Data Sources & Methodology

In the US, perceptions of public safety are inherently tied to perceptions of crime. When perceptions of crime are high, Americans feel unsafe. When perceptions of crime are low, Americans generally feel safe. However, in an effort to heighten public safety and curtail crime, governments have enacted punitive, "tough on crime" policies that have disproportionately targeted non-white groups.

This narrow perspective on public safety overlooks the multifaceted nature of safety. Safety, fundamentally, is freedom from fear—a concept that spans far beyond the threat of physical harm (i.e. violent crime) to include the fear of economic security, health access, educational inequities, housing stability, and social and civic exclusion (Buxton, Fradkin, and Wallace, 2021). While it is true that crime is a salient and critical component of how public safety should be operationalized, it is imperative that it be understood as merely one part of a much broader and encompassing framework.

Unfortunately, freedom from fear has only been operationalized and permitted for white Americans at the expense of violating and infringing on the safety of Black Americans and other marginalized groups. As a consequence, for most Americans it is impossible to detangle pathological understandings of race from how crime is understood and responded to. Stated plainly, American fears of crime are rooted almost entirely in anti-Blackness and xenophobia.

To improve public safety through a restorative approach, it's imperative to move beyond a solely crime-centric perspective. Crime-centric approaches neither help identify restorative ways to reduce violence, nor do they help build a society that combats the cyclical and recurring societal responses that reinforce harmful racial narratives. Here, we propose a working comprehensive public safety framework which leverages the co-produced community public safety frameworks beyond crime. Specifically, the proposed Safety Cities framework here is modeled after the safety domains co-developed by the National Innovation Service and NYC residents (2021) which sought to capture comprehensive measures of safety as identified by individuals most impacted by its absence.

We term this the "Social Determinants of Safety" (SDOS). By adopting this framework to render safety across the US, we aim to center the historical reality that safety has been inequitably distributed to Black and other non-White groups. It is a critical step for allowing communities restructure public safety coalitions and stakeholders that are committed to a restorative rather than punitive response to social problems.  

Safety Domains & Social Determinants of Safety

The Safe Cities framework represents an attempt to redefine and measure public safety moving beyond traditional metrics that equate safety exclusively with crime rates. This initiative is not merely a pivot but a critical leap towards an inclusive and holistic understanding of what it means to live in a safe society. 

Further, inspired by Braveman, Egerter, and Williams (2011), and drawing heavily from the NIS safety model co-produced with community co-researchers (Buxton, Wallace, and Fradkin 2021), we have delineated eight safety domains that comprise the SDOS integral to understanding and improving public safety through an anti-racist lens. These eight proposed domains include the following:

• Crime: Crime, reported and unreported, play a critical role in understanding the threats to physical security. 

• Policing and Incarceration: Responding to social problems with law enforcement and mass incarceration indicates the amount of harm caused by lack of investment into other safety domains.

• Economic Conditions: Providing sufficient access to employment opportunities and capital to secure resources that meet conditions are critical in supporting healthy decision-making.

• Health: Affordable and quality healthcare allows individuals to receive the appropriate care without causing financial stress or health crises that could lead to dangerous situations.

• Housing: Affordable and quality housing is essential for maintaining a safe and stable living environment that can combat homelessness and/or reduce financial stress that could result in other means to secure financial resources.

• Education: Affordable and quality educational opportunities allow for stronger employment opportunities and improved economic conditions. 

• Built Environment & Climate: Community-designed and well-resourced public infrastructure and urban planning can allow for climate resilient and functional environments that support long-term safety.

• Civic & Social Participation. Active engagement in civic life indicates that residents feel included and valued, contributing to a cohesive community that may be more involved in safety planning and problem solving processes. 

The development of this framework has also integrated the critical insights and theories of anti-racist scholars like Ruth Wilson Gilmore (2007) and the foundational tenets of critical race theory (Crenshaw et al., 1995). We recognize that systemic racism not only contributes to disparities in safety but also engenders a differential allocation of fear and vulnerability across society. This framework, therefore, is a declaration of our commitment to dismantle these structures of inequality and to champion policies that are rooted in equity and justice.

