Each year over the last several years I’ve selected a small number of my papers to highlight because they are particular favorites of mine. I list them here, from more recent to older.
Two of my papers, and my book, from 2023 that are favorites of mine:
1) The first paper –with Cynthia Lakon as lead author–is a tale of two pandemics. Our results show that in the first six months of the COVID-19 pandemic, there was considerable evidence of social disparities in health based on the pattern of Covid cases and deaths among residents in U.S. nursing facilities. Nursing facilities with poorer, and more racial minority residents experienced more cases and deaths, and those within more impoverished egohoods with more racial/ethnic heterogeneity experienced more cases and deaths. However, after the first six months, the story completely changed and there was no evidence of such disparities. Why? Although vaccines were important, and we show how they are associated with fewer cases and deaths, they occurred after the change in this pattern. We speculate that it may be due to the adoption and adherence to pandemic-related safety measures by these skilled nursing facilities.
2) The second paper uses novel spatial network data from the American Social Fabric Study to explore how the spatial distribution of residents’ social ties are related to collective efficacy. We use a novel measure of neighborhood safety ties, and show that it is much more strongly related to perceived collective efficacy than is a measure of socializing relationships. We also find that studies focusing only on local ties in the neighborhood are missing part of the picture: more spatially distant ties also matter for fostering collective efficacy.
3) My book–The Spatial Scale of Crime: How Physical and Social Distance Drive the Spatial Location of Crime–describes much of my thinking regarding the spatial distribution of crime. I describe the need for an individual-level model to generate the expected predictions for neighborhood-level theories. I also describe how the spatial movement of people is necessary to account for in models predicting the location of crime.
Five of my papers from 2022 that are favorites of mine:
The first two papers are explicitly longitudinal.
1) The first is with Iris Luo, and we propose that it is not enough to simply ask whether the change in sociodemographic measures are related to changes in levels of crime (that is, a monotonic relationship). Instead, we posit that there can be asymmetric relationships, in which an increase in a measure will increase crime, but a decrease in the measure has little or no effect. And there can be perturbation relationships in which any change in a measure–either increases or decreases–will result in increasing crime. Our study of egohoods over a decade in Los Angeles shows strong evidence for asymmetric and perturbation relationships.
2) The second longitudinal paper is published with Alyssa Chamberlain, and creates a new measure of neighborhood change based on ital flows of residents. Whereas cross-sectional studies measure whether some measure in the neighborhood is related to crime, these are stock measures. Flow measures capture the amount of change, not just net change. We show over two decades that these demographic flow measures capture unique information that provide more insight into how crime changes in census tracts.
3) A third paper takes a very long-range view. Whereas theories posit that racial/ethnic diversity is not only a negative feature of neighborhoods, but that it is also unstable. That is, theories posit that racial heterogeneity is a temporary phenomenon as a neighborhood transitions from one racial group to another. We show in a study of census tracts in Southern California over thirty years that, in fact, there are a nontrivial number of neighborhoods that remain with consistent racial diversity. We term these as persistent racial diversity (PRD) neighborhoods, and show that they in fact do better economically over time compared to a matched set of tracts.
4) Whereas criminologists use geographic profiling to determine the location of a specific offender, we extend this technique as a way to create estimates of possible offenders across the spatial landscape. In this paper with Lyndsay Boggess and Alyssa Chamberlain, we use these estimates to create better forecasts of where crime will occur.
5) In this paper with Loring Thomas and Carter Butts and others published in PNAS, we use spatial networks to help explain the spread of COVID-19 infections. We use data from San Francisco during the first month of the pandemic and demonstrate the counterintuitive result that spatial locations with more cohesion (based on simulated social ties) experienced greater risk of COVID infection. Furthermore, these patterns help explain the racial inequality of infections in the early days of the pandemic.
Four of my papers from 2021 that are favorites of mine:
1) Neighborhood scholars are interested in how residents’ social networks are related to their sense of attachment to the neighborhood. While researchers often only ask about nearby ties (presumably in the neighborhood) in this study led by graduate student Iris Luo we show that long distance ties do not reduce attachment. In fact, they increase neighborhood and city attachment further, although not as strongly as do nearby ties.
2) Several criminological theories posit that features of the built environment impact the location of crime. Testing such theories is often difficult given the lack of data measuring such features. In this paper with the MFI team, we demonstrate an approach that first scrapes street images from google streetview for all streets in a relatively large city, and then uses a machine learning algorithm to classify the built environment characteristics in these features. We show that some of these features–such as the presence of large parking lots, fences, or walls–help explain the locations with high crime concentrations.
