Abstract
In 1979, a group of men incarcerated in New York state (NYS) prisons analyzed the relationship between neighborhoods and incarceration, finding that 75 percent of the NYS prisons’ population came from just seven neighborhoods in New York City (NYC). Inspired by this study, we combine novel archival data with census tract imprisonment data to examine the geography of incarceration in more recent years. Our findings reveal a marked spatial deconcentration of imprisonment since the original study. By 2020, 75 percent of people imprisoned in NYS came from 1,551 census tracts—far exceeding the spatial area covered by the original seven neighborhoods containing about 300 tracts. Using spatial lag regression and cluster analysis, we identify over 20 cities with significantly high imprisonment rate clusters. Despite this geographic shift, some tracts within the original seven neighborhoods continue to have among the highest incarceration rates in the state. Our findings both challenge existing urban theories linking concentrated punishment to urban metros and highlight the durability of punishment vulnerability in place.
Scholarly attention to the geographic pattern of incarceration has increased in recent years, but the first mapping of neighborhood-level imprisonment rates came from within prison walls over 40 years ago. While incarcerated in Green Haven, a maximum-security prison in upstate New York, a group of incarcerated lay researchers led by Lawrence White and Edwin “Eddie” Ellis formed the “Think Tank” and produced an original data analysis of the zip codes and assembly districts of people entering prison in New York state (NYS). Their research began from an anecdotal observation: they often found themselves imprisoned alongside people they knew from home. The Think Tank formalized this social fact and supported it with empirical evidence (Burton 2016). Produced in 1979, and confirmed again in 1990, their research led to the astonishing finding that 75 percent of the state’s prison population came from just seven neighborhoods in New York City (NYC): Harlem, the Lower East Side, Brownsville, Bedford-Stuyvesant, East New York, South Jamaica, and the South and Central Bronx, which together made up only 6 percent of the state’s population (Prisoner’s Alliance with Community 1997).
The Think Tank called their methodology the “Non-Traditional Approach to Criminal and Social Justice.” As a radical form of participatory action research from inside prison, the researchers performed a quantitative analysis that cross-referenced NYS census data with New York Department of Corrections population data to identify the prior neighborhoods of the state’s incarcerated people (Burton 2016). To explain the spatial concentration they identified, they considered a broad set of community-level conditions they termed “crime generative factors” that led to excess criminalization and punishment. What would later be called the “Seven Neighborhoods Study” spurred numerous data and research projects, including the Million Dollar Blocks study of NYC, the Million Dollar Hoods project in Los Angeles County, and studies in Chicago, Tallahassee, and the state of Massachusetts, among other localities (Burch 2014; Clear 2007; Cooper and Lugalia-Hollon 2020; Dhondt 2012; Eason 2010; Fagan and West 2013; Fagan, West, and Holland 2004; Gilmore 2007; Holder et al. 2022; Lynch and Sabol 2004; Lytle Hernandez and Dupuy 2020; Manduca and Sampson 2019; Renauer et al. 2006; Saifee 2022; Sampson and Loeffler 2010; Simes 2021; Simes, Beck, and Eason 2023; Spatial Information Design Lab 2007). In addition, the Think Tank presented their research to legislative conferences with the aim of influencing criminal justice policy reform (Burton 2016).
There are significant theoretical, empirical, and political implications of the original Seven Neighborhoods Study. The Think Tank pointed to state and local divestment in civic infrastructure—such as education, social welfare, and healthcare—and to the simultaneous investment in harsh criminal justice practices as mechanisms driving high incarceration rates in these neighborhoods. To “change the conditions that contribute to crime,” the Think Tank advocated for a redirection of public funds to address root causes: concentrated poverty, inadequate support for families, physical suffering, and “a sense of victimization and despair” (Prisoner’s Alliance with Community 1997:2). In doing so, the Think Tank researchers moved the locus of analysis—and intervention—from individual deviance or deficiencies and toward the social conditions within neighborhoods that produce crime and thus high rates of incarceration. If systems of punishment affect the social, economic, and political life of communities, the Think Tank advocated for reforms that empower communities to decide what initiatives will reduce suffering and expand opportunity across those domains (Prisoner’s Alliance with Community 1997:16).
Long before its recent vogue, the Think Tank identified the importance of tracking neighborhoods of origin, something criminal justice institutions have only recently begun to collect in a systematic way. The Think Tank’s approach underscored the urgency of place-based analysis to identify where incarceration is most concentrated, explain its geographic pattern, and develop policy remedies. In part due to these data limitations, research on mass incarceration and social inequality has historically favored examining national or state-level rates or neighborhoods within single large city case studies. Now that neighborhood data are being collected more accurately and consistently, the original Seven Neighborhoods Study can be meaningfully revisited.
At the time of the Seven Neighborhoods Study, imprisonment rates were spatially concentrated within large urban centers. Recent research has documented a changing geographic pattern of imprisonment, wherein small cities and communities outside of large urban metros now contain neighborhoods with the highest jail and imprisonment rates (Kang-Brown et al. 2023; Simes 2021; Subramanian, Henrichson, and Kang-Brown 2015). Thus, under these conditions, we revisit the original Seven Neighborhoods Study and its theoretical approach. We ask two empirical questions: first, how stable is the ecological structure of imprisonment since the original Seven Neighborhoods Study in the late 1970s? Second, how do spatial and contextual factors explain the uneven distribution of imprisonment across neighborhoods in NYS in the twenty-first century? Answering these questions illuminates how mass incarceration’s 50-year legacy maps onto contemporary neighborhood inequality.
Using data on census tracts and county-level imprisonment rates for the entire state of New York, we structure our analysis in three parts: first, we examine long-run spatial trends in imprisonment using county-level data spanning from 1975 to 2020. Second, we present negative binomial regression models of census tract imprisonment counts to assess contemporary neighborhood-level patterns. Third, we conduct a spatial cluster analysis of predicted 2020 tract-level imprisonment rates to identify concentrations of high imprisonment across NYS.
Legacies of Place-Based Punishment Vulnerability
The prison system has been shown to be deeply embedded in impoverished communities of color (Sampson 2012; Simes 2021), and in recent birth cohorts, incarceration remains a significant part of the way Black men from these communities make their passage through the life course (Gilmore 2007; Knight 2024; Miller 2021; Pettit and Western 2004). The concept of punishment vulnerability (Simes 2021) emphasizes that a small number of places—and all of their community members who may have only indirectly experienced incarceration through family or community membership—experience accumulated and concentrated conditions that render them particularly vulnerable to high rates of incarceration. These conditions span correlated social, cultural, economic, institutional, and political dimensions that work together to entrench punishment vulnerability in place. This in turn significantly diminishes opportunities, well-being, political efficacy, and social cohesion over multiple generations. Punishment vulnerability may be remarkably durable; once established, patterns of disadvantage, segregation, community divestment, and stigmatization reproduce high incarceration rates, even as previous residents vacate and new ones enter (Sampson 2012; Sharkey 2013).
