Homeland Security Risk and Preparedness in Police Agencies: The Insignificance of Actual Risk Factors

Contingency Theory

Contingency theory maintains that organizations are dynamic and rationally adapt to contingencies to achieve fit with their environment and attain effectiveness (Donaldson, 2001). Donaldson’s (2001) structural adaptation to regain fit (SARFIT) model proposes that when organizations are out of fit with their environment, they must change their structure to reestablish fit and regain performance. The theory has been applied to law enforcement organizations (Burruss et al., 2010; Maguire, 2003). In terrorism-related research, contingencies include the risk of a homeland security incident. It would follow that police departments rationally respond to an increased risk of homeland security incidents by taking steps to enhance their homeland security preparedness.

Contingency theory has not received much empirical support when applied to community policing innovations (Zhao, Ren, & Lovrich, 2010) or the creation of gang units (Katz, Maguire, & Roncek, 2002), but it has received considerable support in explaining homeland security preparedness (Burruss et al., 2010; Davis et al., 2004, 2006; Schafer, Burruss, & Giblin, 2009). Although some similarities exist, these past studies have operationalized risk and preparedness slightly differently; nevertheless, perceived risk consistently appears as a significant predictor of homeland security preparedness. This has been demonstrated in national surveys addressing the state of terrorism preparedness among local law enforcement agencies in the United States (Davis et al., 2004, 2006), in small and large municipal agencies across Illinois (Burruss et al., 2010; Schafer et al., 2009), and in a national sample of small municipal agencies (Giblin et al., 2014; for an exception, see Gerber, Cohen, Cannon, Patterson, & Stewart, 2005).

What all of these studies show is that perceived risk is positively associated with preparedness activities. They support contingency theory in that perceived risk (the contingency) is positively associated with preparedness activities (organizational behavior). However, contingency theory has generally only been examined using perceived risk as a proxy measure of actual risk. These two measures of risk, subjective perception and objective reality, are discrete. Only one recent study considers objective risk factors. Using data from the Law Enforcement Management and Administrative Statistics (LEMAS) survey, Roberts et al. (2012) used a composite indicator of built environment vulnerability to predict whether departments had taken five steps to prepare for homeland security incidents. The results indicated that built environment vulnerability was not a significant predictor of homeland security preparedness (Roberts et al., 2012).

Objective Risk

Calculating objective risk for terrorism-related homeland security incidents involves the use of the threat vulnerability consequences (TVC) model. This is the combination of threat (the probability of an attack), vulnerability (if it occurs, the probability of the attack’s success), and consequences (if the attack occurs, the losses that would be incurred through fatalities, injuries, and economic losses; Ezell et al., 2010). In this model, threat is the most difficult to calculate; it takes into account the goals, motives, and capabilities of terrorists, as well as factors about possible targets. The actual probability of an event is measured in probabilistic terms (e.g., 1 in 1,000,000 chance of an attack). However, threat is typically calculated by experts in the intelligence community as the likelihood of any one type of terrorist event (e.g., a conventional explosives attack) relative to any other (e.g., agroterrorism; Ezell et al., 2010). Applying the TVC model to the present study, threat corresponds with perceptions of risk, vulnerability with social and built environment vulnerability, and consequences with the outcomes of past hazards. ¹

Broadly defined, vulnerability is the “susceptibility to damage or harm” (Eakin & Luers, 2006, p. 366). Cutter (2003) describes vulnerability as “those circumstances that put people and places at risk and those conditions that reduce the ability of people and places to respond to environmental threats” (p. 6). Social vulnerability measures the potential for harm to a population. While quantifying social vulnerability is challenging, researchers at the University of South Carolina have created the Social Vulnerability Index (SoVI) based on 32 variables derived from the U.S. Census (Hazards & Vulnerability Research Institute, 2012b). An index score is available for each county in the United States allowing for comparative analyses; higher values on the SoVI are associated with higher levels of social vulnerability.

Myriad variables are included in the SoVI index based on empirical studies linking them with social vulnerability (see Cutter, Boruff, & Shirley, 2003 for an extensive description and list of sources for each concept). For example, the inclusion of gender is premised on research showing that females have a more difficult time recovering from disasters; females typically earn lower wages than their male counterparts and assume a significant role in childcare. Physically disabled, elderly, and youthful populations are also more vulnerable given their potential need for emergency assistance in the event of evacuations. These vulnerabilities are only magnified in the event that childcare and long-term care facilities are physically affected by homeland security events. Ethnically diverse populations may disproportionately suffer if culture and language differences inhibit the distribution of postdisaster funding. Other population characteristics, such as limited poverty and lower employment, are indicators of decreased social vulnerability; the population is presumably better able to absorb the consequences and recover from homeland security incidents.