3) There has long been interest in characterizing the concentration of employment in metropolitan areas. This interest stems from the idea of a monocentric city in which employment is concentrated in a central downtown location. As metropolitan areas grow, characterizing these employment patterns is quite challenging. One particularly thorny issue for researchers comes from attempting to measure employment subcenters, and then using these to characterize the region. The indeterminacy of these subcenters–as noted by many scholars–leads to difficulty in characterizing different areas, especially when they differ considerably in size. In this paper with the MFI team, we introduce a new technique that sidesteps the challenge of defining subcenters. We use insights from the crime concentration literature to create measures that more cleanly capture measures of 1) employment deconcentration, and 2) spatial dispersion.
4) While criminological scholars have often shown that micro-locations with businesses have more crime, other theories posit that some businesses might help reduce crime given that their owners can engage in activities in the community to create more cohesion. In this paper led by former graduate student Young-An Kim we posited that small, locally-owned businesses might be able to reduce crime. We created a 2×2 set of measures of small/large and local/non-local businesses and show evidence in support of this in Southern California.
Four of my papers from 2020 that are favorites of mine:
1) There is a need for research on why some cities have more crime than others. However, to date this research has typically ignored crime occurring at micro- or meso- locations in these cities. While the lack of crime data in smaller geographic units explains this lack of research, in this paper with graduate student Seth Williams we propose a solution in which we simulate the amount of crime occurring in these smaller geographic units, and then assessing why some cities have more crime than others (while accounting for what is occurring in these smaller units). We show that this procedure works quite well. We also show that failing to use this strategy–as nearly all prior research does–can result in biased results. This therefore is a promising strategy for researchers going forward.
2) Researchers often simply aggregate measures of interest to micro- or meso- geographic units and then assess how this is related to levels of crime. But given that persons move about the spatial landscape, what are these aggregations actually capturing? I explore this question in this paper in which I simulate crime events based on a presumption that offenders will offend based on a distance decay effect. I then show some non-obvious conclusions of what our aggregations tell us depending on whether we aggregate to micro-, meso-, or even macro units (cities). I originally presented this paper at the People, Places and Context Symposium held in 2019 at UC Irvine.
3) One pattern that has puzzled the general public about the COVID-19 pandemic is that whereas some people during the first year of the pandemic knew at least one person who contracted the virus–and sometimes many who did– other people not only did not contract the virus, but did not know anyone who did. Whereas some members of this latter group may have concluded that this implied that the virus was simply a hoax and a fabrication, in this study published in PNAS we showed that the expected spatial distribution of networks can explain this pattern. We simulated networks based on a distance decay effect (based on evidence from prior studies showing this) along with a model based on the infection pattern of the virus to demonstrate that this pattern of clustering of the disease can indeed occur, both spatially and temporally.
4) This paper continues my work focusing on neighborhood change. Here I focus on residential mobility that occurs across housing units, and study transitions in which the social characteristics of the new residents differ from those of the exiting residents. Such changes are needed for many types of change in neighborhoods. I then show how features of the built and social environment–including housing unit type, housing age, and length of residence–impact the likelihood of experience more “social distance” during a housing transition. This social distance can occur along dimensions such as race/ethnicity, income, education, etc.
Four of my papers from 2019 that are favorites of mine:
1) Along with graduate student Seth Williams, we survey the literature of spatial criminology in this paper in the Annual Review of Criminology. We define spatial criminology as a distinct perspective in its attempt to measure and theorize explicitly spatial processes and relationships. We discuss three key themes: 1) the use of increasingly smaller geographic units creates even greater need to account for spatial behavior of persons when studying the location of crime; 2) despite the explosion of spatially precise data in recent years, we argue that theory is falling behind in guiding us in analyzing these new forms of data, and therefore inductive approaches may be a useful complement to consider; 3) an important direction for spatial criminology is considering the extent to which micro- and meso-level processes operate invariantly across different macro contexts.
2) In this paper, we study the spatial and temporal patterns of robbery in Southern California street segments. We propose a novel approach to estimate the temporal crime patterns with a nonlinear semi-parametric technique. This avoids the unrealistic assumptions inherent in strategies that chunk time into a priori time periods. Our strategy allowed us to detect such patterns as that segments surrounded by many employees are associated with a reduced robbery risk during the daytime, but not at night. Furthermore, the risk of a robbery is elevated on a high retail segment on weekends during the daytime, and on high restaurant segments into the early evening on weekends.
3) I propose that neighborhood change can best be understood with a bottom-up theoretical model based on residential mobility by households. This approach implies possible nonlinearity in how neighborhoods change. As a basis for this theory, this paper assessed the situations in which a household turnover occurs and there is social distance between the new and the prior residents based on such demographic characteristics as age, education, household income, ownership status, and race/ethnicity. Using the American Housing Survey over a 22 year period, the study finds that transitions in the oldest housing units and for the longest tenured residents result in the greatest amount of social distance between new and prior residents, implying that these transitions are particularly important for fostering neighborhood socio-demographic change.