The Seven Neighborhoods Study was the first to formalize a theory of the relationship between punishment and community, publishing the names of places most vulnerable to punishment. The Think Tank proposed a theory of the direct relationship between a host of community conditions and high levels of punishment:
Our analysis demonstrated a “direct connection” between low income, racially isolated, underserved communities in the “seven neighborhoods,” and racism, racial profiling (as expressed in stop and search reports), financial and banking “redlining,” under achieving schools and poor quality education, majority of single parent families headed by women, high rates of unemployment, excess poverty, substance abuse, public assistance, and an entrenched “under-ground economy” that inevitably leads to encounters with law enforcement that result in prison or death. (Nuleadership 2001)
Under the Think Tank’s theorization, incarceration is an aggregate, community-level condition whereby both imprisoned individuals and the residents of their neighborhood of origin absorb the impact of concentrated patterns of imprisonment. When prisoners and communities are “so inextricably linked” (Prisoner’s Alliance with Community 1997:1), all community members’ economic fortunes and political capacities are at stake. Their analysis called for a tracing of “geographic and demographic changes in both the community and prison population” (Prisoner’s Alliance with Community 1997:11). They hypothesized that “the propelling force in this relationship regards the manifestation of racism, institutional failure, decaying socio-ecological conditions, and the implications of Black/Latino on Black/Latino crime” (Prisoner’s Alliance with Community 1997:12). We expect that some communities have faced decades of and intergenerational exposure to high levels of incarceration. Incarceration rates in neighborhoods would thus be strongly predicted by levels of social vulnerability, housing instability, police enforcement practices, racial segregation, and levels of violence.
The Think Tank theory of the direct relationship between community conditions and incarceration was strongly motivated by trying to understand what they termed “crime generative factors” (Prisoner’s Alliance with Community 1997:1). Thus, we also situate the original Seven Neighborhoods Study within urban and place-based theories of crime (Peterson and Krivo 2010; Sampson 2012; Weisburd 2015, 2018). Studies of the geography of crime and violence show how multiple community-level factors cluster together to produce enduring patterns of crime in urban neighborhoods (Sampson 2012; Sampson and Groves 1989; Sampson, Raudenbush, and Earls 1997; Sharkey 2013; Weisburd 2015). Crime concentrates at very small spatial units, and between 3 and 10 percent of addresses account for over half of crime calls across a number of studies and samples (Weisburd 2015). This concentration of crime in place is highly correlated with racial segregation, poverty, and other social conditions that undermine social cohesion and trust that are the bedrock of safe and peaceful communities (Peterson and Krivo 2010; Sampson 2012; Sharkey 2018). The Think Tank’s place-based theory of incarceration is thus an extension of place-based theories of crime, which posit that entrenched and localized conditions of social disorganization (Sampson and Groves 1989; Shaw and McKay 1942) and the routine activities of neighborhood residents (Cohen and Felson 1979) structure the time, opportunity, and social context in which crime and violence occur in a small number of very localized places. The Think Tank’s novel approach contrasts with beliefs that crime rates alone explain the concentration of high incarceration rates, noting that crime and its criminalization emerged from the interconnected social conditions of disadvantage and segregation (Bell 2020; Du Bois 1899; Sampson 2012; Shaw and McKay 1942; Simes et al. 2023).
Although the Seven Neighborhoods Study was the first to formalize a theory of the relationship between punishment and community, it can also be contextualized in more recent studies of carceral geography (Gilmore 2007; D. Moran 2016), the spatial concentration of imprisonment (Holder et al. 2022; Sampson 2012; Sampson and Loeffler 2010; Simes 2018), and theories of punishment vulnerability (Simes 2021). Contemporary research (Simes et al. 2023) on the spatial pattern of incarceration finds two key trends: (1) intense spatial concentration within-city boundaries, for example, in studies of Chicago, New York, and Tallahassee neighborhoods (Clear 2007; Holder et al. 2022; Sampson and Loeffler 2010) and (2) a broad geographic distribution where the highest imprisonment rates have been found in small cities and rural communities in Massachusetts and North Carolina (Burch 2014; Simes 2021). In the following section, we examine how social change may challenge existing theories of punishment vulnerability and carceral geography that have focused on the conditions of large urban metros like NYC.
Crime, Imprisonment, and Social Change
Recent research has documented significant shifts in the demography and geography of imprisonment. Absolute racial disparities are declining nationally (Muller and Roehrkasse 2021, 2025; National Academies of Sciences, Engineering, and Medicine et al. 2023). High rates are concentrated in small cities and rural counties (Simes 2021; Subramanian et al. 2015), and overall levels of incarceration have declined (Clegg et al. 2024; Phelps and Pager 2016). Whereas in the early decades of mass incarceration, people in prison largely hailed from neighborhoods in large urban areas, in more recent years, this spatial pattern has shifted to small cities and towns in both national and state samples (Simes 2021; Subramanian et al. 2015). In a 2022 study of jail and prison admissions, the Vera Institute of Justice documented high levels of imprisonment in NYS’s small counties (Kang-Brown et al. 2023).
Mass incarceration’s deconcentration from large cities must be understood alongside the well-documented crime decline that took place during the 1990s. Changes in the spatial distribution of crime may explain changes in the spatial distribution of punishment (Sampson 2012; Sampson and Loeffler 2010; Simes 2018). The crime decline took place in the decades following the Seven Neighborhoods Study and is typically associated with changes from 1990 to 2000. NYC’s crime decline was by far the largest of any major urban area in the United States (Sharkey 2018; Zimring 2007). For example, homicide and robbery declined by 70 percent in NYC during this period, a radical change in the level of violence in the city (Zimring 2007). These crime declines in violence and other crimes during the 1990s likely contributed to fewer incarcerations of people from NYC. Research also shows that declines in incarceration lagged behind crime declines for several years, as the incarceration rate remained stable into the mid-2000s (Ghandnoosh and Budd 2024). Although crime declined across the United States in the 1990s, small cities, as well as suburban and rural areas, experienced a comparatively modest decline during this period compared to large cities (Duhart 2000). Furthermore, amid changes to overall levels of violence and other types of crime, research documents the spatial restructuring of poverty, segregation, immigration, politics, and other conditions—owing to changes in the labor force; residential preferences for large urban cities and gentrification; and the political concentration of conservatives outside large cities, which may also contribute to the changing geographic pattern of incarceration (Berube and Kneebone 2006; Chapple 2017; Howell and Timberlake 2013; Hwang 2016; Lichter and Johnson 2020; Lichter et al. 2007; Owens 2012; Simes 2021; Ternullo 2024).
In light of recent shifts in the geography of crime and incarceration, we revisit the Seven Neighborhoods Study to assess social change. We first examine to what extent the geographic structure of imprisonment rates remained stable as neighborhood residents changed over the last 40 years. We then deploy the Think Tank’s Non-Traditional Approach by considering a broad array of community factors in a model of tract-level imprisonment and build on both theirs and more recent work emphasizing segregation, concentrated disadvantage, and housing inequality as key factors in generating high rates of incarceration (Holder et al. 2022; Prisoner’s Alliance with Community 1997; Sampson and Loeffler 2010; Simes 2021). Because the Think Tank identified neighborhood-level patterns at the time of mass incarceration’s emergence, we have a rare opportunity to examine the stability of incarceration from the beginning of mass incarceration, using both novel archival county-level data and newly available data in NYS at the census tract level.
