University of PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

2019-04 Accidental Residential Fires in Calgary: A Theoretical Consideration

Skidmore, Olivia A. B.

Skidmore, O. A. B. (2019). Accidental Residential Fires in Calgary: A Theoretical Consideration (Unpublished master's thesis). University of Calgary, Calgary, AB. http://hdl.handle.net/1880/110270 master thesis

University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca

UNIVERSITY OF CALGARY

Accidental Residential Fires in Calgary: A Theoretical Consideration

by

Olivia A.B. Skidmore

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF MASTER OF ARTS

GRADUATE PROGRAM IN SOCIOLOGY

CALGARY,

APRIL, 2019

© Olivia A.B. Skidmore 2019

Abstract Past research on fire incidence rates has identified strong associations between fire risk and sociodemographic and spatiotemporal factors but has often failed to place these factors within a theoretical framework. Borrowing from the field of criminology, I draw on the framework of routine activities theory (RAT) to explore the potential applicability of the “capable guardianship” aspect of RAT as it applies to fire incidents. I also examine how the relationship between fire incidents and capable guardianship is influenced by neighbourhood disorder.

Using multivariate linear regression, I found that fire incidence rate in Calgary varies based on a community’s level of neighbourhood mobility and household size, as well as level of disorder. However, my findings did not support the hypothesis that fire rates are lower in areas characterized by increased presence of capable guardianship. I conclude that in order to examine mechanisms through which fire events are disrupted or prevented via the presence of capable guardians, research needs to focus on refining and understanding the measure of capable guardianship as it applies to fire incidents.

ii

Acknowledgements I would like to take this opportunity to express my gratitude to all the people who played a role in completion of this thesis. In my case, it truly took a village to help me get to the place where I am today.

I would like to start by thanking my supervisor, Dr. Erin Gibbs Van Brunschot, for encouragement, thoughtful feedback, and never-ending patience. Thank you for taking me and my research on and your guidance along the way. I would not be here if it were not for you!

I also want to thank my graduate program directors, Dr. Fiona Nelson and Dr.

Liza McCoy, and the graduate program administrator, Ines Kubicek. I was not the easiest graduate student and have tried your patience on many occasions with my missed deadlines, absenteeism, and need for extensions. Thank you for seeing me though the program and always being supportive!

Of course, this research would not be possible without the data from the Calgary

Fire Department. I want to thank Deputy Chief Ken Uzeloc for his support and allowing me to use the data. I also want to thank CFD’s Strategic Services team, of which I was a part for over four years. In particular, I want to extend my gratitude to my former supervisor and current friend/sister – Erin Corrigan, who taught me everything I know about research in the fire service and continued to support my journey through graduate school however and whenever she could. Thank you, sister!

Lastly, I would like to thank my children and their father, who put up with my arduous journey to completion long enough! I am finally free on the weekends!

iii

To men and women of the Calgary Fire Department,

and two of my favourite firefighters, A.B. and R.S.,

who will most definitely never read this.

iv

Table of Contents Abstract ...... ii Acknowledgements...... iii Dedication ...... iv Table of Contents ...... v List of Tables ...... vii

CHAPTER ONE: INTRODUCTION ...... 1 Purpose of the Study ...... 2 The Scope ...... 3 Overview of the Thesis ...... 3

CHAPTER TWO: REVIEW OF THE LITERATURE ...... 5 Neighbourhood Disorder ...... 5 “Broken Windows” Theory...... 6 Neighbourhood Disorganization Theory ...... 7 Linking Neighbourhood Disorder to Crime Events...... 9 Routine Activities Theory...... 11 The Concept of Guardianship ...... 13 Distribution of Fire Incidents ...... 16 Neighbourhood-Level Socio-Economic Characteristics and Fires ...... 16 Neighbourhood-Level Spatial and Temporal Features and Fires ...... 23 Conclusion ...... 27

CHAPTER THREE: RESEARCH DESIGN ...... 28 Historical Overview of Calgary and the Calgary Fire Department ...... 29 Overview of Calgary Fire Department Incidents ...... 34 Data Sources ...... 35 Fire Incidents: Data Collection ...... 38 Research Questions ...... 40 Measures ...... 41 Dependent Variable – Neighbourhood Fire Incidence Rate ...... 41 Independent Variable #1 – Capable Guardianship ...... 42 Independent Variable #2 – Social Disorder ...... 44 Control Variables...... 44 Statistical Methodology ...... 45 Conclusion ...... 47

CHAPTER FOUR: RESULTS ...... 48 Descriptive Statistics ...... 48 Bivariate Associations ...... 53

v

Regression Results 1: Guardianship as the Average Weeks Worked in 2010 ...... 56 Regression Results 2: Guardianship as the Average Number of Persons per Household ...... 59 Regression Results 3: Guardianship as the Average Number of Movers ...... 61 Conclusion ...... 62

CHAPTER FIVE: CONCLUSION ...... 63 Discussion of Regression Results ...... 64 Discussion on Capable Guardianship ...... 66 Limitations ...... 70 Conclusion ...... 72

BIBLIOGRAPHY ...... 74 APPENDICES ...... 86 Appendix A: Fire Incident Form...... 86 Appendix B: List of Excluded Neighbourhoods, by Dataset ...... 87 Appendix C: List of Community Name References for the Abbreviations in Figure 1 ...... 91 Appendix D: Variance Inflation Factor: Regression Models ...... 99 Appendix E: Data Use Agreement ...... 100

vi

List of Tables and Figures

Table 1 – Descriptive Statistics of the Dependent, Independent, and Control Variables ...... 49

Figure 1 – Average Fire Incidence Rate Distribution in the Calgary Neighbourhoods, 2012 -

2017 ...... 50

Table 2 – Bivariate Correlations Between the Variables in the Sample ...... 53

Table 3 – Multiple Regression for the Relationship between Guardianship 1 (measured as

“Average weeks worked in 2010”), Disorder, and Fires ...... 57

Table 4 – Multiple Regression for the Relationship between Guardianship 2 (measured as

“Average number of persons per household”), Disorder, and Fires ...... 59

Table 5 – Multiple Regression for the Relationship between Guardianship 3 (measured as

“Neighbourhood Proportion of Movers”), Disorder, and Fires...... 61

vii

CHAPTER ONE: INTRODUCTION

There are few events that have greater potential destructive power in terms of injury, trauma, and economic loss than a fire. In Alberta, fire events annually cost society millions of dollars and are estimated to cause an average of 27 deaths and 235 injuries each year (The

Alberta Fire Commissioner’s Statistical Report, 2013). While fire events can occur in any settings – residential/household, commercial buildings, grasslands – this research focuses on the phenomena that Merrall (2002) calls “anthropogenic accidental dwelling fires”, or AADF.

AADF is defined as “incidents of fire resulting from the interaction of people and their domestic dwelling environment” (Merrall, 2002, p. 1). In fact, the majority of fire incidents are accidental and arise from daily activities such as cooking, smoking, and heating (FEMA, 2011;

Watson, Gamble, and Schofield, 2002).

My thesis considers the fact that AADF phenomena are not random but are rather disproportionately concentrated in areas that share common socioeconomic and environmental characteristics. In the last decade, researchers concerned with spatial and temporal patterns of fire events have identified the types of factors associated with fire risk but have yet to place these factors within a theoretical framework to help further understand the complex situations in which fires occur. However, the field of criminology offers a foundation from which a theory of fire events can be developed. Frameworks such as routine activities theory can be used to gain an understanding of situations that result in fire events. Routine activities theory maintains that crime occurs as a product of normal daily activities. That is, offenders do not go out of their way to commit crime but rather they seize the opportunities that arise from everyday activities, including those arising from the routines of work, school and home life.

From this perspective, a crime occurs when three elements come together in time and space: a

1 suitable target, a motivated offender and an absence of capable guardianship (Cohen and

Felson, 1979). The likelihood of these “situations of coalescence” (Merral, 2002, p. 15), or opportunities, to occur varies across different socio-demographic contexts. Since a large proportion of fire events occurs as unintentional consequences of everyday behaviours, routine activities theory provides a framework to identify situations that produce fire events to inform a more focused approach to allocation of limited fire suppression resources as well as to direct fire prevention activities.

The present study uses City of Calgary data provided by the Calgary Fire Department.

The Calgary Fire Department collects their data in accordance with National Fire Incident

Reporting System (NFIRS), which makes the results replicable and generalizable to other departments that use the same reporting system in similar municipalities. The results of this research are only generalizable in North American context as socioeconomic characteristics discussed in this study as well as fire and crime data collection and trends vary significantly across different geographies. The municipal setting and data collection are discussed further in

Chapter Three.

Purpose of the Study

The main research objective is to explore the potential applicability of the routine activities theory aspect of “capable guardianship” as it applies to fire incidents in order to inform potential future targeted fire prevention education campaigns, mitigation, and emergency response by the Calgary Fire Department. The second research objective is to examine how the relationship between fire incidents and capable guardianship is influenced by neighbourhood disorder. Lastly, another objective of this study is to identify the relationships

2 between anthropogenic fires in Calgary neighbourhoods with socioeconomic factors that characterize those neighbourhoods.

The Scope

This study will focus on anthropogenic fires in the city of Calgary, Alberta. As Merrall

(2002) notes “the term anthropogenic is important as it places people at the center of the dwelling fire problem” (84) as people are usually the ones responsible for fire ignition.

National fire statistics from the UK show that 83% of accidental dwelling fires are caused by the actions of the occupants (Watson et al., 2002). Better understanding of the nature and causes of variation in distribution of fire incidents, by ignition type, neighbourhood characteristics, and time is essential not only to informing the future of fire response and fire safety education initiatives, but also to enhancing the limited existing research on distribution of fire events at the neighbourhood-level. Due to time and data constraints, my thesis does not delve into detailed analysis of all types of fires but rather provides a general overview with the main focus on the relationship between fires, guardianship, and disorder.

Overview of the Thesis

Chapter One outlines the significance of the research and need for theoretical development and frameworks, as well as details the scope of the project. A brief outline of the study area and fire incident data used in this thesis is also provided.

Chapter Two supplies a literature review of the research on dwelling fire incidents, specifically, in relation to geographical and temporal distribution and links to deprivation, housing type and other social, economic and environmental factors. The literature review starts with an examination of the related research theories and concepts developed by or adopted and applied to the study of crime, such as the broken windows theory and neighbourhood social

3 disorganization theory, especially as these relate to spatial and temporal patterns. Routine activities theory is also examined in order to identify the applicability of established theory and methodology to the problem of anthropogenic accidental dwelling fire incidence. One component of routine activities theory – capable guardianship – is applied to accidental dwelling fires and further developed.

The implications of the findings identified in the literature review are then examined through a series of research questions that are posed to test the theoretical premise that accidental dwelling fires are not random, but are rather concentrated in certain areas, based on social and economic factors.

Chapter Three introduces the study area of the City of Calgary and the Calgary Fire

Department as well as provides a detailed description of the data collection and requirements.

This chapter also describes the statistical methodology used to examine the relationships between the dependent, independent, and control variables. Chapter Four provides an overview of the various types of data analyses performed – from the bivariate associations to the ordinary least squares (OLS) multiple regression analysis. This chapter discusses the obtained results as well as the statistical significance of the findings.

Chapter Five summarizes the main research findings, bridging together theoretical and analytical outputs. It also provides a discussion of the challenges, study limitations, unexpected findings and recommendations for further research.

4

CHAPTER TWO: REVIEW OF THE LITERATURE

As mentioned earlier, accidental dwelling fires are not randomly distributed but are rather concentrated in areas of high neighbourhood disorder (Karter and Donner, 1978; Syron, 1972).

Some research also suggests that the association between fires and disorder is through guardianship as areas with high collective efficacy (and, therefore, high capable guardianship rates) are more likely to have low levels of social disorder and fires (Sampson, Raudenbush, and

Earls, 1997). The main focus of my study is the relationship between fires and guardianship via select socioeconomic factors in Calgary neighbourhoods; I also examine the relationship between fires, capable guardianship, and disorder to explore the applicability of elements of criminological theory to fire incidents. My research starts with a review of the available literature on fires, guardianship, and disorder. This chapter provides an overview of routine activities theory, with its emphasis on the concept and measurement of effective guardianship, and neighbourhood disorder theory. This chapter also outlines relevant literature that identifies the role of socio-economic, spatial, and temporal factors in relation to fires. Although I do not test spatiotemporal factors in this study, they are included to provide a more thorough picture of previous research on fire incidents.

Even though the main focus of the research is the link between fires and capable guardianship, I have organized the following literature review slightly differently from the description of the main objectives in Chapter One (and hypotheses in Chapter Three). I wanted a logical flow of the argument to move from most broad to very specific and starting with neighbourhood disorder followed by the overview of RAT (and capable guardianship) which followed by the summary of literature on fires made the most sense to me.

Neighbourhood Disorder

5

The study of neighbourhood disorder has a long history and involves several theoretical perspectives, two of which I will examine here: broken windows theory and social disorganization theory.

In mid-19th century, Henry Mayhew (1862) published London Labour and the London

Poor – a collection of stories acquired through his work as a correspondent at Morning

Chronicles. Aside from collecting anecdotal accounts of the lives of London’s working poor,

Mayhew also examined supplementary data, such as crime statistics and census records. Through this work, the author made an observation that poor living conditions and visual signs of disorder appear to be linked to the perpetuation of poverty.

Nearly a century later, Shaw and McKay (1942) also noted that poverty and neighbourhood deprivation have a strong positive correlation with neighbourhood crime rates.

Since then, the concept of neighbourhood disorder has been examined through multiple lenses, with perhaps the most prominent recent lens being the “broken windows” theory.

“Broken Windows” Theory

According to the “broken windows” theory, conceptualized by Wilson and

Kelling in 1982, “disorder, [if] left unchecked, leads to crime by driving residents indoors and sending a message to would-be offenders that a neighbourhood is out of control”

(Thacher, 2004, p. 381). The idea is that the environment that is well maintained sends a different signal versus an environment that is not maintained as well (characterized by broken windows, graffiti, etc.). The former sends the signal that the area is under the capable guardianship of its residents and where criminal activities will be easily detected; whereas, the latter lacks guardianship and criminal behavior runs little risk of detection.

“The broken window is the first step in a neighbourhood’s slow decline and deterioration

6

[as] the broken window slowly turns into many broken windows” (Adams, 2006, p. 26), leading to the neighbourhood decline which attracts criminal behaviour.

A central idea of the broken windows theory which lead me down the path of guardianship when I was first formulating my research questions was the notion that physical environment conveys the presence of a certain level of collective guardianship, all of which is interpreted by the criminal. Martin Innes and Nigel Fielding (2002) coined the term “signal crimes” to describe just that. The authors suggested that some aspects of disorder speak to the types of activities and social controls, or lack thereof, in the area.

Other literature then ties guardianship to social controls with “broken windows” as a metaphor for the social cohesion of the neighbourhood. In the areas where a sense of social cohesion is strong, whatever is broken gets fixed while in the areas with a weak sense of cohesion repairs do not occur. This creates vulnerability to further social and physical decline and an increased risk of victimization through criminal activity (Kelling and Coles, 1996).

Other research also connects this notion of broken windows and decline with higher crime rates (O’Shea, 2006). However, the focus of this study is not the connection between social disorder and crime, but rather social disorder and its relationship to capable guardianship as it relates to fire. Taylor (1997) suggests that the image of an area affects residents’ attitudes about their property, and, ultimately, their territorial behavior in the form of guardianship. The residents who are attached to their homes are less likely to neglect their dwellings, rendering them capable guardians. The concept of capable guardianship is examined further below

(Routine Activities Theory).

Neighbourhood Disorganization Theory

7

Despite recognition that disorder has tangible negative outcomes for communities and people living in those communities, the definition of disorder has been hard to pinpoint. Thomas and Znaniecki (1920) first mentioned neighbourhood disorder as they examined the lives of immigrants in the United States in the early 20th century. The authors defined social disorder as a

“decrease of the influence of existing social rules of behavior upon individual members of the group” (1920, p. 1128). Sutherland (1934) linked social disorder to crime and argued that through societal change strong kinship ties were replaced by the large network of weak ties thus leading to an influx of criminal activity.

Shaw and McKay (1942) built on previous research in neighbourhood disorder. The authors examined urban centers and the patterns within. The authors observed that areas surrounding downtown had high rates of disorder, with the poor living there in run-down tenements. The delinquency rates in downtown areas was high, despite the fact that there were fairly robust resident turnover rates, making the authors conclude that delinquency was not related to the residents of the area but rather the environmental components of the area.

Skogan (1990) wrote extensively about the process of disorder and decline. He referred to disorder as the perceived lack of order and social controls, such as visible cues that residents see as the signs of disorder. The author points out that the cues of disorder are both physical and social. For Skogan (1990), disorder is tied to petty crimes, such as littering or graffiti, and incivilities, such as noise complaints, and signifies a breakdown in social controls. He ties disorder to guardianship (discussed further below), stating that disorder is “associated with declining trust in neighbors and declining participation in community life” (p. 183).

Other researchers not only refine the definition of physical disorder but also make further distinctions between physical disorder and physical decay. Physical disorder refers to

8 neighbourhood characteristics that arise from the behaviours of residents or visitors (such as graffiti, litter, abandoned cars) whereas neighbourhood decay refers to structural characteristics that can arise from the lack of “institutional investment”, such as burnt-out houses, dilapidated recreational facilities and abandoned buildings (Sampson, 2009; Sampson and Raudenbush,

2004). According to Sampson and Raudenbush (2004) it is important to make a distinction between neighbourhood disorder and decay because physical disorder is “limited to behavioural manifestations that can be conceptually decoupled from structural resources” (p. 326). Physical disorder is not the focus of the present study.

Linking Neighbourhood Disorder to Crime Events

As mentioned previously, one of the objectives of the present research is to examine the applicability of various elements of criminological theories to accidental house fires. Previous research supports the notion of broken windows and decline associated with high crime rates (O’Shea, 2006). Many researchers, including Skogan

(1990), have established that residents see visible signs of disorder as a sign of breakdown of social controls (Skogan and Maxfield, 1981; Taylor and Hale, 1986).

