Abstract

This study examines direct observations of outdoor routine activities to investigate the pathways through which temperatures shape crime. Daily administrative records of crime, weather, and outdoor activity were assembled from 2015 to 2019 in New York City. Mediation analysis (with bootstrapped standard errors) reveals that alterations in routine activities account for a statistically significant, yet modest, proportion of temperature’s relationship with homicides, shootings, assaults, larceny, and public consumption violations. The comparable mediation effects across violent and nonviolent crimes support routine activity theory as an explanatory framework for understanding temperature's impact on crime. The measures introduced here offer a novel approach for testing the theory and suggest other potential applications.

INTRODUCTION

While a large body of scholarship has conclusively established the link between temperature and crime, the exact nature of this relationship remains elusive (Baumer & Wright, 1996; Corcoran & Zahnow, 2022; Kim & Wo, 2022). Some propose that temperature elevates aggression levels in individuals, leading to increased violent crime (Cohn et al., 2004). Others contend that temperature primarily works by reshaping routine activities, increasing opportunities for criminal interactions (Felson & Boba, 2010). Despite decades of research testing these competing explanations through crime patterns alone, very few studies have approached theory-testing by observing behavioral evidence of the hypothesized mechanisms.

  Can routine activity theory account for the link between temperature and crime? The current study seeks to shed light on these processes by empirically measuring routine activities through a novel application of several administrative data sources. Using a mediation analysis framework, this study quantifies shifts in behavior patterns and assesses their role in the temperature-crime relationship, addressing a critical gap in our understanding of how environmental conditions shape criminal behavior.

By comparing violent and nonviolent crimes, with different motivational structures, we aim to identify common and distinct mechanisms underlying temperature-crime relationships. While violent crimes might be influenced by both psychological responses to heat and changes in routine activities, nonviolent crimes like public consumption violations are more purely dependent on opportunity structures created by outdoor social activity. This comparative approach provides a stronger test of whether routine activities mediate the temperature-crime relationship.
  Drawing on comprehensive data from New York City (2015-2019), the study combines administrative records of public transportation usage, outdoor leisure activities, and other public activities with detailed measures of borough-specific crimes and daily weather patterns. Using a panel mediation analysis with bootstrapped standard errors, we provide the first direct test of whether temperature-induced changes in routine activities drive increases in violent and nonviolent crime. By isolating specific mechanisms, this study advances both theoretical understanding and policy responses to temperature-related crime. The next section reviews in greater detail the gaps in prior research explaining the relationship between temperatures and crime, followed by a discussion of the current study’s data and methods, then a close examination of our results and their theoretical and policy implications.

LITERATURE REVIEW

Prevailing explanations

Most prior research has demonstrated that warmer temperatures are associated with increases in crime, with the most prevalent explanations being the heat/aggression theory and the theory of routine activities (Cohn, 1990; Linning et al., 2017). The current study focuses on routine activities. As the following literature review highlights, there is evidence consistent with routine activities leading to changes in crime, but there are challenges in directly observing or testing this mechanism. This leaves open questions about the extent to which routine activities may or may not play a role in the relationship between temperature and crime. Below, we review the two main theories and prior investigations into each of the mediating steps of the processes examined in this study.

Heat/Aggression Theory

The heat/aggression theory, rooted in individual psychology and physiology, posits that elevated temperatures influence individual aggression levels (Simister & Cooper, 2005). Sometimes called the General Aggression Model (Anderson & Bushman, 2002), this perspective draws from a tradition of research on aggression in social psychology. Researchers have hypothesized that higher temperatures can lead to physiological changes, such as increased arousal and discomfort, which in turn may contribute to heightened levels of aggression (Anderson & Anderson, 1984). That route between heat and aggression involves "hostile affect, hostile cognitions, and excitation transfer processes all increas[ing] the likelihood of biased appraisals of ambiguous social events, biased in a hostile direction" (Anderson et al.,1995: 434). In other words, aggression is driven by negative affect (often in a curvilinear manner), with uncomfortable heat being only one source of negative affect. Although not explicitly examined in this study, this mechanism is an important alternative explanation for the link between temperature and crime.

Routine Activity Theory

The second prevailing explanation is routine activity theory, the focus of the current study. Initiated by Cohen and Felson (1979), it focuses on social and situational processes rather than individual psychological processes, positing that crime is a function of patterns of social interaction. In this theory, crime occurs when a motivated offender, a suitable target, and the absence of a capable guardian converge in time and space (Browning et al., 2021). Temperature fluctuations are hypothesized to influence these convergences by altering people's behavior and movement patterns, thus modulating the opportunity structure for crime (Field, 1992; Rotton & Cohn, 2003). This theory as an explanation of the relationship between temperature and crime has some support in diverse contexts (Schinasi & Hamra, 2017; but see Gamble & Hess, 2012).