Data and Methods
Our empirical approach proceeds in three stages. We begin by analyzing county-level data from 1975 to 2020 to characterize long-term spatial trends in imprisonment. We then estimate negative binomial regression models using census tract counts of individuals imprisoned in NYS to assess neighborhood-level correlates. Finally, we generate predicted tract-level imprisonment rates for 2020 and apply spatial cluster methods to identify contemporary high imprisonment clusters across the state. Below, we describe the data used for the current study and our analytic approach. Additionally, we provide descriptive statistics on NYS census tracts.
Spatial Scale, Historical Change, and the Seven Neighborhoods 40 Years Later
When discussing spatial scale, it is important to clarify what we mean by neighborhood. The Think Tank considered the seven geographic areas named in the study to be neighborhoods, so we adopt this perspective. Considerable theoretical and empirical work has examined the neighborhood as both an official unit—as defined by administrative boundaries—and as a socially constructed spatial area shaped by residents’ shared understanding of place (Gieryn 2000; Hunter 1974; Logan 2012). The original seven neighborhoods named by the Think Tank include: Harlem, the Lower East Side, Brownsville, Bedford-Stuyvesant, East New York, South Jamaica, and the South and Central Bronx. We identify census tracts using these neighborhood names in contemporary maps of NYC, acknowledging that neighborhood boundaries are both mutable and contested.
To analyze historical change since 1979, we use county-level data because neighborhood-level data are unavailable before 2010. Counties provide a relatively stable geographic area over time, and data on prison populations are more likely to have been consistently collected across decades. Counties also capture broader structural and policy environments—such as the organization of legal institutions, including courts and community corrections—that shape punishment vulnerability at smaller spatial scales. While county-level data mask within-area heterogeneity, counties offer a valuable opportunity for examining historical change where neighborhood data are unavailable.
We use census tracts as our unit of analysis for more localized imprisonment. The use of census tracts allowed us to identify specific spatial areas within the original seven neighborhoods, while also enabling comparisons with areas outside of NYC. We are particularly interested in examining the broad spatial and contextual patterns of mass incarceration across the rural-urban spectrum. Since not all areas in the state fit neatly into large neighborhood zones comparable to the original seven neighborhoods, using census tracts ensures consistency of measurement statewide.
In addition, regional urbanicity has been a relevant factor in explaining high rates of incarceration, particularly in recent years (Simes 2021; Subramanian et al. 2015). Census tracts are subdivisions of counties and thus nest entirely within counties. Tracts are not always contained within cities, towns, or unincorporated areas and may cut across those spatial boundaries. For this reason, to identify a tract’s level of urbanicity in a regional context, we code census tracts by the urbanicity of the county in which they reside. There are 62 counties in NYS, and in Appendix Table A1, we describe these county designations and examples of which New York cities are contained within each metropolitan or micropolitan/noncore county.
In sum, our main empirical analysis draws on two complementary geographic scales: counties and census tracts. Counties have data on imprisonment covering the years of the original Seven Neighborhoods Study, and they provide a level of analysis that preserves meaningful substate variation while offering the historical depth that the tract-level data lack. Census tracts capture relatively granular variation in imprisonment rates that approximate neighborhood-level processes, including spatial spillover effects. Moreover, census tracts can be fully nested and geocoded within both the named NYC neighborhoods and NYS counties, which allow us to locate the original seven neighborhoods in contemporary data and assess regional variation across the state, respectively. Thus, using both tracts and counties as our main units of analysis allows us to examine the historical trajectories of specific neighborhoods while also identifying broader regional trends over time. This dual-scale design is necessary because key social changes unfold simultaneously at neighborhood and regional levels, and contemporary tract-level data alone cannot capture long-term historical patterns.
Data
Geocoding census tracts to the seven neighborhoods and counties
A first step for the analysis is to identify which census tracts reside within one of the original seven neighborhoods. To do so, we use NYC’s Neighborhood Tabulation Area (NTA) shapefiles generated by the NYC Department of City Planning and shared on the NYC OpenData Portal. The NTA boundaries and their associated names are approximations of recognized NYC neighborhoods. The NTAs in 2010 and 2020 are aggregations of 2010 and 2020 census tracts, respectively. Although neighborhood boundaries and definitions are contested (Coulton et al. 2001), especially in NYC (Buchanan 2023), this approach offers the best available contemporary analogue to the original seven neighborhoods. In three cases (Brownsville, South Jamaica, and the Lower East Side), there are single NTA names in both 2010 and 2020 that clearly correspond to the original neighborhoods. Harlem, Bedford-Stuyvesant, East New York, and the South and Central Bronx are comprised of multiple NTAs that change names from 2010 to 2020. We provide in Appendix Table A2 the set of NTAs selected to correspond to the original seven neighborhoods. Census tracts in each year were coded as being within the original seven neighborhoods and within their given county based on their spatial location for each respective census year. Appendix Figure A shows our approach using South Jamaica as an example.
Prison population data
Data on the prison population come from several sources. First, to study the historical trajectory of punishment’s spatial concentration since the 1979 original report, we examine geographic trends in the county of commitment of the prison population over time using archival data collected from the NYS Archives in Albany, New York (1975–1984) and the Vera Institute of Justice Incarceration Trends Dataset (1988–2020). Crucially, the historical data drawn from the archives covers years when the original Seven Neighborhood Study took place (and prior to the Vera data, which begin in 1988, almost a decade after the original study). We combine these archival and public data sources to allow for an examination of the historical spatial trends in New York. For consistency with the Vera data, we combine prison population data for all five NYC counties into one geographic area. We contacted the original researchers—as well as historians of the Seven Neighborhoods Study—and could not obtain the original records of the study. However, we analyzed additional historical records collected in the NYS Archives, finding 75 to 80 percent of people imprisoned in New York prisons, and Green Haven prison specifically, where the original study took place, came from NYC in the years surrounding the original study in 1979, supporting the original study’s findings.
Census tract prison population counts are from the Prison Policy Initiative (PPI). These data have been made available at each decennial census since NYS ended prison gerrymandering in 2010. Prison gerrymandering is the legislative redistricting practice of counting incarcerated people as residents of the prison’s location on Census Day rather than their home communities, which has been shown to distort political representation (Remster and Kramer 2018, 2023; Williamson and King 2022). The law ending prison-based gerrymandering in New York (Chapter 57 of the Laws of 2010) required the New York Department of Corrections and Community Supervision (NYDOCCS) to share with redistricting officials the home addresses of people in state prisons on Census Day (April 1), so that incarcerated people could be credited to their home communities rather than the prison facility (Wood 2014). The PPI-generated counts of imprisoned people by taking NYS’s redistricting data and subtracting those data from the original Census Bureau redistricting data to produce a file that represented the number of incarcerated people from each census block and then aggregated from blocks to census tracts (Prison Policy Initiative and VOCAL-NY 2020). The PPI reports imprisonment rates with a population denominator that combines the census population with the number of people imprisoned for a given geography.