Cahill and Mulligan (2003) argue that a high degree of social and economic deprivation, population turnover, and ethnic heterogeneity are correlated with neighbourhoods with higher degrees of social disorder. Their article builds on the works of Linsky and Strauss

(1986), Sampson and Groves (1989) and Stark (1996). These researchers note that to understand crime in a meaningful way, the association between demographic, economic, and social variables within certain geographic boundaries needs to be considered.

Tseloni, Osborn, Trickett, and Rose (2002) added family disruption and urbanization as variables contributing to social disorder.

9

Thus, social disorganization theory relates these five factors (demographic, economic, social, family disruption and urbanization) to criminal events. Empirical work has focused on linking crime rates with measurements of poverty (measured using various income levels), unemployment rates, and neighbourhood stability (measured by, for example, the number or percentage of rental units, percentage of single parents, etc.).

Cahill and Mulligan (2003) used ethnic composition (the degree of ethnic mix), education

(percentage of college graduates), and population density, for example, to measure neighbourhood stability and test the applicability of social disorganization theory in

Tucson, Arizona.

Casten and Payne (2008) explain, “Cohen and Felson (1979) noted that studies using social disorganization theory did not explain how changes in social structure generate changes in opportunity for the criminal to commit crime” (p. 397). The authors suggest that it is necessary “to integrate social disorganization theory into an individual opportunity-based theory such as routine activities theory” (p. 397). They suggested broken windows theory as a bridge between social disorganization theory and routine activities theory.

Sampson and Groves (1989) considered social disorganization from a slightly different perspective than their predecessors. These authors sought to explore the disorganization itself as they wanted to know what social disorganization is and if there were any social factors that influenced it. Using two large national surveys of England and Wales, Sampson and Groves (1989) found that neighbourhoods “characterized by sparse friendship networks, unsupervised teenage peer groups, and low organizational participation had disproportionately high rates of crime and delinquency” (p. 799). The

10 findings are in line with what Shaw and McKay theorized in 1942 as the results showed that while disorganization is a result of a neighbourhood’s structural characteristics, it has an effect on both victimization and rates of offending.

Sampson, Raudenbush, and Earls (1997) found that neighbourhoods with high social disorganization are likely to have low collective efficacy. The researchers defined efficacy as the “willingness of local residents to intervene for the common good”

(Sampson, Raudenbush, and Earls, 1997, p. 919). This link between social disorder and its relationship to capable guardianship (and subsequently, fire) is one of the objectives of the present study. I outline below the link between fires and disorder and how this can be examined through perspectives previously applied exclusively to crime. The present study examines the relationship between guardianship and social disorder and how it affects the rate of accidental dwelling fires in Calgary neighbourhoods. We now turn to an overview of routine activities theory with primary focus on the concept of guardianship.

Routine Activities Theory

The main argument of routine activities theory is that three elements must be present in order for crime to occur – a motivated offender, a suitable target and an absence of capable guardianship (Cohen and Felson, 1979). The absence of one or more of these elements means that crime does not occur. This theory suggests that crime is not “a function of offender motivation and proposes that offenders do not go out of their way to find targets; rather daily behavioural regularities bring offenders into contact with suitable targets” (Corcoran, Zahnow, and Higgs, 2016, p. 182). All types of crime, from domestic violence to burglary or terrorism, are

11 said to be addressed using this framework and much of Felson’s subsequent research has applied the routine activity perspective in combination with situational crime prevention.

One of the first research teams to examine the application of routine activities theory to fire incidents was Wuschke, Clare, and Garis (2013). The researchers examined temporal and geographic clustering of residential structure fires in Surrey, British Columbia. Their research builds on and extends explanations of environmental criminological theory to fire events. The researchers analyzed patterns of residential burglary and residential structure fires across time and space, as well as explored the capacity for routine activities and pattern theories to explain the observed trends.

Wuschke et al. (2013) examined burglary and fire incidents by hour of day, day of the week, and month of the year. They found that fires increased between about 1 PM and 8 PM, yet they were evenly distributed across days of the week and only increased slightly in June and

October. In comparison, burglary rates were elevated from 8 AM to 7 PM, were lower on weekends, and slightly elevated in March, August and December.

Wuschke et al. (2013) note that these findings are consistent with previous research and provide an explanation of these trends using the routine activities approach. The authors argue that fires are frequently a direct result of the absence of a capable guardian (which the authors call “incapable guardianship”), with higher risk levels while people are present in their homes, undertaking routine activities such as cooking. The risk is lower when people are at home but asleep. This is the opposite of what routine activities theory tends to predict. This argument illustrates the point that the nature of guardianship varies across various crimes, as appears to do with fires as well. The authors also found that the fire risk was not significantly different for various days of the week or between months of the year.

12

My study does not examine the relationship between suitable targets and motivated offenders, but rather focuses on capable guardianship. There are several reasons for this. First, this study only examines accidental residential fires. There are no ‘unsuitable targets’ as all residential dwellings are at risk of catching fire due to the material nature of those dwellings and the contents inside. Second, since these fires started unintentionally, motivated offenders are absent as well. Some literature, which is discussed in greater detail below, mentions the possibility of looking at sources of ignition (for example, a match or a pot on a stove) as a

“motivated offender” in the instance of fires but this seems to be quite a leap in interpretation considering the definition of motivated offender as described by Cohen and Felson (1979) (see

Corcoran et al., 2016).

Based on these two points, the presence or absence of suitable guardianship becomes the determining factor for the occurrence of fire events. Below is an overview of the concept of suitable guardianship and its application to fire incidents.

The Concept of Guardianship

Over the years, there have been a number of definitions and measurements proposed to adequately capture the concept of capable guardianship. Felson (2006) emphasized the importance of a guardian as someone who “keeps an eye on the potential target of crime. This includes anybody passing by, or anybody assigned to look after people or property. This usually refers to ordinary citizens...” (p. 80). The assumption that Felson made is that guardians must be two things: an individual who is available and able to monitor. The idea is that someone watching is thought likely to deter an offender from committing a crime. However, in recent years, there has been increasing interest in the re-examination of the definition and measurement of the concept of guardianship. Arguments and findings of these studies are described below.

13

Hollis-Peel, Reynald, van Bavel, Elffers, and Welsh (2011) argue that it is guardianship alone that disrupts, directly or indirectly, the interaction between motivated offenders and suitable targets. Their focus was to expand the definition, understanding, and measurement of capable guardianship. My study adopts a somewhat similar approach as the concept of capable guardianship is the only element of routine activities theory under consideration and I specifically examine the physical presence of guardians as the main determinant of capable guardianship.

There are a number of definitions of guardianship used in previous research. For instance, Miethe, Stafford, and Sloane (1990) defined guardianship as household occupancy based on the number of people in the household over the age of 12. Meier and

Miethe (1994) operationalized guardianship as at least two individuals over the age of 16 living in the same household. Stahura and Sloan (1988) measured guardianship through three variables: police employment, police expenditure, and female labor force nonparticipation. Garofalo and Clark (1992) asked respondents about their time spent at home and created proxy measures of guardianship which included presence of a dog and presence of an alarm system. However, Hollis-Peel et al. (2011) argue “the use of measures such as presence of a dog or a security system are outside the purview of the conceptualization of guardianship in terms of human surveillance. Furthermore, the reliance on proxy measures and indicators is problematic in developing a thorough understanding of how the guardianship process operates” (p. 10).

Most often, the research utilized some form of the proxy measure to capture capable guardianship. For instance, Schiebler, Crotts, and Hollinger (1996) measured the number of capable guardians for their study on crimes against tourists in Florida by

14 counting the fulltime law enforcement officers as well as all private security guards using

Florida Department of Law Enforcement and the Florida Department of Licensing data from 1993. The sum of the law enforcement and security guards was divided by the estimated number of visitors to generate “capable guardian rate” (p. 43).

More frequently, guardianship is measured in relation to the household occupancy, or, the amount of time individuals spend in their homes (Coupe and Blake

2006; Miethe and Meier 1990; Wilcox Rountree, and Land 1996). Research also found household occupancy to be negatively associated with the risk of burglary (Miethe,

Stafford, and Long 1987). The longer people are away from their homes, the higher the risk of property crimes as the degree of guardianship decreases.

Some researchers, such as Xie and McDowall (2008), have measured guardianship in terms of residential mobility. The authors found that as neighbourhood residential mobility increases, so does the risk for property crimes. This is due to changes in “crime opportunity structure” because of the reduced capable guardianship. The authors offer several explanations for the phenomena:

“Residential moves sever local ties and weaken social control, so newcomers tend to have fewer friends in the neighborhood who might act as guardians of their homes ... Long-term neighbours are more likely to recognize unusual patterns and unfamiliar people that may be signs of criminal activity. They will also be more likely to intervene, ask questions, and notify the police of clear indications of unusual activities” (Xie and McDowall 2008: 543).

From this brief review, it is evident that guardianship has been studied frequently in relation to various crime events. However, I could find only a single study that used guardianship to explain the distribution of fire incidents. That study is by Wuschke et al.

(2013). As a reminder, the authors argue that “incapable guardianship” is correlated with higher fire rates as people who might not be aware of fire hazards are undertaking

15

potentially fire-causing activities (i.e. cooking, smoking, and so on). The authors also

make an argument that guardianship can be provided by inanimate objects, such as smoke

alarms or sprinkler systems, as these can potentially prevent fires. Unfortunately, I could

not obtain reliable data on smoke detectors or sprinkler systems or any other systems for

fire detection and mitigation, though it would be a worthwhile angle to pursue in future

research. In this study, I focus on the physical presence of individuals to examine the

relationship between capable guardianship and accidental dwelling fires.

I next review the literature on fire events.

Distribution of Fire Incidents

A significant portion of fire research was conducted in the United States in the late 1970s

and 1980s. However, there is a renewed interest in fire events in the research community with

several studies having been published from 2010 onward. The sections below address the

findings of this research and the role of socio-economic factors in the distribution of fire

incidents, followed by a discussion of spatial and temporal factors.

Neighbourhood-level socio-economic characteristics and fire incidents

Early studies have found that residential fires were not inevitable nor were they “acts of

God” but, rather, they were preventable events (Karter and Donner, 1978; Syron, 1972).

Moreover, these studies also showed that the distribution of fires is not random as there was

systematic variation in the nature and severity of fires in different populations. Schaeman, Hall,

Schainblatt, Swartz, and Karter (1977) studied intra-city variations of fires, using census tracts

as their unit of analysis. They considered six separate sets of census data – four cities (Charlotte,

NC, St. Petersburg, FL, San Diego, CA, and Seattle, WA), one county (Fairfax County, VA),

and a combination set (three of the cities and Fairfax County, VA). The authors found that when

16 tested separately, three variables, on average, explained 39 per cent of the variation in fire rates.

These variables were parental presence (percentage of two-parent families with children under

18), under-education (percentage of individuals over the age of 25 who have fewer than 8 years of schooling), and poverty (percentage of people whose income is below the poverty line). The parental presence variable was negatively correlated with the fire rate whereas poverty and under-education were positively associated with the fire rate.

The authors also found that other variables significantly explained the variation in neighbourhood-level fire rates and were positively related: education (percentage of the population with a high school diploma or higher), race (percentage of African-American residents), home ownership (percentage of owner-occupied dwellings), income (percentage of people with income over $15,000), housing density (percentage of homes with more than one person per room), and two interaction variables (education and race, and race and poverty).

Parental presence, income, education and home ownership were found to be negatively correlated with the census tract level fire incident rates, meaning that the fire rates were lower in the census tract areas with higher proportions of the aforementioned variables. Race and housing crowdedness as well as the interaction variables (education and race, and race and poverty), were found to be positively correlated with neighbourhood-level fire rates.

Another study of intra-city variation in fire incident rates was published in 1978 by

Karter and Donner. The two authors examined the socioeconomic as well as building characteristics of five American cities – Syracuse, NY, Kansas City, MO, Newark, NJ, Phoenix,

AZ, and Toledo, OH. Family stability [Schaeman et al. (1977) called this variable “parental presence”] was found to influence fire rates in two out of the five cities. In areas of low family stability, the average fire rates were two times higher in Syracuse and four times higher in

17

Kansas City than the areas with high family stability. In the remaining three cities, areas with high poverty rates experienced double the fire rate than areas of low poverty.

Bertrand and McKenzie (1976) studied the relationship between various socioeconomic, socio-cultural, and socio-demographic variables and the area’s fire rate. The researchers studied the central part of New Orleans, Louisiana, which they divided into five fire demand zones which all had high fire risk characteristics (25 or more fires per annum). These five fire demand zones contained 183 city blocks and 25,000 residents. Roughly 90 per cent of the population in the sample was black; and only 10 per cent of residences were occupied by the owners, a quarter of all households were headed by a single female, and the rate of housing units lacking complete plumbing facilities was 3.3 times greater than the rest of the city. Overall, “the study area [was] rather typical of lower-class sections of larger cities” (Bertrand and McKenzie, 1976, p. 14).

The interviews with the residents and the frontline firefighters who often responded to the area revealed that “human factors” were responsible for the majority of the fires. Smoking in bed, leaving cooking stoves unattended and placing flammable materials close to heaters

(summed by the authors into a category of “attitudes of carelessness”) were determined to be the main contributing factors to the high fire rate in the area, with smoking in bed a main contributor. The interviews with the residents determined that they had knowledge of these behaviours being the main cause of fire yet they continued to practice them with unwavering frequency, partly due to their fatalistic attitudes toward fire, and partly due to their negative views and expectations for the immediate future of their neighbourhoods (Bertrand and

McKenzie, 1976, p. 53-54).

18

Decades later, Xiong, Bruck, and Ball (2017) studied the role of human involvement in non-injury house fires. Their research focused on “how and why human actions or inactions lead to accidental residential fires and [identified] the risk factors most closely associated with those fires caused by human unsafe behaviours” (p. 3). The research was undertaken in two

Australian cities, Melbourne and Victoria, in 2012. The authors conducted 196 interviews selected from a pool of individuals who had experienced a fire between June 2010 and June

2012. They decided on four criteria for their selection: 1) the fire had to have happened on residential property; 2) the fire had to be accidental and not a result of arson; 3) the fire could not have resulted in injury or death; and 4) the fire had to have caused significant damage to the room or building of origin.

The researchers found that nearly half of all residential fires in their sample were caused by the unsafe behaviours of the residents, with more than half of those fires associated with cooking-related activities. Roughly 14 per cent of these fires were determined to be related to negligence (i.e. failing to keep up the maintenance on an appliance or not cleaning it in a timely manner). This finding highlighted the fact that not all unsafe human actions lead to immediate fires; in the case of negligence, it might take years for these activities to cause a fire. Further, the study indicated that not only was the likelihood of fires significantly increased with various unsafe behaviours, but also that individuals involved were “significantly more likely to be mentally ill, physically disabled, asleep, and not in full-time paid employment” (Xiong, Bruck, and Ball, 2017, p. 3), further supporting the arguments in previous research linking disadvantage to high fire risk.

Gunther (1981) also studied the relationship between socioeconomic disadvantage and fire incidents in Toledo, Ohio neighbourhoods. In his study, the author divided the sample into

19 five distinct categories based on income levels and race characteristics – “inner city, low income white, low income black, low income mixed white and black, middle income white, and high income white” (52-53). A single census tract that was “middle income black” was excluded from the study as there were no comparators. Because there was only one census tract that was classified as “middle income black”, the researchers could not compare it to “like” census tracts so it was excluded from the study. Using regression analysis, Gunther was able to show that there is a strong negative relationship between neighbourhood-level median income and fire rates. As median income increased, there was a significant decline in fire rates across all five census tract groups. The author did not specify whether there was an increased use of fire preventative measures such as smoke alarms and sprinklers in higher income groups.

Gunther further divided fires into seven distinct categories to examine the relationship between income and ignition source-specific fire rates. Six out of seven fire rate categories showed a negative correlation with income – suspicious fires, smoking fires, cooking fires, fires started by children, electrical and heating fires. Fires started by appliances were found to have no correlation with income. Fire rates for smoking fires were eight times higher in the inner city than in the high-income census tract areas, and suspicious fires as well as the fires started by children were fourteen times as high in the inner city. Perhaps the most important finding of the study was the relationship between the fire ignition causes and human action, rather than fires being caused by system malfunctions such as electrical short circuit or worn out wires. Gunther concluded that these findings suggest that public education is the best available resource to help reduce the occurrence of fires where human error was the main causal factor (smoking fires, cooking fires, suspicious fires and fires started by children).

20

Chandler (1979) also found significant correlations between increased fire incidence and poverty. His research study was based in Greater London. It was found that areas of high fire call volume had proportionately more deliberate and negligent ignition incidents. His research also supported previous findings by Schaemann et al (1977) and Karter and Donner (1978) on the relationship between parental supervision and fire rate. He found that as the proportion of single parent households increased so did the neighbourhood fire rate.

In 1984, Chandler, along with Chapman and Hallington, conducted another study examining the relationship between fire incidence rate and socioeconomic factors in London and two other urban centers in England. This study found high levels of neighbourhood-level unemployment and social stress were positively correlated with fire incidence rate whereas high levels of social cohesion were negatively correlated with fire risk at the neighbourhood-level.

Munson and Oates (1983) created a model for residential fires using measures of income, owner occupied homes, presence of children, structural conditions, mean temperature in

January, and social tension (which they measured as an interaction variable of percent black and unemployment). Their scope of study was quite broad as they used three samples to test their hypotheses: a 54-city national sample, a random sample of 36 communities in the state of New

Jersey, and a sample from the city of Charlotte, NC. The authors found that income, owner occupation, and temperature significantly explained the variation of fire incidents between the communities. In the end, Munson and Oates concluded that there are a “substantial number of systematic and pervasive relationships between the fire incidence rates and a diverse set of structural and socioeconomic variables” (80).

