol’s complement Ashley Leblanc May 20151505 An analysis of regression and classification

Models in predicting of probability of bankruptcy Elizabeth Jaikes May 20151504 Impact of climate change on wheat

production in Kansas Joshua Howard May 20151503 Crude oil price and inflation in the

United States: 1980-2012 Alan A.D’Auria May 20151502 Microfinance and its effectiveness

on developing nations Josh Apfel May 20151401 Government size and economic growth:

Evidence from selected OECD countries Alice Aleksandrovich May 2014

____________________________________________________

Department of Economics Special Studies Series are materials circulated to stimulate discussion andcritical comment. The analyses and conclusions set forth are those of the authors and do not necessarilyreflect the views of other members of the Department, College of Business, the University of New Havenor its board of Governors. Upon request, single copies of the paper will be provided. References inpublications to Department Special Student Series should be cleared with the individual author to protectthe tentative character of these papers.

The series editor is Professor Esin Cakan. You may contact her via email at [email protected] write to her at:

Department of EconomicsCollege of Business

University of New HavenWest Haven, CT 06516

The Department Senior Thesis Series is administered by the Editor, Esin Cakan and benefits from thecommentary and direction of its board members: Professors A.E. Rodriguez and Kamal Upadhyaya.