These data represent an undercount of all incarcerated people for two reasons: first, these data are restricted to people in the state prison system and thus exclude anyone in local jails or federal prisons. Nationally, just over 50 percent of all incarcerated people are held in state prisons (Sawyer and Wagner 2020). Second, in the 2010 count, 11,807 people could not be reallocated to their home census blocks; about 2,400 were from outside of NY state, 1,276 people did not have an address, and 8,098 addresses were either incomplete or could not be geocoded, reflecting common data quality issues like missing street numbers, nonstandard formatting, or administrative recordkeeping errors (Prison Policy Initiative and VOCAL-NY 2020). Considerable improvements were made to reallocate incarcerated people to their home neighborhoods in 2020, resulting in only 3,465 nonallocations due to errors, missing/unknown addresses, or homelessness (Widra and Encalada-Malinowski 2022). Because the address data for 2020 were collected on April 1, we do not believe the Covid-19 pandemic had a significant impact on the counts of incarcerated people (National Academies of Sciences, Engineering, and Medicine et al. 2020). Finally, note that census tract data estimate prison population counts based on home address, while our county-level data is based on the county that sentenced the individual to prison.
Segregation and neighborhood disadvantage
Prior research on punishment vulnerability, including the original Seven Neighborhoods Study, points to factors related to racial segregation, housing, and concentrated disadvantage in explaining the spatial pattern of imprisonment (Prisoner’s Alliance with Community 1997; Sampson 2012; Simes 2018). Composite measures that address several intersecting community-based conditions are strongly motivated by theories of punishment vulnerability and the Think Tank’s theory of multiple “crime generative factors.” Based on a combination of model fit statistics and tests for multicollinearity, we analyze the following characteristics in relation to tract-level imprisonment: the Social Vulnerability Index (SVI), racialized socioeconomic segregation, residential instability, and housing vacancy. The SVI is comprised of 16 census variables relating to socioeconomic status, household characteristics, racial and ethnic composition, and housing/transportation (Flanagan et al. 2011). The SVI is a standardized percentile ranking, with values ranging from 0 to 1; higher values indicate greater vulnerability. The SVI is provided by the Centers for Disease Control and Prevention and the Agency for Toxic Substances and Disease Registry (CDC/ATSDR).
Data on segregation and housing conditions are derived from the 2006 to 2010 and 2016 to 2020 American Community Survey (ACS) 5-Year Estimates. The Index of Concentration at the Extremes (ICE) measures the spatial concentration of privilege and disadvantage within a census tract by comparing the proportion of individuals who are both racially privileged (e.g., non-Hispanic White) and economically affluent to those who are racially marginalized (e.g., non-Hispanic Black or Latinx) and economically disadvantaged (Krieger et al. 2016). A value of 1 would indicate that all residents are in the privileged group, and a value of −1 indicates that all residents are in the most deprived group. For the models described below, we reverse coded ICE and generated quintiles, where the first quintile represented the most privileged residents and the fifth quintile represented the most deprived residents. Our results compare two to five quintiles with the first quintile (most advantaged). We also examined housing and residential conditions, including residential instability (the proportion of households that moved in the prior year), population density, and the proportion of vacant housing units. Note that racialized economic segregation, residential instability, and housing vacancy are not among the 16 census measures comprising SVI.
We identify the level of census tract urbanicity using the tract’s corresponding county’s urbanicity as defined by the National Center for Health Statistics (NCHS) Urban-Rural Classification. We recoded the NCHS’s six categories into four. Thus, census tracts are coded as being within a county that is large urban (1), suburban (2), small or mid-sized metro (3–4), or rural (5–6).
Crime, arrests, and violence data
The original Seven Neighborhoods Study identified a set of conditions the Think Tank argued were “crime generative,” including an “underground economy,” substance use, and the police response (Nuleadership 2001). To capture these in the contemporary context, we draw on data from two sources. Tract-level crime and arrest statistics are collected by local agencies and are difficult to obtain for all census tracts within a state; NYS has 514 law enforcement agencies that report agency-level data to state and federal reporting programs. Instead, we gathered data from the NYS Division of Criminal Justice Services (DCJS). The DCJS collects data on county-level index crimes and felony drug arrests. Drug arrest rates have been used in prior studies to control for police enforcement in census tracts (Fagan et al. 2004; Simes 2018). To control for tract-level crime, we use an additional measure for 2020 models only (data are not available for 2010), using counts of fatal shootings provided at the XY coordinate by AmericanViolence.org (Sharkey 2024). Counts were aggregated to census tracts to generate a log rate of fatal shootings per 10,000 residents. We also examined rates of nonfatal shootings, and the results were substantively unchanged.
Table 1 provides descriptive summaries of all data used in the analysis of census tract-level imprisonment rates in 2010 and 2020. Each cell provides an average for the given year for all census tracts in NYS, and the standard deviation is reported in parentheses. The average census tract imprisonment rate was 259 per 100,000 residents in 2010 and slightly lower in 2020 (210 per 100,000 residents). Over half of NYS census tracts are within a large urban metropolitan county—NYC, Buffalo, and Rochester, with NYC making up 83 percent of NY census tracts within a large urban county and 44 percent of NYS census tracts overall. Another quarter of tracts are within suburban counties that are contiguous with these large urban counties. The remaining one-fifth of census tracts reside in small or mid-sized urban counties (e.g., counties where Syracuse, Binghamton, and Albany are the county seats) or rural counties. The seven neighborhoods—all within NYC—account for about 6 percent of all NYS census tracts, roughly equivalent to their population share in 2010 and 2020. We note only one significant change from 2010 to 2020—the percentage of people who moved in the prior year. In 2010, approximately one-quarter of occupied housing units reported moving into their housing in the year prior to the ACS survey. By 2020, this number is only 3.6 percent (Table 1). We also checked data from the prior ACS five-year wave (2015–2019); for NYS, 6 percent of households reported moving in the year prior to the survey. This trend is consistent with recent research suggesting that immobility, rather than mobility, has become a significant driver of neighborhood inequality in more recent years (Schmidt 2024).
Descriptive Statistics of New York State Census Tracts, 2010 and 2020.
Note. Standard deviation in parentheses. Data on tract-level fatal shootings is only available for the 2020 analysis.
Analytic Strategy
Estimating county-level spatial clustering in NYS, 1975 to 2020
We assess the spatial concentration of county-level imprisonment rates by calculating a Global Moran’s I from the year 1975 to 2020 using queen contiguity weights (P. A. P. Moran 1948). To incorporate uncertainty, we conducted 1,000 Monte Carlo simulations for each year, generating null distributions under spatial randomness and computing 95 percent confidence intervals centered on the observed Moran’s I values. We used the spdep package in R to calculate county-level Global Moran’s I.