While both theories may help explain temperature effects on violent crime, examining property crime alongside violent crime provides a crucial opportunity to disentangle these mechanisms. Violent crime could plausibly be influenced by both psychological responses to heat and changes in routine activities, but property crime has traditionally been viewed as more strongly driven by opportunity structures (Linning et al., 2017). For instance, moderate temperatures may encourage outdoor activities, increasing exposure and access to targets – key elements of routine activity theory. This theoretical distinction between crime types has led researchers to use property crime as a kind of control case for testing routine activity mechanisms, since property crimes should be less sensitive to heat-induced aggression (Baryshnikova et al., 2021).

The inclusion of nonviolent crimes like unlawful consumption of alcohol and marijuana provides another valuable test of routine activity theory. Unlike property crimes, which occur in both public and private settings, or violent crimes with multiple causal mechanisms, these public order offenses are closely tied to outdoor activities that are sensitive to temperature-driven changes in routine activities (Felson & Boba, 2010). This makes them an ideal additional test case for examining whether routine activities mediate the relationship between temperature and crime.

Prior empirical evidence for the distinctions among crime types and their alignment with routine activity theory has been mixed. Studies quantifying temperature effects have found that both violent and property crimes increase with temperature, though often at different rates. Berman et al. (2020) observed that each 10°C rise in daily mean temperature corresponds to an 11.92% relative increase in violent crime but only a 6.14% increase in property crime rates. Similarly, Ranson (2014) found that temperature increases led to larger effects on violent crimes compared to property crimes, though both showed significant positive relationships. These findings suggest that while property crimes may be less sensitive to temperature than violent crimes, they still exhibit substantial temperature effects that warrant explanation.

The parallel examination of violent and property crimes thus offers two key advantages for testing routine activity theory. First, if routine activities are truly driving the temperature-crime relationship, we would expect to see significant mediation effects for both crime types, since both depend on the convergence of motivated offenders and suitable targets. Second, comparing the magnitude of mediation across crime types can help identify whether routine activities play a relatively larger role in property crimes (as the theory would suggest) or whether the mechanisms operate similarly across crime types. This comparative approach provides a stronger test of routine activity theory than examining either crime type in isolation.

Given this distinction's importance for theory testing, several studies have leveraged property crime data to examine environmental effects on criminal opportunity structures. Doleac and Sanders (2015) found that property crime decreases with increased daylight hours, likely due to improved visibility reducing opportunities for criminal behavior—a finding that aligns with routine activity theory's emphasis on opportunity structures. Similarly, studies of extreme weather events have found that property crimes often show distinct patterns from violent crimes, with some types of property crime actually increasing during periods when routine activities are disrupted (Corcoran & Zahnow, 2022). These findings highlight the value of examining multiple crime types when testing routine activity mechanisms. To better understand these relationships, we turn now to a detailed examination of the evidence for routine activity theory as a mechanism linking temperature to crime.

A weakness of routine activity theory is that it has proven somewhat difficult to directly observe. A few studies have identified scenarios in which logical deductions from the theory would yield distinguishable predictions in crime patterns (Rotton & Cohn, 2003). For example, Hipp et al. (2004) examined locations across multiple climate zones to test distinctive predictions of routine activity theory – they argued that only routine activity theory would predict increased violent crime rates from mid-range temperatures, since the opportunity for interaction would increase in that pleasantly moderate temperature range without the negative stimulus of high-range temperatures. In another study, Cruz et al. (2023) capitalized on divergent predictions for the severity of physical injuries of outdoor crime victims. They found no such evidence of more severe injuries, which they interpreted as more consistent with routine activity theory being the prevailing explanatory mechanism.

Another group of studies in this vein takes advantage of seasonal differences in what the theory would logically predict. For example, Harp and Karnauskas (2018) examined the influence of interannual climate variability on regional violent crime in the U.S. from 1979 to 2016, finding a positive relationship between temperature and violent crime rates. The spatial heterogeneity of this relationship was more pronounced in wintertime compared to summer and fall, aligning with the expectations of routine activity theory. Similarly, in a study of temperature volatility and violent crime in U.S. cities from 2015 to 2021, Thomas and Wolff (2023) found stronger impacts of anomalously warm temperatures in winter than in warmer months, consistent with routine activity theory but not heat/aggression theory.