While social vulnerability characterizes a population of individuals residing in a specific geographic area, physical vulnerabilities are the properties of the built environment that make an area more susceptible to harm. In quantifying the vulnerability of the built environment, researchers have considered variables assessing residential property, commercial and industrial development, lifelines, transportation, infrastructure, and monuments/icons (Borden, Schmidtlein, Emrich, Piegorsch, & Cutter, 2007). Physical vulnerability is addressed in a number ways. Borden et al. (2007; see also Piegorsch, Cutter, & Hardisty, 2007) created a composite index of measures but limited their study to a sample of only large cities. Armas (2008) focused exclusively on one area of interest using a case study methodology. Instead of relying on inventories of the local built environment, Davis et al. (2004) asked survey respondents to identify the number of different types of specific potentially vulnerable targets in their jurisdiction.

Physical vulnerability and social vulnerability have been studied together in past research. Borden et al. (2007) examined the vulnerability of 132 large cities in the United States to environmental hazards. They created indices for socioeconomical vulnerability, built environment vulnerability, and hazard exposure/experience. Adding all three vulnerability index scores together revealed the vulnerability of the cities in relation to one another and indicated that New Orleans was the most vulnerable city in the United States (Borden et al., 2007). Piegorsch et al. (2007) combined the three indices used by others (social, built environment, and hazard vulnerability) into one place-based vulnerability index for 132 large cities in the United States. Using this information, as well as past data on terrorist incidents in the United States, they determined that the place-based vulnerability index was able to significantly predict both whether a terrorist incident occurred and whether it resulted in casualties (Piegorsch et al., 2007).

How individuals perceive risk is somewhat different than the analytical risk assessments performed by experts. The perception of risk can be defined as “the judgments people make when they are asked to characterize and evaluate hazardous activities” (Slovic, 1987, p. 280). While risk was once thought to be completely objective, it is now accepted that risk is actually a subjective judgment and can be influenced by a wide array of factors (Slovic, 2000). The present study examines the relationship between risk—police executives’ opinions of probabilities of homeland security events occurring within their jurisdiction, the objective risk factors of social vulnerability, built environment vulnerability, and experience with past hazards—and preparedness.

Present Study

Past research has found support for the effect of homeland security risk perceptions on preparedness. Increased perceived risk of homeland security incidents is significantly associated with higher levels of preparedness in several studies (Burruss et al., 2010; Davis et al., 2004, 2006; Gerber et al., 2005; Schafer et al., 2009). However, this research has failed to consider, or has only partially considered (Roberts et al., 2012), objective risk factors. By using survey data from small municipal agencies combined with several sources of data outside of the traditional realm of criminal justice research, this study provides a more thorough understanding of the relationship between risk and organizational activities.

Sample

The sampling frame was developed using the 2004 Census of State and Local Law Enforcement Agencies, a commonly used source for identifying organizations comprising the law enforcement population (Bureau of Justice Statistics, 2011). For purposes of the research, a small department was defined as a municipal law enforcement agency employing at least one but no more than 25 full-time sworn officers, a range that covers approximately 78% of departments (Reaves, 2010). ² Each of the 9,708 small agencies fitting these criteria was placed into one of nine rural–urban continuum code categories based on county location. Rural–urban continuum codes were developed by the U.S. Department of Agriculture to distinguish between U.S. counties on the basis of population size and proximity to metropolitan areas (“Measuring Rurality,” n.d.). Stratifying agencies by continuum code recognizes the differences in small agencies, overcoming the misconception that all small agencies are the same and are located in rural communities (see Crank & Wells, 1991; Falcone, Wells, & Weisheit, 2002 for a discussion of this variation). Indeed, nearly half of all small agencies are located in metropolitan counties (see Burruss et al., 2012). Ninety departments were selected from each stratum for a total of 810 agencies. Surveys were mailed to the executive of each agency (e.g., chief, commissioner, officer in charge), requesting participation in the research. The survey instrument addressed a range of topics: perceived risk of homeland security incidents, preparedness activities, interactions with large agency peers, environmental influences, and other topics. Responses were received from 350 agency leaders (34–51 per strata), a response rate of 44.5%. ³ Comparisons between respondents and nonrespondents revealed few differences across 36 different measures (e.g., budget region, functions/responsibilities). Only three significant differences emerged, each related to agency functions (responsibility for first response, arrest of criminal suspects, and processing firearms applications), although effect sizes indicated weak relationships.