Limitations

• Working & Iterating Framework. It is crucial to acknowledge that this framework is in its nascent stages, constituting the first step in a long and ongoing journey towards redefining public safety. While we have made significant strides in identifying key domains and measures, we recognize that this is a living document, subject to continuous refinement and iteration. Our commitment to engaging with communities, incorporating their insights, and adapting the framework in response to emerging needs and challenges is unwavering.

• Complexity and Interlinkages of Domains. The domains identified within the Safety Stats Framework are deeply interlinked, with the relationships between them complex and multifaceted. The interplay of economic security, education, health, housing, and other factors contributes to the overall safety of a community, and these interactions can vary significantly across different jurisdictions. Recognizing and understanding these complexities is paramount, as it enables us to develop targeted interventions and policies that address the unique needs of each community.

• Variability Across Jurisdictions. The framework also acknowledges the inherent variability in how safety is experienced and understood across different jurisdictions. Local contexts, histories, and demographics play a crucial role in shaping the dynamics of safety, necessitating a flexible and adaptable approach. By considering these local nuances, the Safety Stats Framework aims to be a versatile tool, capable of guiding public safety efforts across diverse communities and settings.

• Concerns with Generalizing. Given differences in history, geography, and a range of other factors that influence urban development, there will be significant variation both across, and within cities, of what safety domains are most salient and relevant. In other words, we cannot treat the US, or cities, as a monolith. While we have done our best to take a comprehensive approach to developing safety domains to account for the wide variation in safety, we will continue to work to iterate and improve the framework. Furthermore, we encourage local communities to employ the framework as a starting point but to shape it to make it best suited to their own local communities. For example, issues of housing will not translate the same way from city to city due to variation in access to public housing and other factors and programs in place. 

Procedure for Extracting, Automating, and Structuring Safety Data

To tackle the endeavor of consolidating and rendering safety in municipalities across the US, we divided the data generation and structuring into two components: Physical Security (Crime Data) and Safety data beyond Physical Security (US Census and IPUMS). 

First, we worked to develop a way to extract and automate crime data, representing the Crime Victimization safety domain. For the first phase of Safe Cities, we wanted to start with IPUMS and Census data which is more standardized, comparable, and available than crime data. Thus, given the sensitive nature of the data and the significant limitations and caveats with crime data, we started with first assessing the accuracy, completeness, and quality of the crime data for several dozen municipalities by cross-referencing them with federal, state, local, and other non-governmental data sources. Subsequently, after identifying which cities fit the criteria to be included for the first iteration of Safe Cities, we started to extract and structure safety data for the remaining domains employing IPUMS US Census data. Finally, we believe we have a responsibility to the users of this data platform to best present the data in a way which is true, accurate, contextualized, and not employed in a way which can reinforce harmful narratives. To do this, we made design and research decisions on how to visualize the data. Thus, our approach to extract, automate, and structure safety data was as follows: 

Extraction, Automation, Categorization, and Structuring of Crime

Instead, CZ researchers determined that it would be best to develop in-house data pipelines to each city's open data portal, and develop a process to update, check categorization, and structure monthly crime data. Additionally, the historical FBI crime data was harmonized with these more recently updated city portal data. After harmonizing the data, we developed a process for identifying the first set of cities to focus on for Phase 1 of Safe Cities. 

To determine which cities and department data would be featured in the initial release of Safe Cities, we undertook comparative research. We contrasted annual crime aggregates between the FBI Crime Data Explorer and local level data, focusing particularly on annual homicide totals from 2018-2020. Cities where the three year average differences ranged between 0 and 7% were considered for the site's launch. While we initially started with several dozens of US cities, given significant variation and concerns with accuracy of crime data, we ultimately landed on starting safety stats with nine US cities: 

1. Atlanta, GA

2. Austin, TX 

3. Boston, MA

4. Chicago, IL

5. Denver, CO

6. Detroit, MI

7. Los Angeles, CA

8. New York, NY

9. Phoenix, AZ

10. Seattle, WA

11. Washington, DC

FBI Offense Definitions

The categories employed for all crime categorization on Safe Cities are defined by the FBI. We include part I offenses and exclude part II offenses.

Part I Offenses

• Criminal homicide ― a.) Murder and nonnegligent manslaughter: the willful (nonnegligent) killing of one human being by another. Deaths caused by negligence, attempts to kill, assaults to kill, suicides, and accidental deaths are excluded. The program classifies justifiable homicides separately and limits the definition to: (1) the killing of a felon by a law enforcement officer in the line of duty; or (2) the killing of a felon, during the commission of a felony, by a private citizen. b.) Manslaughter by negligence: the killing of another person through gross negligence. Deaths of persons due to their own negligence, accidental deaths not resulting from gross negligence, and traffic fatalities are not included in the category manslaughter by negligence.

