<<

THE AMERICAN PARAMILITARY UNIT:

AN ANALYSIS OF SWAT AS AN AGGRESSIVE POLICING STRATEGY

by

Brian P. Mello

A Thesis Submitted to the Faculty of

The College for Design and Social Inquiry

In Partial Fulfillment of the Requirements for the Degree of

Master of Science

Florida Atlantic University

Boca Raton, Florida

August 2016 Copyright 2016 by Brian P. Mello

ii

ACKNOWLEDGEMENTS

The author wishes to thank his committee board members for all their help and guidance throughout every step of this extensive process. I would like to especially thank

Dr. Adam Dobrin for his advice, professional insight and continuous support along the way. His genuine interest in the subject and critical thinking helped make this thesis the best it could be.

I am sincerely thankful for all the additional time Dr. Rachel Boba Santos put in throughout this past year. Without her help in creating the conceptual foundation of this research study and her Advanced Research Methods course, this thesis would not have been half as good. Additionally, I would like to thank Dr. William Hauser for his unparalleled assistance throughout the data analysis section. His comprehensive emails and statistical knowledge helped create the best possible analysis of the final thesis database.

I would like to thank my parents for their steadfast support throughout this

Master’s program and thesis development. I genuinely thank them for providing me the opportunity to earn this degree.

iv

ABSTRACT

Author: Brian P. Mello

Title: The American Police Paramilitary Unit: An Analysis of SWAT as

an Aggressive Policing Strategy

Institution: Florida Atlantic University

Thesis Advisor: Dr. Adam Dobrin

Degree: Master of Science

Year: 2016

This research study examines the American paramilitary units know as Special

Weapons and Tactics. Throughout this thesis, the fundamental and foundational research on police paramilitary units is presented and discussed. Additionally, this paper focuses on trends of in American policing. Mainly, these trends are identified by the tremendous rise and normalization of American paramilitary police units since the internal societal of the 1970s. In this study, research is presented on Special

Weapons and Tactics team deployments in the state of Maryland. Four years of data are analyzed focusing on proactive search warrant crime deployments from 2010 through

2013. Several independent variables including, violent crime rates, property crime rates, vice type crime rates and the number of sworn law enforcement officers are examined.

The most important finding of this study is the inverse relationship between proactive vice type arrests and the proactive search warrant SWAT team deployments.

v

THE AMERICAN POLICE PARAMILITARY UNIT:

AN ANALYSIS OF SWAT AS AN AGGRESSIVE POLICING STRATEGY

LIST OF TABLES ...... x

CHAPTER 1: INTRODUCTION ...... 1

Background ...... 1

Problem Statement ...... 2

Thesis Goal ...... 3

Summary ...... 4

CHAPTER 2: LITERATURE REVIEW ...... 7

Introduction ...... 7

History of the Militarization of American Police ...... 8

Historical Evidence ...... 8

Transformation of Law Enforcement ...... 11

Consequences of Militarization ...... 12

The Development of SWAT ...... 13

SWAT Origins ...... 14

Waco, Texas ...... 15

Governmental Funding ...... 16

Implementation of Proactive Paramilitary Officers ...... 16

Patrol Paramilitary Policing...... 17

Police Paramilitary Units ...... 18 vi

Consequences of Militarization ...... 19

Review of the Research ...... 19

Militarism and Militarization...... 20

National Police Paramilitary Unit Activity ...... 21

Characteristics of Police Paramilitary Unit Personnel ...... 23

Modern Police Paramilitary Units ...... 25

The Need for Police Paramilitary Units ...... 26

Conclusions ...... 27

Research Questions...... 29

Research Question 1 ...... 29

Research Question 2 ...... 29

Research Question 3 ...... 29

Research Question 4 ...... 30

CHAPTER 3: METHODOLOGY ...... 31

Introduction ...... 31

Data Collection ...... 31

Maryland SWAT Team Reports ...... 31

Crime in Maryland Reports ...... 33

The United States Census Bureau ...... 34

Database ...... 34

Variables ...... 37

Dependent Variables ...... 38

Independent Variables ...... 39

vii

Analytic Strategy ...... 41

Mixed Linear Models ...... 42

Fixed Effects Models and Random Effects Models ...... 42

CHAPTER 4: DATA ANALYSIS ...... 47

Introduction ...... 47

Descriptive Statistics ...... 47

Descriptive Tables ...... 48

Measures of Central Tendency ...... 52

Research Question 1 ...... 55

Research Question 2 ...... 58

Research Question 3 ...... 60

Research Question 4 ...... 63

Summary of Findings ...... 66

CHAPTER 5: DISCUSSION & CONCLUSIONS ...... 69

Discussion of Findings ...... 69

Policy Implications ...... 71

Limitations ...... 74

Recommendations for Future Research ...... 77

Conclusion ...... 80

BIBLIOGRAPHY ...... 82

viii

LIST OF TABLES

Table 1: Frequency and Percent of Type of Jurisdictions ...... 48

Table 2: Sum of Agencies Part I and Part II SWAT Deployments ...... 49

Table 3: Yearly Part I SWAT Deployments by Agency Type ...... 51

Table 4: Yearly Part II SWAT Deployments by Agency Type ...... 52

Table 5: Measures of Central Tendency ...... 53

Table 6: Part I Crime Rates Random Effects Model ...... 56

Table 7: Part II Arrest Rates Random Effects Model ...... 59

Table 8: Part I Crime Rates Fixed Effects Model ...... 62

Table 9: Part II Arrest Rates Fixed Effects Model ...... 65

ix

CHAPTER 1: INTRODUCTION

Background

Following the initial creation of police paramilitary units in the United States numerous factors have been linked to their tremendous rise in popularity. Ranging from influential historical events, enacted legislation, policies and procedures, American

policing has transitioned into a more militarized entity (Hill & Berger, 2009; Huq &

Muller, 2008). Militarization is the implementation of a belief that show of force and

threat of violence is the most appropriate way to solve problems (Kraska, 2007).

Most notably, the rise and normalization of police paramilitary units in American

society has been attributed to the domestic drug wars initiated in the 1970s (Andreas &

Price, 2001; Huq & Muller, 2008; Kiker, 2015; Kraska, 1999). President Richard Nixon’s

on Drugs” created a need for extreme police paramilitary unit usage in America today. The war on drugs, along with other internal societal “wars,” required law enforcement officials to serve a substantial number of search warrants and arrest warrants

(ACLU, 2014; Kraska & Kappeler, 1997). As a result of the dramatic rise in search warrants for American homes, law enforcement officials needed specialized police units that were highly trained.

By that time the Los Angeles Police Departments’ Special Weapons and Tactics team had already gained national recognition through their highly publicized deployments (Weber, 1999). Following that, police paramilitary programs began being 1

implemented nationwide and after the initiation of the drug wars their popularity and

activity skyrocketed.

Over the past fifty years since the war on drugs, the United States government has

allocated substantial funding and given millions of grade equipment to local

police departments across the nation (ACLU, 2014; Weber, 2014). These materials range

from M16s, grenade launchers and to even armored personnel carriers (Weber, 1999).

Politicians, policy makers and legislation have influenced police peace officers to adopt these militaristic ideals researchers have identified today.

Consequently, the numerous domestic societal “wars” have had many negative

impacts on this nations criminal justice system. One of the most overlooked outcomes is

the rising militarization of policing practices. Police paramilitary programs have been

implemented at every agency level of American policing, and yet little research has been

done to study their impacts, effectiveness or relationship to crime trends. There is

significantly a lack of empirically based research on the programs.

Problem Statement

In terms of studying societal movements such as the war on drugs, criminal

justice researchers typically focus on the changes in sentencing offenders, mass

incarceration trends, or their impacts on the courts. However, before an offender can be

processed by the judicial system, law enforcement officials must first arrest the offender.

The war on drugs pressured law enforcement officers to become more aggressive

towards criminals (Weber, 1999). Militaristic language such as a “war” on drugs pegs

offenders as the adversary and societal enemies of American police officers (Weber,

1999). As a result of this militaristic language and war against criminals, police

2 departments and law enforcement entities across the nation have turned to more of a military style approach.

Police paramilitary units have been implemented into the overwhelming majority of agencies across the nation (ACLU, 2014; Kraska & Cubellis, 1997; Kraska &

Kappeler, 1997). Some of these law enforcement entities have even modified the initial reactive roles of these programs to a new proactive police paramilitary patrol unit. These units are identified as police officers in military style battle dress and semiautomatic rifles, which are now seen performing routine police, work in certain agencies across the nation (Kraska & Cubellis, 1997; Kraska & Kappeler, 1997). Several academics have even labeled these officers as “warrior cops” (Balko, 2014; Kiker, 2015;

Kraska, 2007; Weber, 1999).

Militaristic ideals within a police department can have numerous positive effects on organizational, operational and cultural characteristics. Over militarization in an

American civilian police force can lead to severe consequences for police and the citizens

(Herzog, 2001). Police paramilitary units are called-out for proactive warrant work over sixty percent of the time (ACLU, 2014; Kraska & Cubellis, 1997; Kraska & Kappeler,

1997) and have the ability to infringe on constitutional rights of homeowners and threaten the safety of innocent citizens (ACLU, 2014). Proactive uses of these elite police paramilitary units increase the dangerousness of police work and can influence personnel to be combative towards potential offenders. Research is needed to build empirical evidence in support of these claims.

Thesis Goal

Militarization research on American policing is very limited. Police paramilitary

3 units are viewed as the elite special operations teams of police departments and have a certain degree of secrecy. Law enforcement agencies are protective of their tactics, operations and results. Mainly this is the reason for the lack of empirical studies in this field of study.

Not only is research of militarization trends in American policing limited, they are also quite dated. Many of the most important research studies on these units are from the late 1990s and early 2000s. The majority of these articles only document the tremendous rise and popularity of police paramilitary units across the nation, or their shift to proactive deployments. Only a few actually study statistical relationships and conduct data analysis looking for trends, where the rest focus on qualitative characteristics of police paramilitary unit personnel.

The goal of this thesis is to conduct exploratory research on the topic of SWAT teams and their statistical tendencies. Identifying empirical relationships between police paramilitary units and crime trends. Additionally, the statistical analysis of this thesis will include certain agency characteristics such as their number of sworn law enforcement officers and the type of jurisdiction. The purpose of this thesis is to contribute an empirical study that identifies statistical relationships between Special Weapons and

Tactics teams in Maryland, to the crime trends and jurisdictions attributes.

To conclude, this thesis aims to present implications for future research and practical foundations in the field of SWAT programs. Significant recommendations and conclusions will be offered in this paper.

Summary

4

Today the overwhelming majority of law enforcement agencies across the nation

have specialized units that emulate militaristic ideals. Even though special weapons and

tactics programs are popular in America, little evidence supports any claims of their

effectiveness. The impacts of these units are also unknown on the communities in which

they operate. Police paramilitary units were initially implemented as a reactive function

of an agency and now are used as the perfect tool for agencies to serve proactive search

warrants and arrest an incredible amount of drugs users in their homes.

Following this chapter, this thesis will present and review academic journal

articles on the rise and normalization of militarized police units. It will provide a

complete literature review on the most cited articles in the study of police paramilitary

units. These research articles are the most influential and will address the historical

evidence and legislation that led to their rise in popularity.

Additionally, chapter two will present the creation of special weapons and tactics

teams in America and describe their initial reactive functions in society. Chapter two then

presents the empirical evidence of the rise and normalization of police paramilitary units

in America and highlights their new proactive functions. Qualitative studies of police

paramilitary unit personnel will be present in chapter two as well. The conclusion of chapter two will focus on the need for specialized units in policing and the impacts that these units have in American society.

Chapter three will describe the final thesis database. This section will present the three sources of information of the final data collection. Background information, along with the sampling methods of all three sources will be provided in chapter three.

Furthermore, this chapter will describe the format and procedures of the database and

5

additionally describe the two dependent variables and five independent variables used for

data analysis. Chapter three concludes with a presentation of the data analytic strategy

and how such models were used for each research question.

The following chapter is about how the final database was analyzed. Chapter four

focuses on the statistical qualities of the thesis database. First, this chapter will present

the basic descriptive statistics for variables and their measures of central tendency. Next,

the chapter will give a detailed presentation of each research question’s analysis. These subsections will provide the thesis with data analysis results and highlight several statistically significant relationships. The conclusion of this chapter will be a summary of important statistical findings.

The final chapter of this thesis will highlight the implications of the research done. Chapter five will focus on the presentation of the most important findings, discuss policy implication and describe limitations of the study. Specific recommendations for future research will highlight this section and provide a new direction of the study of police paramilitary units. In conclusion, this thesis will provide the reader with a full understanding of militarization in American policing, along with a description and presentation of how this thesis was conducted.

6

CHAPTER 2: LITERATURE REVIEW

Introduction

An increase of militarization in American policing is linked to the several

historical events that altered the criminal justice system and its beliefs. As a result, new policies and procedures that lacked evidence-based research unintentionally altered the criminal justice system and therefore, American policing.

A review of research involving militarization trends in American policing will be

presented in this chapter. Key historical evidence will inform the reader about how and

why militarization has increased. These events have inspired law enforcement entities to

portray a new philosophy towards criminal adversaries. Additionally, common terminology used in militarization publications will be defined for the readers’ education.

Following historical implications, this chapter will focus on national trends depicting the rise and normalization of police paramilitary units. These characteristics have transformed over the past few decades into new styles of policing across the country. Police paramilitary units have been studied on the national and local levels, and the conclusions of that research will be presented.

The purpose of this chapter is to educate and inform readers by presenting research involving militarization trends in American policing. Though this literature review is comprehensive, there is still little research about police paramilitary units.

These units are incredibly popular and yet, academia does not fully understand their impacts on society, their effectiveness in practice or their relationship to crime rates. 7

History of the Militarization of American Police

The rise and normalization of militarization in American policing was influenced by significant historical events. The evolution of police officers has transformed the lone cop-on-a-beat style of the early 1900s into a modern show-of-force police officer. When compared to today, law enforcement organizations have also evolved into a more militarized entity (Kraska, 2007). Presented in this section are the historical events that changed American policing (Kraska, 2007; Hill & Berger, 2009; Huq & Muller, 2008).

Historical Evidence

As the Cold War between the United States and the Soviet Union started to diminish, America lost its communist adversary and foreign enemy (Andreas & Price,

2001; Kiker, 2015; Kraska, 1999). Around the same time crime rates across the United

States soared and public fear increased. American policymakers and politicians in turn began developing “wars” against internal societal problems (Andreas & Price, 2001;

Kiker, 2015; Kraska, 1999). These “wars” were deemed internal because they targeted specific problems inside American territory.

These politicized movements were called the “war on poverty,” “war on drugs,”

“war on crime,” and “war on terror” (Huq & Miller, 2008). Some researchers believe that politicians initiated these powerful “war” metaphors as a way to become elected (Huq &

Miller, 2008). Manipulation of public fear through political candidates shifted national ideals, and when elected integrated the goal of incarceration for domestic problem offenders (Campbell & Campbell, 2010; Huq & Muller, 2008; Kraska, 2007; McCulloch,

8

2004). Metaphorical societal “wars” were developed as a tool for politicians, to sway power in American society.