Estimating the tract-level spatial context of imprisonment in the twenty-first-century NYS
Because census tract boundaries changed significantly from 2010 to 2020, we analyze imprisonment rates for each year separately. To stabilize rates, we remove tracts with fewer than 1,000 total residents in each year of the analysis. In 2010, this removed 192 census tracts from the analysis (3.9 percent of tracts), and in 2020, this removed 250 census tracts from the analysis (4.6 percent of tracts). We conducted the same analysis, removing tracts with fewer than 500 total population, and the results were substantively unchanged.
Adopting the Think Tank’s Non-Traditional Approach to studying punishment vulnerability, we examine “crime generative factors,” estimating census tract-level counts of imprisoned people as a function of these tract- and county-level conditions. For census tract i, we fit the following negative binomial regression to the count of imprisoned people, Yi,
where the regression contains an offset term for the tract total population, P, and thus the coefficients can be interpreted as the association of the predictors with the log imprisonment rate. Predictors include a spatial lag of the dependent variable,
Spatial cluster analysis of regression predictions
As a final analysis, we extract the predicted values from our 2020 census tract-level negative binomial regression model of imprisonment rates that include a control for tract-level fatal firearm violence (but exclude spatial controls) to analyze the spatial concentration of imprisonment using Local Indicators of Spatial Association (LISA) methods (Anselin 1988) using the spdep package in R. LISA methods, which assess the degree of local spatial autocorrelation, allow us to identify clusters of high or low imprisonment rates and spatial outliers where a tract’s imprisonment rate differs significantly from its neighbors.
In sum, we aim to present a historical account of the spatial distribution of incarceration inspired by the Think Tank’s Seven Neighborhoods Study. We first examine historical trends in global spatial autocorrelation of imprisonment at the county level from 1975 to 2020. Next, by predicting imprisonment rates for each census tract based on demographic, socioeconomic, violence, and housing factors, we generate a detailed spatial representation of the predicted distribution of imprisonment across tracts. By integrating the spatial cluster analysis with our model predictions, we can identify and visualize the spatial dynamics of imprisonment across the state, revealing localized concentrations of incarceration that are shaped by neighborhood characteristics and broader regional conditions in the twenty-first century.
Results
We divide our results into three parts. First, we analyze spatial trends inclusive of the entire period of study (1975–2020) using county-level data. Second, we report results from negative binomial regression models of census tract counts of people imprisoned in NYS. Third, we present a spatial cluster analysis of predicted values from a model estimating tract-level imprisonment in 2020, identifying high imprisonment clusters in contemporary NYS.
County-level Spatial Clustering in NYS, 1975 to 2020
First, we examine statewide historical trends in imprisonment rates, using counties as the spatial unit of analysis. We anchor our analysis in the period inclusive of the original study in 1979 and the reanalysis performed by the Think Tank in 1990, which was later published in The New York Times. Figure 1 displays four maps of NYS counties for 1979, 1990, 2000, 2010, and 2020. The maps show, all on a common scale, the tremendous growth in incarceration during the 45-year period. The 1979 and especially the 1990 and 2000 maps show NYC counties have the highest imprisonment rates in the state. The 2010 and 2020 maps show that the highest rates of imprisonment are outside of NYC.
County-level imprisonment rates in New York state, 1979 to 2020.
Figure 2 plots the Global Moran’s I of county-level imprisonment rates for 1975 to 2020, with some years excluded due to data availability. The years of the Seven Neighborhood’s Study (1979 and 1990) are highlighted in blue. Estimates include 95 percent confidence intervals based on 1,000 Monte Carlo permutations. Higher values (range: −1 to 1) indicate greater spatial clustering of imprisonment rates. Gray shaded areas indicate statistically significant spatial clustering. The resulting plot shows a clear decline in spatial autocorrelation over time, with high and statistically significant clustering from 1975 to 2005, followed by weaker and statistically insignificant clustering afterwards. The Global Moran’s I point estimate increases again in 2018, but based on the maps in Figure 1, this is due to spatial clustering outside of NYC.
Global Moran’s I of county-level imprisonment rates in New York state, 1975 to 2020.
These results reveal that imprisonment rates in New York have become increasingly spatially dispersed away from NYC. Although we confirm the concentrated clustering identified by the Think Tank at earlier time points (1979 and 1990), there has been a profound change since. Indeed, the data provide empirical evidence of the spatial deconcentration of mass incarceration over recent decades, with imprisonment spreading beyond NYC to other parts of the state. Thus, over the lifespan of mass incarceration, there has been a significant macrolevel spatial shift in the location of high rates of imprisonment, from concentrating in NYC in the earlier periods of mass incarceration to the more recent period characterized by a nascent broadening to small and mid-sized counties (see Appendix Table A1 for description of small/mid-sized metros).
Modeling Imprisonment within the Seven Neighborhoods in the Twenty-First Century
Figure 3 displays three maps. The left map indicates in blue the location of the seven neighborhoods (using the 2020 NTA boundaries). The middle map displays rates of imprisonment in NYC’s NTAs in 2010, and the right map displays rates of imprisonment in 2020 NTAs. The map suggests that over 40 years after the original Seven Neighborhoods Study, imprisonment remains spatially concentrated within and surrounding the original seven neighborhoods, with particularly noticeable clusters in East Harlem and Central Brooklyn. This pattern is particularly pronounced in 2010, but similar in 2020 as well.
Imprisonment rates in New York City and the seven neighborhoods, 2010 and 2020.
In 1979, the Think Tank found that 75 percent of people incarcerated in NYS came from seven neighborhoods in NYC that, in 1979, only represented 12 percent of the state population at the time of the study (Prisoner’s Alliance with Community 1997), sending people to prison at a level over six times their population share. Our initial descriptive analysis finds a continuing overrepresentation of residents of the original seven neighborhoods in both 2010 and 2020. In 2010, the seven neighborhoods accounted for about 6.3 percent of the state’s population (Table 1), but over 18.4 percent of the state’s prison population—an imprisoned population nearly three times its population share. In 2020, we find tracts within the seven neighborhoods comprise 5.7 percent of the NYS population (Table 1), but 14.9 percent of the prison population.
Despite this overrepresentation, we observe a substantial shift in the spatial pattern of imprisonment in New York in these more recent data. The seven neighborhoods comprise about 300 census tracts in NYS (approximately 6 percent of all census tracts). By 2010, 75 percent of the prison population came from 1,330 tracts (27 percent of 2010 census tracts). In 2020, 75 percent of the prison population came from 1,551 census tracts (29 percent of 2020 census tracts). The spatial concentration of imprisonment has diffused to roughly five times the number of tracts since the original study, which is supported by our county-level findings (Figure 2).
We model rates of census tract imprisonment in relation to social vulnerability, segregation, housing, drug criminalization and violence, and spatial conditions. Table 2 reports results from negative binomial regression analyses of imprisonment in 2010 and 2020. For each wave of data, we first estimate the log rate of imprisonment in relation to tract-level SVI percentile, racialized economic segregation quintiles (ICE), criminalized violence, felony drug arrests, and housing disadvantage. Models 4 and 5 introduce a tract-level measure of fatal firearm violence, restricted to 2020 because of data availability. We then examine how spatial conditions influence estimates of imprisonment, including whether a tract resides in one of the original seven neighborhoods (1/0), the tract’s county-level urbanicity (with large urban counties as the reference group), and the average log rate of imprisonment in surrounding tracts (i.e., a spatially lagged dependent variable in Models 2 and 5).