Murrey, Pitts Jr., Smith, and Hollman (1987: 68) conducted a cross-sectional factor analysis at the state level. They examined four factors – “general climate” (including violent and

21 property crimes, percent of non-white residents, population density, percent urbanized, and proportion of children living with both parents), “socioeconomic/income structure” (median household income), “economic climate” (business failures and unemployment), and “social structure” (divorce rate and percentage of the population under the age of 24 years). They concluded that general climate and socioeconomic structure factors are significant in determining the number of fires at the state level.

Wallace and Wallace (1984) argued that with changes in the level of fire protection introduced in the 1970s in New York, the scale of fires became more severe, leading to heightened damage and mass displacement, as buildings became uninhabitable. This phenomenon forced people to move in with their friends or relatives which caused overcrowding, creating an even higher fire rate. The authors argue that successive waves of displacement and fires led to disruption of social networks, higher rates of homelessness, and large-scale abandonment.

Additionally, Wallace (1981) put forth a theory about fire rate being one of the components of neighbourhood decline. He wrote that when a structure is damaged by fire, it may be abandoned due to the fact that the fire event “may trigger withdrawal of maintenance from others on that block by absentee landlords in preparation of abandonment. This in itself is sufficient to cause more structural fires, many of which occur in badly damaged piles of rubbish, poorly maintained boilers, or deteriorating electrical systems” (437).

This research on the relationship between fire damage and abandonment was reiterated in a study conducted by Burchell and Sternleib (1978). The authors conducted an eight-year study of 567 buildings in Newark, NJ. The main research objective was to isolate the conditions correlated with abandonment. In general, the study found that abandonment was likely to occur

22 in areas with a high concentration of minorities, low home ownership, high crime rates, and low income. The fact that Sternleib and Burchell considered fire incidence rate as a predictor of building abandonment years before Wallace’s research is remarkable as, prior to this, the linkage between fires and abandonment has not been explored in the literature.

Neighbourhood-level spatial and temporal features and fire incidents

Social and economic characteristics of the neighbourhood were not the only determinants of fire risk for the residents. During the last decades, research on fire incidents, however scarce, indicates that fires are not distributed uniformly across space and time. Urban spaces are often spatially fragmented due to differences in socioeconomic characteristics (Chhetri, Stimson, and

Western, 2006). Research on temporal, spatial, and spatiotemporal distribution of structural fire incidents occurring in Toronto, Ontario, revealed that the downtown area had the highest risk for all types of fires, with the exception of fire started by children playing (Asgary, Ghaffari,

Levy, 2010). The authors also found the variation of fire intensity to be highly concentrated on the weekends (between Friday and Sunday), during late afternoon to early evening, with the lowest concentration of fires during the night. Their findings are corroborated by several other sources.

Bruck (2001) conducted a review of the literature on the topic of responsiveness to smoke and fire alarms and found that “age, sleep deprivation, signal frequency, background noise, hearing loss, time of night, stage of sleep, sex differences, dream incorporation, depression, signal meaningfulness, sleeping tablets, alcohol and marijuana impacted responsiveness to alarms” (p. 623). Bruck found that most of fires that do not involve injury or fatality occur during the day between 08:00 a.m. and 20:00 p.m., however, fires occurring at night tend to

23 pose more threat to life. This led to the conclusion that sleep is a life-threatening risk factor when it comes to residential fires.

These findings were supported by the Aherns (2017) National Fire Protection Association study which brought up several interesting points about temporal distribution of residential fires. Aherns (2017) found that, in fact, most of the fires happen around dinnertime, specifically, between 17:00 and 20:00, with roughly around 20 per cent of residential structure fires occurring during night-time (23:00 and 07:00). Night-time fires, however, caused more than half (52 per cent) of all house fire deaths.

These distributions of fire events can be attributed to the activities that people normally perform while at home during these times: cooking, a leading cause of residential structure fires; smoking; and other activities. In the last decade of the 20th century, the most common ignition source of residential fires was heating-related equipment. This category includes stoves, space heaters, fireplaces, chimneys, and hot water heaters (Baker and Adams, 1993; Ahrens, 2017). However, in recent years, fires started by cooking activities became the most common fire incident type, followed by the fires started in bedrooms (Aherns, 2017).

Ducic and Ghezzo (1980) examined the epidemiology of accidental fires in residential dwellings in Montreal. The authors divided fires into two categories –

“serious” (this was determined by the fire department and had to have sustained considerable damage to the property) and “minor” (any other accidental home fires, such as a pot on a stove or an oven fire). The distinction was made to examine the circumstances surrounding substantial destruction. They found that in both categories, fires usually started in areas of the house where people spend the most time such as the

24 bedroom, which was the most common room of origin of the fire ignition. The causes of fire, according to these authors, are directly related to the activities of the residents – smoking, cooking, lighting candles. On average, people who experienced fire events were younger, had more children, smoked in bed more often, and stayed at home less during the day. The authors also were able to conclude that home fires had a characteristic seasonal pattern (more fires occurred during colder months) and tended to occur in areas with a higher concentration of older homes, signifying the fact that fires were distributed unevenly across temporal and spatial dimensions.

Gilliam (1985) examined the role of people’s interactions with their environment in relation to fire incidents using data collected in Highland Park, Michigan, from 1970 to 1977.

The author looked at the relationship between specific categories of fire incidents such as cooking, arson, smoking, electrical, and other. The author found that fires caused by cooking activities and smoking-related fires were more common in households that rented their accommodations whereas electrical and appliance fires were more common in owner-occupied housing. These relationships between housing tenure status and fire risk are particularly important in understanding how to tailor intervention and education campaigns to mitigate the risks of a particular area.

Goodsman and Mason (1985) focused on housing type as a placeholder for fire risk in two boroughs – Tameside and Gateshead, UK. In their study, housing types were categorized as houses and flats, and by private, council, or housing association. The study uses “rateable value of properties” as a proxy for likely social class of the occupants. The key finding of the study is that council flats were most likely to experience fire and fires with casualties. They found that

37% of Gateshead dwellings experiencing fires during the study period (151 households) had

25 had some contact with social services over the past twelve months prior to the fire. The authors suggest that the value of the research lies in the intervention initiatives through application of research by providing fire safety education to the residents at an increased risk of fire.

Jennings (1999) identified housing tenure and quality, variance in community prevalence of smoke detectors present in housing units, social conditions promoting smoking and alcohol use, lack of support for single parents, and educational underachievement as main contributing factors to neighbourhood-level fire risk. In 2003, Holborn, Nolan, and Golt pointed out that association does not infer causation. The authors suggest that association between certain socioeconomic characteristics and fire risk is a reflection of the association between social deprivation and other risk factors indicating that the vulnerable groups in society (the elderly, persons with mental disabilities, those physically impaired, and those suffering from alcohol and drug addiction) are at most risk. Chandler et al. (1984), building on their previous research, have found that the average U.K. resident living in a detached house has a significantly lower chance of experiencing a fire than someone who is unemployed, living in overcrowded, shared accommodation.

While considering the variation among neighbourhood-level fire rates, multiple studies have also found a strong positive correlation between alcohol consumption and fire risk.

Consumption of alcohol increases the risk for causing a fire often due to forgotten or dropped smoking materials (Holleyhead, 1999), as well as from unattended cooking (Duncanson, 2001).

It also contributes to failure to awaken to the sound of a smoke alarm (Bruck, 2001) and to successfully follow emergency escape plans (Miller, 2005), leading to increased likelihood of injury and death. Alcohol impairment has also been found to contribute to a rise in irrational

26 behaviors during the fire event, such as moving towards the fire instead of away from it and attempting to fight fire (Miller, 2005).

Conclusion

This chapter outlined several criminological theories – broken windows, social disorganization and routine activities theories, with emphasis on the concept of capable guardianship – as well as literature on structure fires. It is evident that research conducted on fire rates undertaken to date is lacking theoretical grounding. This thesis aims to address this gap. The following chapter sets out the methodological details involved in situating fire incidents in social disorganization theory and application of capable guardianship to explain patterns of distribution of fires in the city of Calgary. Prior to the analysis, the next chapter includes an historical overview of the city of Calgary as it relates to the creation of socioeconomic and spatial conditions within the geographic boundaries of the city. At the same time, the overview of the Calgary Fire Department is provided along with the historical progression of their responses to fire events.

27

CHAPTER THREE: RESEARCH DESIGN

Chapter Two provides an overview of the range of relationships between socioeconomic, spatial, and temporal characteristics of the neighbourhoods in terms of disorder and fire incidence rate. Some of the studies have even examined fire rates through the lens of capable guardianship. However, no study has looked at the effects of neighbourhood disorder on the relationship between capable guardianship and fires.

Previous research suggests that with increased rates of neighbourhood disorder, rates of capable guardianship decrease. This, in turn, would have an impact on neighbourhood-level fire incidents. Since the present study seeks to determine ways in which neighbourhood disorder impacts capable guardianship and fire incidence, multiple data sources must be considered.

Below are the descriptions of three data sources used to measure disorder, guardianship, and fires.

This chapter not only provides an overview of data sources and hypotheses, but also situates the research questions in the context of the Calgary Fire Department (by providing its history and present-day emergency and non-emergency activities) and delves into the nuances of the collection of fire-specific data. The topic of fires is not widely researched, and, therefore, is not well understood outside of the fire industry. This thesis offers a unique opportunity to showcase data collection undertaken by a fire department in general and the Calgary Fire

Department, in particular. A description of the data precedes how measures were constructed and what was included or excluded. This chapter concludes with a discussion of the key variables to be addressed in the analysis of fire incidence as well as the methodology used to examine this relationship.

28

Historical Overview of Calgary and the Calgary Fire Department

As mentioned previously, this study is based on data for Calgary, Alberta. Calgary was selected for several reasons: it is a city of sufficient size to ensure an adequate number of fires and some comparability to other major urban centers in Canada; the

Calgary Fire Department participates in the National Fire Incident Reporting System

(NFIRS) which means the fire records are standardized and comparable; and lastly, the

Calgary Fire Department was supportive of this research and willing to provide the data.

This brief history of Calgary’s development is drawn primarily from Stephanie

White’s Unbuilt Calgary (2012) and The Elbow: A River in the Life of the City (2010) by

John Gilpin. The overview focuses on events which contributed to the built environment and socioeconomic status of its residents. The history of the Calgary Fire Department and its own growth, challenges, and goals was based on facts from Yours for Life: 125 Years of Courage, Compassion, and Service from the Calgary Fire Department (Monchuk,

2011) as well as the Calgary Fire Department museum website.

Before delving into the historical overview of the city of Calgary as a municipality, it is important to acknowledge that the present-day city occupies the traditional territory of the Siksika (Blackfoot), Kainai (Blood), Piikani (Peigan), Nakoda

(Stoney), and Tsuu T’ina (Sarcee) peoples who “frequented the confluence of the Bow and Elbow Rivers long before it became a haunt of the North-West Mounted Police in

1875” (Gilpin, 2010, p. 5).

In 1875, a detachment of the North-West Mounted Police was built. The main goal of then-known Fort Brisebois (named after the F Troop Commander Ephrem A.

Brisebois) was to better control the whiskey trade in the region. The original location of

29 the fort was selected at the mid-point between Fort Macleod and Fort Edmonton.

Commander Brisebois, however, who was not known for following orders, instead relocated the fort to the place of convergence of Elbow and Bow Rivers (Gilpin, 2012).

In the summer of 1875, in recognition of his ancestral connections to the Scottish

Bay, Colonel James F. Macleod renamed Fort Brisebois to Fort Calgary after Calgary

Bay on Scotland's Isle of Mull. This is the reason the city is called Calgary and one of the major roadways connecting south to north is called Macleod Trail (White, 2012).

A few years after, in 1883, the railway was extended to Calgary to provide a connection between Canadian East and West. With its construction, the new transportation system brought hopeful ranchers from all over the country and even beyond. Shortly after that, in 1884, Calgary was incorporated as the Town of Calgary with a population of 506 people (The City of Calgary website, Historical information).

The next year, in 1885, 22 volunteers were organized into the first Calgary fire department, known as the Calgary Hook, Ladder, and Bucket Corp. By the end of that year, the fire department not only had a trained (for the times) personnel but also “a chemical fire engine, ladders, and eight hand-pumped water wells” (Monchuk, 2011, p.

6). The first fire station was constructed shortly after that (probably as result of the Great

Fire of Calgary of 1886).

The story is that the Great Fire of Calgary was the launching platform of the young fire department into the age of modernized firefighting arsenal. During the fire of

1886, strong wind made it difficult to fight the fire and contributed to its spread to 14 buildings (“four hotels, three warehouses, a couple of saloons and a handful of stores”

(Monchuk, 2011, p. 7), leaving behind damage of over $100,000. Dozens of volunteers

30 showed up to help, but the flames could not be extinguished with rubber buckets and wooden ladders (Calgary Fire Department Museum website, 2018).

The city already had a chemical engine which proved to be of little use during the

Great Fire. After the fact, the town officials decided to invest in a new $4,000 (that is close to $110,000 in today’s money) “steam-powered fire engine from the J. D. Ronald

Company of Brussels, Ontario. It truly would have been the cutting-edge technology for the times – a horse-drawn engine and two hand-pulled hose reels with 2,000 feet of hose” as in those early days the firefighters pulled the equipment themselves as they ran to fires

(Calgary Fire Department Museum website, 2018). By the 1890s, the Fire Department had its own herd of horses to pull the apparatus, which numbered 21 animals by 1911

(Calgary Fire Department Museum website, 2018).

Even in its early years, the fire department responded to a variety of emergency calls, not just the fires. In June 1897, when the Bow River overflowed, the firefighters were the ones to evacuate the residents living near the banks of the river and received $75 from the city in recognition of their efforts (Calgary Fire Department Museum website,

2018).

During the 1910s, the department was thrust into the 20th century by becoming the first in Western Canada to use motorized equipment as well as to install a prototype of the early alarm alerting system (today, it is known as 9-1-1). Cappy Smart, the Fire Chief at the time, was a strong proponent of fire prevention well before it became a norm. He introduced the first electrical fire alarm system in 1902 that included roughly twenty street alarm boxes and a dozen bells connected to the homes of firefighters directly. By

31

1910, the number of fire alarm boxes installed around the city grew to 56 (Calgary Fire

Department Museum website, 2018).

In 1914, oil was discovered in the surrounding foothills of Calgary and the origins of a boomtown were established (Gilpin, 2010). Alberta’s oil industry has continued since those early years to impact the economy, employment, population and housing within the city. However, the boom and bust economy of the region as well as WWI which was getting underway in 1914, subsequent economic downturn, and WWII that followed shortly after in 1939, all took a toll on the department’s resources and personnel

(Calgary Fire Department Museum website, 2018).

After that period, things started to look up for the department and with 855 fire calls in 1946, the city was looking to modernize yet again, this time – their alarm and dispatch system. The first dispatch center was built a few years later and “featured a central office switchboard connected to more than 200 individual fire alarms and eight fire stations” (Calgary Fire Department Museum website, 2018).

Over the course of the 20th century, the fire department saw great changes and growth in their numbers. The Fire Prevention Bureau was created in 1951, expanding work beyond the front-line firefighting to include public education and compliance with fire codes. The emphasis was not solely on safety of the citizens but also on the firefighters with a bigger focus on firefighter training (to establish best practices and increase personal safety during emergencies), and as a result, CFD opened its firefighter training academy in 1955 (Calgary Fire Department Museum website, 2018).

In the 1970s, the Fire Department became responsible for answering emergency calls after the city officially adopted the universal 9-1-1 emergency number. In 2006, this

32 responsibility transferred to the city’s Public Safety Communications, which today oversees emergency and non-emergency call answering and dispatch for the Fire

Department, Calgary Police Service and Alberta Health Services Emergency Medical

Services. Emergency Medical Services (EMS) joined the Fire Department in 1971 and engines started to respond to all emergency medical calls providing first aid and CPR until an ambulance with trained paramedics would arrive. This service delivery ended in

1983 when EMS separated from the fire department (the service moved out of the city altogether in 2009 when the province of Alberta took over responsibility for the delivery of emergency medical services) (Calgary Fire Department Museum website, 2018).

Today, Calgary is the largest city in Alberta and the third largest municipality and fifth largest metropolitan area in Canada (Census Canada, 2016). In the early 21st century,

Calgary saw a significant and rapid population growth and many corporations moved their corporate offices into the downtown core (Calgary Economic Development Annual

Report, 2017). This created some challenges for the Calgary Fire Department as the urban sprawl spread rapidly and the volume of incidents increased (The Calgary Fire

Department website, 2018).

While Calgary’s economy fared well during the global economic downturn of

2008, the same was not the case in late 2014/early 2015 when Calgary (along with the rest of the Alberta) entered into a recession due to the plunging prices of oil (Economic

Outlook, 2015-2020). Compared to the rest of the province, Calgary struggled to rebound from the economic downturn and continues to experience a slow recovery (Calgary and

Area Employment Forecast Report, 2018).

33

Overview of Calgary Fire Department Incidents

Presently, there are over 1,300 firefighters who are delivering fire and community services to 1.25 million citizens (Calgary Fire Department annual report, 2017) – a huge leap from the 22 volunteers serving just over 500 people in 1885. Along with the growing population and service area, the scope of the fire department has expanded as well.

Firefighters are now responding to an even greater variety of calls – from public service assistance calls, such as a cat in a tree, to car crashes, to medical assistance calls, to fire calls. In 2017, the front-line staff of the Calgary Fire Department responded to

62,764 calls, of which 25% per cent were either confirmed or suspected fires (Calgary

Fire Department Annual Report, 2017).

When considering the impact of fires the Calgary Fire Department reports three measures: dollar loss from fires, flame spread, and civilian injuries and deaths. According to the 2017 Fire Department Annual Report, there was an estimated loss from fire of

$57.8 million; the fire spread was limited to the room or object of origin 67 per cent of the time; and there were 33 civilian injuries and two deaths as the result of a fire.