The key figures responsible for the creation and implementation of these societal internal wars are presented below. President Lyndon B. Johnson introduced the concept of “war on crime” during his debates with presidential candidate Barry Goldwater in

1964 (Huq & Muller, 2008). By the end of President Johnson’s presidency crime rates soared, racial riots were highly televised and civil unrest was a public fear in America

(Huq & Muller, 2008).

Adding to those whom initiated these domestic “wars,” in 1970 President Richard

Nixon’s State of the Union Address continued the war on crime and began the “war on drugs” (Huq & Muller, 2008). Both of these wars are linked to sentencing reforms requiring harsher sentences and the dramatic rise of mass incarceration within the United

States (Radosh, 2008; Spohn, 2009). Following Nixon, there were no new societal “wars” until the Reagan years.

President Ronald Reagan and President George H. Bush intensified the war on drugs in the 1990s as congress and the media fueled the drug war hysteria to all time heights (Huq & Muller, 2008). President Reagan’s policies eventually led Congress to amending old legislation and enacting new laws (Hill & Berger, 2009). These changes in legislation created a need for more militarized police units with greater weaponry and tactics.

The Posse Comitatus Act of 1878, which enacted a “clear separation between internal law enforcement conducted by civilian police and external security functions carried out by the military” was significantly weakened by these changes in policy

9

(Kraska, 2001, p.6). Researchers identified this erosion of the Posse Comitatus Act as

allowing the United States military involvement in matters (Hill &

Berger, 2009; Kraska, 2007). Since then, the military has planned and helped law

enforcement officials in the execution of numerous tactical deployments (Kraska, 1999;

Weber, 1999).

Furthermore, The Military Cooperation with Law Enforcement Officials Act of

1981 was additionally passed during President Reagan’s administration (Hill & Berger,

2009). The new legislation allowed military forces and law enforcement officers to work

together to combat the growing illicit drug trade. It also permitted shared intelligence and

equipment, and incited intimate cooperation between law enforcement and military

operations (Hill & Berger, 2009). According to Hill and Berger (2009), the 1981

Cooperation Act intensified the countrywide trend of law enforcement agencies using

military objectives, methods and equipment. The result was another rising degree of

militarization among American law enforcement entities.

President George W. Bush launched the most recent societal war. The “war on terror” is a result of the September 11, 2001 terrorist attacks. This new war drastically advanced domestic security policies for Americans (Huq & Muller, 2008; McCulloch,

2004). In addition, this war led to law enforcement officials gaining vast amounts of military grade weapons (Weber, 1999).

Propaganda, public fear and militaristic language have increased the militaristic ideals of this nation (Andreas & Price, 2001; Kiker, 2015; Kraska, 1999; Huq & Muller,

2008). Lead to new legislation, immense government spending, and evidence-lacking crime control procedures that potentially limit American civil rights. As a result of these

10 historical events, harsher sentences were given to offenders and mass incarceration became the norm (Radosh, 2008; Spohn, 2009).

Each metaphoric war and enacted legislation increased the need for police, judges, prosecutors and the criminal justice system to be tough on criminals. The criminal justice system felt these pressures and influenced the police to further emulate the U.S.

Military (Campbell & Campbell, 2010; Hill & Berger, 2009; Kiker, 2015; Kraska, 2007).

Transformation of Law Enforcement

When studying militarization in American policing one must be able to measure these characteristics. The way of measuring militarization within an agency can be done with the “continuum of militarization” (Kraska, 2007). Simply put, the continuum is a tool that allows researchers and practitioners to determine the degree a particular entity is militarized. The continuum is used to nominally rank militarization based on four separate characteristics (Kraska, 2007). The four indicators are material, cultural, organizational, and operational.

Material indicators include the extent of weaponry (e.g. automatic weapons, armored personnel carriers) and equipment/technology (Kraska, 2007). Cultural indicators suggest the extent to which the agency uses martial language and military style appearance (e.g. Battle Dress Uniforms)(Kraska, 2007). Organizational indicators are measured by martial arrangements, such as command and control centers, organizational accountability programs (e.g. Compstat) and normalized elite units (Kraska, 2007).

Operational indicators represent the extent that the operation models the military in terms of intelligence gathering, supervision, high-risk situations, and zero-tolerance initiatives

(Kraska, 2007).

11

A rise in any of these indicators will indicate to researchers or practitioners that the specific agency has undergone a transition towards militarization. When comparing the decades prior to the war on drugs to today, most agencies in the United States would reflect a dramatic increase in these indicators. This continuum shift that spans over generations reveals an apparent rise in militarization in American policing.

Militaristic language is also related to the rise and normalization of militarization in American policing. Besides societal “wars,” policing ideologies changed by pegging offenders as enemies (Weber, 1999). Instead of a potential offender being perceived by police as an American citizen, they were viewed as an adversary and the reason for public fear. This in turn transformed some units into what Weber would call “warrior cops”

(1999). Warrior cops are described this way because of their battle dress uniforms

(BDU’s), heavy weaponry (machine guns), and impression of physical force (Balko,

2014; Kiker, 2015; Kraska, 2007; Weber, 1999).

As a result of these historical policies lacking supporting academic evidence,

American law enforcement agencies were encouraged to increase the degree to which they emulate the U.S. military, thus creating these warrior cop impressions (Balko, 2014;

Kiker, 2015; Kraska, 2007; Weber, 1999). Many police departments have blurred the distinctions between military soldiers and police officers through shifts in ideology, appearance, technology, organizational programs and operational techniques since the

1970s (Andreas & Price, 2001; Beede, 2008; Campbell & Campbell, 2010; Herzog, 2001;

Kraska, 2007; Kraska & Paulsen, 1997; Weiss, 2011).

Consequences of Militarization

12

In 2014, the ACLU presented a report concentrating on the safety issues and constitutional rights of American citizens that are threatened by this police practice. The legality of forced entry into citizens’ homes was viewed as a consequence of the increased number of search warrants (ACLU, 2014). The ACLU (2014) also argues that these units present risks through lacking standards, oversight and transparency.

Militarized policies and procedures within law enforcement agencies additionally contradict the growing values of community based policing (ACLU, 2014; Bull & Stratta,

1995; Hill, Beger & Zanetti, 2007; Kraska, 2007).

The majority of researchers have concluded that the military paradigm has become increasingly integrated into the day-to-day operations of law enforcement agencies through particular programs, procedures and the subcultures aligned with police officers (Kraska, 2007). Many criminal justice training academics and organizations fear a further shift towards total militarization (ACLU, 2014; Balko, 2014; Jefferson, 1990;

Kraska, 2007).

The civilian police force, at its inception, was created to implement officers in a way to that would promote a peaceful society (Weber, 1999). Therefore, the actions and values of American law enforcement officers should be noticeably different than those of wartime soldiers (Andreas & Price, 2001; Campbell & Campbell, 2010; Weber, 1999;

Weiss, 2011). Today, police officers are increasingly using military tactics for domestic problems, while military soldiers police foreign populations. Research suggests that there is an opportunity for potential role convergence if these trends continue. Militarization in

American policing is best portrayed through the creation and implementation of police paramilitary units.

13

The Development of SWAT

Before American policing altered its degree of militarization through political agendas and new legislation, the first Special Weapons and Tactics (SWAT) team was created. The purpose of this section is to provide background information on the origins and development of SWAT in America, along with its initial purpose.

SWAT Origins

According to Weber (1999) in August of 1966, a shooter at the University of

Texas shot 46 people, killing 15 of them from atop a 32-story clock tower. Shortly after this incident, the Los Angeles Police Department (LAPD) formed America’s first SWAT team to deal with these types of high-risk situations that normal police officer are not trained to handle (Weber, 1999). The acronym SWAT was implemented to describe their elite force.

In the years following SWAT’s conception, the LAPD SWAT team was tested through the Black Panthers in 1969 and Symbionese Liberation in 1973 (Weber,

1999). Having successfully deployed against these organizations the SWAT program, acquired national prestige. This in turn, increased media coverage lead to numerous large law enforcement agencies to adopt the SWAT protocol and procedures (Weber, 1999).

SWAT was initially created to deal with high-risk reactive situations that routine police are not qualified to handle such as situations involving hostages, terrorists,

“manic” , civil riots, and barricaded suspects (Kraska, 2007; Kraska & Kappeler,

1997). In the beginning these units were deployed in small portions and limited to only large urban police departments (Kraska, 2007). It was not until the drug wars of the

1980s that law enforcement agencies had to serve unparalleled search warrants and

14 increased arrest warrants (Kraska & Kappeler, 1997). As a result, these policies created a need for SWAT in the majority of police departments.

Today the usage of police paramilitary units has endured a complete shift. Kraska and colleagues have documented an incredible rise in the popularity of these units across the nation in both small, and medium to large police departments (Kraska & Cubellis,

1997; Kraska & Kappeler, 1997). In these reports Kraska and colleagues also documented a shift to proactive deployments (Kraska & Cubellis, 1997; Kraska &

Kappeler, 1997). A proactive deployment allows military style units to raid private residences and arrest American citizens. Many academics and organizations warn readers about the potential consequences of allowing further militarization of the American civilian police.

Waco, Texas

Since the beginning of SWAT in America, this most renowned law enforcement operation was heavily influenced by the military. This operation was the highly covered

51-day siege in Waco, Texas (Kraska, 1999; Weber, 1999). The results of this raid would provide researchers with an extraordinary case study documenting the military and law enforcement roles blurring. Waco highlighted the potential consequences that could occur when police officials and military commanders collaborate.

In 1993, preparation for what would be called a “disastrous raid,” the U.S.

Military trained law enforcement agents with Army Special Forces at Fort Hood, Texas

(Kraska, 1999; Weber, 1999). Additionally, Delta Force commanders advised high- ranking law enforcement officers on planning the siege. Completion of the Waco raid resulted in the highest number of civilian deaths ever from law enforcement intervention

15

(Weber, 1999). This case study, along with the MOVE and Ruby Ridge incidents, are examples of potential consequences of converging societal roles (Kraska & Kappeler,

1997).

Governmental Funding

Increasing trends of militarization in American policing can be represented through the actions of the United States government. According to Weber, Congress has encouraged the U.S. military to provide intelligence, equipment and training to local law enforcement agencies for over 20 years (1999). Additionally, in a two year span the

Department of Defense gave police departments 1.2 million pieces of military equipment, ranging from grenade launchers and M-16s to armored personnel carriers (Weber, 1999).

In addition, the ACLU report claims that federal programs created motivation for state and local law enforcement agencies to obtain and use military equipment (2014).

The federal government has historically provided incentives to militarize police (ACLU,

2014). Through use of the same weapons, similar uniforms, governmental authority to use physical force and current proactive deployments, police paramilitary units are reproducing characteristics of formidable U.S Military special operations teams.

Implementation of Proactive Paramilitary Officers

Militarization research in law enforcement agencies over the past few decades has focused on two aspects. Law enforcement organizations in America have become increasingly militarized in two ways—patrol paramilitary policing and police paramilitary units (PPUs)(Kraska, 2007; Kraska & Cubellis, 1997; Kraska & Kappeler,

1997). The focal point of militarized police research is the use of pro-activity in the field.

16

Patrol paramilitary policing and PPUs are typically called out to create proactive arrests, instead of the typical police function of reacting to criminal activity.

Patrol Paramilitary Policing

Patrol paramilitary policing refers to police officers, in full SWAT gear, conducting routine patrol work (Kraska, 2007). Equipment helps to identify these PPUs, ranging from battle dress uniforms (BDUs), heavy weaponry (e.g. automatic rifles) and an appearance of military-style force. PPUs comprise of specially trained police officers equipped with the most dangerous weapons and effective technology available to police agencies (Campbell & Campbell, 2010; Hill & Berger, 2009; Kraska & Kappeler, 1997;

Weber, 1999).

According to the study of Kraska’s national PPU trends, patrol paramilitary policing has become a norm in over twenty percent of law enforcement agencies (Kraska,

2007; Kraska & Cubellis, 1997; Kraska & Kappeler, 1997). Research suggests that paramilitary policing patrol officers are currently being deployed to prevent and arrest criminals in high crime areas (Bull & Stratta, 1995; Kraska, 2007; Kraska & Paulsen,

1997). Law enforcement agencies employ this technique by conducting street sweeps in such neighborhoods, in order to commonly practice a weed and seed approach to improving communities (Hill & Berger, 2009; Kraska, 2007; Kraska & Paulsen, 1997).

The weed and seed method is a frequent use of PPUs that targets high crime neighborhoods by removing dangerous offenders (DOJ, 2004; Kraska & Paulsen, 1997;

Lilley, 2015). This approach has been correlated to a decline in robbery, burglary and vehicle theft (Lilley, 2015). A significant consequence to this show of force by law enforcement officials is an enhancing of citizens’ fear of criminal activity (Bridenball &

17

Jesilow, 2005). Additionally, as a result of using police paramilitary patrols there is a decrease in perceived police legitimacy by the affected neighborhoods’ citizens (Bull &

Stratta, 1995; Phillips, 2014).

The implementation of the weed and seed program portrays a proactive aggressive policing strategy that routinely uses patrol paramilitary officers. Research suggests a decrease in some crime trends, but it also shows how these units can negatively affect the neighborhood and perceived legitimacy of the police.

Police Paramilitary Units

Although there is no universal name of these elite police units, the international

term used is police paramilitary unit (Hill & Beger, 2009; Rantatalo, 2012). However, in

America and Anglo-Saxon countries (e.g. United Kingdom and Australia) these units are

frequently referred to as Special Weapons and Tactics (SWAT), and occasionally

identified as Emergency Response Teams (ERT) or Special Patrol Groups (SPG)(Hill &

Berger, 2009).

According to Kraska, the appearance, tactics, operations, weaponry and culture of

these units are closely modeled after elite military Special Forces units, such as the Navy

SEALs and Army Rangers (2007). Not only are police units modeled after specialized

military units, PPU officers are commonly trained by those elite military personnel

(Campbell & Campbell, 2010; Hill & Berger, 2009; Kraska & Kappeler, 1997; Weber,

1999). Police paramilitary officers are depicted as the elite special operations agents of

police forces.

These elite police officers are typically used as a proactive entity. Statistics from

several studies show that these police paramilitary units are overwhelmingly used for

18

proactive warrant work (ACLU 2014; Kraska & Cubellis, 1997; Kraska & Kappeler,

1997).

Consequences of Militarization

Shifting the degree of militarization in American police departments has

transformed the community friendly cop-on-a-beat to the “show of force” patrolling paramilitary police units (Kraska & Cubellis, 1997). Continuation of this trend could result in an overly aggressive criminal justice-industrial complex that limits civilians’

constitutional rights (ACLU, 2014; Kraska, 2007).