Negative Binomial Regression Analysis of Imprisoned Population within New York State Census Tracts, 2010 and 2020.
Note. Unstandardized coefficients. Robust standard errors in parentheses. Results from ICE Quartiles 2 and 3 suppressed.
p < .05. **p < .01. ***p < .001.
Social vulnerability and racialized economic segregation are strongly associated with the rate of imprisonment in a census tract. For example, tracts residing in areas with the greatest levels of racialized economic segregation (quintile 5) have over five times the level of imprisonment as places with the lowest levels of segregation (Model 2: exp(1.765) = 5.842). The SVI, encompassing a wide range of social and economic hardships, is significantly associated with higher imprisonment across all model specifications. Residential instability and higher levels of vacant housing are also significantly associated with the imprisonment rate in 2010 and 2020 census tracts in NYS. For example, a 20 percent increase in the census tract share of recent movers is associated with a 30.8 percent higher rate of imprisonment (Model 5: exp(1.341*.2) = 1.308). Population density (logged) is not significant across the models.
The estimated relationship between county-level crime and arrest rates and the tract-level imprisonment rate varies by model specification, time period, and unit of analysis. For example, in 2010, county-level drug arrest rates were negatively associated with imprisonment rates in tracts, but in 2020, the relationship was positive and significant. This could reflect changing levels of drug enforcement at the county level across the two time periods. These aggregate measures, however, attenuate the estimated associations between crime and criminalization and imprisonment rates. Tract-level fatal violence, as measured by the rate of fatal shootings in the census tract (Models 4 and 5), is significantly associated with higher levels of imprisonment: a 20 percent change in the log fatal shooting rate in census tracts is associated with a 5 percent increase in the tract-level imprisonment rate in 2020.
Models 2 and 5 introduce a set of spatial conditions that test theories related to urbanicity and proximity to neighborhoods identified by the original Seven Neighborhoods Study. First, there is evidence of significant spatial spillover of imprisonment in a statewide analysis—that is, a given tract’s level of imprisonment is significantly influenced by rates in neighboring tracts. In Model 5, a 20 percent increase in the log average imprisonment rate of contiguous tracts is associated with a 4 percent increase in a given tract’s expected imprisonment rate. In a statewide analysis and holding constant county and tract-level violence, drug arrests, housing disadvantage, segregation, social vulnerability, and spatial clustering of imprisonment, we find that in 2010, the seven neighborhoods had a 34 percent higher imprisonment rate compared to all other tracts in NYS (exp(.291) = 1.337). Holding constant the same neighborhood conditions in 2020, tracts within the original seven neighborhoods had an approximately 10 percent higher imprisonment rate compared to all other census tracts in NYS (exp(.0999) = 1.105). Despite the changing geography of imprisonment, we find evidence that over four decades later, the seven neighborhoods remain significantly impacted by imprisonment compared to tracts with similar conditions. We also find evidence that the expected imprisonment rate within the original seven neighborhoods, net of controls, has declined from 2010 to 2020.
A second set of spatial conditions explores the census tract’s level of urbanicity, measured by the county-level classification of large urban, suburban, small/mid, and rural counties. With large urban counties as the reference category and holding constant tract-level disadvantage, racial segregation, and housing inequality, we find that census tracts in small to mid-sized counties had imprisonment rates that were between 50 percent and over 200 percent greater than tracts in large counties in 2010 and 2020, respectively. In 2020, census tracts within rural counties had about two times the rate of imprisonment compared to tracts within large urban counties (exp(.788) = 2.198). A census tract in rural Montgomery County has an observed imprisonment rate of over 1,400 per 100,000 residents, compared to the average rate of 209 per 100,000 residents for all NYS census tracts. Four of the census tracts among the top 10 highest imprisonment rates are in Onondaga County, where Syracuse is the county seat. Thus, in a large state with over 4,500 census tracts, in both 2010 and 2020, we find evidence of significantly higher rates of imprisonment outside of NYC, net of controls.
As a sensitivity analysis, we examine whether key predictors of 2020 tract-level imprisonment rates show similar associations if we model for unobserved county-level heterogeneity through a multilevel model with random intercepts for county. In Appendix Table A3, we report these results and show that the main conditions of social vulnerability, racialized economic segregation, tract-level violence, and the tracts within the original seven neighborhoods remain strongly associated with higher imprisonment in 2020.
Identifying Tract-level Imprisonment Spatial Clusters in 2020
In the following set of analyses, we map the predicted values from Model 4 (which contain a tract-level control for firearm violence but does not introduce spatial controls) to identify spatial clusters within NYS.
Figure 4 displays six maps of the counties (and cities) with the greatest number of high imprisonment-rate clusters, using LISA (Anselin 1995). In each map, colors represent different types of spatial relationships between a tract and its neighboring areas. Red indicates “high-high” clusters, where tracts with high predicted imprisonment rates are surrounded by other tracts with similarly high rates, indicating significant spatial clustering of incarceration. Blue represents “low-low” clusters, where tracts with low predicted imprisonment rates are adjacent to other low imprisonment rate tracts. Light red (or pink) identifies “high-low” outliers, where a high imprisonment tract is surrounded by low imprisonment tracts, highlighting potential anomalies or localized spikes. Light blue identifies “low-high” outliers, where a low imprisonment tract is surrounded by high imprisonment areas, suggesting resistance to the neighboring trend. Light gray marks areas with no statistically significant spatial autocorrelation, indicating that the imprisonment rate in these tracts does not follow a clear spatial pattern relative to their neighbors.
Local Indicators of Spatial Association (LISA) of predicted imprisonment rates in New York state census tracts in 2020.
First, in the top left map of the five counties comprising NYC, we identify that the concentration of high rates of imprisonment focuses mainly in Central Brooklyn, parts of Harlem, the Bronx, and South Jamaica in Queens. By 2020, the Lower East Side (one of the original seven neighborhoods) will no longer be part of a significant cluster of high imprisonment rates. We note that within NYC, there are significant low-low clusters throughout Manhattan and some areas of Brooklyn and Queens, demonstrating high levels of within-city inequality in neighborhood imprisonment rates.
The other five plots show the counties in which the cities of Rochester, Buffalo, Syracuse, Mt. Vernon, and Binghamton reside. Here we see more regional inequality, where tracts within mid to large-sized cities are largely within a high-high cluster, but surrounding tracts in suburbs and exurbs are more uniformly within low-low clusters (with the exception of Broome County in the lower right corner).
The Think Tank found that 75 percent of people imprisoned in NYS came from just seven neighborhoods within NYC. Our results examining spatial clusters in 2020 show a dramatic change since the original study. By 2020, 43 percent of people imprisoned in New York came from 787 census tracts identified as within a “high-high” cluster. In Appendix Table A4, we report 23 cities containing the centroids of at least four census tracts located within a significantly high and contiguous cluster of prison population.