However, fires are no longer the exclusive “bread and butter” of the Calgary Fire

Department – 45 per cent of their calls are what CFD calls “critical medical interventions” (these are medical assistance calls where firefighters provide basic life support (BLS) to the patients), which have steadily increased year after year (Calgary

Fire Department Annual Report, 2017). The role of firefighters continues to expand with increasing calls to motor vehicle collisions and for public service assistance.

The following sections describe data sources as well as the data collection process undertaken by the firefighters at the scenes of emergencies.

34

Data Sources

This thesis contains data from four different sources, each of which are aimed at different aspects of this study. First, the National Household Survey is used for its measures of “Guardianship 2” (neighbourhood mobility), immigration status, education level, and income. Second, from the Census of Canada, we draw measures of

“Guardianship 1” (average number of weeks worked in 2010) and “Guardianship 3”

(average number of people per household). Third, from the City of Calgary Police

Service we have gleaned measures of disorder, and, finally, from the City of Calgary Fire

Department we have data on the fire events.

Data from the National Household Survey (NHS, 2011) is included to supplement the information provided by the Census data and is used to provide a context for guardianship and socio-economic controls as the Census contained few usable measures.

A random sample of 4.5 million dwellings was selected for the NHS. This is slightly less than one-third (30%) of all private dwellings in Canada in 2011. The survey provides social and economic information, covering such topics as: immigration, citizenship, place of birth, ethnic origin, visible minorities, religion, Aboriginal peoples, labour, education, place of work, commuting to work, mobility and migration, language of work, income, earnings, housing and shelter costs. Data are provided for selected standard geographic areas including Canada, provinces and territories, census divisions, census subdivisions, census metropolitan areas and census agglomerations, and federal electoral districts. The

NHS is a sample survey. The sample size ensures a uniform dissemination probability for small areas and small populations. The NHS sample was selected from the 2011 Census of Population dwelling list. The data collection ran from May to August of 2011

35

(Statistics Canada website, 2016). Present research uses the portion of the NHS pertaining to the city of Calgary, where the data was aggregated to the neighbourhood- level.

The Census was collected by Statistics Canada in 2011. The census data used in this research was aggregated to the neighbourhood-level for the city of Calgary and these neighbourhoods match those from the NHS. The definition of neighbourhood used in this thesis match the geographic boundaries of the neighbourhoods as provided and determined by the City of Calgary. Therefore, census tract data was aggregated to the neighbourhood-level where the neighbourhood was determined by the boundaries drawn by the city. The Census enumerates everyone living in Canada; this includes Canadian citizens, both native-born and naturalized, landed immigrants and non-permanent residents and members of their families living with them in Canada. The Census also counts Canadian citizens and landed immigrants who are temporarily outside of Canada on census day (Statistics Canada website, 2016). The 2011 census included all the people alive at midnight between May 15 and 16, 2011. The census contains such information about the population such as family and household composition, various measures of average income, visible minority status, and information on First Nations peoples, as well as employment statistics, education, and sector of employment.

The Calgary Police Service (CPS) crime data is used to calculate neighbourhood- level social disorder. The City of Calgary Police Crime Report data was collected from

2012 to 2016 and contains information on various crimes. The crime count is based on the most serious violation per incident. Incidents are assigned to the neighbourhood using the geographic location of where the crime was committed and match those used in the

36

Census and NHS. The categories for crimes available are non-domestic assault, commercial robbery, street robbery, violence (other, non-domestic), residential break and enter, commercial break and enter, theft of vehicle, theft from vehicle, social disorder, and physical disorder. The disorder variables have been already grouped by their call types by the Calgary Police Service. The physical disorder variable contained fire incidents and since it was not possible to eliminate fire incidents as counted by the

Calgary Police Service this variable was dropped.

The Calgary Fire Department fire incidents data was used to create neighbourhood- level variables for accidental residential fires. Fire data was collected between 2005 and

2016. For this research, individual fire incident data were aggregated to the level of the neighbourhood using the addresses at which the incidents have occurred. The process of assigning addresses to a specific area is called geocoding and is performed through

ArcGIS software. Geocoding is a transformation of a description of a location, such as a pair of coordinates, an address, or a name of a place, to a location on the earth's surface.

The locations are output as geographic features with attributes, which can be used for mapping or spatial analysis (ArcGIS Pro website, 2018). In this case, all of the addresses assigned to all the incidents in the dataset were geocoded to the neighbourhood-level

(neighbourhoods match those from the other sources of data).

Not all fire incidents could be geocoded as the addresses in Fire Records

Management System, or FireRMS, are not always usable (this was due, for example, to part of the address missing, the address recorded as an intersection, the incident happened on a major highway such as Deerfoot Trail or Stoney Trail, or the incident happened at an

37

LRT station, and so on). Five per cent of all fire incidents could not be attached to a specific neighbourhood (n=1,067, or five per cent).

The full dataset also contained all dwelling fire incidents including those that were the result of malicious or deliberate acts of ignition. These arson fires accounted for 11 per cent of all fires; this proportion is in line with the rates for deliberate ignition dwelling fire incidents across Alberta (Alberta Fire Commissioner’s Statistical Report,

2013). For my research, the incidents of deliberate ignition were excluded as the focus of the research is on anthropogenic accidental dwelling fires.

Fire Incidents: Data Collection

The previous section highlighted the wide-ranging emergency activities performed by the Calgary Fire Department personnel. For each of these activities, the incident commander, who is most often a company officer (either a Fire Captain or a Fire

Lieutenant), must fill out an incident report at the end of the call upon returning to the station. All of the call types require a report to be completed for a number of reasons, including: emergency activities statistics; Freedom of Information and Protection of

Privacy requests from the members of public; legal documentation for cases which end up in court; requests for information from media outlets; and billing purposes, to name a few.

The Calgary Fire Department is legally required to report all fires to the Alberta

Office of the Fire Commissioner which reviews all fires in the province of Alberta. Based on the requirements set out by the Alberta Office of the Fire Commissioner, there are a number of mandatory fields, such as the information on the building (types of roof shingles, siding, windows and so on), information on owners, and even the weather at the

38 time of the event that incident commanders have to complete, making fire reports the most onerous of the tasks as an incident commander.

It was mentioned previously that the Calgary Fire Department collects their data in accordance with the National Fire Incident Reporting System (NFIRS) guidelines.

These guidelines are based on a President’s Commission on Fire Prevention and Control report, published in 1972, “America Burning”, and is a first of its kind, in-depth report on what was then coined as “the fire problem” in the United States. This report became the reason for the establishment of the National Fire Prevention and Control Administration.

One of the main objectives for the team was to establish a system to collect national data on fires which, in turn, led to the establishment of the National Fire Incident Reporting

System (NFIRS).

Originally, NFIRS was developed as a means of assessing the nature and scope of the fire problem in the United States. However, soon enough, NFIRS became a country- wide fire reporting system with all the states, more than 40 major cities, and around

15,000 fire departments (and counting) taking part (FEMA NFIRS User Guide, 2015).

On an annual basis, more than half a million of fire incidents and more than five million of non-fire incidents are added to the database, making it the world’s largest database of fire departments’ emergency scene incidents (FEMA NFIRS User Guide, 2015).

The Calgary Fire Department is part of the network using the system, as it built in

NFIRS codes into their own custom record management system, FDM RMS1. Within

FDM RMS, there are “kiosks” which house various types of information about the incidents. The main kiosk contains all the time stamps for the first dispatched, first

1 FDM RMS is the brand name of the fire incidents records management software supplied by the FDM Analytics where FDM stand for the first letters of the last names of the owners. 39

“enroute” (the first apparatus to leave the station after being dispatched), first arriving apparatus on scene, last apparatus to clear from the incident.

Other kiosks have yet more information, depending on the type of the event.

Much of the information is auto-populated into the FDM RMS directly from the

Computer Aided Dispatch (CAD) used at Calgary 9-1-1. In general, the FDM RMS receives data in one of two ways – data is either transferred from the CAD system or data is entered manually. Presently, some information is not held in CAD, thus requiring manual input from the incident commanders to satisfy the legal requirements set by the

Alberta Fire Commissioners Office.

When it comes to fire reports, most of the information must be filled out manually by the incident commander as it is not carried over from CAD. The summary tab, or the main kiosk, is one of the few areas that are auto-populated from CAD. The information contained in this kiosk is described above.

The information required to be filled out by the incident commander ranges from the description of the place of the incident (i.e., the structure), to construction materials used, the involved individuals, to what kind of weather the day of the event. Appendix A provides an example of the fire report (which is publicly available through the Office of the Fire Commissioner) and illustrates the required fields for data collection on fire calls.

Research Questions

Although several studies have illustrated different relationships between fire incidents and socioeconomic and spatiotemporal characteristics at the neighbourhood- level (Munson and Oates, 1983), or differing relationships between fires and neighbourhood disorder (Karter and Donner, 1978; Syron, 1972), these studies have not

40 considered the impact of capable guardianship on such relationships. First, my research studies whether fires are more frequent in areas characterized by low capable guardianship. Second, my study asks whether the relationship between fires and capable guardianship are influenced by the presence of social disorder.

Below are the hypotheses examined in the present study:

H1: As capable guardianship decreases, fires increase;

H2: As neighbourhood social disorder increases, capable guardianship decreases;

Measures

This research project focuses on the relationship between fire incidence rate and capable guardianship and how this relationship is influenced by neighbourhood social disorder. The following measures shed light on these relationships.

Dependent Variable #1: Neighbourhood Fire Incidence Rate

The primary focus of this research is accidental dwelling fires. The main dependent variable is the average number of fire incidents at the neighbourhood-level for the years 2012 to 2017 per 1,000 households. The reason for using averages and not a single year (2012, for example) is because there are very few fires that occur, in general; and the results can vary greatly year over year. In order to account for this variation, I have chosen to use the multi-year average and to normalize the variable further, the rate per 1,000 households was created.

Now, what makes a fire “accidental”? Previously, it was mentioned that fires are assigned into various categories by either Fire Captains or Fire Investigators depending on where they started, who started them, and how much the fire spread. For the purpose of this study, only fires that have started unintentionally are considered. The proportion

41 of fires used in this study is only a portion of all fires that happened in Calgary. The majority of fires are classified as having “undetermined” causes of fire and could not be used for analysis. Therefore, this analysis is bound to the parameters of the available data.

The fire incident data provided by the CFD includes the years 2005 to 2017.

However, while the fire data might have been available for 2005 to 2011, the CPS did not start reporting on social disorder variable until 2012. Therefore, in order to keep with the time ordering of all the variables in the model, fire incidents included in the variable are only those fires from 2012 to 2017.

Independent Variable #1: Capable Guardianship

There has been some debate in recent literature regarding the measurement that should be used to capture the concept of capable guardianship (e.g., Reynald 2011). This research, however, uses the definition originally supplied by Cohen and Felson (1979) and later expanded by Felson (1995). That definition emphasized the role of the guardians as individuals who are able to watch over the potential crime targets and by virtue of being present are able to deter the would-be offenders from committing a crime.

Felson (1995) writes that “[t]he “capable guardian” against crime serves by simple presence to prevent crime and by absence to make crime more likely” (53). The author goes on to give an example that “a retired person at home might well discourage daytime burglary of his or her own home or even the home next door. Conversely, someone working away from home during the day contributes by that absence to a greater risk of burglary” (Felson, 1995: 53). Guardianship should therefore be greatest the more people are present. Thus, I have decided to use in my thesis three different measures to explore the definition of capable guardianship. The measures I am using are – 1) the

42 neighbourhood-level as the average number of weeks worked in a year; 2) average number of people per household; and 3) neighbourhood mobility. I also considered using the proportion of people who work from home or neighbourhood proportion of those who did not work at all (either by choice or otherwise) but both seemed to be discriminatory variables, specific to certain populations; therefore, I decided against it as, when plotted, both showed a very large skew toward the lower end of the distribution. Therefore, based on the previous work of Xie and MaDowall (2008), Miethe et al. (1990) and others, I have decided to use other proxy measures for guardianship that speak to the amount of time people spend in their homes, along with neighbourhood stability. Neither Census nor

NHS provide data for the weekly hours worked, therefore, I used the measure that was most closely associated with that, which is the average weeks worked in 2010, at the neighbourhood-level.

Guardianship 1 speaks to the amount of time people spend in their homes and is measured as the neighbourhood-level average number of weeks worked in 2010 as reported by the residents.

Guardianship 2 is a measure of physical presence and is computed as the average number of persons in private household as measured by Census Canada (2011).

Guardianship 3 is a measure of neighbourhood mobility and is determined by the proportion of the population who lived in the same residence on the reference day (May

10, 2011), as they did five years before (May 10, 2006) as measured by Census Canada

(2011).

I had also considered using the neighbourhood-level proportion of people working from home as a measure of guardianship. Upon further consideration, however, I decided

43 against it as while people might be physically present in the house, they are not necessarily available as capable guardians as they may be occupied with their work.

Independent Variables #2: Social Disorder

In most of the neighbourhood disorder research, disorder includes both social and physical disorder (Skogan, 1990; Sampson and Raudenbush, 2004). While CPS does measure both physical and social disorder and provides publicly available data for both, the use of the physical disorder variable was not possible due to the fact that CPS includes “fires” as one of the components of its physical disorder variable. Since no actual data was given for those fires it was not possible to eliminate fires from this measure to avoid collinearity with the dependent variables.

Thus, social disorder is measured as the average annual number of all social disorder incidents at the neighbourhood-level for the years 2012 to 2017 per 1,000 households. This measurement mirrors that of the dependent variable as described above as the logic for the construction of the disorder variable is very much the same as that for the fires. The incidents are assigned to the social disorder category by the Calgary Police.

These incident types include: “Drunk”, “Disturbance”, “Indecent Act”, “Juvenile

Complaint”, “Landlord/tenant”,” Mental health concern”, “Neighbour dispute”, “Party complaint”,” Prowler”, “Suspicious person”, “Threats”, “Drugs”,” Noise complaint”,”

Possible gunshots”, “Unwanted guest/patron”, “Prostitution”, “Speeder”, and “Suspicious

Auto”.

Control Variables:

The literature on fires focusses on other characteristics of neighbourhoods; some of which are used as controls in this study. First is the proportion of recent immigrants –

44 defined as the percentage of residents who have immigrated to Canada prior to 2005, at the neighbourhood-level. This variable is based on the date of the Census which was collected in 2011. Second, education level is calculated as the neighbourhood-level proportion of the population who has completed at least high school level education. This includes those individuals who have acquired a trade, have some level of post-secondary, have completed post-secondary and beyond.

Finally, neighbourhood-level low income is measured as the proportion of residents in the bottom decile of adjusted after-tax family income. There are two reasons why I chose to use this specific measure as opposed to adjusted income. The first reason is pragmatic, the variable was readily available through the NHS. The second reason is theoretical as I wanted to capture “economic disadvantage” at the neighbourhood-level.

According to Statistics Canada, “the deciles divide the population ranked by size of adjusted after-tax family income into 10 groups of equal size. The population in the bottom decile is the one who falls in the lower 10 percent of the adjusted after-tax family income distribution” (Statistics Canada website, 2018).

Statistical Methodology

In order to address whether the fire incidence rate varies across differing levels of capable guardianship and social disorder across Calgary neighbourhoods, individual fire incidents were aggregated to the neighbourhood-level (based on the neighbourhood boundary layer in ArcGIS provided by The City of Calgary).

To test the hypotheses, the variables are analyzed using descriptive statistics, bivariate correlations, and multivariate regression. First, the descriptive statistics for the control variables, fire rate, social disorder, and guardianship measures were examined.

45

Second, the bivariate correlations between all of the independent variables and between the independent variables #1 (capable guardianship) and #2 (social disorder) and the dependent variable (fire rate) were tested. These were re-run separately for each of the three measures of guardianship.

The last component of the analysis is the multivariate regression analysis. The multivariate regression analyses were conducted for the neighbourhood-level fire incidence rate using capable guardianship and social disorder (both at the neighbourhood- level) as the independent variables.

Model 1 explores the relationship between the control variables (percentage of recent immigrants, percentage of people with at least completed high school education, and percentage of low-income households, all at the neighbourhood-level) and the dependent variable (fire incidents).

Model 2 explores the relationship between the independent variable #1 (capable guardianship) and the dependent variable (fire incidents), as well as the control variables from Model 1. Model 2 has three versions of the analysis as I am testing various guardianship measures. Therefore, models are labeled Model 2.1 (for guardianship 1 – average number of weeks worked in 2010), Model 2.2 tests guardianship # 2 (average number of people per household), Model 2.3 examines guardianship #3 (neighbourhood mobility).

Model 3 explores the relationship between the independent variable #2 (social disorder) and the dependent variable (fire incidents), as well as the independent variable

#1 (capable guardianship) from Model 2 (all three versions) and the control variables from Model 1.

46

After assessing all of the variables in all three of the models for collinearity, the variables were included in their entirety as none of the variance inflation factors (VIFs) exceeded 2.5 (Allison, 1999). The following results section provides measures of variance explained (R-squares) for each of the models as well as the beta coefficients.

Conclusion

This chapter outlined the methodological considerations for my research. In particular, the chapter elaborated on the historical overview of the city and the fire department as well as the detailed description of the data collection and requirements.

Outside of the fire industry, there is little understanding of the methodology of fire data.

This chapter also provided the research questions and statistical methodology that is used

(and elaborated on) to examine the relationship between the independent variables and the dependent variable as well as the control variables. Chapter Four provides a detailed description of the findings.

47

CHAPTER FOUR: RESULTS

The previous chapter addressed the methodological considerations and outlined the variables used in the analysis. This chapter will address the main research hypotheses:

Does a higher level of disorder at the neighbourhood-level result in higher rates of accidental residential fires? Are capable guardianship levels higher in the neighbourhoods with low levels of social disorder? And, are fires less frequent in the areas with high levels of capable guardianship and low levels of social disorder? This chapter provides the detailed description of the variables used (distribution as well as bivariate associations) along with the overview of the three models briefly mentioned in Chapter

Three.