According to Herzog, the process of over-militarizing American police forces will

ultimately led to the revolution of demilitarization (2001). Through his research overseas

studying the demilitarization process of the once highly militarized Israeli National

Police (INP), Herzog has concluded that the willingness to change from within the police

organizational structure is a major obstacle (2001).

Considering that American police agencies do not emulate the INPs centralized

governmental police, altering American policing would ultimately present far more

difficulty. Unlike the INP, municipal, county and state agencies all operate individually

with limited centralized pressures. Changing the culture of American policing to limit

militarization could only come through legislation, political intervention and a willing

law enforcement community.

Review of the Research

Following historical evidence that highlighted the rise and normalization of

paramilitary forces within American policing, this section provides original academic

19

research. Review of the research will present militarization academic studies, along with

the conclusions surrounding police paramilitary units/officers in America by academics.

Research on the militarization movement in American policing is limited.

Additionally, the majority of articles are from the late 1990s and early 2000s. Historical

events have created the degree of militarization today and shaped the countries culture

along with altering citizens’ attitudes. Few articles have collected data on these changes, and even fewer have conducted data analysis while looking for statistical trends.

Most of the research involving police paramilitary units focuses on the rise and

normalization of their usage. Some researchers have distinguished the need for studying

PPU personnel and the sub-cultural themes involved with such programs. Furthermore,

academics have examined the use of force used by SWAT teams and conducted

descriptive studies and process evaluations of how these units train. The review of

statistical research on police paramilitary units in American policing is presented.

Militarism and Militarization

Leading research on this subject focuses on two concepts, militarism and

militarization (Kraska, 2007). Kraska’s research identifies militarism as an ideology.

Militarism believes that the best approach to solve a problem is through use of force and

threat of violence (Kraska, 2007). This approach exhibits strength and power by the

“military industrial-complex” as a primary problem-solving tool to gain political authority, and as a consequence, glorifies military power, hardware and technology

(Kraska, 2007; Kraska & Kappeler, 1997; McCulloch, 2004).

Militarization is the implementation of militarism and defined as “the process of arming, organizing, planning, training for, threatening, and sometimes implementing

20

violent conflict” (Kraska, 2007, p.3). Total militarization for example would be a

criminal justice system that completely adopted and implemented the show-of-force ideology. Militarism and militarization are concepts that define and determine the degree to which military-type involvement is perceived in modern policing.

National Police Paramilitary Unit Activity

In 1997, Kraska and colleagues published two articles on the growth and normalization of paramilitary policing in America. Through separate surveys, they documented and presented findings on increased “call-outs” and the new usage of PPUs

(Kraska & Cubellis, 1997; Kraska & Kappeler, 1997). Kraska defines call-outs as any event requiring the deployment of a PPU (Kraska & Cubellis, 1997).

Data and research obtained by these surveys provide the most important results in the study of American police paramilitary units thus far. One survey collected information from law enforcement agencies serving jurisdictions over 50,000 citizens

(Kraska & Kappeler, 1997), while the other collected similar statistics from agencies serving 25,000 to 50,000 citizens (Kraska & Cubellis, 1997).

Kraska and Kappeler (1997) conducted an independently funded national survey of medium to large law enforcement agencies. The researchers conclude from their results that PPUs have become a norm in U. S. policing. That is, in 1982, about 59 percent of the 548 departments surveyed had a PPU, and by 1995 over 89 percent had a unit (Kraska & Kappeler, 1997). In addition, they found that the total number of call outs from 1980 to 1995 increased by over 538 percent (Kraska & Kappeler, 1997).

In particular, the researchers showed that of the total 25,201 call-outs, 76 percent

(19,125) were for proactive warrant work (Kraska & Kappeler, 1997). The majorities

21

were for drug raids identified as high-risk warrants. While the original goal of

implementing the use of PPU was to reactively respond to potentially violent situations,

these units were found to routinely deploy for proactive cases involving drug users and

drug dealers (Kraska & Kappeler, 1997).

In addition, the researchers found that PPUs were being increasingly used for

proactive patrol. In 1982, only 24 of these departments deployed PPUs for proactive foot

patrols, but in 1995 the number rose to 94 agencies (Kraska & Kappeler, 1997). Kraska’s

research shows the increase of almost four times as many agencies with proactive patrols.

Kraska and Cubellis (1997) did similar research on small law enforcement

agencies in America. Findings were consistent with Kraska and Kappeler (1997).

According to the researchers, out of 552 departments over 65 percent had a PPU. Only 20

percent of those departments had a PPU at the beginning of 1980 (Kraska & Cubellis,

1997). Additionally, in 1980 there were a total of 220 call-outs, and by 1995 that number rose 1,589 percent to 3,715 deployments (Kraska & Cubellis, 1997).

Similarly, 66 percent of these call-outs were for proactive warrants and 17 percent of the departments used the unit for a proactive foot patrol. Through their survey, Kraska and Cubellis (1997) were also able to identify that 93 percent of these units were classified as part-time, and over 80 percent of them had MP5 machine guns, tactical semiautomatic shotguns, rifles and night vision equipment.

Clearly, the rise and normalization of PPUs in America increased substantially

from 1985 to 1995. By the end of 1995, over 75 percent of law enforcement agencies had

a PPU (Kraska & Cubellis, 1997; Kraska & Kappeler, 1997). The returned surveys

combined to document a 939 percent increase (Kraska & Cubellis, 1997; Kraska &

22

Kappeler, 1997). Additionally, the most notable statistic reported by these surveys was

that of the law enforcement agencies that replied, over twenty percent used their PPUs for

proactive patrols (Kraska & Cubellis, 1997).

More recently, in 2014, the American Civil Liberties Union published a report that showed 79 percent of PPU deployments were for search warrants and 62 percent of those were drug searches (ACLU, 2014). The ACLU uniquely studied and documented the victims and targets of these raids. The authors concluded that 39 percent were black,

11 percent Latino and 20 percent white, with 30 percent unknown. The report suggests that police paramilitary units in America primarily target and impact people of color

(ACLU, 2014).

When comparing Kraska’s surveys in the late 1990s to the ACLU report in 2014 the numbers suggest stabilization in proactive usage. All three reports suggest that 66 to

79 percent of PPU deployments were for proactive warrant work (ACLU, 2014; Kraska

& Cubellis, 1997; Kraska & Kappeler, 1997). Although this is still an overwhelming majority, it suggests that the militarization movement in policing may have stabilized.

Characteristics of Police Paramilitary Unit Personnel

Another feature of studying police paramilitary units focuses on the individuals themselves. PPU personnel have different characteristics than typical police and they possess specific qualities. The purpose of understanding the nature of PPU personnel is to shed light on the subculture of these units.

Qualitative research suggests four interrelated cultural themes within police paramilitary units. These four themes are “the fifth branch of the military,” “hyper- dangerousness,” “pleasures of paramilitary policing,” and “constructing an elite military

23

status,” (Kraska & Paulsen, 1997, p.261). Cultural themes were gathered through in-

depth interviews and participant-observations to study the sub-cultural attitudes, norms and values of a typical PPU officer (Kraska & Paulsen, 1997). In conclusion, the ideology of militarism is deeply embedded in American PPUs.

These four themes portray PPUs as being culturally aggressive. This research suggests that paramilitary police officers believe they are elite and dangerous with militaristic values. Through identifying these themes researchers have presented a subculture within law enforcement agencies that seeks confrontation. As a consequence,

PPU personnel can be viewed as warrior cops (Balko, 2014; Kiker, 2015; Kraska, 2007;

Weber, 1999).

A characteristic identified in PPUs is the hyper-masculinity (Dodge, Valcore &

Gomez, 2011). This is additionally evidenced-by cultural themes, personnel mentality, as well as the low participation rate of women in PPUs, when compared to the agency as a whole (Gaylean, Wherry & Young, 2009; Kraska & Paulsen, 1997). Police paramilitary units are typically teams consisting of men because of the dangerous and potentially violent subculture. Thus, Dodge et al. conclude that elite tactical teams’ patriarchal subculture and themes of violence may explain the absence of women within their programs (2011). Their research allows academics to further understand PPU personnel and what kind of individuals these units attract.

Along with typical characteristics of individuals on PPUs, research has been conducted to identify certain attributes that are present in those personnel who are characterized as “successful” PPU members. According to Super (1995), those who were successful as PPU personnel were “task focused,” “ambitious,” “inclined to take charge,”

24

“tough minded,” “self-sufficient,” and “emotionally independent” (Super, 1995, p.61).

This qualitative research alludes to the mentality of successful PPU members and furthers

academic understanding of PPU personnel.

Further analysis of SWAT team members conducted by Gaylean et al. sheds light

on the members’ values towards mental health professionals. The study concluded that

SWAT team members do not value the post-incident counseling and debriefing services of mental health professionals as highly as other contributions (Gaylean et al., 2009).

Although, Gaylean et al. (2009) suggests that individual SWAT team members do value these services differently, based on particular demographic variables and previous

personal/professional experiences.

Modern Police Paramilitary Units

According to the multi-method study conducted by Klinger and Rojeck, SWAT

teams around the nation are primarily deployed to resolve high-risk situations (2008).

This suggests that the original purpose of SWAT units is still the most common form of

implementation.

In addition to resolving high-risk situations, PPUs are able to address another issue in society. PPUs are used to fill the gap between military and police. Hill et al.

(2007) states that PPUs are capable of enforcing laws within this “security gap.” Not only are these units providing a societal service that military soldiers and police officers cannot address, these PPUs are doing so with limited force, and rarely impose the use of lethal force (Klinger & Rojeck, 2008).

Williams and Westall’s research supports this claim through their study of the use of force by SWAT in Glynn County, Georgia. According to their research, there is no

25 statistically significant relationship between the number of use of force reports by SWAT officers and non-SWAT officers in the Glynn County Police Department (Williams &

Westall, 2003). Suggesting that these units are not only filling a need, but they are doing so with limited force that is typical in any police situation.

On the other hand, the report published by the ACLU (2014) concludes that using violent tactics and equipment increases the risk of physical harm and property damage.

Through studying national data on SWAT team deployments, Klinger and Rojeck (2008) also suggest that SWAT teams performed how they are trained, and should be encouraged to train as much as the budget and time will allow.

Clark, Jackson, Schaefer and Sharpe’s (2000) research demonstrate just how inconsistent and diverse training standards and programs are between small and larger departments across the nation. From located at military institutions to colleges, amount of hours trained, types of training and other factors, Clark and colleagues describe an unregulated and conflicting SWAT standards (Clark et al., 2000). These unregulated

SWAT training standards portray SWAT programs as a governmental program without regulations. The ACLU report also stresses the need for more standards and oversight of police paramilitary units (2014).

The Need for Police Paramilitary Units

PPUs are needed in society to deal with high-risk events that regular police officers are untrained to respond. Intense training of SWAT officers needs to be properly overseen to ensure proper techniques and use of force protocol. In order to successfully deal with these situations the training standards of the potentially most lethal police force must be governmentally regulated for the safety of citizens.

26

Over the years some academics have published contrasting ideas about the rise in militarization in American policing. Kraska and colleagues that support the claim of militarization have been challenged by a few researchers who believe that PPUs and paramilitary policing tactics are just the next evolutionary step in policing (den Heyer,

2014; Waddington, 1993; Waddington 1999).

These academics believe the militaristic subcultures studied by Kraska and colleagues are unsubstantiated (den Heyer, 2014; Waddington, 1993, 1999). According to

Kraska’s adversaries, those who suggest there have been a rise in the militarization of civilian police are basing their conclusions on little empirical evidence (den Heyer, 2014;

Waddington, 1993, 1999). On the other hand, Kraska and colleagues have published additional journal articles in order to refute these claims and build upon their study of militarization (Kappeler & Kraska, 2015; Kraska, 1999; Sarao, 2014).

Conclusions

In review of the research presented, there are more authors and articles in support of the assertion that American police have been increasing the degree to which they are militarized (ACLU, 2014; Balko, 2006; Balko, 2014; Hill & Beger, 2009; Kappeler &

Kraska, 2015; Kiker, 2015; Kraska, 2007; Kraska & Cubellis, 1997; Kraska & Kappeler,

1997; Kraska & Paulsen, 1997; Weber, 1999). Waddington (1993; 1999) and den Heyer

(2014) are the few to suggest policing is not. Although the is validated as occurring, Kraska and colleagues identify that certain aspects of the military model are in fact useful for policing.

Some positive virtues of the military model include, discipline, honor, control, competence, and a form of patriotism (Kraska & Cubellis, 1997). Additionally, Kraska

27

concedes that SWAT is an “essential reactive function” (Kraska, 2007, p. 11). He claims

that the use of SWAT should remain limited to the reactive and high-risk situations to

which they were first created. Limiting the use of police paramilitary units should protect

American police forces from the needed demilitarization process that the Israeli National

Police is attempting (Herzog, 2001).

Furthermore, some researchers suggest that Intelligence-Led Policing would

increase successful and safe PPU deployments (Bull & Stratta, 1995; Phillips, 2014).

Increasing safe and effective SWAT deployments will also increase the perceived police

legitimacy and lead to proper expectations. To further study police paramilitary units would need to increase transparency (ALCU, 2014). Law enforcement agencies are governmental entities and their success, or lack their of, should be disclosed to the public at large. Departments should be held accountable for their operations and the impacts

those operations have on the community.

For certain, there is a lack of research about police paramilitary units. Academia

does not have statistical insight into the reasons PPUs are used, the units’ effectiveness,

or their relationship to crime trends. By lacking research academics do not know what

variables are likely to increase, or decrease, the likelihood of SWAT team deployments

within a jurisdiction.

Research must identify if agencies with high crime trends rely on PPUs at a

higher rate as a way of aggressively combating offenders. Police paramilitary units are

firmly entrenched into law enforcement agencies and have not been studied enough.

Potential correlations between SWAT usage and crime, agency or jurisdictions have not

been examined, thus leaving academia in the dark about these programs societal impacts.

28

Research Questions

This section presents specific research questions, which are the purpose of this thesis. Along with the four research questions, each subsection will have a brief explanation of why the goal of this thesis is to identify statistically significant relationships between the variables.

Research Question 1

How are Part I violent incidents, Part I property crime incidents, the type of agency and number of sworn officers predictive of Part I crime search warrant SWAT team deployments over time and across agencies? The purpose of this research question is to determine the statistical relationships between Part I incidents and SWAT team usage while accounting for both time and space. Research question 1 portrays how proactive and aggressive law enforcement agencies are towards Part I crime through the use of SWAT teams.

Research Question 2

How are Part II vice arrests, the type of agency and number of sworn officers predictive of Part II crime search warrant SWAT team deployments over time and across agencies? The goal of this research question is to determine whether agencies that tend to be proactively aggressive towards vice type crimes, are also proactively using SWAT teams to combat crime. Policing vice type crimes to make arrests are typically done proactively just like search warrant SWAT work. This question also accounts for both time and agency.