4) This paper continues my research interest in how crime brings about neighborhood change. In this paper, we focus on the role crime plays in impacting businesses. We use annual data over a number of years to study how crime impacts business survival, business mobility decisions, and the destination locations of businesses in the subsequent year. Using business data from Reference USA (Infogroup 2015) and crime data from the Southern California Crime Study (SCCS) we assess this question for neighborhoods across cities in the Southern California region. We find that in general, higher violent and property crime are significantly associated with both business failure and mobility, and that higher crime in a destination neighborhood reduces the likelihood that a business locates there.
Four papers of mine from 2018 that are favorites of mine:
1) Although much research has posited that residents in neighborhoods can impact the level of crime through informal social control action, very little research has been able to test the individual-level mechanisms of these theories. This paper is therefore an important contribution as it uses longitudinal data of residents nested in neighborhoods to assess whether perceptions of disorder impact residents’ perceptions of the level of collective efficacy in the neighborhood, and then whether that impacts actual behavior to improve the neighborhood.
2) This paper extends the collective efficacy literature by considering why residents in some neighborhoods disagree about the level of collective efficacy. The study shows that the level of social distance in the neighborhood (measured as egohoods) reduces this level of agreement. Importantly, it is a general measure of social distance (based on several socio-demographic measures) that reduces this agreement, whereas simply measuring difference based on income or race/ethnicity does not reduce agreement.
3) The co-location in space and time of offenders and targets is posited to increase the possibility of crime at a location, and yet measuring the presence of persons at a location is difficult and data intensive. In this paper, we use geolocated and temporally precise twitter information as a proxy for the number of persons at a location during a particular hour of the day, and show that this has a robust positive relationship with the level of different types of crime in a sample of blocks in Southern California.
4) Filtering theory from housing economics posits that as housing ages, it filters down to lower income households. In this paper we build on this basic insight of the aging of housing to posit that such aging housing will tend to become more dilapidated over time, and hence result in more physical disorder at the location, and potentially more crime. Our results show that indeed, street segments with older housing tend to have more crime than streets with newer housing, even controlling for the usual measures of locations, including socio-economic status. We find that this positive effect tends to level off at particularly older age ranges, implying that there may be a gentrification effect in which older housing is then renovated.
Four papers of mine from 2017 that are favorites of mine:
1) This paper co-authored with James Wo and Young-an Kim considered how the relationship between the micro-context and crime can vary across different macro contexts. Building on the insights of an earlier paper with Aaron Roussell, we chose four cities cross-classified based on population in the micro-environment, and population in the broader macro-environment. The results demonstrate that the assumption that micro-crime patterns will necessarily generalize across different macro environments does not seem to hold.
2) This paper co-authored with Nick Branic considered both fast and slow dynamics in neighborhoods, and the consequences for levels of crime. We use annual HMDA data to capture year-to-year changes in neighborhoods, and show that the pace of change in neighborhoods (based on changing racial composition or economic composition) has important consequences for changes in crime rates over the decade. We term this temporal nonlinearity to capture neighborhoods in which the change along a particular dimension occurs primarily in either the early or latter part of the decade. The velocity of this demographic change appears important for understanding changing crime levels.
3) This paper co-authored with Kevin Kane and Jae Hong Kim uses a recently developed technique— kernel regularized least squares (KRLS)—that allows for non-parametric estimation of relationships between various measures of mixing in neighborhoods and the change in average household income from 2000 to 2010 in Southern California. KRLS is a machine learning technique that also detects nonlinear interactions between measures in the model. We develop a new Stata ado package that allows us to plot the key detected nonlinear interactions in the model. We refer to these combinations as a “recipe” for economic growth in a neighborhood over the decade.
4) This paper co-authored with Kevin Kane extends the macro literature on city crime rates in several fashions. First, it adopts an explicitly longitudinal view in studying changes in city characteristics and changes in crime rates over decades. Second, it takes into account long-term change by estimating the models on four separate decades. Third, it takes into account the broader region in which these cities are located and shows that the population growth and economic vibrancy of the larger region has consequences for how crime changes in the cities within those regions. Thus, spatial scale is important even in macro criminology research.
Three papers of mine from 2016 that are favorites of mine:
1) This paper proposes a new general theory of the spatial patterning of crime. It explicitly incorporates offenders into the model. Rather than trying to model all movements of offenders, targets, and guardians, it utilizes the key insight of the principle of least effort in proposing that spatial patterns of persons will typically exhibit a distance decay effect. It uses this insight to build a model that helps understand the spatial distribution of crime.