This chapter begins with, first, the sample description that provides an overview of each of the variables used for analysis with the inclusion of the means, distribution of values, and standard deviations. Second, the bivariate associations between the dependent variable, independent variables, and the control variables are examined. Lastly, the results of the multivariate regression analysis are included. This stage was split into three models which examined the effect of various variables on the dependent variable independent of each other as well as together.

Descriptive Statistics

Starting with the description of the sample, Table 1 provides the summaries, at the neighbourhood-level, of the variables used in the models, representing the data from 2011

Census, 2011 National Household Survey, 2012-2017 CFD data, and 2012-2017 CPS data. The number of the neighbourhoods with available data varied by the dataset

(Census 2011 contained 200 neighbourhoods, NHS 2011 had data for 195

48 neighbourhoods, CFD data was available for 263 neighbourhoods, and CPS provided data for 287 neighbourhoods/areas), and after combining all four of the data sources, the dataset used in the analysis contained 188 neighbourhoods in the city of Calgary. The list of the neighbourhoods and areas that were excluded is available in the appendices under

Appendix B.

Table 1 – Descriptive Statistics of the Dependent, Independent, and Control Variables (N=188)

Standard Minimum Maximum Mean Deviation

Accidental Fire Rate 0.24 19.4 1.6 1.9

Recent Immigrants 0.00% 33.80% 5.50% 4.40%

Education 60.30% 100.00% 85.90% 7.50%

Low Income 0.00% 57.10% 8.80% 5.80% Capable Guardianship 37.6 48.7 44.6 1.5 Weeks worked in a year Average Household 1.3 4 2.5 0.5 Size Neighbourhood 17.60% 100.00% 48.10% 17.70% Mobility

Social Disorder Rate 31.5 4416.7 220.0 371.0

In order to illustrate the geography and the variables described below, I have included a map of Calgary as Figure 1 which is overlaid with the fire incidence distribution. I have divided into three equal categories for fire rates – low (0.2 -6.5), moderate (6.6 – 12.9), and high (13.0 – 19.4), as a general overview of the fire problem in

49 the city of Calgary. These figures are per 100, 000 population. The reference table for the abbreviations is available in the Appendix C.

Figure 1 – Average Fire Incidence Rate Distribution in the Calgary Neighbourhoods, 2012 -2017 (N = 188)

50

The average neighbourhood-level proportion of economic disadvantage is 8.8 per cent; however, it ranges considerably from 0.0 per cent (in several neighbourhoods including

Mayfair, Bayview, and Britannia) to 57.1 per cent (in the University of Calgary neighbourhood). Along with having the highest proportion of people living in the bottom decile of the adjusted after-tax family income, the University of Calgary neighbourhood also had the highest proportion of residents who have achieved at least high school level of education (which is expected of the post-secondary institution district). The lowest proportion of residents with high school education was located in the neighbourhood of Forest Lawn at

60.3 per cent.

When it comes to immigration status, Calgary has a relatively large recent immigrant population as, on average, 5.5 per cent of Calgarians came to Calgary from other countries within the last five years; meanwhile, the average proportion of recent immigrants in Canada was at 3.5 per cent (Census 2011). The proportion of recent immigrants in Calgary neighbourhoods ranged from 0.0 per cent (in a number of neighbourhoods, including but not limited to Greenwood/Greenbriar, Bayview, Christie

Park, and Mahogany) to 33.8 per cent in the University of Calgary (due to the large population of foreign students).

The yearly average rate for neighbourhood social disorder was 220 incidents per

1,000 households. The lowest social disorder rate was observed in the neighbourhood of

Point McKay with an annual average of 31.5 incidents per 1,000 households (followed closely by Hamptons where the social disorder rate was 31.6). The highest average disorder rate was recorded in Shepard Industrial with 4416 incidents per 1,000 households. High social disorder rate in Shepard Industrial stems from the fact that the

51 large number of people travel through the area daily as it contains many industrial and commercial properties, attracting the opportunities for social disorder crimes (i.e. disturbance or vandalism). I made a decision to keep it in the dataset as it does have 30 residential households and does have a significant number of accidental residential fires.

When it comes to capable guardianship, first, I examined the average number of weeks worked in a year – the lower the number of average number of weeks worked, the higher the chance of individuals being at home and available to provide guardianship.

The variable ranged from 37.6 weeks (highest rates of guardianship) in the University of

Calgary district (likely due to the high number of students and retirees in the area) to 48.7 weeks (lowest rates of guardianship) in Mahogany. On average, however, Calgarians worked 44.6 weeks in 2010.

The average number of people in households in Calgary in 2011 was 2.5, which is in line with the Canada-wide average. In general, since the middle of the 20th century, the average household size in Canada has been in a steady decline. This is mainly due to the decline in fertility (the average number of children per household Canada-wide was 1.1 in

2011 and 1.0 in Calgary) as well as the rise of one-person households as individuals

(particularly, women) choose to delay or forgo marriage and childbearing (Statistics

Canada, “Canadian Demographics at a Glance”, 2014)

In the case of the neighbourhood mobility, the University of Calgary neighbourhood had the highest neighbourhood-level percentage in the city of individuals who have moved in the last five years at 100.0%. The neighbourhood of Bayview had the lowest proportion of those who have moved within the last five years at 17.6 per cent

(with city-average being 47.0 per cent).

52

As for fire incidents, the variation was great between neighbourhoods – it ranged from 0.24 fires per 1,000 households (in Lincoln Park) to 19.4 fires per 1,000 households

(in Forest Lawn) (standard deviation = 1.9). On average, however, the neighbourhoods across Calgary experienced a yearly 1.6 fires per 1,000 households. Overall, in the last five years Calgary has been experiencing a steady decline in fires due to changes in building code as well as public safety education and awareness campaigns put forth by the Calgary Fire Department. However, there are still areas in the city where the fire incidents continue to happen with unyielding frequency.

Bivariate Associations

Now, to further explore the relationship between the variables, I have run the bivariate correlations. Table 2 summarizes the results of the bivariate associations that were produced between all the variables in the model as well as all three guardianship variables.

Table 2 - Bivariate Correlations between the Variables in the Sample (N=188)

Guardianship 1 Guardianship 2 Guardianship 3: Low Disorder Weeks Household Movers Immigrants Education Income Worked Size

Fire Rate 0.584** -0.093 0.020 0.288** 0.161* -0.075 0.285**

Disorder -0.124 -0.242** 0.256** 0.095 -0.238** 0.401**

Guardianship 1: 0.076 0.212** -0.155* 0.042 -0.381** Weeks Worked Guardianship 2: -0.271** 0.045 -0.153* -0.367** Household Size Guardianship 3: 0.464** 0.188** 0.351** Movers

53

Immigrants -0.096 0.570**

Education -0.191**

** The correlation is significant at 0.01 alpha level (2-tailed) *The correlation is significant at 0.05 alpha level (2-tailed)

From the table above, it’s easy to see that the variables are strongly associated with one another. As the table shows, at an alpha level of 0.01, neighbourhood-level fire incidence rate is positively correlated with disorder, neighbourhood mobility, and low income. This is unsurprising as disorder and fires are closely linked with CPS identifying fires as one of the variables of physical disorder and literature linking neighbourhood mobility and low income to neighbourhood decay.

At an alpha-level of 0.01, neighbourhood disorder is positively associated with the neighbourhood-level percentage of movers (Guardianship 2) and low income. Once again, this confirms the findings from the literature. At that same alpha level, disorder was negatively correlated with education and average number of people per household

(Guardianship 3). When we think about where crime tends to occur these relationships are hardly surprising. Downtown and surrounding areas, as well as the areas close to major public transportation routes, experience high crime rates (based on CPS data).

These are also the areas with high proportions of multi-family housing where single- person households and low-income (and education) are more frequently found as opposed to the suburbs, dense with single detached dwellings housing families.

Lastly, I wanted to mention the correlations between three of the measures of guardianship with other variables. Guardianship 1 was negatively correlated with (at an alpha level of 0.01) the neighbourhood-level percentage of low-income population. In

54 this case, guardianship should be less as the average number of weeks worked goes up as residents are away from their homes more often and are less likely to be available in case a fire occurs. Therefore, the negative relationship between low income and guardianship means that in low income areas guardianship is actually high as people work fewer weeks, on average. Already, it is evident that average number of weeks worked during the year might not be a suitable variable to measure capable guardianship as not only does it present difficulties in interpretation, it also seems to relate more to socio- economic status of residents as opposed to capable guardianship. When I chose it, I wanted to explore several measures of guardianship as there is no consensus on its measurement in the literature reviewed. In the following chapter, I expand on the debate in the literature on the measurement and possible reasons for difficulties with narrowing it down when it comes to capable guardianship.

Next, when it comes to Guardianship 2, at the alpha level 0.01, it is also negatively correlated with low income as well as the neighbourhood-level proportion of movers (Guardianship 3). This means that households with more people in them are less likely to have low income status (presumably because more people are able to contribute to that household income) and are unlikely to have moved in the last five years. Even though this measure of guardianship is not significant for fire incidents, it is significant for social disorder, meaning that it is not an unsuitable way to measure guardianship although it presents interpretation challenges. It just might be that the average number of people per household is an unsuitable measurement of the guardianship for fire events.

Lastly, Guardianship 3 is significantly correlated with all of the variables in the model – negatively with the average household size (Guardianship 2) and positively with

55 disorder, fires, average weeks worked (Guardianship 1), immigrants, education, and low income. This is the most promising measure of guardianship as, based on the bivariate associations alone, it captures the essence of what I am trying to measure – environments where there are individuals present who have the capability to interfere in case there is a fire event. However, bivariate associations alone cannot establish further connections.

In order to add to this picture, and to examine the relationships between the variables in the models further, I next run the regression analysis, which is described below.

Regression Results 1: Guardianship as the Average Weeks Worked in 2010

Table 3 displays the results of the ordinary least squares (OLS) multiple regression models that explores the first set of relationships between guardianship and fires. I ran the regressions for this relationship in stages to compare the observed relationship between guardianship and fires to the observed relationships between guardianship and fires accounting for neighbourhood disorder. In order to make that comparison, the first stage involved a multivariate regression model that includes the control variables correlated with fire rates as suggested by previous research (and discussed in the literature review section of this thesis). The second stage involved examining the relationship between guardianship and fires, accounting for the control variables. Lastly, the third model added neighbourhood disorder to complete the understanding of the relationship between guardianship and fires. I have run Models 2 and 3 thrice (they are entitled Models 2.1, 2.2, and 2.3 and Models 3.1, 3.2, and 3.3), once for each of the guardianship variables.

56

Just as with the bivariate associations, I am starting my discussion with the analysis which includes a guardianship variable measured by the average number of weeks worked in 2010.

Table 3 – Multiple Regression for the Relationship between Guardianship 1 (measured as “Average weeks worked in 2010”), Disorder, and Fires

Model 1 Model 2.1 Model 3.1

Constant 2.018 (1.093) 5.698 (3.284) 6.836 (3.218)

Immigration Status 0.034 (0.024) 0.036 (0.024) 0.027 (0.023)

Education -1.321 (1.226) -1.471 (1.231) -1.162 (1.203)

Low Income 0.043* (0.018) 0.034 (0.020) 0.014 (0.020)

Capable Guardianship -0.078 (0.066) -0.112 (0.065)

Social Disorder 0.002***(0.001)

N 188 188 188

R² 0.098 0.104 0.155

Note: unstandardized regression coefficients are shown, in parenthesis – standard errors *p < 0.05 **p < 0.01 *** p < 0.001

The regression results present the effect of independent variables as well as the control variables on the dependent variable (fires). Along with significance, all variables were tested for collinearity. According to Allison (1999), when running a multivariate regression, tolerance (1 minus R² for each independent/control variable) needs to be taken into the consideration. Thus, only variables with low tolerances (below 2.5) were included. Even though the commonly accepted threshold for VIFs is often 10 (Cohen,

Cohen, West, and Aiken, 2003), I have purposefully adopted a very conservative 2.5 VIF

57 as my cut off. My research is unique in its premise, therefore, I wanted to be certain that the variables were measuring unique concepts, so I could establish their effects on fires and not run into problems with interpretation of these relationships.

Therefore, VIF values were requested for all variables in the model (they came

back within acceptable tolerances). This indicated that multicollinearity was not an issue

in my model and I could proceed with the regression analysis. I have included the table

for VIF values of the variables in the Appendix D.

When it comes to the interpretation of the results, to keep in line with the

literature, only those relationships that were significant at alpha level of 0.01 or lower

are being discussed in the sections below (however, the table does show various levels

of significance – 0.05, 0.01 and 0.001).

Looking at the results of the regression 1, there are no significant relationships in

Model 1 and 2.1. When it comes to Model 3.1, the only significant relationship is between social disorder and fires. The model shows that the neighbourhood-level fire incidence rate is positively correlated with the neighbourhood social disorder (b=0.002,

S.E. = 0.001). Overall, the variables used in Model 3.1 explain 15.5 per cent of the variance in the fire incidence rate in the city of Calgary. The addition of social disorder improved the fit of the model by 4.1 per cent. These results closely align with previous research on disorder and its strong association with fire rate (ex. Wallace 1981) as well as the fact that fire and disorder are closely related with some research and agencies (such as

CPS) viewing fire as a form of disorder.

The purpose of this study is to explore the relationship between fire rate, capable guardianship, and disorder at the neighbourhood-level. The measurement used for

58 capable guardianship in Part One of the regression analysis did not support the hypothesis that guardianship was lower in areas of high social disorder and fire incidence rate; therefore, I tested the second measure of guardianship –average household size.

Regression Results 2: Guardianship as the Average Number of Persons per

Household

For this analysis, I have re-run the OLS regressions from Models 2.1 and 3.1

(named fittingly, Models 2.2 and 3.2 from here onward). Table 4 presents the results and can be found directly below.

Table 4 – Multiple Regression for the Relationship between Guardianship 2 (measured as “Average number of persons per household”), Disorder, and Fires Model 2.2 Model 3.2 Constant -0.227 (1.360) -1.826 (1.358)

Immigration Status 0.013 (0.025) 0.009 (0.024)

Education -0.365 (1.257) -0.549 (1.224)

Low Income 0.072***(0.021) 0.064** (0.020)

Capable Guardianship 0.514**(0.191) 0.749*** (0.191)

Social Disorder 0.002***(0.001)

N 188 188

R² 0.132 0.208

Note: unstandardized regression coefficients are shown, in parenthesis – standard errors *p < 0.05 **p < 0.01 *** p < 0.001

As can be seen from the Table 4, the effect of capable guardianship in Models 2.2 and 3.2 is significantly correlated with the fire incidence rate.

59

Now, examining Model 2.2, the same independent variables (which became control variables here) were used as for Model 1 (see Table 3) as well as the capable guardianship independent variable (measured as the average number of persons per household). Overall, the variables used in Model 2.2 explained 13.2 per cent of variance in neighbourhood-level fire incidence rate in Calgary.

When controlling for low income, immigration status, and education, capable guardianship is significantly (at alpha level 0.01) associated with the fire rate. The relationship between the two variables (capable guardianship and fires) is positive

(b=0.514, S.E.= 0.191), meaning that as the household size increases at the neighbourhood level, so too does the fire rate in those neighbourhoods. Intuitively, this too makes sense as more people does not mean better guardianship level for fires. As number of people increases so does the potential for fire hazards – through cooking, smoking materials, candles. Some research relates overcrowding to numerous negative outcomes for people living in those conditions, including elevated risk for fire-related injuries and fatalities (see Chandler et al. 1984, Krieger and Higgins, 2002). It is apparent that “average number of persons per household” is also not the best measure of guardianship, but it could be used to determine fire risk (as one of the control variables or an independent variable for a separate set of research questions).

This measure of guardianship also does not support the hypothesis being tested – areas of low capable guardianship will have higher fire rate and higher social disorder.

The findings from Models 2.2 and 3.2 illustrate the need for more work in area of capable guardianship as well as a better theoretical understanding of what exactly makes guardians capable of diverting a crime or fire event. I engage in a more thorough

60 discussion of this issue in the final chapter. Below is the final set of the regression results with guardianship measured as neighbourhood mobility.

Regression Results 3: Guardianship as the Average Number of Movers

Research that I have reviewed in previous chapters suggested that guardianship can be measured through the concept of neighbourhood cohesion (Chandler et al. 1984). I have decided to test that theory and included neighbourhood mobility as my third measure of the concept of capable guardianship. The results for the Models 2.3 and 3.3 are presented in Table 5 below.

Table 5 – Multiple Regression for the Relationship between Guardianship 3 (measured as “Neighbourhood Proportion of Movers”), Disorder, and Fires Model 2.3 Model 3.3

Constant 2.228*(1.097) 1.774 (1.089)

Immigration Status 0.020 (0.026) 0.017 (0.025)

Education -1.928 (1.285) -1.406 (1.276)

Low Income 0.038*(0.019) 0.025 (0.019)

Capable Guardianship 0.908 (0.599) 0.609 (0.597)

Social Disorder 0.001**(0.001)

N 188 188

R² 0.109 0.148

Note: unstandardized regression coefficients are shown, in parenthesis – standard errors *p < 0.05 **p < 0.01 *** p < 0.001

Overall, all of the variables in Model 2.3 explain 10.9 per cent of the variance in the neighbourhood-level fire incidence rate. Here, capable guardianship is not

61 significantly associated with the neighbourhood level fire incidence rate.

Model 3.3 explains 14.8 per cent of the variance in fire rate and the only significant association is between social disorder and fires. Controlling for other variables in the model, social disorder is positively associated with the neighbourhood-level fire rate (b=0.001, S.E. = 0.001).

To summarize the regression results from all three sets of models, I found that capable guardianship is a difficult concept to capture as none of the regression results yield a result that could be interpreted by theory (or at least, not the ones I have considered in this research), or that were obviously logical or significant. I did find fire incidents to be positively associated with low income and social disorder.

Conclusion

This chapter provides the detailed descriptions of the findings from the bivariate associations and multiple regression analyses. It also explored three different measures of capable guardianship as suggested by previous research in the field of criminology.