Research Question 3

29

How are Part I violent incidents, Part I property crime incidents, the type of agency and number of sworn officers predictive of Part I crime search warrant SWAT team deployments over time for an agency? This question is quite similar to the first question, yet it does not account for agency. By running this analysis the statistics will portray how agencies change, their SWAT deployments and Part I incidents over time.

Research Question 4

How are Part II vice arrests, the type of agency and number of sworn officers predictive of Part II crime search warrant SWAT team deployments over time for an agency? Additionally, this question is much like the second question, but it also does not account for agency. This analysis shows the tendencies of agencies over time. Both Part

II arrests research questions have the greatest likelihood of producing statistically significant relationships. This is due to the proactive nature of both the vice type style arrests and search warrant SWAT deployments.

30

CHAPTER 3: METHODOLOGY

Introduction

The reason for this chapter is to inform the reader on how statistical data was collected, where it was collected from, what variables were used and the methodological steps taken for data analysis. Data collected for this research study came from several sources.

First, the Maryland Governor’s Office of Crime Control and Prevention (GOCCP) collected data spanning five fiscal years from 2010 through 2014, which focused on

SWAT team usage and results. The second source is also from the GOCCP. The Crime in

Maryland, Crime Report is published detailing crime statistics and agency information. To finish, the populations of agency jurisdictions were collected from the U.

S. Census. Together the information collected created a final database used for this research study.

This chapter begins by describing each data source and explaining how the data was collected. This chapter will also present and explain each variable used, and the reason why the variable was included. Finally, the analytical strategies used in this study to reveal trends and statistical significance will be discussed.

Data Collection

The purpose of this section is to provide insight into how and why the data was collected by these organizations. Detailed information about each separate source of data will be presented. Each subsection will address data collection techniques and objectives. 31

Additionally, the unit of analysis for this thesis is the combination of agency plus year.

Maryland SWAT Team Reports

The first set of data comes from Maryland’s SWAT Team Reports published by

the GOCCP. In 2009, Maryland passed Senate Bill 447 (SB 447) requiring all law

enforcement entities to record and report information on each SWAT deployment.

SB 447 was enacted because a SWAT team unknowingly raided Berwyn Heights

Mayor Cheye Calvo’s house and his dogs were shot (ACLU, 2015; Balko, 2015). Senate

Bill 447 collected very unique data for five fiscal years, but the Maryland SWAT Team

Reporting bill was “sunsetted” in June 2014 (ACLU, 2015).

This legislation held each responsible for data collection, which was entered into a pre-formatted Excel spreadsheet and then submitted to the

GOCCP. Following the data collection by the GOCCP, the Maryland Statistical Analysis

Center publishes the results into a SWAT Data Report. This occurred every fiscal year from 2010 to 2014. This unique legislation and statistical analysis increased transparency throughout Maryland about SWAT team deployments.

Maryland SWAT Team Reports data collection include:

• The number of times the SWAT team was “activated and deployed” by the

agency in the previous six months;

• The name of the county and/or municipal corporation and the zip code(s) of the

location(s) where SWAT team was deployed for each activation;

• The reason(s) for each activation and deployment;

• The legal authority (e.g. type of warrant, emergency, exigent circumstances, etc.)

for each activation and deployment;

32

• The result or outcome of each deployment including whether arrests were made,

and how many, whether property was seized, whether forcible entry was made,

whether any weapon was discharged by a SWAT team member, whether any

person or domestic animal was injured or killed by a SWAT team member (by

whatever means). MD Code Ann., Public Safety Art., § 3-507 (B), and whether

any SWAT Team officer was injured by a person or an animal during the

deployment.

The GOCCPs Maryland SWAT Team Reports are the backbone of this thesis. A unique data collection allowed research to be done on police paramilitary unit activity in

Maryland.

Crime in Maryland Reports

Since 1975 Maryland has published the Crime in Maryland reports. This is in accordance with Public Safety Article, Section 2, Subsection 307 and 308 of the

Annotated Code of Maryland. The Maryland Department of State Police submits this publishing to the Governor. Maryland’s uniform crime reports are then sent and published by the Federal Bureau of Investigations (FBI).

Inside Crime in Maryland is a statistical presentation of crime incidents, arrests, law enforcement officers killed or assaulted, and law enforcement employee data. By publishing their own Uniform Crime Reports (UCR), Maryland has presented this information in much greater depth than the Federal Bureau of Investigations UCR reports. The data for each county, city, and agency is much more detailed. Additionally,

Crime in Maryland combines and contrasts some of these statistics within their expansive publication for a better look at law enforcement agencies across the state.

33

Crime in Maryland is the second source for this thesis and these yearly

publications provided crime reports and law enforcement employee information needed

for the final database. These statistics were essential to the data analysis in order to compare crime trends to the usage of police paramilitary units in Maryland.

The United States Census Bureau

The third and final source of data for this thesis database came from the U. S.

Census Bureau. In 2010, the U. S. Census collected and dispersed population totals for

the country. Population totals for states, counties and cities are also included in these

reports.

Thesis data collection started for the year 2010, which used exact population totals. Considering these population totals are only taken every ten years, the following

years of populations for this database are the U. S. Census’ estimated populations.

Estimated populations from the U. S. Census Bureau were imputed into the thesis

database for years 2011 through 2013.

The purpose of collecting this data was to compare agency jurisdictional size to

the other variables collected. Additionally, the population totals were used to create the

rates for the final data analysis. Using rates instead of counts helped compare agencies

more uniformly. Conducting data analysis in the form of rates allowed for more populous

jurisdictions to be compared evenly to jurisdictions with smaller populations.

Database

Since there are three separate sources used in this data collection, specific

statistics were chosen and manipulated for the analysis. From each of these sources

variables were taken and inputted into the thesis database. In this section, the samples

34

chosen will be presented.

Maryland SWAT Team Reports provided data on police paramilitary units and

five years of every SWAT deployment in Maryland. Senate Bill 447 created the legal

requirement for each agency to collect this information.

As a result of Senate Bill 447, this thesis was provided the universe of cases for

Maryland’s SWAT team deployments. Every single SWAT deployment in Maryland

from 2010 through 2014 was recorded and reported to this source. From these cases, a

selection of the deployments was chosen for specific data analysis in order to gauge the

proactive nature of each agency and how they respond to crime trends.

Maryland’s SWAT team records were chosen for this thesis because the data is

extremely unique. The ability to conduct data analysis on years of SWAT team

deployments led the way for this exploratory research study.

The Maryland SWAT Team Reports were collected by fiscal years (July through

June) because of this the years of 2009 and 2014 were incomplete, and thus discarded. As

a result, the data focused on only the four full calendar years of SWAT deployments,

which are in the same distribution as the uniform crime reports.

Next, the data was considered and additional subtractions were completed. For the purpose of this study, the database was restricted to only proactive search warrants as a legal authority. Each year search warrants made up the overwhelming majority of these deployments, about 89 to 93 percent, and it made logical sense to focus on the proactive majority of cases. Additionally, the only data selected were for Part I and Part II call-outs.

This allowed the different deployments to be compared to specific crime trends in the data analysis.

35

Furthermore, the data was restricted to only county and municipal agencies.

Before this limitation, there were only two state level agencies with deployments and the

analysis for them would be flawed by comparison. Also, the data was limited to only law

enforcement agencies and not correctional agencies. This final subtraction rid the

database of two agencies that would have skewed the analysis with little (or no) crime

statistics, but a number of SWAT deployments.

Originally, the Maryland SWAT team dataset consisted of the fiscal years 2010

through 2014, providing 8,249 call-outs. By the end of these restrictions the data provided 5,649 call-outs over four calendar years, for thirty-five separate agencies, of

only Part I and Part II crime search warrant deployments. This data distribution created

the foundation for the panel design data analysis.

The Crime in Maryland publication provides transparency for law enforcement

entities. It presents numerous statistics on every agency in the state. For the function of

this thesis, the aggregate counts gathered from Crime in Maryland focused on crime

trends and law enforcement personnel. These statistics provided the database the ability

to compare crime trends and agencies personnel to SWAT deployments. All of these

counts were gathered for the years 2010 through 2013, for only the thirty-five agencies

that met this thesis’ qualifications for SWAT deployments during that time frame.

First, Part I violent incidents were gathered from the Crime in Maryland reports.

These incidents included murder, rape, robbery and aggravated assault. Additionally, the

Part I property incidents were also gathered, which included breaking or entering, larceny

theft, and motor vehicle theft. Both violent incidents and property incidents were

gathered separately and then combined later for the final dataset. Part I incidents will be

36

compared to the search warrant SWAT deployments from the first data source.

Next, Part II arrests were also collected from Crime in Maryland. This thesis focused on only proactive policing vice type arrests. These arrests included prostitution, narcotics, driving under the influence, liquor laws, disorderly conduct and vagrancy. Vice type crimes were gathered to portray how proactive an agency is, which can then be compared to the proactive nature of the SWAT search warrant deployments from the first source.

Last of all, Crime in Maryland provided the number of sworn law enforcement employees for each of the agencies. The collection of these counts was used to gauge the size of the agency. The number of sworn law enforcement officers was perceived as crucial to comparing agencies to each other.

The sampling method for the final data source focused on agency jurisdictions. As a result of the data, the unit of analysis chosen was the agency plus year. Populations were collected for all of the thirty-five agencies that met the SWAT call-out criteria.

Populations included were for counties and municipalities. Maryland jurisdictional populations for the four years were then entered into the database and used for rates formulas. Jurisdictional populations were integral to the data analysis for this thesis to compare the size of agencies more uniformly.

Variables

The purpose of this section is to describe the final database and present all the variables used in this study. Dependent and independent variables will be discussed in the following subsections. The final database consisted of two dependent variables and five independent variables. Most variables were converted into rates using the jurisdictions

37 populations.

The final database was constructed into a long format. Long format is when each agency plus year is a separate case. For example, Baltimore Police Department (BPD) had four cases in the database, one for each of the years studied (i.e. BPD 2010, BPD

2011, BPD 2012, and BPD 2013). Long format is preferred for panel design data analysis.

Throughout the process of constructing the database, Microsoft Excel was used and then once the final variables were entered it was converted into IBM SPSS. For running the tests and performing data analysis the statistical package STATA.

Dependent Variables

The dependent variables were used to measure the law enforcement agencies aggressive policing tendencies. SWAT search warrant deployments are a proactive policing strategy, where police are actively pursuing criminals instead of reacting to a situation. Proactive warrant deployments are the overwhelming majority of SWAT call- outs.

Research questions 1, 2, 3, and 4 hypothesized that proactive SWAT deployments were to be the result of higher crime rates. The analysis’ goal is to show that as crime increases, so does aggressive policing agencies through search warrant SWAT team deployments.

In the final database there are two dependent variables. They are SWAT search warrant deployments for Part I offenses (SWPO) and SWAT search warrant deployments for Part II offenses (SWPT). Both of these dependent variables were collected through the Maryland GOCCP SWAT Team Reporting database.

38

The dependent variables were collected from the original Maryland SWAT Team

Reporting data collection through a cross tabulation table yearly totals for each agency

and for each reason of deployment. Therefore, search warrant SWAT deployments in

Maryland were separated by agency and year for Part I and Part II crimes. Thus, giving

the final database the deployment counts for the unit of analysis (agency per year) for the

statistics.

Dependent variables were then converted into rates using the jurisdictional

population. For example, SWAT search warrant deployments for Part I offenses rate

equation was; SWAT Part I Deployments Rate = (SWAT Part I Deployments/Population)

* 100,000. This formula was used for both dependent variables in this analysis. The goal

of using a rate instead of totals was to account for an agencies jurisdictional population.

Independent Variables

The independent variables in this study were focused around predicting SWAT

deployment usage. As previously stated, this study assumed that a higher rate of SWAT

deployments was a result of higher crime rates. The was assumption that as crime rates rose, law enforcement agencies would react by increasing proactive SWAT team deployments in order to combat criminal activities in their jurisdictions. Thus, the independent variables are the jurisdiction’s crime rates.

Two additional independent variables were included in the regression model.

They were the type of jurisdiction and number of sworn law enforcement officers. Each independent variable will be discussed further.

First, just like the dependent variables, most of the independent variables were converted into rates using the jurisdictional population. The same formula, Xrt =

39

(X/POP) * 100,000, was used. Using rates instead of counts allowed for a jurisdictions

size to be included in the analysis.

The independent variables for all Part I crime rate analysis were separated into

two categories, violent crime incidents and property crime incidents. Total violent crimes

were a combination of all murder, rape, robbery and aggravated assault incidents. The

second category, property crime total, was the sum of breaking or entering, larceny theft,

and motor vehicle theft incidents reported in the Crime in Maryland reports. These counts

were all the Part I offenses published in the reports.

Part I crime rates for data analysis reflected reported incidents of crimes. These

are crimes that were reported to law enforcement agencies. On the other hand, the Part II

figures were collected from the number of arrests and converted into rate form. Part II

arrest rates were used because of the inability to accurately count Part II crimes in

Maryland. Part I crime incidents are more easily counted because of victims and

evidence. Yet, Part II crimes are most often viewed as “victimless” and therefore not

often reported. Thus, Part II crimes are more easily recorded as arrests, instead of

incidents.

Part II arrest counts were collected similarly. These independent variables were

also collected from the Crime in Maryland reports, but instead of using every possible

Part II offense, this database only focused on arrests that portray a proactive policing entity. As a result, the final database reflected how many Part II arrests occurred in the jurisdiction, as a proxy for how aggressive the agency was towards vice type crimes.

Total vice arrests within a jurisdiction was calculated by the sum of prostitution, narcotics, driving under the influence, liquor laws, disorderly conduct, and vagrancy

40 arrests. These are vice type offenses because they typically do not have victims. The parties involved are typically willing, but the more a law enforcement agency proactively pursues these offenders portrays an agency that has a more aggressive policing culture.

The assumption that a higher degree of aggressive policing behavior is correlated to more proactive search warrant SWAT deployments is the basis of this thesis.

In addition to the jurisdiction’s crime rates, the database included two more independent variables. Firstly, the number of sworn law enforcement officers rate was incorporated to show the statistical relationship between the size of the agencies personnel and the use of SWAT.

To finish, type of jurisdiction was the final independent variable in this thesis database. Type of jurisdiction distinguished between county agencies and city agencies that deployed SWAT units. This variable is the only independent variable that was not a rate.

Analytic Strategy

The purpose of this section is to discuss the statistics and why mixed linear models were used. Using random effects models and fixed effects models for data analysis was the result of examining the final database. Random effects models and fixed effects models were chosen because of two limitations in the thesis data collection, length of time and number of cases.

The final data collection contained a time element. Unfortunately, it was not long enough for a true time series strategy. Maryland’s SWAT Team Reports only collected data for five fiscal years (which only totaled four calendar years) and that made a time lag equation unreasonable. Analytical strategies were severely restricted due to the lack of an

41

extended time period where SWAT data was collected.