Dependent Variable

The dependent variable of interest is preparedness, steps taken by law enforcement agencies to prevent, respond to, and recover from homeland security incidents. Survey respondents were presented with brief descriptions of 13 steps or activities designed to enhance preparedness and asked to indicate which were practiced by their organizations. ⁴ While this is a common approach for collecting preparedness data, there is less agreement on its actual measurement in analytical models; some researchers examine the steps individually (e.g., Davis et al., 2004, Roberts et al., 2012), employ factor analysis to reduce the number of items to a single preparedness factor (e.g., Pelfrey, 2007), or sum the items into a single additive index (e.g., Burruss et al., 2010; Randol, 2012). In the present study, survey responses were dichotomously coded (1 indicating adopted or practicing policy/activity; 0 indicating absence of policy/activity) and summed into an index ranging from 0 to 13 (α = .815).

Independent Variables

Thirty-five separate indicators were gathered to measure various aspects of the TVC model of risk, including perceived risk, hazards, built environment vulnerability, and social vulnerability. A series of factor analyses (principal components extraction with varimax rotation to aid in interpretability) was used to reduce the number of items into a smaller number of latent components. Rather than enter all 35 items simultaneously, items were entered into separate analyses according to theoretical expectations and the empirical literature (Borden et al., 2007; Burruss et al., 2010; Cutter et al., 2003). In each instance where a factor analysis is reported, factor scores were saved in statistical software for use in the regression and path analyses reported later. ⁵

If a threat is measured according to the probability of a homeland security incident occurring, it is necessary to know something about the “intention and capability of an adversary to undertake actions that would be detrimental to an asset or population” or the likelihood of some type of natural or man-made disaster (Cox, 2008, p. 1750; Willis, 2007). Given the inherent difficulties in directly measuring these probabilities, police scholars have relied on respondent assessments of event likelihoods, a procedure followed in the present study. Survey respondents were asked to assess the likelihood that a terrorist attack would occur within their jurisdiction within the next 5 years. Responses to six items, each reflecting a different type of terrorist attack (chemical, biological, radiological, conventional explosive, cyber-terrorism, and military weapon), were scored on a scale from 0 (not at all likely) to 10 (very likely). The scores were factor analyzed, producing a single perceived risk component accounting for 66.3% of the variance in the indicators (see Table 1).

OPEN IN VIEWER

Table 1.

Six-Factor Analysis Solutions, With Indicators and Data Sources.

Analysis	Components	Indicators	Source	Range of factor loadings	Percent of variance explained by factor
1	Perceived risk	Likelihood of chemical, biological, radiological, conventional explosive, cyber-terrorism, or military weapon attack	Survey	0.757–0.876	66.3
	Total variance explained		66.3
2	TRANS1: Bridges	Number of highway bridges, rail bridges	Hazus-MH	0.819–0.859	27.8
	TRANS2: Water transportation	Number of ferry facilities, ports	Hazus-MH	0.800–0.892	24.7
	TRANS3: Mass transportation	Number of airports, bus facilities	Hazus-MH	0.631–0.928	22.1
	Total variance explained		74.5
3	LIFE1: Other	Number of fire departments, schools, police departments, wastewater facilities, hospital beds, potable water facilities, electric power facilities, emergency operations centers	Hazus-MH	0.581–0.875	46.6
	LIFE2: Oil/gas	Number of gas facilities, oil facilities	Hazus-MH	0.691–0.829	15.8
	Total variance explained		62.4
4	Commercial/industrial development	Number of private, nonfarm establishments; manufacturing establishments, hazardous materials facilities	U.S. Census and Hazus-MH	0.913–0.948	87.5
	Total variance explained		87.5
5	Residential infrastructure	Housing density, number of building permits	U.S. Census	0.928–0.928	86.2
	Total variance explained		86.2
6	HAZ1: Casualty	Injuries per 10,000; fatalities per 10,000	SHELDUS	0.829–0.847	25.9
	HAZ2: Monetary loss	Property damage as % of county domestic product; crop damage as % of county domestic product	SHELDUS	0.752–0.892	23.5
	HAZ3: Frequency/diversity	Incidents per year; diversity of incidents	SHELDUS	0.865–0.885	26.3
Total variance explained		75.6