Chapter Five provides a detailed overview of the challenges encountered during the analysis stage as well as engages in a discussion of the results and linkages to past research and possible implications and considerations for future exploration. Chapter

Five also answers the research questions posed earlier in this thesis. Lastly, I discuss limitations of my research and provide suggestions for further consideration.

62

CHAPTER FIVE: CONCLUSION

The primary goal of this research was to explore the patterning of the fire incidence rates across Calgary neighbourhoods controlling for differing levels of social disorder and capable guardianship. Exploring the routine activities theory framework as it applies to accidental dwelling fires, my thesis examined the definition of capable guardianship, which is discussed in detail in this chapter.

This project attempted to fill a gap in the literature by considering not only the applicability of capable guardianship to fires but also the effects of social disorder on the relationship between accidental dwelling fires and capable guardianship, at the neighbourhood- level. By comparing the relationship between community-level fires with capable guardianship and social disorder, this thesis aimed to provide insight into the mechanisms through which capable guardianship and social disorder affect fire rates at the neighbourhood level. These research questions required an in-depth consideration of the definitions and understandings of capable guardianship.

While the results of this research are far from what I expected, the findings do contribute to developing knowledge within the field of fire events, as well as contribute to direction for further research and policy development. This chapter begins with, first, an overview of the results of the multivariate regression analysis described in Chapter

Four. I then provide a thorough examination of the issues related to the measurement of capable guardianship. Lastly, I conclude with the limitations of my study as well as suggestions for future research and possible real-life implementations of the findings.

63

Regression Results

Firstly, I would like to note that the results obtained by the present study should be approached with caution. While I have strived to normalize the variables the best I could by using the rates per household or the proportions, due to the fact that all incident reports are completed by individual fire personnel, there remains a margin for error in terms of their initial coding of fires. Overall, my regression results have confirmed a strong, positive association between neighbourhood economic disadvantage, fire rates, and disorder.

The variable for low income was significant in all three models that I tested, which indicates that the relationship between income and fires is stable across various iterations of the tested models – this relationship remained positive and significant regardless of the measurement of capable guardianship. While the research questions did not focus on the relationship between fire rate and income, this finding suggests an avenue for further exploration as it supports previous research (Bertrand and McKenzie, 1976; Gunther, 1981; Chandler, 1979; Munson and

Oates, 1983; Goodman and Mason, 1985). What is also remarkable is the fact that the relationship between economic disadvantage and fire rates seems to have remained unchanged in the last 30 plus years. While correlation does not equal causation, and some groups within society are more likely to fall into the low-income category (such as elderly or people with disabilities), the finding does suggest that there are common factors that drive the fire rates.

Based on the literature review, there are two major factors associated with low income – housing type and certain behaviours. Researchers have found that low income is associated with poor housing conditions (Chandler et al., 1984; Rhoades, Wenzel, Golinelli, Tucker, Kennedy, Green,

Zhou, 2011) as well as behaviours that would increase the risk for fire events such as smoking

(Gilliam, 1985; Ducic and Ghezzo, 1980), and consumption of alcohol and drugs (Jefferis,

64

Manor, and Power, 2007; Cerda, Johnson-Lawrence, and Galea, 2011; Redonnet, Chollet,

Fombonne, Bowes, and Melchior, 2012).

Poor housing conditions are also often associated with a lack of properly installed and working smoke alarms (Aherns, 2017). Several studies have looked at the differences in fire events across various income groups and found that areas characterized by economic disadvantage had low rates of uptake for the installation and proper maintenance of smoke alarms and obviously would not benefit from them (DiGuiseppi, Roberts, Wade, Sculpher,

Edwards, Godward, Pan, and Slater, 2002). Therefore, those living in poverty might be disproportionately represented in the fire incident sample as they have a greater chance of experiencing a larger fire due to the fact that smoke and fire alarms have not warned them of the presence of smoke in the house thus allowing the fire to spread beyond the object of origin.

Another significant association found in my study is between fire rate and social disorder.

Similar to low income, the relationship between social disorder and fire incidence remained positive and significant across all three models. This means that areas of high social disorder have also experienced high fire incidence rate, net of other variables in the models. This finding supported one of the hypotheses that I set out to test – that fire incidence rates would be higher in areas of high social disorder. In fact, as mentioned previously, fire and disorder (either social, as tested in this thesis, or physical) are closely linked with each other. Research by Wuschke et al.

(2013) found that fires and burglaries occur within the same spaces only at different times, making residents in properties susceptible to crime victimization and also subject to further victimization through involvement in a fire event. These areas are also likely to have high levels of low-income households, creating an environment which continues to perpetuate disadvantage

65 and continuously re-victimizes its residents. Further exploration of the factors necessary to break the cycle is necessary.

When I set out to write this thesis, I hypothesized that this cycle could be broken through the presence of capable guardianship in areas that have high social disorder and fire rates. The section below discusses the findings of the three models in regard to capable guardianship as well as the issues of measuring this construct.

Capable Guardianship

The reason why capable guardianship requires its own discussion is due to the lack of significant findings that may be related to numerous issues in defining the concept. I did not start writing this thesis with the idea of examining the definition of capable guardianship and had assumed it would be a more straightforward undertaking. However, through my literature review as well as the analysis stages of my project, I have concluded that defining what exactly constitutes capable guardianship is the key to determining the relationship between fires and disorder. I decided to test three different measures of guardianship (as they were measured in previous studies on crime, discussed in the literature review section) and their applicability to fires. As a reminder, these three measures were the average number of weeks worked in 2010

(Guardianship 1), average household size (Guardianship 2) and neighbourhood proportion of residents who have moved within the last five years (Guardianship 3).

The analysis found that measures of Guardianship 1 and 3 were not significantly associated with fire rates in Calgary, and that Guardianship 2 was significantly and positively associated with both fire rate and disorder. This relationship was the opposite of my hypothesis which was that areas of low guardianship would have higher fire rates and higher social disorder.

In fact, the results point to the opposite: guardianship, as defined by the average number of

66 people per household, is higher in areas of higher fire rate and higher social disorder. This means that while average numbers of people per household might be an appropriate measure for capable guardianship as it relates to crime prevention, it is not a suitable measure of capable guardianship when it comes to fire events.

Overall, I struggled to find a measure of guardianship that was appropriate to use in relation to fire incidents. Just as Reynald (2011) in her book Guarding Against Crime:

Measuring Guardianship Within Routine activities theory set out to “investigate the mechanisms underlying the concept of guardianship and to test the assumptions about how it functions as a crime prevention technique” (p. 17), I strived to do the same with fire events.

Firstly, capable guardianship must be specific to target places (such as various spaces around the house) and times (people had to be present at home as much as possible) in order to be considered effective, or “capable”. The main idea behind the measure when it was conceived by Felson and Cohen in 1979 was the assumption that supervision or monitoring of a suitable target – be it a property, an object, or a person – would deter a motivated offender from committing a crime against it. Therefore, I thought that selecting a measure involving the amount of time people spent in their homes would speak to that principle and would be a sufficient measure for capable guardianship. This is how the definition for Guardianship 1 was conceived.

Secondly, the literature suggested that the volume of capable guardians could prove to be the deciding factor between well-guarded and unguarded properties. Therefore, I have taken my cue from the research by Miethe et al. (1990) who measured capable guardianship as an average number of people per household and used the same measure to apply to fire incidents. As I mentioned about, while the relationship between this measure of guardianship was significant, it is not the right measure when it comes to fire incidents as the direction was the opposite as that

67 for crime. For crime, a greater number of people is a form of guardianship, while greater numbers of people with regard to fires is, essentially, an opportunity (hazard) for fires.

Lastly, the measure of neighbourhood mobility (as measured by the proportion of residents in a neighbourhood who have moved within the last five years) was aimed at addressing overall neighbourhood ability to monitor for signs of potential fire threat. I found no significance for the relationship between fire rate and this measure of guardianship (labeled as

Guardianship 3). Again, this could be due to this measurement being captured at the neighbourhood level. I am testing only the accidental residential fires, therefore, by definition, these fires had to originate from within the home or in extremely close proximity to the house

(such as deck or fence fires). Thus, in hindsight, neighbourhood mobility does not necessarily speak to the household proportion of capable guardians, nor does it speak to the fact that these fires were inside homes and, therefore, possibly immune to neighbourhood intervention or prevention.

This brings me to another issue that I encountered when it came to measuring capable guardianship – level of measurement. Fires are household-centric – they originate within the home, they spread only within a short distance of the house of origin, if at all, and they typically only affect the residents of that household. Capable guardianship, as I have measured it, is not based on the household, but rather the entire neighbourhood. I had to make assumptions about the types of characteristics, neighbourhood-wide, that would speak to the level of guardianship at the household level. In part, the issues I have faced during my thesis when it comes to the capable guardianship concept were the result of the level of measurement I was using.

There are a number of ways to operationalize the concept of capable guardianship (as mentioned previously in Chapter Two) and, ideally, capable guardianship should be an

68 observable concept with the individual households that could be aggregated to the neighbourhood level. Reynald (2011) proposed a new model for measuring capable guardianship

– Guardianship in Action (GIA). In her model, which she applied to crime events, the author argues that it is the intensity of guardianship that creates a disruption in crime event. There are four stages of the model (zero to three) which address the degree to which people are able to interfere in case there is a crime present. Stage 0 is a base model where there are no available guardians (therefore, the intensity of guardianship is nonexistent). At Stage 1, there are guardians available (or can be visibly observed on the property but they are not engaging in any of the monitoring or other preventative measures). At Stage 2 guardians are available and they are supervising, or monitoring their surroundings; and, at Stage 3, these guardians are not only available and monitoring but are also intervening should the crime event take place. How exactly this might work for fire prevention may present challenges but may be finessed to work to understand fire incidents.

Reynald (2011) also argues for the distinction between intentional guardians (residents of those homes they are “guarding”) and unintentional guardians (neighbours or passers-by in the areas who happen upon a crime and fall into one of the previously described stages). This distinction between types of guardians is particularly important in studying fire events.

Guardianship has to be intentional to count as guardianship as people have to be aware of the fire hazards around them and act on actively preventing fires from occurring (such as taking pots off hot burners before leaving the kitchen, moving tea towels away from the stove, etc.). Therefore, future research on fire incidents should consider utilizing GIA as the model “expos[ed] some of the previously neglected factors that enable residents to become capable of preventing [fires] in

69 their residential places” (Reynalds, 2011, p. 51). This would serve a dual purpose – to validate the model and to generate a measure of guardianship applicable to both crime and fire events.

This brings me to a discussion of the limitations of my research.

Limitations

Along with the issues of the measurement of capable guardianship, my thesis has other limitations. First, there are several limitations when it comes to the data used in this thesis. First and foremost, fire data contains a high degree of subjectivity in classification of the fire incidents, along with a lack of reliability among the incident commanders. This is due to varying standards in commander training (such as training taken at different times and under different requirements), personal perceptions of what constitutes a certain type of fire as opposed to another type of fire, and available information at the time that the report is being filled out.

Further tests of the reliability of the data as well as its validity would be in order to ensure that the results are accurate, precise and replicable.

Another limitation of the study is the time-ordering of the variables in the model. The data sets that were merged were all collected at different points in time (CFD data was from

2012 to 2017 as was the CPS data; whereas, the Census and NHS data was collected in 2010). In future studies, the time ordering of the data collection of the various measures should be as consistent as possible, especially in the municipalities where large socio-economic shifts are occurring as they did in Calgary with the decline of the economy into recession starting in 2014 and continuing for a number of years. These societal shifts may have contributed to demographic changes in the population and, therefore, the relationship between socio-demographic factors, fires and disorder.

70

Also, due to the small number of annual fires, I was compelled to use the five-year average to avoid the skew in the numbers. I have also attempted to include as many neighbourhoods in the sample as possible. Year over year, there is a great fluctuation in the neighbourhood fire rate, particularly among the areas with low fire rates. The five-year average allowed me to include all of them, even though their rates were very low. In the future, a more robust method needs to be explored in order to ensure accuracy and precision. The same is true for the social disorder variable.

When it came to the measurement of the socio-economic neighbourhood characteristics, I have chosen to follow the same logic of previous studies and use the measure for the low-income status of the residents. However, with the establishment of the positive and significant relationship between low income and fires, future studies need to establish a better understanding of various types of economic controls and their relationship with the fire rate.

Lastly, my study did not include any of the spatial or temporal factors which have been found to influence fire rates (Aherns, 2007). Future research needs to address this gap as the variance between daytime and nighttime, seasonal changes in fire rate patterns, as well as spatial characteristics, such as neighbourhood decline and abandoned buildings, could not only be significant in determining the frequency of fires but also impact severity and damages (either monetary or life).

In the future, to address some of the limitations listed above, the research should include a qualitative component. Firstly, it would add to the analysis but showcasing the variability within the community as not every household has the same probability for the occurrence of a fire event. Secondly, since fires happen at the level of household and affect mainly the residents of those dwellings the data needs to reflect that. Therefore, in order to understand the root causes

71 of fire (for example, lack of awareness of fire hazards, unsafe cooking practices or hazardous behaviours such as smoking in bed) or the impact of the fires on the lives of the residents as well as the changes in behavior, interviews or questionnaires filled out by the affected individuals would be greatly beneficial to the advancement of fire incidence research.

Conclusion

In conclusion, this thesis examined the relationship between capable guardianship, social disorder, and accidental dwelling fires in Calgary neighbourhoods. This thesis attempted to fill a gap in the literature by situating fires within a theoretical framework and having an in- depth look into the definitions and understandings of capable guardianship. There were many instances during this research where I felt that lack of results was a sign of failure, but now, I think of it as a confirmation that further exploration of the guardianship concept is necessary.

The research on guardianship is very limited, as is fire incidence research. This thesis contributes to these areas and expands the understanding of the ways in which guardianship could be examined in relation to fire events. The GIA model proposed by Danielle Reynald

(2011) may add traction to the definition of guardianship (as an observable action or inaction) and provide a better understanding of the mechanisms by which capable guardianship is constructed, especially when it comes to fire events.

The findings of my research contribute to the development of knowledge within the field of fire incidence research (which has been studied through the lens of geography, sociology, public health, and public policy) and has the potential to influence policy development, with real-life implications such as targeted resource allocation and deployment as well as fire safety education campaigns.

72

Overall, this study has confirmed the need for further research on guardianship. The exploration of associated factors such as disorder that contribute to patterns of fire rates directs us toward the idea that, presently, fires may have less to do with humans and more to do with fire detection devices and building codes. Unlike crime, fire incidents have been a neglected area of study with a limited number of researchers working in the field.

73

BIBLIOGRAPHY

Adams, J. The “Broken Windows” Theory. Canada: UBC, 2006.

Ahrens, M. “Home Structure Fires”. Report produced for the Fire Analysis and Research

Division, National Fire Protection Association, Quincy: MA. 2007.

---. “Home Fires Involving Cooking Equipment”. Report produced for the Fire Analysis and

Research Division. National Fire Protection Association. 2017.

Alberta Fire Commissioner's Office. “Alberta Fire Commissioner’s Statistical Report”. Alberta

Fire Commissioner's statistical report [serial online]. 2013.

Allison, P. D. How Can Multicollinearity be Diagnosed. Thousand Oaks, CA: Pine Forge Press.

1999.

Asgary, A., Ghaffari, A., Levy, J. “Spatial and Temporal Analyses of Structural Fire Incidents

and Their Causes: A Case of Toronto, Canada”. Fire Safety Journal 24 (2010): 44–57.

Baker, D.E., Adams, P. “Residential Fire Detection”. University Extension, University of

Missouri-Columbia: Columbia, 1993.

Bertrand, A.L., McKenzie, L.S. “The Human Factor in High Fire Risk Urban Residential Areas:

A Pilot Study in New Orleans, LA”. National Fire Prevention and Control

Administration, 1976.

Brantingham, P.J., Brantingham, P.L. Environmental Criminology, second ed., Waveland Press,

Inc., Prospect Heights, Illinois, 1991.

74

---. “Environment, Routine, and Situation: Toward a Pattern Theory of Crime” in: R.V. Choice,

Clarke, M. Felson (Eds.), Routine Activity and Rational, Transaction Publishers, New

Jersey: NY (1993): 259–294.

---. “Nodes, Paths and Edges: Considerations on the Complexity of Crime and the Physical

Environment”, Journal of Environmental Psychology 13 (1993): 3–28.

---. “The Rules of Crime Pattern Theory” in: R. Wortley, L. Mazerolle (Eds.), Environmental

Criminology and Crime Analysis, Willan Publishing, Devon, U.K., 2008.

Brantingham, P.J., Dyerson, D.A., Brantingham. “Crime Seen Through a Cone of Resolution”.

American Behavioral Scientist 20 (1976): 261– 327

Bruck, D. “The Who, What, Where and Why of Waking to Smoke Alarms: A Review”. Fire

Safety Journal 36 (2001): 623-639.

Burchell, R.W., Sternlieb, G. “Planning Theory in the 1980's: A Search for Future Directions”.

Center for Urban Policy Research, New Brunswick, N.J: Rutgers University, 1978.

Cahill, M.E., Mulligan, G.F. “The Determinants of Crime in Tucson, Arizona”. Urban

Geography 24 (2003): 582 – 610.

Casten, J.A. and Payne, B.K. “The Influence of Perceptions of Social Disorder and Victimization

on Business Owners’ Decisions to Use Guardianship Strategies”. Journal of Criminal

Justice 36 (5) (2008): 396-402.

Cerdá, M., Johnson-Lawrence, V.D., Galea, S. “Lifetime Income Patterns and Alcohol

Consumption: Investigating the Association Between Long- and Short-term Income

Trajectories and Drinking”. Social Science and Medicine 73(8) (2011): 1178–1185.

75

Chandler, S. E. “The Incidence of Residential Fires in London: The Effect of Housing and Other

Social Factors”. Information Paper IP 20/79. Building Research Establishment, Fire

Research Station, Borehamwood, 1979.

Chandler, S. E., Chapman, A., and Hallington, S. J. “Fire Incidence, Housing, and Social

Conditions—The Urban Situation in Britain”. Fire Prevention 172 (1984): 19-28.