Additionally, the final number of cases limited the thesis database. In Maryland’s

SWAT Team Reports only thirty-five agencies conducted the specific search warrant

deployments that met this thesis’ criteria. With this limited number of agencies, many

statistical analysis models were not reasonable.

In order to account for these limitations, the unit of analysis chosen for the

analytic strategy was agency per year (e.g. Baltimore Police Department (BPD) 2010).

By combining these two variables the number of cases increased significantly to a total of

140, instead of thirty-five. As a result, each agency had four separate cases in the final

database (e.g. BPD 2010, BPD 2011, BPD 2012 & BPD 2013). This adjustment allowed

the data collection to be manipulated with a statistical mixed linear model (i.e. random

effects models and fixed effects models). These models will be further explained in the

following subsections.

Mixed Linear Models

Mixed linear models refer to models that include both fixed and random effects.

According to Pratt (2000), in the study of criminology and criminal justice many statistical models contain multiple effect size estimates and therefore create a lack of statistical independence. Random and fixed effects models were developed to address this issue. Instead of over-generalizing, mixed linear models assume that each case has

individual characteristics (Pratt, 2000). Therefore each unit of analysis has its own corresponding linear equation. Mixed linear models estimate the mean of these effects.

Fixed Effect Models and Random Effects Models

Fixed effects models and random effects models are quite similar, although there

42 are several important distinctions. The most important difference between the models is that a random effects model acknowledges possible outside variables (i.e. assumes changes and variations), while the fixed effects model controls for all outside variables

(i.e. assumes constants)(Brame, Bushway & Paternoster, 1999; Pratt, 2000).

According to Brame et al. (1999, p. 605) “the goal of fixed-effect estimation is to estimate a term for each individual in the data that takes all of the stable features of the process under study (i.e., criminal behavior) into account.” By identifying this term (i.e. fixed effect) the data set is then characterized by its distribution. Following taking into account the fixed effect, the only remaining variation is due to change over time (Brame et al., 1999). This provides the analysis the strongest control over individual differences for the agencies in this thesis.

Random effects models are more popular in criminology and criminal justice research than fixed effects (Brame et al., 1999). Random effects models intend to allow for multiple individual-specific effects that are randomly included in the meta-analysis

(Brame et al., 1999; Pratt, 2000). Additionally, random effects models control for potential biases that arise during estimation across size estimates (Pratt, 2000). Thus, the use of random effects models in criminology and criminal justice are often chosen because of the assumptions that variables with correlations to criminal behavior vary

(Brame et al., 1999).

Considering the thesis’ panel design database and its limitations, the best choice of statistical model for this thesis was the random effects model (i.e. random intercepts model). According to Brame et al. the most common use of this analysis techniques are for panel models (1999).

43

In this thesis, the use of random effects models and fixed effects models were to

analyze the data in two specific ways. The purpose of using a random effects model was

to compare units across agencies and time, while acknowledging for possible outside variables. The goal of using random effects models is to generalize through analysis of a range of scenarios. Secondly, the purpose of using the fixed effects models was to assess variance within a single agency over time, controlling for all outside variables. Fixed

effects models are stricter and lack the ability to account for variations or estimate the

effect of variables that are fixed over time.

These two models are similar, but are also conceptually different. The use of both

random effects and fixed effects models to analyze the thesis’ database allowed for

statistical analysis of agencies on one hand, and agencies over time on the other. Random

and fixed effects models allowed for a complete understanding of the statistical

relationships between SWAT search warrant deployments, crime trends, law enforcement

personnel, type of agencies, and population size of jurisdictions.

Random effects models and fixed effects models yield similar statistics outputs.

When these hierarchical linear models are performed using STATA, the statistical

package will present tables that include a coefficient, standard error term, z-value, 2

tailed p-value and the 95% confidence interval for the coefficient.

Each of these statistical values presents the empirical findings of the mixed linear

models. First, the coefficient is a value assigned that is used in predicting the dependent

variable from the independent variable (UCLA, 2016a). Coefficients show how much the

dependent variable is affected and which direction the correlation is headed (positive or

negative).

44

Next, mixed linear models present the standard error term assigned to the

variable. The standard error term presents a statistic portraying the variance of the model

and showing the difference between the clusters (Raudenbush & Bryk, 2002). Standard error terms are correlated to the coefficients and used to identify statistically significant correlations (UCLA, 2016a).

From these standard error terms, empirical analysis will identify the z-value and

p-value and present statistically significant correlations. The z-value and p-value are both statistics testing the null hypothesis of the equation (UCLA, 2016a). According to

Wasserstein and Lazar (2016), a p-value is the probability that under the specific statistical model performed, the statistical summary of the data would be either equal or greater than the observed value. Subsequently, if the p-value is less than 0.05 then the correlation is statistically significant and able to be generalized to the population at large.

Furthermore, the mixed linear model in STATA produces the 95% confidence intervals for the coefficients. Confidence intervals are very useful in identifying how high and how low the population value of the parameter might be (UCLA, 2016a).

Another statistic presented in mixed linear models is the Rho variable. Rho, also known as the intraclass regression coefficient, shows the correlation of cases from within a cluster (UCLA, 2016b). Rho is a way of estimating a fraction of the overall variance from a cluster.

Finally, the last statistic value presented through the mixed linear models is the term R-squared. R-squared is the proportion of variance in the dependent variable that is explained by the independent variables value (UCLA, 2016a). Simply, R-squared is an overall term used to measure the strength of association between the two variables.

45

In conclusion, based on the data statistics available and sources relevant this thesis will use mixed linear models to produce empirically based evidence that suggests correlation between the dependent and independent variables. Data analysis using random effects models and fixed effects models will be presented in greater depth throughout chapter four. Chapter four will additionally focus on identifying and explaining the correlation statistics, which were defined above, and apply them to the relationships between search warrant SWAT usage in Maryland and the law enforcement agencies specific characteristics.

46

CHAPTER 4: DATA ANALYSIS

Introduction

The purpose of this chapter is to present the statistical results from data analysis

of the final thesis database. This chapter will begin by focusing on the basic descriptive

statistics of the data collection. A frequency table will be first, followed by a sum of

SWAT deployments, yearly distributions and then tables of measures of central tendency for the dependent and independent variables. Explanation of these tables will allow for a better understanding of the databases statistics.

Furthermore, the data analysis from hierarchical linear regression models of the mixed linear models will be described. Within this study random effects models and fixed effects models tested hierarchical linear regression for research questions one through four. Data analysis for this study was run in the software statistical program STATA.

This chapter will present the data analysis and highlight any statistically significant results.

Descriptive Statistics

Descriptive statistics were computed by simple data analysis of the final data

collection. The first subsection will show how frequently a city agency or county agency

conducted a search warrant SWAT deployment. Furthermore, the following table will

provide a sum of the type of agencies Part I and Part II SWAT deployments. Next, the

yearly distributions of SWAT deployments will be presented. Then the measures of

central tendency of each dependent and independent variable will be discussed. Measures 47

of central tendency presented will include mean, median, minimum, maximum, range and

standard deviation. Following the descriptive statistics section will be a presentation of

data analysis for each research question.

Descriptive Tables

This subsection focuses on the frequency of data in the final database. Table 1

presents a clear understanding of which case types most commonly deployed SWAT

teams for search warrants in the state of Maryland. In Maryland, the three types of law

enforcement agencies that fit the criteria for this study were city police departments,

county police departments and county offices.

The frequency table presented data showing that city police departments

accounted for 72 agencies (51.4 percent), county police departments for 20 (14.3 percent), and county sheriff offices for 48 (34.3 percent). This distribution showed that city police departments were the most common type of agency in the study, followed by county sheriff offices and then county police departments. Table 1 provides the reader with a basic understanding of the most common types of agencies that deployed SWAT units for search warrants. The total number of cases in this study totaled 140.

Table 1.

Frequency and Percent of Type of Jurisdiction Frequency Percent City Police 72 51.4 County Police 20 14.3 County Sheriff 48 34.3 Total 140 100 n=140

Another way of looking at type of agency data showed a very different pattern of

statistics. In Table 2, instead of studying the frequency of the agency cases, the table 48 shows the sum of their deployments. Table 2 represents which type of agency deployed

SWAT teams for Part I and Part II crimes the most often. This table gives a much clearer representation of agency type tendencies. Of the total 5,646 SWAT deployments in

Maryland that met the criteria for this thesis, Part I deployments accounted for 2,495 and

Part II deployments totaled 3,151.

Each of these totals has been separated into subdivisions that represent city police departments, county police departments, and county sheriff offices trends. Part I SWAT deployments show that city police departments totaled only 439 of the call-outs and county sheriff offices accounted for 507.

Although Table 1 shows that these are the most common types of cases, they do not represent the sum of deployments. Table 2 shows that county police departments, which totaled only 14.3 percent of the agency cases, reported the most individual Part I deployments at 1,549. The distribution of sums was the purpose for adding this table.

The following tables not only present counts, but also percentages of the deployments. Table 2 provides percentages for each agency (left percentage) and the percentages for each type of deployment (right percentage). For example, Part I SWAT deployments accounted for 29% of all city police deployments, while city police departments also accounted for 18% of all Part I SWAT deployments.

Table 2. Sum of Agencies Part I and Part II SWAT Deployments Part I SWAT Part II SWAT Total Deployments Deployments Type of City Police 439(29%)(18%) 1,073(71%)(34%) 1,512(100%)(27%) Agency County Police 1,549(49%)(62%) 1,589(51%)(50%) 3,138(100%)(56%) County Sheriff 507(51%)(20%) 489(49%)(16%) 996(100%)(18%) Total 2,495(44%)(100%) 3,151(56%)(100%) 5,646(100%)(100%) n=5,646 49

Additionally, Part II deployments in Maryland suggest that city police department numbers rose to 1,073 deployments. This dramatic increase shows a propensity to target vice type crimes at the municipal level. County level statistics remained quite stable compared to Part I deployments. County police departments totaled 1,589 (50%) and

county sheriff offices reported 489 (16%). These statistics show that city police are more

likely to deploy their SWAT units for Part II offenses, while county levels seem to

distribute somewhat evenly. The purpose of Table 2 was to present certain statistics from

the same thesis database and show how simple frequencies can be misleading.

Furthermore, Tables 3 and 4 presented the yearly distributions of the thesis database. Table 3 provided the yearly counts and overall totals for Part I SWAT deployments of city police, county police and county sheriff, while table 4 did the same for Part II SWAT deployments.

Table 3’s statistics show that county police deployed the most for Part I offenses every year, and that city police deployed the least every year but one (2011). County

police totals also shows that their SWAT deployments were much higher than the other two agency types combined. Table 3 suggests that county police departments are the most likely agencies to deploy SWAT teams for Part I crimes.

Additionally, the yearly distributions showed a range of consistency across all

types of agencies and also their totals. In conclusion, there was minimal variation

between the yearly counts of SWAT teams deployed for Part I crime offenses throughout

Maryland from 2010 through 2013.

50

Table 3.

Yearly Part I SWAT Deployments by Agency Type Part I SWAT Deployments 2010 2011 2012 2013 Total Type of City Police 124(21%) 119(18%) 97(16%) 99(15%) 439(18%) Agency County Police 325(55%) 424(66%) 411(67%) 389(60%) 1,549(62%) County Sheriff 137(23%) 101(16%) 108(18%) 161(25%) 507(20%) Total 586(100%) 644(100%) 616(100%) 649(100%) 2,495(100%) n= 2,495

Table 4 then presents the same distribution, although this time it focuses on the

Part II SWAT deployments throughout Maryland. Much like table 3, every year county police departments were the most likely to deploy SWAT units for Part II offenses and additionally county police totaled more than the other two types of agencies over that four-year span.

Unlike Table 3, Table 4 reveals that county sheriff offices were the least likely to deploy SWAT teams for Part II offenses each of the four years. County sheriff offices only totaled 489 deployments over the four-year span, while city police totaled at 1,073

(34%) and county police totaled 1,589 (50%).

51

Table 4.

Yearly Part II SWAT Deployments by Agency Type Part II SWAT Deployments 2010 2011 2012 2013 Total Type of City Police 234(33%) 322(38%) 259(35%) 258(30%) 1,073(34%) Agency County Police 350(50%) 406(48%) 374(50%) 459(54%) 1,589(50%) County Sheriff 119(17%) 120(14%) 112(15%) 138(16%) 489(16%) Total 703(100%) 848(100%) 745(100%) 855(100%) 3,151(100%) n= 3,151

For the final databases analysis, types of agency jurisdictions were manipulated to

yield more fairly even results. City police accounted for barely over fifty percent, so

county police and county sheriffs were combined to represent a unified county

jurisdiction. By doing so the data analysis compared only two types of agencies and the

totals were fairly even. City police department jurisdictions accounted for 51.4 percent of

the agencies (n=72). County jurisdictions accounted for 48.6 percent of the agencies

(n=68). Overall (n=140), the distribution between municipal and county agencies

provided this thesis with the ability to generalize more similarly.

Measures of Central Tendency

The purpose of this subsection is to present the measures of central tendency for

dependent and independent variables used in data analysis. All but one of the final

variables is a rate. Type of Jurisdiction is not a rate because it is a dummy variable

representing either county or city agencies. The measures of central tendency in Table 5 include mean, median, minimum, maximum, range, and the standard deviation.

52

Table 5.

Measures of Central Tendency for Dependent and Independent Variables Violent Property VICE Number SWAT SWAT Type of Crime Crime Arrests of Sworn Part I Part II Jurisdiction Rate Rate Rate Officer Deploym Deploym Rate ent Rate ent Rate Mean 462.289 3,345.393 2,118.430 237.169 12.173 27.705 0.486 Median 404.609 2,722.837 1,127.730 210.122 4.334 17.989 0.000 Min 34.887 319.841 204.000 61.008 0.000 0.000 0.000 Max 1,752.137 20,441.600 32,896.000 1,538.462 203.478 209.234 1.000 Range 1,717.250 20,121.750 32,692.000 1,477.454 203.478 209.234 1.000 SD 369.482 3,230.042 4,838.605 224.259 24.686 37.470 0.502

For the independent variable rates of violent crime the results present a minimum

of 34.887 and a maximum of 1,752.137, which is a range of 1,717.25. The average of

these figures is 462.289 with a standard deviation of 369.482. The median of the violent

crime rates was 404.609. These statistics show a large distribution of violent crime rates that may contain outliers. Violent crime is typically the least likely type of crime and these statistics confirm that for the agencies in Maryland that deploy SWAT teams.

Next, for the independent variable rates of property crime the measures of central

tendency show a minimum of 319.841 and a maximum of 20,441.6, suggesting a range of

20,121.75. Property crime rates are on average 3,345.393 with a standard deviation of

3,230.042. Furthermore, the median of Maryland agencies was 2,722.837. As a weakness

this distribution was rightward skewed.