While prior research has provided circumstantial evidence supporting routine activity theory, no studies have directly tested whether observed changes in everyday activities mediate the relationship between temperature and crime. The current study attempts to address this gap. To examine whether such activities might mediate the relationship between temperature and crime, it is important to first establish the theoretical mechanisms connecting temperatures, activities, and crime. The following sections review and synthesize prior research on the two stages of the possible mediation mechanisms: first, the impact of temperature on the particular routine activities examined in this study, and second, the impact of those routine activities on crime.

Temperature and routine activities: Public transit and outdoor recreation

Public Transit

The relationship between temperature and public transit usage has been well-documented across various contexts. A comprehensive literature review of 54 studies found that warmer temperatures have consistently corresponded to higher uses of urban public transit (Bocker et al., 2013). While these patterns manifest across different transportation modes, the dynamics and magnitude of effects can vary by transit type.

Traditional public transit systems like subways and buses show clear temperature sensitivity, with studies consistently finding that higher temperatures lead to increased ridership across different transit modes (Guo et al., 2007). The effect of temperature on public transit ridership, while statistically significant, has been found to be less pronounced than the impact of precipitation (Cools et al., 2010). In Washington state, for example, rainfall emerged as the dominant driver of transit ridership across all seasons (Stover & McCormack, 2012).

Taxi services, while serving different populations and purposes than traditional public transit, show similar temperature sensitivity in urban environments. Multiple studies have documented positive associations between temperature and taxi usage (Liu et al., 2020). A time series analysis of New York City Taxi & Limousine Commission data specifically demonstrated that taxi usage increases with maximum daily temperatures (Willis & Tranos, 2021). This pattern suggests that taxi services, like other forms of urban transportation, reflect broader patterns of temperature-driven mobility, with higher temperatures being associated with greater levels of mobility.

Across both traditional public transit and taxi services, researchers have identified important dynamics in these relationships. The temperature effect appears particularly strong for discretionary or occasional transit compared to non-discretionary transit, such as commuting to work (Bocker et al., 2013). This pattern holds true across transportation modes (especially important in a city like New York where multiple forms of public transit are widely used), suggesting that temperature's influence on urban mobility extends beyond any particular mode of transit.

Outdoor recreation activities

Warmer temperatures increase public outdoor recreation and leisure activities across diverse urban contexts. Bocket et al.'s (2013) literature review documented this relationship across multiple cities globally (e.g., San Francisco, Chicago, Montreal, Athens, Taichung City). In the city of Knoxville, Tennessee, Wolff and Fitzhugh (2011) used local infrared sensors in a popular urban greenway to study the effects of weather conditions on outdoor physical recreation activity (such as jogging, cycling, and anything else involving people physically moving), finding a positive relationship between warmer temperatures and outdoor urban physical activities. A study in New York City found that park and street noise complaints varied substantially across the five boroughs due to different urban morphology and local patterns of behavior, and that across all boroughs, these complaints were lowest in winter, which the authors attributed to lower temperatures plausibly "causing fewer outdoor activities, such as walking in the street and attending outdoor retail markets" (Tong & Kang, 2021). Overall, there is strong evidence across multiple contexts of a robust relationship between temperature and outdoor urban activity.

Routine activities and crime: Public transit and outdoor recreation

Public Transit

For public transit usage to mediate the relationship between temperature and crime, we must consider how different forms of urban transportation might influence crime opportunities. Transportation hubs, particularly subway and bus stations, can function as nodes where routine activities converge, potentially making these areas crime generators or attractors by concentrating both potential offenders and victims (Brantingham & Brantingham, 1993). These locations might simultaneously experience increased guardianship from higher passenger volumes and other observers who could potentially intervene in crimes (Cohen & Felson, 1979; Zahnow, 2023).

Research has consistently shown that activity around train stations and bus stops can increase opportunities for criminal incidents (Ceccato et al., 2022; Irvin-Erickson & La Vigne 2015). A study of a Philadelphia public transit strike found that closing subway stations reduced violent crime by up to 38% in station-adjacent areas (Wu & Ridgeway, 2021). The impact of transit hubs on crime has also been shown to vary between warmer and cooler months (Szkola et al., 2020).

Studies focusing specifically on ridership levels suggest a positive relationship with crime opportunities (Zhang et al., 2022). For instance, research in New York City found positive associations between subway ridership and certain types of crimes like robbery, though the relationship varied by crime type, with some evidence suggesting that higher passenger volumes might enhance guardianship for certain offenses (Li & Kim, 2023).

The relationship between taxi usage and crime, while less studied, offers a complementary perspective on urban mobility and crime opportunities. Taxi services, through both pickup and dropoff patterns, can indicate movement between different activity nodes and overall urban mobility patterns. While the criminogenic effects of taxi usage have not been the focus of prior empirical research, taxi activity patterns can serve as a proxy for broader mobility trends, potentially indicating areas where both potential offenders and targets concentrate.