Chhetri, P., Corcoran, J., Stimson, R. “Exploring the Spatio-temporal Dynamics of Fire

Incidence and the Influence of Socio-economic Status: A Case Study from South East

Queensland, Australia”. Spatial Science 54.1 (2009): 79–91.

Chhetri, P., Corcoran, J., Stimson, R.J., Inbakaran, R. “Modeling Potential Socio-economic

Determinants of Building Fires in South East Queensland”. Geographical Research 48.1

(2010): 75–85.

Calgary Economic Development. Annual Report 2017. Accessed March 2018 at

https://www.calgaryeconomicdevelopment.com/ar2017/pdf/ced-annual-report.pdf.

City of Calgary. Economic Outlook, 2015 – 2020. Accessed October 2018 at

http://www.calgary.ca/cfod/finance/Documents/Corporate-Economics/Calgary-and-

Region-Economic-Outlook/Calgary-and-Region-Economic-Outlook-2015-Spring.pdf.

City of Calgary Fire Department. Incident Report, 2005 – 2017. Windows Excel files. Accessed

January – December 2018.

City of Calgary Fire Department. 2017 Annual Report, Accessed March 2018 at

http://www.calgary.ca/CSPS/Fire/Documents/Annual-

Report/CalgaryFireDepartment2017AR.pdf.

76

City of Calgary Fire Department Museum website. Accessed October 2017 – October 2018 at

www.calgaryfiremuseum.com.

City of Calgary Police. Crime Report, 2009 - 2017. Web. Accessed January 2018.

Corcoran, J., Higgs, G., Brunsdon, C., Ware, A. “The Use of Co-maps to Explore the Spatial and

Temporal Dynamics of Fire Incidents: A Case Study in South Wales, United Kingdom”.

The Professional Geographer 59.4 (2007): 521–536.

Corcoran, J., Higgs, G., Higginson, A. “Fire Incidence in Metropolitan Areas: A Comparative

Study of Brisbane (Australia) and Cardiff (United Kingdom)”. Applied Geography 31

(2011): 65–75.

Corcoran, J., Higgs, G., Brunsdon, C., Ware, A., Norman, P. “The Use of Spatial Analytical

Techniques to Explore Patterns of Fire Incidence: A South Wales Case Study”.

Computers, Environment and Urban Systems 31 (2010): 623–647.

Corcoran, J., Zahnow, R. and Higgs, G. “Using Routine Activity Theory to Inform a Conceptual

Understanding of the Geography of Fire Events”. Geoforum 75 (2016): 180-85.

Cohen, J., Cohen, P., West, S.G., Aiken, L.S. Applied multiple regression/correlation analysis

for the behavioral sciences. Mahwah, NJ.: Lawrence Erlbaum Associates, Inc. 2003.

Cohen, L.E., Felson, M. “Social Change and Crime Rate Trends: A Routine Activity Approach”.

American Sociological Review 44 (1979): 588–605.

Coupe, T., and Blake, L. Daylight and darkness targeting strategies and the risks of being seen at

residential burglaries. Criminology 44.2 (2006): 431- 464.

77

DiGuiseppi, C., Roberts, I., Wade, A., Sculpher, M., Edwards, P., Godward, C., Pan, H., Slater,

S. “Incidence of Fires and Related Injuries after Giving out Free Smoke Alarms: Cluster

Randomised Controlled Trial”. BMJ (Clinical research ed.) 325.7371 (2002): 995 – 997.

Ducic, S., and Ghezzo, H. “Epidemiology of Accidental Home Fires in Montreal”. Accident

Analysis and Prevention 12 (1980): 68-69.

Duncanson, M. “Cooking, Alcohol and Unintentional Fatal Fires in New Zealand Homes 1991-

1997”. Produced for the New Zealand Fire Service Commission Research Report

Number 16. 2002.

Felson, M., Poulsen, E. “Simple Indicators of Crime by Time of Day”. International Journal of

Forecasting 19 (2003): 595–601.

Felson, M. “Routine Activities and Crime Prevention in the Developing Metropolis”.

Criminology 25.4 (1987): 911–931.

---. Crime and Everyday Life: Insight and Implications for Society. Thousand Oaks, CA: Pine

Forge Press. 1994.

Federal Emergency Management Agency. NFIRS User Guide, 2015. Accessed October 2018 at

https://www.usfa.fema.gov/downloads/pdf/nfirs/NFIRS_Complete_Reference_Guide_20

15.pdf.

Garofalo, J., Clark, D. “Guardianship and Residential Burglary”. Justice Quarterly 9.3 (1992):

443-63.

78

Gilliam, G. “An Analysis of Residential Fire Patterns in Highland Park, Michigan 1970 and 1977

Using Individual Unit, Block and Census Tract Data”. Master’s Thesis. Detroit, MI:

Wayne State University, 1985.

Gilpin, J. F. The Elbow: A River in the Life of the City. Second Edition, Revised and Updated.

ed. 2017.

Government of Alberta. Calgary and area employment forecast report, 2018 Winter. Accessed

January 2019 at https://open.alberta.ca/publications/calgary-and-area-employment-

forecast-report#summary.

Goodsman, R.W. and Mason, F. “Housing Factors and Fires in Two Metropolitan Boroughs”.

Unpublished fire research note. Building Research Establishment, Borehamwood. 1985.

Gunther, P. “Fire-Caused Patterns for Different Socioeconomic Neighbourhoods in Toledo,

Ohio”. Fire Journal 75 (1981): 52 -59.

Holborn, P.G., Nolan, P.F., Golt, J. “An Analysis of Fatal Unintentional Dwelling Fire

Investigated by London Fire Brigade Between 1996 and 2000”. Fire Safety Journal 38.1

(2003): 1-42.

Holleyhead, R. “Ignition of Solid Materials and Furniture by Lighted Cigarettes. A Review”.

Science & Justice 39 (1999): 75-102.

Hollis-Peel, M.E., Reynald, E., Bavel, D. Elffers, M., and Welsh, M. “Guardianship for Crime

Prevention: A Critical Review of the Literature”. Crime, Law and Social Change 56.1

(2011): 53-70.

79

Innes, M., Fielding, N. “From Community to Communicative Policing” “Signal Crimes: and the

Problem of Public Reassurance”. Sociological Research Online 7(2) (2002): 1-12.

Jefferis, B. J., Manor, O., and Power, C. “Social Gradients in Binge Drinking and Abstaining:

Trends in a Cohort of British Adults”. Journal of Epidemiology and Community Health

61.2 (2007):150–153.

Jennings, C. “Urban Residential Fires: An Empirical Analysis of Building Stock and

Socioeconomic Characteristics for Memphis, Tennessee”. Unpublished Doctoral Thesis

in Faculty of the Graduate School of Cornell University, Cornell University. 1996.

---. “Socioeconomic Characteristics and Their Relationship to Fire Incidence: A Review of the

Literature”. Fire Technology 35.1 (1999): 7–34.

Karter, M.J., Donner, A. “Fire rates and Census Characteristics – An Analytical Approach”.

National Fire Protection Association (1977): 4-22.

Kelling, G. L., Coles, C.M. Fixing Broken Windows: Restoring Order and Reducing Crime in

Our Communities. New York, NY: Free Press. 1996. Print.

Kelling, G. L., Wilson, J.Q. “Broken Windows: The police and

Neighborhood Safety”. Atlantic Monthly 249.3 (1982):29–38.

Krieger, J. and Higgins, D.L. “Housing and Health: Time Again for Public Health Action”.

American Journal of Public Health 92.5 (2002): 758-68.

Linsky, A., and Strays, M. Social Stress in the United States: Links to Regional Patterns in

Crime and Illness. Dover, MA: Auburn House. 1986.

Mayhew, H. Lonodon Labour and the London Poor, IV (1861 – 1862). Charles Griffin, 1862.

80

Merrall, S. “Anthropogenic Accidental Dwelling Fire: Incident Distribution, Theory and the Fire

Service”. PhD Thesis, University of Liverpool. 2002.

Miethe, T., Meier, R. Crime and its social context: Toward an integrated theory of offenders,

victims, and situations. Albany: The State University of New York Press. 1994.

Miethe, T., Stafford, M.C., Long, S.J. “Social Differentiation in Criminal Victimization: A Test

of Routine Activities/Lifestyle Theories”. American Sociological Review 52.2 (1987):

184 – 194.

Miethe, T., Stafford, M.C., Sloane, D. “Lifestyle Changes and Risks of Criminal Victimization”.

Journal of Quantitative Criminology 6.4 (1990): 357-376.

Miller, I. Human Behaviour Contributing to Unintentional Residential Fire Deaths 1997-2003.

New Zealand Fire Service Commission Research Report Number 47. 2005.

Monchuk, J. Yours for Life: 125 Years of Courage, Compassion and Service from the Calgary

Fire Department. 2010.

Munson. M.J. and Oates, W.E. “Community Characteristics and the Incidence of Fire: An

Empirical Analysis”. In The Social and Economic Consequences of Residential Fires.

Chester Rapkin, ed. Lexington, Ma.: D.C. Health and co. 1983.

Murrey, J., Jr. Pitts, R.E., Smith, D.A., and Hollman, K.W. “The Relationship Between Selected

Socioeconomic Variables and Measures of Arson: A Cross-Sectional Study”. Fire

Technology 23.1 (1987): 60-73.

O’Shea, T. C. “Physical Deterioration, Disorder, and Crime”. Criminal Justice Policy Review

17.2 (2006): 173–187.

81

Redonnet, B., Chollet, A., Fombonne, E., Bowes, L., and Melchior, M. “Tobacco, Alcohol,

Cannabis and Other Illegal Drug Use among Young Adults: The Socioeconomic

Context”. Drug and Alcohol Dependence 121.3 (2012): 231-39.

Reynald, D.M. “Guardians on Guardianship: Factors Affecting the Willingness to Supervise, the

Ability to Detect Potential Offenders, and the Willingness to Intervene”. Journal of

Research in Crime and Delinquency 47.3 (2010): 358-390.

---. Guarding against Crime: Measuring Guardianship within Routine Activity Theory. Farnham,

Surrey, England; Burlington, VT: Ashgate. 2011.

Rhoades, H., Wenzel, S.L., Golinelli, D., Tucker, J.S., Kennedy, D.P., Green, H.D., Zhou, A.

“The Social Context of Homeless Men’s Substance Use”. Drug and Alcohol Dependence

118.2 (2011):320–325.

Sampson, R.J. and Groves, W.B. “Community Structure and Crime: Testing Social-

Disorganization Theory”. American Journal of Sociology 94 (1989): 774 – 802.

Sampson, R.J., Raudenbush, S., Earls, F. “Neighbourhoods and Violent Crime: A Multilevel

Study of Collective Efficacy”. Science 277.5328 (1997): 918-924.

Schaeman, P., Hall, J.R., Schainblatt, A.H., Swartz, J.A., Karter, M.J. Procedures for improving

the measurement of local fire protection effectiveness. Quincy, MA: National Fire

Protection Association. 1977.

Schiebler, S., J. Crotts, and Hollinger, R. “Florida Tourists’ Vulnerability to Crime”. In Tourism,

Crime and International Security Issues, ed. A. Pizam and Y. Mansfeld. Chichester,

UK: Wiley (1996): 37-50.

82

Shaw, C. R. and McKay, H.D. Juvenile delinquency and urban areas; A study of rates of

delinquents in relation to differential characteristics of local communities in American

cities. Chicago: University of Chicago Press.1942.

Skogan, W. G. Disorder and Decline: Crime and the Spiral of Decay in American

Neighborhoods. New York: Free. 1990.

Skogan, W. and Maxfield, M. Coping with Crime. Beverly Hills, CA: Sage. 1981.

SPSS for Windows. [Statistical Software]. Chicago: SPSS Inc. 2018.

Stahura, J. M., and Sloan, J.J. “Urban Stratification of Places, Routine Activities and Suburban

Crime Rates”. Social Forces 66.4 (1988): 1102-1118.

Stark, R. “Deviant Places: A Theory of the Ecology of Crime in Readings in Contemporary

Criminological Theory”. eds P. Cordella and L. J. Seigel. Boston: Northeastern

University Press (1996): 128-142.

Statistics Canada Website. City of Calgary Individuals File, 2011 Census and National

Household Survey (Public Use Microdata Files). Ottawa: Statistics Canada. Accessed

November 2018 and October 2016.

Sternlieb, G.and Burchell, R.W. Residential Abandonment: The Tenement Landlord Revisited,

New Brunswick, N.J.: Center for Urban Policy Research, Rutgers University. 1973.

Sutherland, E.H. Principles of Criminology. Chicago, Philadelphia, J.B. Lippincott Co. 1934.

Syron, R. “An Analysis of the Collapse of the Normal Market for Fire Insurance in Substandard

Urban Core Areas”. Research report 49. Boston, Ma.: Federal Reserve Bank of Boston.

(1972):159 – 171.

83

Taylor, R. “Social Order and Disorder of Street Blocks and Neighborhoods: Ecology,

Microecology, and the Systemic Model of Social Disorganization”. Journal of Research

in Crime and Delinquency 34.1 (1997): 113–155.

Taylor, R., and Hale, M. “Testing Alternative Models of Fear of Crime”. Journal of Criminal

Law & Criminology 77.1 (1986): 151 – 189.

Thacher, D. “Order Maintenance Reconsidered: Moving beyond Strong Causal Reasoning”. The

Journal of Criminal Law and Criminology 94.2 (2004): 381– 414.

Thomas, W.I., Znaniecki, F. The Polish Peasant in Europe and America (1918-1920), Vol. 2.

New York: Dover. 1958 (reprint).

Tseloni, A., Osborn, D.R., Trickett, A., Rose, K. “Modelling Property Crime Using the British

Crime Survey: What Have We Learnt?” British Journal of Criminology 42 (2002): 109 –

128.

U.S. Fire Administration. “Socioeconomic Factors and the Incidence of Fire”. Federal

Emergency Management Agency, United States Fire Administration, National Fire Data

Center. 1997.

Wallace, D., and Wallace, R. “Structural Fire as an Urban Parasite: Density Dependence of

Structural Fire in New York City, and Its Implications”. Environment and Planning 16

(1984): 249–260.

Wallace, R. “Fire Service Productivity and the New York City Fire Crisis: 1968-1979”. Human

Ecology 9.4 (1981): 433 – 464.

84

Wallace, R., and Wallace, D. “The Coming Crisis of Public Health in the Suburbs”. The Mitbank

Quarterly 71.4 (1993): 543 – 564.

Watson, L., Gamble, J., Schofield, R. Fire Statistics United Kingdom 2000, Department of

Transport Local Government and the Regions bulletin. Home Office Research,

Development and Statistics Directorate, London. 2002.

White, Stephanie. Unbuilt Calgary: A History of the City That Might Have Been. Toronto:

Dundurn. 2012.

Wilcox Rountree, P., and Land, K. “Perceived Risk versus Fear of Crime: Empirical Evidence of

Conceptually Distinct Reactions in Survey Data”. Social Forces 74.4 (1996): 1353 –

1376.

Wuschke, K., Clare, J., and Garis, L. “Temporal and Geographic Clustering of Residential

Structure Fires: A Theoretical Platform for Targeted Fire Prevention”. Fire Safety

Journal 62 (2013): 1 – 10.

Xie, M., and Mcdowall, D. “The Effects of Residential Turnover on Household

Victimization”. Criminology 46.3 (2008): 539 – 575.

Xiong, L., Bruck, D., and Ball, M. “Preventing Accidental Residential Fires: The Role of Human

Involvement in Non‐injury House Fires”. Fire and Materials 41.1 (2017): 3-16.