Additionally, for the independent variable rates of vice part II arrests the measures

of central tendency show a minimum of 204 and a maximum of 32,896, providing a

range of 32,692. The statistics of vice arrests of Maryland agencies average 2,118.43 with

a standard deviation of 4,838.605. As a result, the agencies median vice arrests rates was 53

1,127.73. The variable distribution for part II arrests was also rightward skewed.

For the independent variable rates of sworn law enforcement officers the minimum was 61.008 and the maximum was 1538.462, which yields a range of

1,477.454. The mean of sworn law enforcement personnel rate is 237.169, providing a standard deviation of 224.259. In addition, the agencies median for this variable was

210.122. As a result of these statistics, the data revealed a rightward skewed distribution.

One dependent variable rates of Part I search warrant SWAT deployments for this database has a minimum of 0 and a maximum of 203.478, which is a range of 203.478.

The average rates of Part I search warrant SWAT deployments in Maryland are 12.173, with a standard deviation of 24.686. Additionally, the agencies median deployments were

4.334. Part I search warrant SWAT deployments’ distribution revealed another rightward skewed distribution.

The dependent variable rates of Part II search warrant SWAT deployments provided a minimum of 0 and a maximum of 209.234, providing a range of 209.234. The mean of this variable rate was 27.705, with a standard deviation of 37.470. Part II search warrant SWAT deployments in Maryland also had a distribution that presented a median of 37.470.

The final independent variable used in the data analysis formula was the type of jurisdiction. This variable divided the agencies into either city or county agencies. Type of jurisdiction shows a minimum of 0 and a maximum of 1, which yields a range of 1.

Zero was the dummy variable for city, while 1 corresponded to county agencies.

Moreover, the mean of this variable was 0.486.

54

In analysis of these dependent and independent variable rates show that the

majority of them portray skewed distributions. Comparing large jurisdictions like Prince

George County and Baltimore County to cities like Berlin, Chestertown and Ocean City

are going to present problems. The majority of the means and standard deviations suggest

distributions that are significantly skewed to the right. These curves suggest that there are significant outliers as large agencies, even when controlling for jurisdictional populations in the form of rates.

Research Question 1

How are Part I violent incidents, Part I property crime incidents, the type of agency and number of sworn officers predictive of Part I crime search warrant SWAT team deployments over time and across agencies? This research question was analyzed using a random effects model for Part I crime.

The null hypothesis states that there is no relationship between the numbers of

Part I violent incidents, Part I property crime incidents and the number of Part I search warrant SWAT team deployments over time and across agencies. The alternative hypothesis states that there is a relationship between the number of Part I violent incidents, Part I property crime incidents and the number of Part I search warrant SWAT team deployments over time and across agencies; in that the more Part I violent incident and Part I property crime incidents, the more Part I search warrant SWAT team deployments an agency conducts.

The random effects model for this research question yielded no statistically significant relationships. Data analysis results suggests that neither Part I violent incidents, nor Part II property crime incidents are significantly correlated to Part I search

55 warrant SWAT team deployments. As a result, the thesis retains the null hypothesis for research question one.

Table 6.

Part I Crime Rates Random Effects Model SWAT Part I Coefficient Standard Z-Value P-Value [95% Confidence Deployments Error Interval] Violent crime -0.0050 0.0119 -0.42 0.673 -0.0282 0.0182 rate Property -0.0002 0.0022 -0.11 0.911 -0.0045 0.0041 crime rate Type of -4.8679 7.0033 -0.70 0.487 -18.5942 8.8583 Jurisdiction Sworn -0.0056 0.0289 -0.19 0.846 -0.0622 0.0510 Officers Rate Overall 18.9966 7.5342 2.52 0.012 4.2298 33.7633 Intercept Rho 0.2045 (Fraction of variance due to u_i) P-value * .10 significant; ** .05 significant; *** .01 significant R-sq: Within= 0.0069; Between= 0.0245; Overall= 0.0113

Table 6 shows the coefficient, standard error term, z-value, p-value and 95% confidence interval for several independent variables. Although there are no statistically significant correlations in Table 6, the data analysis statistics are worth presenting.

The independent variable violent crime rates provided a regression coefficient of -

.005, with a standard error term of 0.0119. This shows a weak negative correlation to the dependent variable Part I SWAT deployments. Violent crime rates also produced a z- value of -0.42 and a p-value of 0.673, which is statistically insignificant. In terms of the

95% confidence level, the majority of violent crime rates for Part I SWAT deployments fell between -0.0282 and 0.0182.

Independent variable property crime rates provided a regression coefficient of

-0.0002 and a standard error term of 0.0022. Again, this is a weak negative correlation to

Part I SWAT deployments. The random effects model for property crime rates produced 56

a z-value of -0.11 and a p-value of 0.911. Both of these terms are statistically

insignificant findings. Last of all, the independent variable property crime rates provided

a 95% confidence interval between the figures -0.0045 and 0.0041.

Next, is the independent variable and dummy variable type of jurisdiction, this

variable produced the lowest regression coefficient of -4.8679 and a standard error term

of 7.0033. These findings show a strong negative correlation and the second highest error

term in the table. Furthermore, this model presented a z-value of -0.70 and a p-value of

0.487. Neither of these values are statistically significant. The 95% confidence intervals for type of jurisdiction fall between -18.5942 and 8.8583.

The independent variable number of sworn officers rate for random effects model

shows a regression coefficient of -0.0056 and a standard error term of 0.0289. Table 6

presents a weak negative correlation between this independent variable and the dependent

variable Part I SWAT deployments. Number of sworn officers rate measures a z-value of

-0.19 and a p-value of 0.846, which are both statistically insignificant. Additionally, the

random effects model provided the 95% confidence intervals of -0.0622 and 0.0510.

Moreover, the overall intercept was the final term to produce the same statistics.

Overall intercept (or constant term) provided the highest positive regression coefficient of

18.9966 and the highest standard error term of 7.5342. The overall intercept also shows a z-value of 2.52 and a p-value of 0.012. Although this p-value is not statistically significant, the p-value here shows a possible correlation between the overall intercept and the dependent variable of Part I SWAT deployments. Table 6 also shows 95%

confidence intervals of 4.2298 and 33.7633.

57

Finally, the last terms portrayed in Table 6 measure the variances within the

random effects model. The Rho term for this random effects model is 0.2045.

Furthermore, there are three r-squared terms produced from this model. R-squared within

0.0069, between 0.0245 and overall 0.0113 represent the correlation variance for this

model.

Research Question 2

How are Part II vice arrests, the type of agency and number of sworn officers predictive of Part II crime search warrant SWAT team deployments over time and across agencies? This research question was analyzed using a random effects model for Part II vice arrests.

The null hypothesis states that there is no relationship between the number of Part

II vice arrests and the number of Part II search warrant SWAT team deployments over time and across agencies. The alternative hypothesis states that there is a relationship between the number of Part II vice arrests and the number of Part II search warrant

SWAT team deployments over time and across agencies; in that the more Part II vice arrests, the more Part II search warrant SWAT team deployments an agency conducts.

Research question two analysis yielded two statistically significant relationships.

First, is the vice arrest rate. As one vice arrest increment rate unit increases, SWAT Part

II deployments decreases by 0.00067 (p-value of 0.001). This relationship was unexpected, because one would assume that the more proactive arrests an agency makes the more SWAT deployments they would have. These results suggest a statistically significant inverse relationship between vice type arrests and SWAT deployments when comparing agencies across time and department.

58

Secondly, the number of sworn officers was also statistically significant. An increase of sworn officer rate results in SWAT Part II deployments increases by 0.2066 (p-value of 0.000). This relationship was expected because the bigger the agency the more likely it would be to have and use SWAT. Other than those two variables there were no more statistically significant results drawn from the part two-arrest rates random effects model.

Table 7.

Part II Arrest Rates Random Effects Model SWAT Part II Coefficient Standard Z-Value P-Value [95% Confidence Deployments Error Interval] VICE Arrests -0.0067 0.0020 -3.28 0.001 *** -0.0107 -0.0027 Rate Type of -4.6723 11.6460 -0.40 0.688 -27.4981 18.1534 Jurisdiction Sworn Officers 0.2066 0.0487 4.24 0.000 *** 0.1111 0.3021 Rate Overall -4.8754 13.0493 -0.37 0.709 -30.4515 20.7006 Intercept Rho 0.6596 (Fraction of variance due to u_i) P-value * .10 significant; ** .05 significant; *** .01 significant R-sq: Within= 0.1710; Between= 0.2199; Overall= 0.2018

Table 7 produced other statistical values for the independent variables through the random effects model. These findings are explained and presented below.

First, independent variable vice type arrest rates were studied in contrast to the dependant variable Part II SWAT deployments. Table 7 produces a regression coefficient of -0.0067 and a standard error term of 0.0020. Standard error term for vice type arrest rate was the lowest error term for this table. Vice type arrest rates z-value was -3.28 and p-value was 0.001. This significant p-value has already been discussed. Additionally, the

95% confidence intervals were identified as -0.0107 and -0.0027.

Next, type of jurisdiction produced a regression coefficient of -4.6723 and a standard error term of 11.6460. This independent variable also provided this table with 59

the z-value of -0.40 and p-value of 0.688. Neither of the findings is statistically

significant. The 95% confidence intervals for this variable are -27.4981 and 18.1534.

Random effects model measuring independent variable number of sworn officers

rate and dependent variable of Part II SWAT deployments measured a regression

coefficient of 0.2066 and a standard error term of 0.0487. Table 7 shows a z-value of 4.24

and a p-value of 0.000. This p-value is statistically significant and previously discussed.

The 95% confidence intervals for this variable are 0.1111 and 0.3021.

Overall intercept is the final term that has these similar statistics. Table 7 presents

a regression coefficient of -4.8754 and a standard error term of 13.0493 when compared

to Part II SWAT deployments. The overall intercept has the lowest regression coefficient

and highest standard error term for this model. Moreover, the z-value is -0.37 and p-value

is 0.709, both of which are statistically insignificant. This term provided the tables largest

95% confidence interval gap at -30.4515 to 20.7006.

To conclude, the final figures to discuss are those that portray the levels of variance for this random effects model. The fraction of variance, or Rho, for this model was 0.6596. Additionally, several r-squared values were provided by the data analysis. R-

squared within was 0.1710, between was 0.2199 and overall 0.2018.

Research Question 3

How are Part I violent incidents, Part I property crime incidents, the type of agency and number of sworn officers predictive of Part I crime search warrant SWAT team deployments over time for an agency? This research question was analyzed using a fixed effects model for Part I crime.

60

The null hypothesis for research question three states that there is no relationship

between the number of Part I violent incidents, Part I property crime incidents and the

number of Part I search warrant SWAT team deployments over time for an agency.

Additionally, the alternative hypothesis states that there is a relationship between the

number of Part I violent incidents, Part I property crime incidents and the number of Part

I search warrant SWAT team deployments over time for an agency; in that the more Part

I violent incident and Part I property crime incidents, the more Part I search warrant

SWAT team deployments an agency conducts.

Research question three yielded no statistically significant relationships from the fixed effects data analysis model. The results show that neither Part I violent incidents, nor Part II property crime incidents are significantly correlated to Part I search warrant

SWAT team deployments. As a result, the thesis retains the null hypothesis for research

question three.

The first measure noted in table 8 is the lack of type of jurisdiction values because

of collinearity. By using a fixed effects model, the data analysis measured within

agencies, not across like the random effects models. This means that fixed effects model

measured each agency over the four years and therefore, did not compare across agencies

so there would be no differences over agency type.

61

Table 8.

Part I Crime Rates Fixed Effects Model SWAT Part I Coefficient Standard Z-Value P-Value [95% Confidence Deployments Error Interval] Violent crime -0.0495 0.0313 -1.58 0.117 -0.1116 0.0127 rate Property crime 0.0010 0.0053 0.18 0.858 -0.0096 0.0115 rate Type of 0 (Omitted) Jurisdiction Sworn Officers 0.2020 0.1382 1.46 0.147 -0.0721 0.4761 Rate Overall Intercept -16.0459 30.0875 -0.53 0.595 -75.7244 43.6326 Rho 0.7873 (Fraction of variance due to u_i) R-sq: Within= 0.0328; Between= 0.0275; Overall= 0.0077 P-value * .10 significant; ** .05 significant; *** .01 significant

Table 8’s first independent variable is violent crime rate from the Part I search

warrant SWAT deployments crime rates fixed effects model analysis. Data analysis

presented violent crime rates with a regression coefficient of -0.0495 and a standard error

of 0.0313. Violent crime rates also produced a z-value of -1.58 and a p-value of 0.117.

This independent variable measured 95% confidence interval values of -0.1116 and

0.0127.

Next, property crime rates measured a regression coefficient of 0.0010 and a

standard error term of 0.0053. The standard error term was the lowest of table 8. Fixed

effects analysis produced a z-value of 0.18 and a p-value of 0.858. Neither of these terms is statistically significant. Additionally, the fixed effects model for Part I crimes provided

95% confidence intervals of -0.0096 and 0.0115.

The next independent variable for table 8 is the number of sworn officers rate.

Table 8 shows a regression coefficient of 0.2020 and a standard error term of 0.1382. The

regression coefficient was the highest for table 8. Furthermore, this analysis produced a

62

z-value of 1.46 and a p-value of 0.147. These terms identify that the correlation between

the number of sworn officers rate and Part I search warrant SWAT deployments is not

statistically significant. Fixed effects analysis also presented the 95% confidence

intervals of -0.0721 and 0.4761.

The final variable that presented the similar terms was the overall intercept. The

overall intercept produced a regression coefficient of -16.0459 and a standard error term

of 30.0875. The regression coefficient was the lowest of table 8 and the standard error

term was the highest of table 8. Moreover, the overall intercept produced a z-value of -

0.53 and a p-value of 0.595. Both of these findings are statistically insignificant. To finish, the 95% confidence intervals for this measure were -75.7244 and 43.6326, which is the largest range in table 8.

Finally, the last group of measures reported in table 8 focuses on the variance of the fixed effects model analysis. Rho presented a fraction of variance of 0.7873.

Additionally, r-squared reported variance on three levels. R-squared presented terms of variance from within (0.0328), between (0.0275) and overall (0.0077).

Research Question 4

Research question four asks: How are Part II vice arrests, the type of agency and number of sworn officers predictive of Part II crime search warrant SWAT team deployments over time for an agency? This research question was analyzed using a fixed effects model for Part II vice arrests.

This research questions null hypothesis states that there is no relationship between the number of Part II vice arrests and the number of Part II search warrant SWAT team deployments over time for an agency. On the other hand, the alternative hypothesis states

63

that there is a relationship between the number of Part II vice arrests and the number of

Part II search warrant SWAT team deployments over time for an agency; in that the more

Part II vice arrests, the more Part II search warrant SWAT team deployments an agency

conducts.