The next set of measures captured aspects of a location’s vulnerability, typically its infrastructure and assets, to homeland security incidents (Borden et al., 2007; Cutter et al., 2003; Piegorsch et al., 2007; Willis, 2007). ⁶ In the present study, vulnerability is captured using measures derived from Borden et al.’s (2007) research on physical and social vulnerability. They described the built environment as including transportation infrastructure, lifelines, commercial and industrial development, and residential property. ⁷ Data for the built environment came from both the U.S. Census Bureau and, in the case of transportation, lifelines, and a portion of the commercial development measures, the Federal Emergency Management Agency’s (FEMA) Hazus-MH software program (2012). Hazus is designed to allow governments and emergency management personnel to estimate losses associated with floods, hurricanes, and earthquakes. Within the program are databases documenting critical infrastructure. Although the software is current, data contained within the software program are collected at different times.

The built environment measures represent nonnormalized counts; they were not converted to rates to take into account interjurisdictional differences in population sizes. Comments from Illinois police leaders indicate that the number of targets, rather than the ratio of targets to the population, is an important consideration in homeland security preparedness:

[The town is] surrounded by major interstates. The major line of the Burlington Northern Santa Fe, three rails pass through town. I expect that over the next 3 years some form of terrorist act will occur.

We don’t have the emergency radio equipment needed with [the redacted] Chemical Depot just 20 miles away!

We are a small community…we only have several businesses and 2 part-time officers…we don’t really have many conceivable possible terrorist targets but that doesn’t mean we shouldn’t plan for the worst.

We interact with the CN [Canadian National] and Metro BNSF [Burlington Northern Santa Fe] and Amtrak police departments because of the 5 rail lines that bisect our community.

I have 2 rail systems, large grain and fertilizer operations in my jurisdictions (as quoted in Schafer et al., 2009).

A county’s transportation infrastructure was measured using multiple indicators: the number of airports, bus facilities, ferry facilities, highway bridges, rail bridges, and ports. A factor analysis (Analysis 2) revealed a three-component structure comprising bridges, water transportation, and mass transit (see Table 1). Lifelines included fire departments, schools, police departments, wastewater facilities, potable water facilities, electric power plants, natural gas facilities, oil facilities, emergency operations centers, and the total number of hospital beds. As shown in Analysis 3 (see Table 1), gas and oil facilities loaded onto one factor (gas/oil) and the remaining indicators loaded onto a second factor (other lifelines). The commercial and industrial infrastructure was captured with three indicators: the number of private, nonfarm establishments in 2009, the number of manufacturing establishments in 2007, and the number of hazardous materials facilities. All items loaded onto a single commercial/industrial factor (see Analysis 4 in Table 1). The residential factor included two items: housing unit density per 100 square miles (2010) and building permits issued (2011). The residential factor accounted for 86.2% of the variance in the indicators (Analysis 5 in Table 1).

Social vulnerability was measured using Cutter et al.’s (2003) SoVI developed nearly a decade ago. A latent construct comprising 31 separate county-level socioeconomic indicators, the SoVI captures characteristics of the population across seven areas: race and class, extreme wealth, elderly residents, Hispanic ethnicity, care-dependent females, Native American ethnicity, and service industry employment (Hazards and Vulnerability Research Institute, 2010). SoVI scores were gathered for each of the law enforcement agencies in the sample from the Hazards and Vulnerability Research Institute at the University of South Carolina. SoVI scores, as they are the product of a factor analysis of 31 indicators, are useful for comparative analysis only; the score by itself is unitless; its value comes when examined relative to other scores. Descriptive statistics for the SoVI or its indicators are not reported here.

The remaining measures were included to capture the likelihood of consequences, or magnitude of damage, of homeland security incidents (Willis, 2007). Data were extracted from the Spatial Hazard Events and Losses Database for the United States (SHELDUS) maintained by the University of South Carolina’s Hazards and Vulnerability Research Institute (2012a). The SHELDUS database includes information on five decades worth of natural events producing injury, death, and monetary loss organized at the county level. ⁸ Six indicators derived from Borden et al. (2007) were used from the SHELDUS data covering the period 2001 through 2010. Magnitude and intensity measures included the number of deaths and number of injuries during the period per 10,000 county residents and the amount of property and crop loss during the period as a percentage of the county’s 2006 domestic product. Temporal spacing indicated the number of events within each county during the 10-year period, while the diversity of incidents counted the number of different types of incidents occurring during the period out of the 18 recorded in the SHELDUS database. The factor analysis yielded a three-component solution where injuries and fatalities loaded onto a single casualty factor, crop and property damage form a monetary loss factor, and the temporal/diversity indicators combine into a frequency/diversity factor (see Table 1). It should be noted that the measures of consequences are limited in that they only measure the consequences of natural hazards rather than accidents or terrorist incidents.