• Rape ― The penetration, no matter how slight, of the vagina or anus with any body part or object, or oral penetration by a sex organ of another person, without the consent of the victim.

• Robbery ― The taking or attempting to take anything of value from the care, custody, or control of a person or persons by force or threat of force or violence and/or by putting the victim in fear.

• Aggravated assault ― An unlawful attack by one person upon another for the purpose of inflicting severe or aggravated bodily injury. This type of assault usually is accompanied by the use of a weapon or by means likely to produce death or great bodily harm. Simple assaults are excluded.

• Burglary (breaking or entering) ― The unlawful entry of a structure to commit a felony or a theft. Attempted forcible entry is included.

• Larceny-theft (except motor vehicle theft) ― The unlawful taking, carrying, leading, or riding away of property from the possession or constructive possession of another. Examples are thefts of bicycles, motor vehicle parts and accessories, shoplifting, pocket-picking, or the stealing of any property or article that is not taken by force and violence or by fraud. Attempted larcenies are included. Embezzlement, confidence games, forgery, check fraud, etc., are excluded.

• Motor vehicle theft ― The theft or attempted theft of a motor vehicle. A motor vehicle is self-propelled and runs on land surface and not on rails. Motorboats, construction equipment, airplanes, and farming equipment are specifically excluded from this category.

• Arson ― Any willful or malicious burning or attempt to burn, with or without intent to defraud, a dwelling house, public building, motor vehicle or aircraft, personal property of another, etc.

• Human Trafficking, commercial sex acts ― Inducing a person by force, fraud, or coercion to participate in commercial sex acts, or in which the person induced to perform such act(s) has not attained 18 years of age.

• Human Trafficking, involuntary servitude ― The obtaining of a person(s) through recruitment, harboring, transportation, or provision, and subjecting such persons by force, fraud, or coercion into involuntary servitude, peonage, debt bondage, or slavery (not to include commercial sex acts).

The Part II offenses, for which only arrest data are collected, are:

• Other assaults (simple) ― Assaults and attempted assaults where no weapon was used or no serious or aggravated injury resulted to the victim. Stalking, intimidation, coercion, and hazing are included.

• Forgery and counterfeiting ― The altering, copying, or imitating of something, without authority or right, with the intent to deceive or defraud by passing the copy or thing altered or imitated as that which is original or genuine; or the selling, buying, or possession of an altered, copied, or imitated thing with the intent to deceive or defraud. Attempts are included.

• Fraud ― The intentional perversion of the truth for the purpose of inducing another person or other entity in reliance upon it to part with something of value or to surrender a legal right. Fraudulent conversion, obtaining of money or property by false pretenses, confidence games, and bad checks, except forgeries and counterfeiting, are included.

• Embezzlement ― The unlawful misappropriation or misapplication by an offender to his/her own use or purpose of money, property, or some other thing of value entrusted to his/her care, custody, or control.

• Stolen property: buying, receiving, possessing ― Buying, receiving, possessing, selling, concealing, or transporting any property with the knowledge that it has been unlawfully taken, as by burglary, embezzlement, fraud, larceny, robbery, etc. Attempts are included.

• Vandalism ― To willfully or maliciously destroy, injure, disfigure, or deface any public or private property, real or personal, without the consent of the owner or person having custody or control by cutting, tearing, breaking, marking, painting, drawing, covering with filth, or any other such means as may be specified by local law. Attempts are included.

• Weapons: carrying, possessing, etc. ― The violation of laws or ordinances prohibiting the manufacture, sale, purchase, transportation, possession, concealment, or use of firearms, cutting instruments, explosives, incendiary devices, or other deadly weapons. Attempts are included.

• Prostitution and commercialized vice ― The unlawful promotion of or participation in sexual activities for profit.

• Sex offenses (except rape, prostitution, and commercialized vice) — Offenses against chastity, common decency, morals, and the like.