Research question four yielded two statistically significant relationships. The first

was the number of vice arrest rates. As one vice arrest rate unit increases, Part II search

warrant SWAT deployment rate decreases by 0.0189 (p-value of 0.00). These results suggest a statistically significant inverse relationship between vice type arrests and

SWAT deployments when comparing each agency across time. Just like research question two Part II search warrant SWAT deployments did not increase with vice type crime rates. Additionally, the constant term, or overall intercept also resulted in a statistically significant correlation (p-value of 0.017). This correlation is positive with a

77.2077 coefficient.

As a result of research question fours fixed effects model data analysis the thesis rejects the null hypothesis. There is a statistically significant relationship between vice type crime rates and part two search warrant SWAT deployment rates, yet the direction of the relationship was not predicted.

The first aspect of Table 9 that stands out was that type of jurisdiction was omitted

once again. Similarly to Table 8, Table 9 omitted type of jurisdiction because of collinearity. Fixed effects focused on within agencies and not across, therefore there were no changes in agency types.

64

Table 9.

Part II Arrest Rates Fixed Effects Model SWAT Part II Coefficient Standard Z-Value P-Value [95% Confidence Deployments Error Interval] VICE Arrests -0.0189 0.0037 -5.10 0.00 *** -0.0263 -0.0116 Rate Type of 0 (Omitted) Jurisdiction Sworn Officers -0.0395 0.1141 -0.35 0.730 -0.2658 0.1868 Rate Overall 77.2077 31.8913 2.42 0.017 ** 13.9587 140.4566 Intercept Rho 0.9771 (Fraction of variance due to u_i) R-sq: Within= 0.2433; Between= 0.2511; Overall= 0.1713 P-value * .10 significant; ** .05 significant; *** .01 significant

Data analysis shown in Table 9 reveals that vice type arrest rates had a regression

coefficient of -0.0189 to Part II search warrant SWAT deployments. Vice type arrest rates also produced a standard error term of 0.0037. This standard error term was the lowest presented in table 9 and so was the regression coefficient. Additionally, this independent variable measured a z-value of -5.10 and p-value of 0.00. This p-value is statistically significant and was previously discussed. The 95% confidence intervals for vice type arrests rates were -0.0263 and -0.0116.

Next, independent variable number of sworn law enforcement officers rate was measured in this fixed effects model analysis. Sworn officers rate produced a regression coefficient of -0.0395 and a standard error term of 0.1141. Table 9 shows a z-value of -

0.35 and a p-value of 0.730. Both of these figures are statistically insignificant. Fixed effects model for Part II search warrant SWAT deployments provided 95% confidence intervals of -0.2658 and 0.1868 for the number of sworn officers rate.

65

The final measurement from table 9 that are similar to the previous independent variable was the overall intercept. Overall intercept revealed a regression coefficient of

77.2077 and a standard error term of 31.8913. Both of these figures were the highest

values of table 9. Table 9 also shows a z-value of 2.42 and a p-value of 0.017. This

reflects a statistically significant correlation that was discussed previously. Fixed effects

analysis produced 95% confidence intervals of 13.9587 and 140.4566, which was the

largest range of table 9.

To conclude, several other terms revealed statistical trends about the variance of the fixed effects analysis. Fraction of variance, or Rho, produced a term of 0.9771. This fraction of variance is extremely high. Furthermore, the data analysis revealed several r-

squared values. Table 9 provides an r-squared value from within (0.2433), between

(0.2511) and overall (0.1713).

Summary of Findings

To conclude, specific statistical trends from the final database are discussed and revisited from the preceding sections. The most important material gathered from data analysis will be highlighted in this section.

Chapter four began by presenting several tables with simple statistical measures.

Table one focused on the cases in the dataset and presented a frequency and percent table of the types of jurisdictions. This table showed the most common unit was the city police departments. Further analysis in table twos’ sum of the agencies Part I and Part II SWAT deployments suggested different results. After being the smallest percent in table one, table two showed that county police departments actually totaled the most individual deployments at 3,138 of the total 5,646 deployments. The purpose of table three was to

66

provide a yearly distribution of Part I and Part II SWAT deployments. This table showed that 2013 was the highest distribution with 1,504 deployments.

Following tables one through three was the measures of central tendency table.

This table provided specific measures for the two dependent variables along with the five independent variables. In review of this table, specific variable were skewed to the right.

Large county jurisdictions were significant outliers and skewed the distribution curve.

In addition, tables five and six tested the mixed linear model random effects to identifying statistical correlations over time and across agencies. Table five analyzed Part

I crime rates using the random effects model and identified no statistically significant relationships. On the other hand, table six identified two statistically significant relationships. The first was vice type arrest rates were inversely correlated to Part II search warrant SWAT deployments. Additionally, table six suggests that the number of sworn law enforcement officers is statistically significant and is positively correlated.

Finally, Tables 7 and 8 tested the mixed linear model fixed effects to find statistically significant correlations over the yearly of an agency. Table seven tested for correlations between Part I crime rates and Part I search warrant SWAT deployments, finding no statistically significant correlations. The final table of this thesis was for the fixed effects model of vice type Part II arrest rates and Part II search warrant SWAT deployments. Table eight showed two statistically significant effects. Like table six, table eight showed a negatively significant correlation between vice type arrest rates and

SWAT deployments. Also, table eight shows that the constant term was statistically

significant and positive.

67

Chapter four provided empirical evidence into the study of police paramilitary units in America. Maryland’s SWAT deployments, crime rates and agency characteristics produced limited statically significant correlations. Although the statistical results did not match expectations, the analysis can be used as a building block for future research.

Chapter five will discuss the implications of these results in more depth. Also, it will discuss the limitations, policy implications and provide future recommendations for police paramilitary research.

68

CHAPTER 5: DISCUSSION & CONCLUSION

Discussion of the Findings

The purpose of this section is to further understand the statistical results from chapter four. Of the four mixed linear models, two were random effects models and two were fixed effects models. From these four statistical models data analysis identified four statistically significant relationships. These statistically significant relationships will be discussed in greater detail and their implications will additionally be presented.

Table 7 presented the random effects model, which analyzed Part II vice arrest rates and Part II search warrant SWAT deployments in Maryland. From this model two significant correlations were identified. The first statistically significant correlation suggests that as vice arrests increase, SWAT deployments conversely decrease. This relationship can possibly be explained by agencies working higher caseloads, and therefore do not have time to deploy SWAT as often as they would like.

Another explanation could be that law enforcement agencies do not feel pressured into excessively using SWAT for vice type crimes. That is they may not have the resources or manpower to send out SWAT units after a certain number of deployments.

This explanation describes a type of threshold, or maximum limit to deploying SWAT.

This result, a significant negative correlation, was not hypothesized and unexpected from proactive vice type arrests. This relationship will be further discussed later on.

Additionally, Table 7 suggests that the number of sworn law enforcement officers is statistically significant and positively correlated to Part II search warrant SWAT 69 deployments. This relationship is logical and was expected. As a law enforcement entity increases in size, there is more funding available for programs such as SWAT. Also, as the number of personnel increases, it is expected that so will the number of willing persons whom are interested and qualified for training special units such as SWAT. The larger and more diversified an agency becomes, allows that agency the ability to implement and conduct many different programs. A statistically significant positive correlation between the number of sworn law enforcement officers and the rate of Part II search warrant deployments was an unsurprising outcome.

Furthermore, through the fixed effects model analyzing Part II vice arrest rates and Part II search warrant SWAT deployments, two more statistically significant correlations were presented in Table 9. Vice arrests and the constant term both revealed significant correlations.

Vice arrests were once again negatively correlated to SWAT team deployments in

Maryland. As previously stated, this result was not only unexpected, but the opposite of what this thesis hypothesized. This negative correlation between vice type arrests and

Part II search warrant SWAT deployments will be further discussed in depth.

Finally, to revisit the negative correlation between Part II vice arrests and Part II search warrant deployment, for both random effects (Table 7) and fixed effects (Table 9), these correlations are the most empirically significant and influential findings of this thesis study. Logical assumption suggests that as vice type arrests increase (which would be a predictor to a proactive subculture within a specific agency) so should SWAT deployments (which would also be deemed predictive to a proactive subculture).

The vice arrest variables in this thesis study were specifically chosen as a way to

70

gauge the proactive nature/subculture of an agency. This thesis hypothesized that if an agency were more proactive towards crimes such as prostitution, narcotics, liquor law violations, driving under the influence, disorderly conduct and vagrancy, then the agency would most likely be more prone to proactive deployments of SWAT. This thesis hypothesized a positive statistically significant correlation by relating the aggressive proactive policing strategies of an agency, to their willingness to deploy SWAT for proactive deployments.

The fact that these vice arrests suggest a negative correlation to Part II SWAT deployments is revealing. Perhaps agencies did not have the ability to deploy SWAT as much as they wanted to or needed. Possibly, agencies did not feel the need for SWAT and carried out more search warrants with routine police officers. Perhaps a large portion of these vice arrests reflects petty crimes, more than the high-risk SWAT needed search warrants.

This study should be replicated, to see if these trends are common in other agencies, since these findings suggest a negative correlation, but only for the state of

Maryland. Other states with harsher sentencing practices or more militaristic police values may indeed reveal significant positive correlations between search warrant SWAT deployments and proactive vice type crime arrests.

Policy Implications

Empirical findings from this study and reviewing of the research on police paramilitary units in America reveals the need for several policy changes. The conclusion is that the most effective and significant way to influence police paramilitary units is through new legislation. For example, the Maryland SWAT Team Reports were enacted

71

legislature that required the collection and dissemination of every SWAT team

deployment in the state. As a guide, Maryland’s legislation should be duplicated across

the country.

Specific policy implications can be concluded from this research study. American

law enforcement entities are self-governed, without a central command structure. In order for significant change across all these entities, Congress needs to enact legislature to

govern PPUs in several ways. These policy changes will provide a number of

propositions to the study and practice of police paramilitary units. Last of all, educational programs should be mandatory for PPU personnel to educate them on the dangers of emulating elite military teams.

To begin with, the study of PPUs must start with policies focused on data collection and dissemination of results. The first major attempt to do this was the initial implementation of Maryland’s SWAT Team Reports by Senate Bill 447. The Maryland

SWAT team-reporting program should be reproduced across the country and re-

implemented in Maryland. So far one other state (Utah) has begun collecting similar

SWAT information and more should follow (Balko, 2015). The importance of data

collection and data analysis on PPUs across the country is to reveal certain trends and for

police practitioners to adjust their programs based on information and results. To collect a

significant amount of this data, policymakers, academics and practitioners must be

pressured by legislature.

One important distinction of this policy is that unlike Compstat, which was

created and implemented by the New York Police Department, these policies must be

enacted into law (Walsh, 2001). Compstat was not enacted legislature, and without the

72

pressures of the law this led to numerous agencies falsified crime statistics and reports

(Walsh, 2001). In Maryland, Senate Bill 447 enacted SWAT team reports by passing

legislation. It is suggested that by enacted the data collection into law, with criminal

consequences for manipulated records, there will be far less falsified reports than

programs such as Compstat.

By doing so, a significant amount of trustworthy data can be collected on police

paramilitary units. This data can then be used to compare crime rates, arrest rates,

jurisdictional size, and also study the impacts that these programs have on the

surrounding communities.

Furthermore, a number of implications for policy are derived from the literature

review, and not the thesis’ original research. The stigma of secrecy among SWAT units

must be removed. Mainly this must occur because the program is a governmentally

funded program. Since SWAT programs are governmentally funded they should be

transparent and the public at large should know their impacts and results, for better or

worse. Increased transparency can portray a confident and community friendly law

enforcement agency. Furthermore, a transparent police department is not viewed as an

adversary to the common citizen.

Moreover, significant policies must be passed to implement standardized training

and increase oversight standards for SWAT units. Since police agencies across the United

States are not centrally governed, SWAT standards and training are extremely diverse.

By enacting training standards for all SWAT personnel will increase the consistency of their training and could save the lives of officers and offenders.

One way to implement this strategy would be to pass legislation requiring each

73

state to have a training school and course that all SWAT personnel in the state must pass.

Governmental oversight will increase along with the increasing of consistency in SWAT

training. Standardized training increases professionalism of SWAT officers and increases

oversight of such programs, which will lead to more effective and safe deployments.

Finally, the last policy implication would be to create educational programs that teach police paramilitary unit personnel that there are significant dangers with emulating

Navy SEALs and Army Rangers. If SWAT team members were briefed on the consequences of viewing criminals as wartime enemies, incidents like the Waco raid could be prevented. These educational sessions would not try to persuade PPU members that criminals are friends, but its goal would be to limit the subcultures militaristic ideals and try to remind PPU personnel that targets are still American citizens and presumed innocent until proven guilty.

Limitations

In development of this thesis, several limitations severely narrowed the range of

empirical research that could be conducted. The majority of these limitations focused

around the size and type of the data collected. Other limitations focused on the state of

Maryland itself, along with its policies, politicians, demographics and polarity of citizens.

Significant limitations to this study of police paramilitary units will be presented and

discussed in the following paragraphs.

The first set of limitations for this study focused on the lack of SWAT data. After

examination of the SWAT Team Reporting data two issues were identified. The data

significantly lacked a long enough time element and it did not provide a large enough

amount of cases.

74

First, the SWAT Team Reports provided what initially seemed to be six years of deployments. Unfortunately, the data was collected for fiscal years and therefore two half years of call outs had to be discarded. By cutting these two half years of data, the final thesis database lost potentially a few thousand search warrant SWAT deployments. In order to compare the SWAT deployments to the UCRs crime rates and arrest rates the data needed to be full annual years and not fiscal years.

Additionally, losing these two half years worth of data made the final database only four full years. This significantly limited the types of data analysis that could be conducted. With only four years of data a true time series analysis was not an option, thus leading to the mixed linear models of fixed effects and random effects. If the data spanned a longer time span a true time series analysis would have been very effective at revealing correlations.

Secondly, another limitation from the data was the apparent lack of cases. After downsizing the database to only Part I and Part II search warrant SWAT deployments only thirty-five separate agencies remained. This low number severely limited data analysis, which is why the units of analysis had to be adjusted to agency plus year. This multiplied the final database by four and provided the thesis with 140 cases. This adjustment was needed to run any significant data analysis models. If the data represented more agencies and years the data analysis could possibly have been able to identify more statistically significant relationships.

Furthermore, another limitation of this research study was the states political affiliations and demographics. Maryland would be considered a strongly “blue” state.

Over the past ten presidential elections their electoral votes have gone to the democratic

75

candidate eight times. A democratic allegiance within a small state such as Maryland

makes generalizations difficult to compare to the rest of the country. Specifically states

such as California, Texas and Florida, which are highly punitive and mostly conservative,

cannot be expected to produce the same SWAT data analysis results as Maryland’s

democratic policies.

Based on the ranking tables from United States Census Bureau’s 2010 statistical

abstract, Maryland is significantly ranked in several demographic areas. Firstly,

Maryland is ranked 9th in violent crime (U.S. Census Bureau, 2015). This high ranking can be interpreted to show the need for law enforcement to be aggressively proactive.