A measure of the county’s 2010 resident population (logged) as determined by the U.S. Census was also included in several analyses later. Descriptive statistics for observed indicators are shown in Table 2. A correlation matrix for all measures is included in Table 3.

OPEN IN VIEWER

Table 2.

Descriptive Statistics for Observed Indicators (Independent Variables).

Indicator	Mean	SD	Indicator	Mean	SD
PERCEIVED RISK	LIFELINES
Chemical	2.44	2.43	Fire departments	14.34	16.89
Biological	1.97	2.05	Schools	58.38	109.64
Radiological	1.72	2.04	Police departments	10.58	13.91
Conventional explosive	2.72	2.35	Wastewater facilities	11.59	28.76
Cyber-terrorism	2.76	2.51	Hospital beds	433.08	1,125.07
Military weapon	1.84	2.05	Potable water facilities	1.67	6.72
HAZARDS	Electric power facilities	2.00	3.82
Injury rate per 10,000	1.69	4.87	Emergency operations centers	0.94	3.04
Fatality rate per 10,000	0.30	0.69	Gas facilities	2.09	13.46
Property damage	2.93	12.53	Oil facilities	0.44	1.58
Crop damage	0.77	3.43	COMMERCIAL
Incidents per year	5.83	5.37	Nonfarm establishments	3,017.30	7,007.96
Diversity of incidents	6.65	1.66	Manufacturing	149.07	333.23
TRANSPORTATION	Hazmat facilities	54.05	163.13
Highway bridges	195.27	226.58	RESIDENTIAL
Rail bridges	3.47	8.12	Housing density	8,913.87	20,589.63
Ferry facilities	0.18	0.73	Building permits	266.59	1,049.35
Ports	4.44	20.41	OTHER
Airports	1.83	4.24	Dams	27.67	44.84
Bus facilities	1.40	2.59	Nuclear	0.05	0.22
	Population, 2010	125,433.42	296,326.75

OPEN IN VIEWER

Table 3.

Correlation Matrix for All Measures Included in Analyses.^a

(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
(1) Preparedness	1.000
(2) PERCEIVED RISK	0.258	1.000
(3) HAZ1	−0.099	−0.072	1.000
(4) HAZ2	−0.135	−0.069	0.000	1.000
(5) HAZ3	0.086	0.069	0.000	0.000	1.000
(6) SoVI	−0.141	−0.023	0.177	0.218	−0.281	1.000
(7) TRANS1	0.083	0.081	−0.151	−0.228	0.320	−0.270	1.000
(8) TRANS2	0.080	0.027	−0.047	−0.058	0.143	−0.077	0.000	1.000
(9) TRANS3	0.143	0.037	−0.132	−0.216	0.198	−0.085	0.000	0.000	1.000
(10) LIFE1	0.181	0.049	−0.220	−0.345	0.452	−0.483	0.585	0.255	0.576	1.000
(11) LIFE2	0.079	0.075	0.002	−0.105	0.205	0.052	0.083	0.116	0.258	0.000	1.000
(12) COMMERCIAL	0.202	0.080	−0.246	−0.364	0.468	−0.596	0.527	0.237	0.478	0.859	0.232	1.000
(13) RESIDENTIAL	0.234	0.088	−0.266	−0.351	0.420	−0.620	0.490	0.257	0.352	0.803	0.183	0.896	1.000
(14) Population (logged)	0.220	0.079	−0.248	−0.390	0.497	−0.538	0.559	0.221	0.511	0.866	0.270	0.946	0.914	1.000
(15) COMMRES b	0.220	0.085	−0.261	−0.369	0.461	−0.621	0.526	0.251	0.440	0.857	0.218	0.984	0.961	0.958	1.000

Note. ^aMeasures derived from factor analyses are in all caps.

b

COMMRES comprises all indicators included in the COMMERCIAL and RESIDENTIAL factors. It is discussed in more detail in the text.