85

APPENDICES

Appendix A

Fire Incident Form

86

Appendix B

List of Excluded Neighbourhoods, by Dataset

CPS CFD Census NHS 01B 01B 01F 01F 01C 01C 02E 10A 01E 01D 10A 12A 01F 01E 10C LEGACY 01G 01F 12A SHAWNEE SLOPES 01H 02B 12C 02A 02C FOOTHILLS 02B 02F FOREST LAWN INDUSTRIAL 02C 02K LEGACY 02E 02L SHAWNEE SLOPES 02F 03I 02K 05C 02L 06A 03H 06B 03I 09D 03S 09H 03V 09K 03W 10A 05B 10B 05C 12A 06A 12B 06B 12C 09H 12I 09K 12J 09N 13F 10A 13I 10B 14U 10C 14V 12A ALYTH/BONNYBROOK 12B AURORA BUSINESS PARK 12C BURNS INDUSTRIAL 12I CALGARY INTERNATIONAL AIRPORT

87

12J CANADA OLYMPIC PARK 13A DEERFOOT BUSINESS CENTRE 13B EAST FAIRVIEW INDUSTRIAL 13C EAST SHEPARD INDUSTRIAL 13D EASTFIELD 13E FAIRVIEW INDUSTRIAL 13F FISH CREEK PARK 13G FOOTHILLS 13H FOREST LAWN INDUSTRIAL 14U FRANKLIN 14V GLENMORE PARK ALYTH/BONNYBROOK GOLDEN TRIANGLE AURORA BUSINESS GREAT PLAINS PARK BURNS INDUSTRIAL GREENVIEW INDUSTRIAL PARK CALGARY HIGHFIELD INTERNATIONAL AIRPORT CANADA OLYMPIC HORIZON PARK CARRINGTON MANCHESTER INDUSTRIAL CFB - CURRIE MAYLAND CITYSCAPE MCCALL DEERFOOT BUSINESS MERIDIAN CENTRE EAST FAIRVIEW NOLAN HILL INDUSTRIAL EAST SHEPARD NORTH AIRWAYS INDUSTRIAL EASTFIELD NOSE HILL PARK FAIRVIEW OGDEN SHOPS INDUSTRIAL FISH CREEK PARK PEGASUS FOOTHILLS REDSTONE FOREST LAWN ROYAL VISTA INDUSTRIAL

88

FRANKLIN SADDLE RIDGE INDUSTRIAL GLENDEER BUSINESS SECTION 23 PARK GLENMORE PARK SETON GOLDEN TRIANGLE SKYLINE EAST GREAT PLAINS SKYLINE WEST GREENVIEW SOUTH AIRWAYS INDUSTRIAL PARK HIGHFIELD SOUTH FOOTHILLS HORIZON STARFIELD LEGACY STONEY 1 LIVINGSTON STONEY 2 MANCHESTER STONEY 4 INDUSTRIAL MAYLAND SUNRIDGE MCCALL VALLEYFIELD MERIDIAN WESTWINDS NOLAN HILL NORTH AIRWAYS NOSE HILL PARK OGDEN SHOPS PEGASUS REDSTONE ROYAL VISTA SADDLE RIDGE INDUSTRIAL SECTION 23 SETON SHAWNEE SLOPES SKYLINE EAST SKYLINE WEST SOUTH AIRWAYS SOUTH FOOTHILLS STARFIELD STONEGATE LANDING STONEY 1 STONEY 2 STONEY 3 STONEY 4 SUNRIDGE VALLEYFIELD

89

WESTWINDS

90

Appendix C

List of Community Name References for the Abbreviations in Figure 1

Community Abbreviation Community Name 01B 01B 01C 01C 01H 01H 01I 01I 01K 01K 02B 02B 02C 02C 02E 02E 02F 02F 02K 02K 02L 02L 03D 03D 03W 03W 05D 05D 05E 05E 05F 05F 05G 05G 06A 06A 06B 06B 06C 06C 09D 09D 09H 09H 09K 09K 09O 09O 09P 09P 09Q 09Q 10D 10D 10E 10E 12A 12A 12B 12B 12C 12C 12J 12J 12K 12K 12L 12L 13A 13A 13B 13B 13C 13C

91

13D 13D 13E 13E 13F 13F 13G 13G 13H 13H 13I 13I 13J 13J 13L 13L ABB ABBEYDALE ABP AURORA BUSINESS PARK ACA ACADIA ALB ALBERT PARK/RADISSON HEIGHTS ALT ALTADORE APP APPLEWOOD PARK ARB ARBOUR LAKE ASP ASPEN WOODS AUB AUBURN BAY AYB ALYTH/BONNYBROOK BDO BONAVISTA DOWNS BED BEDDINGTON HEIGHTS BEL BEL-AIRE BLM BELMONT BLN BELTLINE BNF BANFF TRAIL BNK BANKVIEW BRA BRAESIDE BRE BRENTWOOD BRI BRIDLEWOOD BRT BRITANNIA BUR BURNS INDUSTRIAL BVD BELVEDERE BYV BAYVIEW CAM CAMBRIAN HEIGHTS CAN CANYON MEADOWS CAP CAPITOL HILL CAR CARRINGTON CAS CASTLERIDGE CED CEDARBRAE CGR COUGAR RIDGE CHA CHAPARRAL CHK CHINOOK PARK

92

CHN CHINATOWN CHR CHRISTIE PARK CHV COUNTRY HILLS VILLAGE CHW CHARLESWOOD CIA CALGARY INTERNATIONAL AIRPORT CIT CITADEL CLI CLIFF BUNGALOW CNS CORNERSTONE COA COACH HILL COL COLLINGWOOD COP CANADA OLYMPIC PARK COR CORAL SPRINGS COU COUNTRY HILLS COV COVENTRY HILLS CPF COPPERFIELD CRE CRESCENT HEIGHTS CRM CRESTMONT CSC CITYSCAPE CUR CURRIE BARRACKS DAL DALHOUSIE DBC DEERFOOT BUSINESS CENTRE DIA DIAMOND COVE DIS DISCOVERY RIDGE DNC DOWNTOWN COMMERCIAL CORE DNE DOWNTOWN EAST VILLAGE DNW DOWNTOWN WEST END DOV DOVER DRG DEER RIDGE DRN DEER RUN EAG EAGLE RIDGE EAU EAU CLAIRE EDG EDGEMONT EFI EASTFIELD EFV EAST FAIRVIEW INDUSTRIAL EPK ELBOW PARK ERI ERIN WOODS ERL ERLTON ESH EAST SHEPARD INDUSTRIAL EVE EVERGREEN EVN EVANSTON EYA ELBOYA

93

FAI FAIRVIEW FAL FALCONRIDGE FHI FOOTHILLS FHT FOREST HEIGHTS FLI FOREST LAWN INDUSTRIAL FLN FOREST LAWN FRA FRANKLIN FVI FAIRVIEW INDUSTRIAL GAG GARRISON GREEN GAW GARRISON WOODS GBK GLENBROOK GBP GLENDEER BUSINESS PARK GDL GLENDALE GLA GLAMORGAN GPI GREAT PLAINS GPK GLENMORE PARK GRE GREENWOOD/GREENBRIAR GRI GREENVIEW INDUSTRIAL PARK GRV GREENVIEW GTI GOLDEN TRIANGLE HAR HARVEST HILLS HAW HAWKWOOD HAY HAYSBORO HID HIDDEN VALLEY HIF HIGHFIELD HIL HILLHURST HIW HIGHWOOD HKS HOTCHKISS HOR HORIZON HOU HOUNSFIELD HEIGHTS/BRIAR HILL HPK HIGHLAND PARK HSD HOMESTEAD HSN HASKAYNE HUN HUNTINGTON HILLS KCA KINCORA KEL KELVIN GROVE KIL KILLARNEY/GLENGARRY KIN KINGSLAND KSH KEYSTONE HILLS LEG LEGACY LIV LIVINGSTON

94

LKB LAKE BONAVISTA LKV LAKEVIEW LMR LOWER MOUNT ROYAL LPK LINCOLN PARK MAC MACEWAN GLEN MAF MAYFAIR MAH MAHOGANY MAL MAYLAND HEIGHTS MAN MANCHESTER MCI MCCALL MCK MCKENZIE LAKE MCT MCKENZIE TOWNE MDH MEDICINE HILL MEA MEADOWLARK PARK MER MERIDIAN MID MIDNAPORE MIS MISSION MLI MAYLAND MLR MILLRISE MNI MANCHESTER INDUSTRIAL MOP MOUNT PLEASANT MOR MONTEREY PARK MPK MARLBOROUGH PARK MPL MAPLE RIDGE MRL MARLBOROUGH MRT MARTINDALE NAW NORTH AIRWAYS NEB NEW BRIGHTON NGM NORTH GLENMORE PARK NHU NORTH HAVEN UPPER NHV NORTH HAVEN NOL NOLAN HILL NPK NOSE HILL PARK OAK OAKRIDGE OGD OGDEN OSH OGDEN SHOPS PAL PALLISER PAN PANORAMA HILLS PAT PATTERSON PEG PEGASUS PEN PENBROOKE MEADOWS

95

PIN PINERIDGE PKD PARKDALE PKH PARKHILL PKL PARKLAND POI POINT MCKAY PUM PUMP HILL QLD QUEENSLAND QPK QUEENS PARK VILLAGE RAM RAMSAY RAN RANCHLANDS RCK RDL ROSEDALE RED RED CARPET REN RENFREW RIC RICHMOND RID RIDEAU PARK RMT ROSEMONT ROC ROCKY RIDGE ROX ROXBORO ROY ROYAL OAK RSN REDSTONE RUN RUNDLE RUT RUTLAND PARK RVW RANGEVIEW RYV ROYAL VISTA S23 SECTION 23 SAD SADDLE RIDGE SAN SANDSTONE VALLEY SAW SOUTH AIRWAYS SCA SCARBORO SCE SCENIC ACRES SDC SUNDANCE SET SETON SFH SOUTH FOOTHILLS SGH SAGE HILL SGL STONEGATE LANDING SHG SHAGANAPPI SHI SHEPARD INDUSTRIAL SHN SHAWNESSY SHS SHAWNEE SLOPES SHW SHERWOOD

96

SIG SIGNAL HILL SIL SILVER SPRINGS SKE SKYLINE EAST SKR SKYVIEW RANCH SKW SKYLINE WEST SNA SUNALTA SOC SOUTH CALGARY SOM SOMERSET SOV SOUTHVIEW SOW SOUTHWOOD SPH SPRINGBANK HILL SPR SPRUCE CLIFF SRI SADDLE RIDGE INDUSTRIAL SSD SUNNYSIDE SSW SCARBORO/ SUNALTA WEST ST1 STONEY 1 ST2 STONEY 2 ST3 STONEY 3 ST4 STONEY 4 STA ST. ANDREWS HEIGHTS STD STARFIELD STR STRATHCONA PARK SUN SUNRIDGE SVO SILVERADO TAR TARADALE TEM TEMPLE THO THORNCLIFFE TUS TUSCANY TUX TUXEDO PARK UMR UPPER MOUNT ROYAL UND UNIVERSITY DISTRICT UNI UNIVERSITY HEIGHTS UOC UNIVERSITY OF CALGARY VAF VALLEYFIELD VAL VALLEY RIDGE VAR VARSITY VIS VISTA HEIGHTS WAL WALDEN WBN WOODBINE WES WESTWINDS WGT WESTGATE

97

WHI WHITEHORN WHL WEST HILLHURST WIL WILLOW PARK WIN WINSTON HEIGHTS/MOUNTVIEW WLD WILDWOOD WND WINDSOR PARK WOO WOODLANDS WSP WEST SPRINGS YKV YORKVILLE

98

Appendix D

Variance Inflation Factor: Regression Models

Regression Model Model 1 Model Model Model Model Model Model Variables VIFs 2.1 3.1 2.2 3.2 2.3 3.3 VIFs VIFs VIFs VIFs VIFs VIFs Immigration Status 1.482 1.484 1.502 1.659 1.735 1.730 1.732 Education 1.039 1.050 1.056 1.129 1.166 1.150 1.175 Low Income 1.512 1.727 1.906 2.047 2.065 1.552 1.651 Capable Guardianship 1.185 1.215 1.400 1.535 1.445 1.493 Social Disorder 1.249 1.335 1.258

99

Appendix E

DATA USE AGREEMENT BETWEEN

The City of Calgary (the “Holder”)

and

Olivia Skidmore (the “Recipient”)

This Data Use Agreement is made and entered into on November_24, 2016 by and between The City of Calgary, hereafter “Holder” and Olivia Skidmore, hereafter “Recipient.”

WHEREAS the Recipient and the Holder wish to enter into this agreement (the “Agreement”) pursuant to which the Holder will provide data to the Recipient to permit the Recipient to conduct the study outlined in accordance with the terms of this Agreement; The Recipient and the Holder agree as follows: Permitted Uses and Disclosures: 1.1 Except as otherwise specified herein, Recipient may make all uses and disclosures of the Limited Data Set necessary to conduct the study outlined in Schedule A (the “Research Project”).

1.2 In addition to the Recipient, the individuals, or classes or individuals, who are permitted to use or receive the Limited Data Set for purposes of the Research Project include: Graduate Student Supervisor (current or future) of Olivia Skidmore (the “Supervisor”), on condition that that the Supervisor(s) is/are not part of the Recipient’s workplace and is/are not directly involved in the Research Project. The Recipient shall enter into a data agreement with the other classes of persons before the release of the Limited Data Sets.

Recipient Responsibilities 1.3 Recipient will not use or disclose the Limited Data Set for any purpose other than permitted by this Agreement pertaining to the Research Project or as required by law;

1.4 Recipient will use appropriate administrative, physical and technical safeguards to prevent use or disclosure of the Limited Data Set other than as provided for by this Agreement;

1.5 Recipient will report to the Holder any use or disclosure of the Limited Data Set not provided for by this Agreement of which the Recipient becomes aware within 15 days of becoming aware of such use or disclosure;

100

1.6 Recipient will ensure that any agent, including a subcontractor, to whom it provides the Limited Data Set, agrees to the same restrictions and conditions that apply through this Agreement to the Recipient with respect to the Limited Data Set;

1.7 Recipient will not identify the information contained in the Limited Data Set; and

1.8 Recipient will not contact the individuals whose information might be contained in the Limited Data Set.

2. Term and Termination 2.1 The terms of this Agreement shall be effective as of _November 24, 2016, and shall remain in effect until the Limited Data Set provided to the Recipient is destroyed or returned to the Holder.

2.2 FOIP-The parties acknowledge that this Agreement and the relationship between the Holder and the Recipient are subject to the provisions of the Freedom of Information and Protection of Privacy Act (Alberta) ("FOIP"), if applicable.

2.3 Upon the Holder’s knowledge of a material breach of this Agreement by the Recipient, the Holder shall provide an opportunity for Recipient to cure the breach or end the violation. If efforts to cure the breach or end the violation are not successful within the reasonable time period specified by the Holder, the Holder shall discontinue disclosure to the Recipient and report the problem to the Conjoint Faculties Research Ethics Board (CFREB) or its designee. The Holder shall immediately discontinue disclosure of the Limited Data Set to the Recipient if the Holder determines cure of the breach is not possible.

General Provisions 2.4 Recipient and Holder understand and agree that any other individuals or parties are not intended to be third party beneficiaries of this Agreement.

2.5 This Agreement shall not be assigned by Recipient without the prior written consent of the Holder.

2.6 Each party agrees that it will be responsible for its own acts and the results thereof to the extent authorized by law and shall not be responsible for the acts of the other party or the results thereof.

IN WITNESS WHEREOF, the parties hereto execute this agreement as follows:

Date: 12/05/2016 By: Ken Uzeloc (Calgary Fire Department representative)

101

By: Olivia Skidmore Date: 12/05/2016 (Master of Arts Student, Department of Sociology, University of Calgary)

SCHEDULE A THE RESEARCH STUDY

Overview

The goals of this research project are threefold: to contribute to the existing body of literature, to enhance Calgary Fire Department (CFD) community-level risk assessment, and to impact policy. First, this study will have a direct impact on understanding and improving risk-based methodology for the Calgary Fire Department to utilize in conjunction with the enhanced dynamic deployment model, the fire risk index for planning (currently under development) and allocation of community safety resources. Secondly, the aim is to extend the literature on the occurrence and distribution of fire incidents at the neighbourhood – level by examining them through the prism of environmental criminology theories. Finally, current research intends to influence policies and guidelines regarding outcome measurements and planning best practices at the national and international level.

Research Objectives

This study will focus on the anthropogenic, or accidental, fire issues relating to the city of Calgary, Alberta. The national fire statistics in the UK show that 83% of accidental dwelling fires are initiated by the actions or omissions of the occupants of the dwellings (Watson, 2002). Better understanding of the nature and causes of variance in distribution of fire incidents, by ignition type, neighbourhood characteristics, and time is essential to not only inform the future of fire response and fire safety education initiatives, but also, to enhancing the limited existing research on distribution of fire events at the neighbourhood – level.

Therefore, the main research objective of this study is to identify the relationship between anthropogenic fires in Calgary neighbourhoods and socio-economic and environmental factors. The second research objective is to explore the potential applicability of the routine activity theory to fire incidents in order to inform future targeted fire prevention education, mitigation, and emergency response in the city of Calgary.

102

Justification (implications of knowing the answer)

I believe that incorporating the factors above will lead to better outcomes for customers and a more efficient allocation of public finances. It will also provide a basis for targeted prevention and education programming, possibly borrowed or adapted from the criminological research and helping to inform work already underway through the Community Safety division and the Regional partnerships.

Finally, the end goal is that this research will begin to inform policy on a broader scale. If I can demonstrate that there is an evidence-based planning and education campaigns that yield better results all around, then perhaps prevention initiatives and resource deployment can be changed to reflect this reality, not just in Calgary, but in other municipalities as well. The research will position the City of Calgary at the forefront of fire service planning.

Theoretical Framework and Prior Research

A full literature review will be a part of the research project. The initial scan showed that the significant portion of fire research has been conducted in the United States in the 1970s and 1980s. However, there seems to be a renewed interest in fire events in the research community with a number of studies having been published from 2010 forward. This literature review will aim to provide a comprehensive look not only at the research conducted on distribution of fire incidents, but also on role of human behavior in initial ignition and an overview of the environmental criminology theories. Firstly, the literature review will address the role of socio- economic factors in distribution of fire incidents, followed by spatial and temporal factors. Secondly, the general overview of the criminological environmental theory will be provided as it offers a background to theoretical platform being tested in this research. Lastly, routine activity theory literature will be examined, specifically as it has been applied to the fire incidents.

Methodology/approach

I envision a three-phase project:

Phase I -Data collection and cleaning (1-6 months) I will collect several years (ideally ten) of basic fire incident data, including actions taken and damage associated with the incident. Only incidents that can be geocoded will be used in the analysis. Final analysis will include incidents aggregated to the neighbourhood – level and merged with the National Household Survey (2011) dataset, in order to interpret relationship and infer causation. During this phase, the full literature

103 review will be conducted as well. No personal information or other identifying characteristics will be made available in this dataset to the Researcher.

Phase II -Analysis (2-6 months) From the collected data, I will determine the relationship between fire rate and neighbourhood-level socio-economic and environmental characteristics. Next, I will run ordinal least squares (OLS) regression to determine the size of effect of individual characteristics on the neighborhood-level fire rate. Lastly, I will run a linear regression to determine whether there is an increased fire risk based on the temporal and spatial factors, such as time of day and location of the source of ignition.

Phase 3 –Interpretation of the results and write-up of the final report (12+ months) I will interpret the results of the analysis conducted in Phase 2 and ground them in the environmental criminology theory. I will also demonstrate how the routine activity theory can be used to inform the targeted fire prevention and education approach.

Budget

The project is estimated to require 1,000 person-hours to complete. The monetary cost to the department is $0.00 as it is a part of the program fulfillment for the Master of Arts degree in Sociology at the University of Calgary and is self-funded by the graduate student. The sole researcher is Olivia Skidmore and all work related to the project will be completed by her, outside of her official role with the Calgary Fire Department as the Research and Development Analyst.

Bibliography Watson, L., et al, (2002), Fire Statistics United Kingdom 2000, Department of Transport Local Government and the Regions bulletin, Home Office Research, Development and Statistics Directorate, London.

104