On a positive note, Maryland’s resident population ranks 19th, with a ranking of

6th in both annual pay and personal income per capita (U.S. Census Bureau, 2015).

Maryland also ranks 4th on the list for persons 25 years old and over with a bachelors

degree or higher (U.S. Census Bureau, 2015). These demographics show a population

that is paid more and have more educated than the majority of the country, and because

of that the correlation results of the data may not be accurately generalized for the rest of

the country.

Additionally, Maryland is a very polarized community, which significantly limits

the data analysis results. Of the twenty-three counties, Baltimore County and Prince

George County crime statistics are immensely higher than the rest of the state. This conclusion is best represented by the skewed data from this thesis.

By running an analysis comparing very high population and crime counties to other low population and crime counties can be very misleading. Even using rates the way this thesis did can present limitations. The rural and urban areas in Maryland are

76 quite different and would produce conflicting results that could not be generalized to the countries population. Furthermore, by comparing such dissimilar counties in the same data analysis could create an average, which takes away from each counties individual characteristics.

Specific and unique demographics of Maryland considerably influence the states beliefs in crime fighting. Political affiliations within the state significantly affect

Maryland’s policies, laws, and criminal justice system. The distinctive combinations of factors make Maryland unique and therefore limit its ability to generalize for the

American population. To summarize, although this study resulted in statistically significant correlations in Maryland, the process must be reproduced in other states to properly compare SWAT usages in America.

Recommendations for Future Research

The purpose of this thesis was to provide empirically based evidence that sheds light onto the study of police paramilitary units in America. Outdated, small sample sizes and qualitative characteristics studies have limited prior research on PPUs.

Recommendations specific to this study focus on the performed data analytics on search warrant SWAT deployment in Maryland from 2010 through 2013. Correlations between these deployments and Maryland’s crime rates, arrest rates, jurisdictional population and number of sworn law enforcement officers presented several statistically significant relationships that can be generalized to the American population at large. This section will discuss how, why and what future recommendations should be implemented for the study and practice of police paramilitary units.

Furthermore, several recommendations are presented that focus on innovating and

77

creating new types of SWAT research studies. PPU research needs to strive to further

identify relationships between SWAT usage and the surrounding characteristics of the

agency. Empirical correlations between crime trends, arrest rates, jurisdictional

populations and SWAT deployments can influence legislation, policies and procedures.

SWAT programs are already extremely popular in America and their impacts on different

communities can be quite insightful into American policing. In short, empirical studies

must replicate this thesis’ formula to identify national trends in SWAT usage.

Next, a major innovation on the study of SWAT should be to use GIS mapping

systems to plot where the deployments took place. SWAT raids should be studied in

relationship to criminal hot spots and the types of communities that they impact.

Potentially identifying a statistically significant relationship between agencies SWAT

deployments and a cities criminal hot spot would be a colossal accomplishment in the study of PPUs.

By discovering the most likely places SWAT teams are used could allow police to

proactively combat such crimes before the SWAT team is actually needed. For example,

if SWAT is constantly deployed to a specific hot spot for low-level drug arrests, then possibly implementing social programs or a style of broken windows policing to that area could decrease the need of SWAT raids needed. Identification of a common location or reason for deployments could be addressed by another strategy such as problem-oriented policing.

Furthermore, implementing GIS mapping of SWAT deployments can shed light on whom these units typically target and how centrally located they are. Data collection on the race, gender, age, education level, social class and other characteristics can be

78

extremely valuable to the study of PPUs impacts on society. By collecting this

information researchers could possibly identify other correlations to the criminal justice system overwhelmingly targeting minorities. Targets of the most elite and dangerous police unit should be classified and statistical trends must be identified to further research on the subject.

One of the most recent innovations being tested in American policing is the use of body worn cameras. Implementing body cameras to SWAT personnel throughout their deployments could have extreme positive effects on the officers’ course of action and safety. By using these cameras SWAT officers can show on video where and why they used certain tactics and force. Not only will these cameras hold responsible to their degree of force used, it will protect them from possible lawsuits. If an officer can show exactly where and why he/she used a certain amount of force, it can protect them if the victim takes a case to court.

Body cameras have numerous impacts on routine police officers, but the intensity and dangerousness of SWAT raids provides SWAT personnel with the ultimate reason to document and be able to explain their actions. Thus, body worn cameras by SWAT personnel would not only protect the officer, it will increase the safety of innocent bystanders. To finish, recording deployments could be extremely helpful in training of

SWAT officers. Officers could then re-watch the raid and look for mistakes to correct.

Much like athletes studying film before a game, SWAT officers could perfect their techniques and relive situations to perform at a higher level on the next deployment.

In addition, SWAT is a potentially lethal and hyper dangerous program in modern

American policing and therefore the use of it should be conducted with an excessively

79

limited risk to innocent bystanders. A co-implementation of the SWAT protocol and

procedures with intelligence-led policing strategies could increase safety and limit the

risk to non-targets. Furthermore, co-implementing intelligence led policing with SWAT units should lead to a higher percentage of successful SWAT raids. An increased amount of intelligence about the targets, neighborhood and community could decrease the errors of SWAT teams, which have been highly publicized.

Conclusion

Police paramilitary units are extremely popular across the United States and with them come specific stigmas, subcultures and beliefs. All of which, if unchecked, can drastically influence areas of American policing into becoming a more militarized entity.

PPUs have the ability to raid residential homes with lethal force and commonly do so to serve an immense amount of search warrants, which are typically for low-level drugs users ever since the societal drug wars of the 1970s. American citizens civil rights have been infringed by these policies and procedures.

Moreover, standardized training and lack governmental oversight do not control the hyper-dangerous natures of these units. Police paramilitary officers are commonly armed with M16s, grenade launchers or sniper rifles and transported in armored personnel carriers. These potentially lethal units are now typically used for more proactive deployments than the reactive deployments for which they were initially created. Transitioning to these proactive deployments, such as search warrants, suggests an increasing aggressive nature of law enforcement entities in American.

Police paramilitary units if not used properly, can cause more harm than good. In addition to standardized training and governmental oversight, these units’ personnel need

80

to be educated and influenced to adopt beliefs not associated with militarization. SWAT

members who view criminals as the adversary could lead to severe and potentially lethal

consequences.

In conclusion, further examination of the societal causes and impacts surrounding

police paramilitary units in America needs to become a priority in criminology. The surge

of PPU research of the late 1990s and early 2000s must be taken to the next level to

further learn about these units. Articles only documenting the rise and normalization of police paramilitary units are no longer good enough.

Empirical evidence suggesting certain trends correlated to police paramilitary units and their effectiveness, impacts and reasons for deployments must be studied on a national level. Future police paramilitary practices must be guided by the safety of

citizens and not by politically motivated policymakers. Police paramilitary units such as

the American Special Weapons And Tactics teams are potentially too dangerous for

researchers and practitioners to not fully understand their societal impacts.

81

BIBLIOGRAPHY

American Civil Liberties Union (ACLU). (2014). War comes home: The excessive

militarization of American policing. New York, NY: ACLU Foundation.

American Civil Liberties Union. (2015). 2015 State legislation priorities. ACLU of

Maryland. Retrieved from http://www.aclu-

md.org/our_work/2015_policy_priorities

Andreas, P., & Price, R. (2001). From war fighting to crime fighting: Transforming the

American national security state. International Studies Review, 3(3), 31-52.

Balko, R. (2006). Overkill: The rise of paramilitary police raids in America. Washington,

D.C.: Cato Institute.

Balko, R. (2014). Rise of the warrior cop: The militarization of America’s police forces.

New York, NY: Public Affairs.

Balko, R. (2015, March 19). Maryland legislature mulls new SWAT transparency bill.

The Washington Post. Retrieved from https://www.washingtonpost.com/news/the-

watch/wp/2015/03/19/maryland-legislature-mulls-new--transparency-bill/

Beede, B. R. (2008). The roles of paramilitary and militarized police. Journal of Political

and Military Sociology, 36, 53-63.

Brame, R., Bushway, S., & Paternoster, R. (1999). On the use of panel research designs

and random effects models to investigate static and dynamic theories of criminal

offending. Criminology, 37(3), 599-642.

82

Bridenball, B., & Jesilow, P. (2005). Weeding criminals or planting fear: An evaluation

of a weed and seed project. Criminal Justice Review, 30, 64-89.

Bull, D., & Stratta, E. (1995). Police-community consultative committees: A response to

paramilitary policing? Journal of Sociology, 31(3), 67-82.

Campbell, D. J., & Campbell, K. M. (2010). Soldiers as police officers/police officers as

soldiers: Role evolution and revolution in the United States. Armed Forces &

Society, 36(2), 327-350.

Clark, J. G., Jackson, M. S., Schaefer, P. M., & Sharpe, E. G. (2000). Training SWAT

teams: Implications for improving tactical units. Journal of Criminal Justice,

28(5), 407-413.

den Heyer, G. (2014). Mayberry revisited: A review of the influence of police

paramilitary units on policing. Policing and Society, 24(3), 346-361.

Department of Justice. (2004). The weed and seed strategy. Washington, D.C.: Office of

Justice Programs, Community Capacity Development Office.

Dodge, M., Valcore, L., & Gomez, F. (2011). Women on SWAT teams: separate but

equal? Policing: An International Journal of Police Strategies & Management,

34(4) 699-712.

Galyean, K. D., Wherry, J. N., & Young, A. T. (2009). Valuation of services offered by

mental health professionals in SWAT team members: A study of the Lubbock,

Texas SWAT team. Journal of Police and Criminal Psychology, 24, 51-58.

Herzog, S. (2001). Militarization and demilitarization processes in the Israeli and

American police forces: Organizational and social aspects. Policing and Society,

11(2), 181-208.

83

Hill, S., & Beger, R. (2009). A paramilitary policing juggernaut. Social Justice, 36, 25-

40.

Hill, S. M., Beger, R. R., & Zanetti, J. M., II. (2007). Plugging the security gap or

springing a leak: Questioning the growth of paramilitary policing in US domestic

and foreign policy. Democracy and Security, 3(3), 301-321.

Huq, A. Z., & Muller, C. (2008). The war on crime as a precursor to the war on terror.

International Journal of Law, Crime and Justice, 36(4), 215-229.

Jefferson, T. (1990). The case against paramilitary policing. Philadelphia, PA: Open

University Press.

Kappeler, V. E., & Kraska, P. B. (2015). Normalising police militarization, living in

denial. Policing and Society, 25(3), 268-275.

Kiker, C. R., III. (2015). From Mayberry to Ferguson: The militarization of American

policing equipment, culture and mission. Washington and Lee Law Review, 71(4),

282-298.

Klinger, D. A., & Rojek, J. (2008). Multi-method study of special weapons and tactics

teams. Washington, D.C.: National Institute of Justice.

Kraska, P. B. (1999). Questioning the militarization of U.S. police: Critical versus

advocacy scholarship. Policing and Society, 9(2), 141-155.

Kraska, P. B. (2001). Militarizing the American criminal justice system: The changing

roles of the armed forces and the police. Boston, MA: Northeastern University

Press.

Kraska, P. B. (2007). Militarization and policing: Its relevance to 21st century police.

Policing: A Journal of Policy and Practice, 1(4), 501-513.

84

Kraska, P. B., & Cubellis, J. T. (1997). Militarizing Mayberry and beyond: Making sense

of American paramilitary policing. Justice Quarterly, 14(4), 607-629.

Kraska, P. B., & Kappeler, V. E. (1997). Militarizing American police: The rise and

normalization of paramilitary units. Social Problems, 44, 1-18.

Kraska, P. B., & Paulsen, D. J. (1997). Grounded research into U. S. paramilitary

policing: Forging the iron fist inside the velvet glove. Policing and Society, 7(4),

253-270.

Lilley, D. (2015). The weed and seed program: A nationwide analysis of crime outcomes.

Criminal Justice Policy Review, 26(5), 423-447.

McCulloch, J. (2004). Blue , khaki police and the cavalry on the new American

frontier: Critical criminology for the 21st century. Critical Criminology, 12(3),

309-326.

Pratt, T. C. (2000). Random and fixed effects models in meta-analysis in criminal justice

and criminology: Choosing the appropriate method. Journal of Crime and Justice,

23, 65-79.

Phillips, S. W. (2014). Myths, militarism and the police patrol rifle. Policing and Society,

DOI:10.1080/10439463.2014.922088.

Radenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and

data analysis methods (2nd Ed.). Thousand Oaks, CA: Sage Publications, Inc.

Radosh, P. F. (2008). War on drugs: Gender and race inequities in crime control

strategies. Criminal Justice Studies, 21(2), 167-178.

Rantatalo, O. (2012). The miscellany of militaristic policing: A literature review. Journal

of Policing, Intelligence and Counter , 7(1), 51-65.

85

Sarao, S. (2014). The job of SWAT: Understanding the need for Special Weapons and

Tactics capabilities. Harvard Kennedy School Review, 14, 80-82.

Spohn, C. (2009). How judges decide? The search for fairness and justice in punishment.

Thousand Oaks, CA: Sage Publications, Inc.

Super, J. T. (1995). Psychological characteristics of successful SWAT/Tactical Response

Team personnel. Journal of Police and Criminal Psychology, 10(3), 60-63.

UCLA. (2016a). Stata annotated output logistic regression analysis. Institute for Digital

Research and Education. Retrieved from:

http://www.ats.ucla.edu/stat/stata/output/stata_logistic.htm

UCLA. (2016b). Stata library: Analyzing correlated (clustered) data. Institute for Digital

Research and Education. Retrieved From:

http://www.ats.ucla.edu/stat/stata/library/cpsu.htm

United States Census Bureau. (2015, September 3). 2010 Statistical abstract: State

Rankings. Retrieved from:

https://www.census.gov/library/publications/2009/compendia/statab/129ed/rankin

gs.html

Wasserstein, R. L., & Lazar, N. A. (2016). The ASA’s statement on p-values: Context,

process, and purpose. The American Statistician,

DOI:10.1080/00031305.2016.114108

Waddington, P. A. J. (1993). ‘The case against paramilitary policing’ considered. British

Journal of Criminology, 33(3), 353-373.

Waddington, P. A. J. (1999). Swatting police paramilitarism: A comment on Kraska and

Paulsen. Policing and Society, 9(2), 125-140.

86

Walsh, W. F. (2001). Compstat: An analysis of an emerging police managerial paradigm.

Policing: An International Journal of Police Strategies and Management, 24(3),

347-362.

Weber, D. C. (1999). Warrior cops: The ominous growth of paramilitarism in American

police departments. Washington D.C.: Cato Institute.

Weiss, T. (2011). The blurring border between the police and the military: A debate

without foundations. Cooperation and Conflict, 46(3), 396-405.

Williams, J. J., & Westall, D. (2003). SWAT and non-SWAT police officers and the use

of force. Journal of Criminal Justice, 31(5), 469-474.

87