Discussion
Places fundamentally shape life chances and well-being, often carrying intergenerational implications. A significant share of research on durable neighborhood inequality focuses on the racial and economic trajectories of places. However, after a half-century of mass incarceration, places became vulnerable to high incarceration rates, and those conditions of concentrated punishment may have also contributed to place-based stratification. In an analysis of novel historical data on county-level imprisonment and newly available data on imprisonment at the census tract level, we revisit the 1979 Seven Neighborhoods Study, a rich sociological account of the community-level factors that generate place-based punishment vulnerability. Produced by a group of men incarcerated in Green Haven Prison in New York, they found 75 percent of the prison population came from seven neighborhoods in NYC, despite only 12 percent of the state’s population living in these neighborhoods at the time of their study.
Our article reports several empirical findings. First, we identify a dramatic shift in the spatial patterning of imprisonment in NYS over the past four decades. By 2020, 75 percent of the state’s incarcerated population came from five times as many census tracts as in the late 1970s—now concentrated in smaller and mid-sized metros across upstate New York. At the county level, we observe a stark decline in spatial clustering over time: while clustering was high during the period of the original study (Moran’s I = 0.6), owing to the extreme concentration in NYC, county-level spatial autocorrelation was not statistically significant beginning in 2005, rising again by 2018 due to clustering outside of NYC. Regression analyses controlling for social vulnerability, racial segregation, housing inequality, and spatial characteristics show that the census tracts with the highest predicted imprisonment rates are no longer found exclusively in urban cores, but in small cities, suburban areas, and rural communities across the state. Cluster analysis of the predicted values from these regressions confirms that new high-rate clusters have emerged in Rochester, Buffalo, and more than 20 smaller cities outside of NYC.
At the same time, we find evidence of durable punishment vulnerability. Tracts within the original seven neighborhoods still have predicted imprisonment rates that are 10 to 35 percent higher than other tracts in NYS, even after adjusting for the same covariates. Although these cross-sectional models have limited explanatory power, we interpret their remaining significant associations as suggestive evidence of an enduring legacy of neighborhood punishment vulnerability within the original seven neighborhoods (Simes 2021). This legacy may extend beyond social and economic conditions to include persistent stigma and reputational effects, as well as the intergenerational histories of incarceration within families and social networks, shaping long-term exposure and vulnerability to punishment. While some areas of the original seven neighborhoods remain targets of criminal justice policy, the cluster analysis also shows that some parts of the seven neighborhoods—particularly tracts in the Lower East Side—are no longer statistically significant hotspots of imprisonment.
Taken together, these findings contribute to theories of urban inequality, mass incarceration, and social change. Mass incarceration was a dramatic social transformation that led to a historic rise in the prevalence of imprisonment, concentrated in impoverished communities of color (Shannon et al. 2017; Travis, Western, and Redburn 2014). In the twenty-first century, mass incarceration may be changing in ways related to the social, economic, and political conditions of local places (Muller and Roehrkasse 2021; National Academies of Sciences, Engineering, and Medicine et al. 2023; Simes 2021; Ternullo 2024). As crime and incarceration increasingly shift across geographies—and remain high in many small cities and rural areas—there is a pressing need for new criminological and sociological theories of place, crime, and punishment, as many theories address urban social processes, particularly in large cities (Clear 2007; Clear, Rose, and Ryder 2001; Sampson and Loeffler 2010). Such theories must grapple with the social, political, and economic forces driving these changes, including political polarization, regional economic displacement, and the diversification of communities outside major urban centers. More broadly, mixed-methods research that examines the lived conditions of small cities will be essential for deepening our understanding of how place shapes crime and how policy responds to it. Furthermore, we situate these findings in the research literature on neighborhood change (Hwang 2016; Owens 2012; Sampson 2012), whereby gentrification and economic development have reshaped urban neighborhoods that previously were impacted by mass incarceration (Golash-Boza 2023).
Our study has theoretical implications related to research documenting the durability of neighborhood inequality in research on poverty, violence, and racial segregation (Sampson 2012; Sharkey 2013; Sharkey and Faber 2014). Our analysis provides evidence of durable inequality in neighborhood imprisonment rates even in the context of the spatial reorganization of the prison population’s home communities. Rather than viewing incarceration as a single outcome of a criminal justice process, a theory of the legacy of punishment vulnerability draws attention to the historical and intergenerational patterns of inequality that may reinforce and solidify neighborhood reputations and public divestment, adding a new dimension to sociological understandings of why certain neighborhoods experience enduring inequality. In 2020, census tracts within the South and Central Bronx and neighborhoods in Central Brooklyn and Jamaica, Queens—all 60 percent or more Black or Latinx—still exist in significantly high clusters of incarceration 40 years after the original study. Our results also suggest that new sites of high incarceration rates may experience similar historical legacies without more scholarly and policy attention to those areas.
Our research demonstrates the significance of the Non-Traditional Approach to Criminal and Social Justice put forth by the Think Tank, which theorized the ecological linkage between neighborhood and prison long before it was widely discussed in sociology and criminology, while simultaneously not restricting their analysis to NYC. Community-level conditions of social vulnerability, housing inequality, and racial segregation are key to understanding incarceration and neighborhoods, even as the spatial concentration of imprisonment rates has dispersed to a broad set of communities outside of NYC. This suggests that it is not urbanicity per se that drives incarceration, but rather it is a combination of local social, economic, and political conditions (neighborhood, city, and county) that drive punishment vulnerability (Simes 2021). To the extent we continue to emphasize large urban contexts within this body of research, we may obscure the role of these local conditions in predicting imprisonment. More engagement with scholarship focused on small cities, suburbs, and rural areas may point to new mechanisms of place that drive high imprisonment rates in local areas.
What theories could account for this spatial shift in high imprisonment rates? First, in the example of New York, by the mid-2000s, there were significant declines in the state prison population overall, which were found to be driven largely by reforms aimed at reducing felony convictions within NYC (Austin and Jacobson 2013). In a fractured criminal justice system largely run by local and county officials, policy innovation and implementation happen in localized ways (Pfaff 2017). Another theory suggests that local nonelected leaders play a significant role in policy change (Levine 2016). If local organizations focused on police reform and decarceration advocacy aimed their efforts at social change within large cities, advocacy-driven reform may not have diffused to smaller cities and towns upstate. For example, in the case of environmental advocacy, strong networks and coalitions in small cities and rural areas were key to amplifying advocacy efforts via resource sharing; where these organizational linkages were sparse, communities were more isolated, underserved, and at risk for environmental impacts (Irwin and Pischke 2016). Moreover, one study of small city and rural social service organizations finds that localized gaps in a broad range of services drove organizations to engage in direct collaboration with police, parole, and jails (Simes and Tichenor 2022). Finally, the intense gentrification of urban neighborhoods produced a spatial shift in the demography of large cities and thus the location of policing and punishment. In a study of Washington, D.C., Golash-Boza (2023) argues that mass incarceration transformed urban cities, removing large numbers of Black residents through extreme levels of policing and punishment; thus, the gentrification of urban neighborhoods represents a continuation of long-standing practices of displacement and exclusion characteristic of both segregation and mass imprisonment (Golash-Boza 2023). Under these conditions, large cities are no longer the locus of criminalization and punishment as wealthy and white populations have moved in to claim urban spaces. Examining the dynamics of punishment vulnerability emerging from locally driven policy areas, including neighborhood redevelopment, could further test this hypothesis.