2) This paper published with Young-an Kim raises cautionary insights about the recently popular “law of crime concentration”. First, it uses a large sample of cities in Southern California to demonstrate that there is empirically more variability in the level of concentration across cities than one would reasonably expect for such a law. Second, it raises the methodological question of whether concentration should be measured across macro units of widely varying sizes. Third, it highlights the statistical challenge of measuring crime concentration in that researchers often pose a baseline assumption of a uniform distribution of crime when in fact a Poisson distribution is more reasonable. Fourth, it highlights the crucial need to consider the temporal assumption employed when measuring high crime locations. And finally, it points out the peculiar theoretical implications if a law of crime concentration indeed exists, which have been given virtually no consideration by scholars.
3) A paper published with Rebecca Wickes (freely available) uses two innovative measures of 1) residents’ assessments of neighborhood ethnic minorities; and 2) the extent of social ties between members of the same ethnic group compared to chance. The measure of residents’ bias towards seeing more minorities in the neighborhood than actually exists was first developed in an earlier paper of ours looking at residents’ perceptions of disorder. This current paper views the relationship of this measure to residents’ perceptions of social capital (measured by social cohesion, place attachment and neighboring). The measure of proportion of ties within and across groups accounts for the ethnic composition of the neighborhood; prior research typically fails to account for this. We use insights from an earlier paper of mine on intergroup violence that adjusted for neighborhood racial/ethnic composition. We find that residents who perceive more minorities in their neighborhood, who have more or fewer ties with members of the other ethnic group than expected by chance, or who live in neighborhoods with more inter-group ties than would be expected report lower levels of social capital.
Three papers of mine from 2015 that are favorites of mine:
1) This paper theoretically considers that collective efficacy develops over time. It develops two key principles. First, the notion of updating: persons re-evaluate their perception of the level of collective efficacy regarding some task based on feedback information. In the neighborhoods literature, this implies that crime and disorder would be expected to impact residents’ perception of collective efficacy. Second, the notion of uncertainty: residents in neighborhoods with few incivilities will typically be uncertain about the level of collective efficacy. An implication is that an incivility event has the potential to greatly change the perceived level of collective efficacy in such neighborhoods, which has rarely been considered.
2) A paper published with Wouter Steenbeek considers how the dynamic processes of neighborhoods may operate differently for various types of crime. It considers three dimensions of crimes: the violent/nonviolent nature; the public/private nature; the relative frequency. The paper considers how the mechanisms of social disorganization theory likely operate differently for various types of crime. And it considers that the feedback effect on residents’ perceptions and behaviors likely will be different over various types of crime.
3) A paper published with Alyssa Chamberlain explores the spatial distribution of crime at a larger scale. Using data for neighborhoods (census tracts) in 79 cities, it asks whether relative inequality—measured as the difference in concentrated disadvantage between a neighborhood and the surrounding neighborhoods—impacts levels of crime. It also explores the effect of relative inequality measured by the ratio of concentrated disadvantage in a neighborhood to the city in which it is located, and finds that this larger context also has important effects.
A fun paper from 2014 that I worked on:
A paper that I published with colleagues down under (Rebecca Wickes and Jonathan Corcoran) explores the intersection between the physical environment and the social environment. We propose the notions of social holes and wedges, and suggest that they can impact the social interactions among residents. A consequence of these social holes in wedges is that they impact the social porosity within and between neighborhoods, which has consequences for the level of cohesion in neighborhoods. We demonstrate these principles with an empirical sample: read about it here [freely available through open access].
Three papers from 2013 that I worked on that are fun:
1) Adam Boessen and I proposed a new measure of neighborhoods: egohoods. This approach is a radical shift away from nearly all prior research that defines “neighborhoods” as having non-overlapping boundaries. We instead think of neighborhoods as being overlapping concentric circles. The paper is here. The code to make your own egohoods is here.
2) Measuring the networks of all residents in a neighborhood is difficult, much less the networks of all residents in a city. From a project I am working on with Carter Butts and other colleagues, we developed a novel approach that simulates the network of ties among all residents in a city. We use the approach to simulate the networks of five cities, and then construct several key network measures to capture the structure of the network. We then find that these simulated network properties actually do a good job predicting the micro-location of crime in cities! Sound fantasmagorical? Read about it here.
3) Understanding the spatial scale of population density and the consequences for crime has bedeviled researchers since the time of Louis Wirth. A challenge is distinguishing between population size and density. Aaron Roussell and I conceive of these as micro-population density and macro-population density. We then propose novel measures of each of these: population density exposure to capture micro-density, and a measure of population within a 20 mile radius to capture macro-density. We point out that implications of routine activities theory are that these should have nonlinear interaction effects on city level crime rates. We demonstrate that this is empirically the case in both 1990 and 2000: read the paper here. The online appendix is here.
The Orange Crush: The Squeezing of Orange County’s Middle Class. This study views trends in Orange County, CA cities and neighborhoods over a forty year period (1960-2000). This report was done for the Center on Inequality and Social Justice, and the full report can be found here.
More recently, the MFI did a similar updated report that covered the entire Southern California region, from 1970-2018. This can be seen here.