• Drug abuse violations ― The violation of laws prohibiting the production, distribution, and/or use of certain controlled substances. The unlawful cultivation, manufacture, distribution, sale, purchase, use, possession, transportation, or importation of any controlled drug or narcotic substance. Arrests for violations of state and local laws, specifically those relating to the unlawful possession, sale, use, growing, manufacturing, and making of narcotic drugs. The following drug categories are specified: opium or cocaine and their derivatives (morphine, heroin, codeine); marijuana; synthetic narcotics―manufactured narcotics that can cause true addiction (Demerol, methadone); and dangerous nonnarcotic drugs (barbiturates, Benzedrine).

• Gambling ― To unlawfully bet or wager money or something else of value; assist, promote, or operate a game of chance for money or some other stake; possess or transmit wagering information; manufacture, sell, purchase, possess, or transport gambling equipment, devices, or goods; or tamper with the outcome of a sporting event or contest to gain a gambling advantage.

• Offenses against the family and children ― Unlawful nonviolent acts by a family member (or legal guardian) that threaten the physical, mental, or economic well-being or morals of another family member and that are not classifiable as other offenses, such as assault or sex offenses. Attempts are included.

• Driving under the influence ― Driving or operating a motor vehicle or common carrier while mentally or physically impaired as the result of consuming an alcoholic beverage or using a drug or narcotic.

• Liquor laws ― The violation of state or local laws or ordinances prohibiting the manufacture, sale, purchase, transportation, possession, or use of alcoholic beverages, not including driving under the influence and drunkenness. Federal violations are excluded.

• Drunkenness ― To drink alcoholic beverages to the extent that one’s mental faculties and physical coordination are substantially impaired. Driving under the influence is excluded.

• Disorderly conduct ― Any behavior that tends to disturb the public peace or decorum, scandalize the community, or shock the public sense of morality.

• Vagrancy ― The violation of a court order, regulation, ordinance, or law requiring the withdrawal of persons from the streets or other specified areas; prohibiting persons from remaining in an area or place in an idle or aimless manner; or prohibiting persons from going from place to place without visible means of support.

• All other offenses ― All violations of state or local laws not specifically identified as Part I or Part II offenses, except traffic violations.

• Suspicion ― Arrested for no specific offense and released without formal charges being placed.

• Curfew and loitering laws (persons under age 18) ― Violations by juveniles of local curfew or loitering ordinances.

Extraction, Automation, and Structuring of Safety Data beyond Crime

Given the lack of standardized and validated frameworks for categorizing local-level safety data (i.e. 911, etc), we started with the US Census's ACS data (via IPUMS), providing city-level socioeconomic metrics and population counts disaggregated by race and ethnicity from 1980 onwards. Policing & Incarceration data includes fatal police violence by city, year, and victim race, sourced from Mapping Police Violence.

The largest single source of safety data beyond crime is the US Census’s American Community Survey (ACS), which we access via IPUMS. IPUMS offers data at the geographic level of Public Use Microdata Areas (PUMAs). We selected PUMAs that most closely matched the PLACE Census variable. For most cities and years, the PLACE and matched PUMAs overlap exactly. However there are some exceptions such as Austin where the census city boundaries shifted between the 2010 and 2020 census. 

We extracted data for all available years from 1980 onwards. ACS data are published on an annual basis, so we redownload this data from IPUMS and update them on the platform once a year.

These data include (with their variable names in parentheses): poverty rate (POVERTY), public assistance rate (INCWELFR), health insurance rate (HCOVANY), high school education rate (EDUC), average household income (HHINCOME), average rental burden (via RENTGRS), and unemployment rate (EMPSTAT).

In addition to being presented as city-level metrics for each year, all of these variables are further disaggregated, based on the RACE and HISPAN variables, into the following categories: AAPI, Black, Hispanic, White, and Other/Multi.

For both crime data and safety data beyond crime, some metrics require population counts for each city and year. For these numbers, we source them from the US Census’s American Community Survey (ACS), which we access directly via the Census API. We used the PLACE Census variable to specify the geographic unit in which the data is collated, which corresponds to the city level. These counts are in both aggregate form, and disaggregated by race and Hispanic/Latinx origin.

Additionally, we source eviction data from the Eviction Lab. We use the city-level data that is currently updated on an ongoing basis. Of these cities in active tracking, there are 4 which overlap with Safe Cities' cities: Austin, Boston, New York, and Phoenix.