Future research in sociology, public health, and related fields can continue the tradition of the Think Tank to examine the effects of legacies of punishment vulnerability and the consequences of the geographic shifts in incarceration for communities. For example, future research could develop a richer account of the mechanisms of place that link community conditions to imprisonment, and furthermore, link high levels of imprisonment with health and well-being. Future studies could examine the outcomes of residents within and surrounding the original seven neighborhoods to understand how legacies of punishment vulnerability may be harmful to human health, political participation, and other outcomes. Sociologists should attend to the trajectories of punishment vulnerability over time across a variety of community-level conditions, but especially after 50 years of mass incarceration in the United States. Both qualitative and quantitative research should simultaneously consider histories of policing and incarceration experienced by the same places over time, as well as emergent high rates of policing and incarceration in small cities and towns. Finally, this research has implications for conducting community-based research. The far-reaching theoretical, empirical, and policy implications of the original Seven Neighborhoods Study underscore the significant contributions of people directly affected by the criminal justice system for developing research and policy reforms to end mass incarceration. We thus advocate for a research practice that acknowledges that people who currently are or have been confined in prisons have distinct expertise and a critical perspective on how criminalization and punishment operate in U.S. society, the potential effects of exposure to imprisonment, and interventions that could address and mitigate those impacts (Clair 2021; Farrell et al. 2021). The Think Tank’s Non-Traditional Approach represents a novel theoretical tradition grounded in a radical form of participatory action research that could be replicated in future research designs aiming to study incarceration and its effects.
Decarceration’s proponents advocate for a reckoning with high rates of incarceration as an effective policy for ensuring community safety, if, after decades of incarceration, these same places continue to be beset by violence and then are highly surveilled and punished. As one advocate states, “communities have been historically deprived resources and then criminalized in their struggle to survive” (Mack 2021). Moreover, we note evidence of enduring relative racial disparity in imprisonment. The original Seven Neighborhoods Study found that “approximately 85% [of the NY State prison population] is composed of Black and Latino prisoners (Black 50 percent, Latino 35 %)” despite being only 28% of the state’s population (Prisoner’s Alliance with Community 1997). As of January 2020, while the NYS prison population has decreased to levels observed in the 1980s, relative racial disparities persist: Black NYS residents still comprise about half of the prison population but only about 14 percent of the state population (NYDOCCS 2020).
There are important limitations of the current analysis. Our study is limited to state prison populations and thus excludes people serving time in county jails or federal prisons. Our findings are limited to New York, and there may be important state and regional differences in the spatial concentration of incarceration. The PPI data are among the first publicly available resources on tract-level imprisonment, but the data have limitations worth noting. Single years of tract-level imprisonment data (2010 and 2020) can provide unstable estimates at the census tract level. PPI also documents undercounting of imprisonment; 20.6 percent of the incarcerated population was not counted in these geographic data in 2010 (Prison Policy Initiative and VOCAL-NY 2020). However, significant improvements were made to the data collection and reporting in 2020, with only 8.2 percent of incarcerated individuals not recorded in the 2020 geographic dataset (Widra and Encalada-Malinowski 2022). Given that these estimates are derived from overall counts of imprisoned populations, we cannot estimate race-specific rates using these data.
Even with these novel census tract-level data, separate cross-sectional models for 2010 and 2020 cannot capture the temporal dynamics of census tracts in the way that a historical census tract-level data series would. The lack of recordkeeping of the prior addresses of people entering prison greatly limits our ability to assess the effects of key policy interventions, demographic shifts, or neighborhood redevelopment over time. Moreover, tract-level measures of drug arrests are not available for 2010 and 2020 statewide, and violence or crime data are not available at the tract level for 2010. Finally, imprisonment may indeed intensify neighborhood disadvantage, violence, and population instability, creating a feedback loop. To address these challenges, we advocate for longitudinal geographic data collection and reporting on crime, criminalization, and punishment, so that large samples of places may be studied in analyses where policy shocks and historical change can be assessed.
Despite these data limitations, our analysis provides evidence of both durability and change in the original seven neighborhoods. The Think Tank identified neighborhoods with exceptionally high incarceration rates in the late 1970s and 1990s, and county-level trends reveal the extreme concentration of imprisonment in NYC through the early 1990s. Our contemporary tract-level analysis shows where—and to what extent—these historical patterns have shifted. Together, these findings highlight the evolving yet persistent geography of punishment in New York. In our revisiting of the original Seven Neighborhoods Study, we conclude by noting that the findings in this study could be used to draw attention to the needs and challenges of a diverse set of neighborhoods toward the research and policy goal of understanding the drivers of neighborhood punishment vulnerability and assessing the social changes underfoot in the case of mass incarceration (Phelps and Pager 2016; Simes 2021). Research and policy should address mass incarceration’s continuation in some areas, and emergence in others, through a broad strategy of research and advocacy that includes all geographies.
Footnotes
Appendix
New York State Cities and Towns with High Imprisonment Clusters (N Census Tracts ≥ 4), 2020.
| City/town | Total tracts in high-high clusters |
|---|---|
| New York City | 410 |
| Rochester | 60 |
| Buffalo | 49 |
| Syracuse | 37 |
| Mount Vernon | 17 |
| Binghamton | 14 |
| Utica | 14 |
| Elmira | 13 |
| Albany | 12 |
| Niagara Falls | 12 |
| Cheektowaga | 10 |
| Schenectady | 10 |
| Newburgh | 8 |
| Yonkers | 7 |
| Jamestown | 6 |
| Southport | 6 |
| Irondequoit | 5 |
| Monticello village | 5 |
| Amsterdam | 4 |
| Johnson village | 4 |
| Kiamesha Lake | 4 |
| Poughkeepsie | 4 |
| Rotterdam | 4 |
Acknowledgements
The authors would like to thank Jaquelyn Jahn, Jonathan Mijs, Jane Pryma, Monica Bell, Matthew Desmond, Tessa Desmond, Reuben Miller, Patrick Sharkey, Bruce Western, Cati Connell, Brenden Beck, Japonica Brown-Saracino, and the anonymous reviewers for helpful feedback on earlier drafts.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by NSF CAREER Grant #2237580 from the National Science Foundation (JTS) and a Clinical and Translational Science Award (KL2TR004421) from the Icahn School of Medicine at Mount Sinai; funding from the Criminal Justice Research Training Program at Brown/Lifespan (R25DA037190 Pilot Award), the Distinguished Scholar Award, and a Resource and Education Core (REC) award through the Pepper Center at the Icahn School of Medicine at Mount Sinai (5P30AG028741) (LH).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.