Outdoor recreation activities

Park usage and other outdoor recreation activities have been shown in at least some studies to be associated with higher crime rates (McCord & Houser, 2017), though the dynamics are complex. Regarding the current study, routine activities in park areas and streets/sidewalks can also be considered nodes and pathways where potential offenders and victims may gather, forming crime generators or attractors. For urban park usage, in a study of 218 cities in California, Kim and Hipp (2018) found that areas near parks experienced higher levels of aggravated assault, robbery, and disorder than areas not near parks. Other studies have demonstrated that parks draw people from nearby areas, so the characteristics of and social processes in the nearby area shape the relationship between parks and victimization (Boessen & Hipp, 2018). Some prior research has even demonstrated a negative relationship between park usage and crime (Schertz et al., 2021). In balance, a synthetic literature review found mixed results, concluding that "regarding parks, there are insufficient studies to reflect on the impact” on crime (Shepley et al., 2019). However, these mixed findings suggest at least a potential mechanism linking park usage and crime.

Regarding other outdoor urban activities, there has also been mixed evidence consistent and inconsistent with a relationship with outdoor crime. Some studies have shown that an increase in urban foot traffic and other outdoor activities is associated with increased opportunities for victimization (Brantingham & Brantingham 1993; Kim & Hipp, 2020). For example, Kim and Hipp (2023) analyzed both the vitality and diversity of street social activity in New York City, finding a crime-enhancing association with both violent and property crime. Other studies have shown outdoor activities are associated with a decrease in crime because of an increase in capable guardians and informal social control (Wo, 2019). A study in Chicago and New York City found a negative association between street activity and crime – a 5% increase in street activity was associated with a reduction of 9% in violent crime in Chicago and a 2.7% decrease in violent crime in New York (Schertz et al., 2021). Overall, the evidence is mixed about a general relationship in either direction.

Research Gap

Based on this review of prior research, if routine activity theory explains the relationship between temperature and crime, we would expect to observe specific alterations in daily behavior patterns on warmer days. These changes might manifest in several ways: individuals may adapt their leisure activities, shifting from indoor to outdoor venues; transportation patterns might change as people use different modes of transit; and overall urban mobility patterns could shift as people spend more time in public spaces. Each of these behavioral adaptations could theoretically increase opportunities for criminal interactions by bringing more potential offenders and victims into contact in situations with potentially reduced guardianship.

While previous studies have made valuable contributions through indirect tests of routine activity theory – such as examining seasonal variations, comparing climate zones, or analyzing injury patterns – there remains a crucial gap in directly observing and measuring the hypothesized behavioral mechanisms. The empirical evidence reviewed above suggests both that temperature influences urban routine activities (through transit usage and outdoor recreation) and that these activities may affect crime opportunities, but no study has yet connected these pieces to empirically observe the full mediation pathway. This represents a significant limitation in our understanding of how temperature affects crime rates.

The current study addresses this gap by providing the first empirical observation of these proposed behavioral mechanisms through careful measurement of actual activity patterns. By combining multiple administrative data sources that capture different aspects of urban routine activities, we can assess whether changes in observable behavior patterns mediate the relationship between temperature and crime. This approach moves beyond inferential tests to empirically examine the mechanisms proposed by routine activity theory, potentially resolving long-standing questions about how temperature influences criminal behavior. In doing so, the current study makes a substantial contribution to this perennial body of research, by attempting to directly observe what had hitherto been merely deduced.

CURRENT STUDY

To test whether routine activities mediate the relationship between temperature and crime, this study leverages comprehensive administrative data from New York City (2015-2019). We create an Activity Index that combines multiple indicators of urban routine activities, including public transportation usage, and other outdoor urban behaviors. This novel approach allows us to empirically measure how temperature affects these activities and whether such changes in behavior patterns subsequently influence crime rates.

By examining both violent and nonviolent crimes, we create multiple tests of the routine activities mechanism. While violent crimes might be influenced by both psychological responses to heat and changes in routine activities, nonviolent crimes like public consumption violations are more purely dependent on opportunity structures created by outdoor social activity. This multi-crime approach provides a more robust test of whether routine activities mediate the temperature-crime relationship.

The study employs a panel mediation analysis with bootstrapped standard errors to assess two key pathways: (1) the relationship between temperature and the Activity Index, and (2) the relationship between the Activity Index and both violent and nonviolent crime, controlling for temperature. By examining these pathways simultaneously, we can quantify the extent to which changes in routine activities explain the temperature-crime relationship. This methodological approach provides the first direct empirical test of routine activity theory as an explanation for temperature's effect on crime. The following sections detail our data sources, analytical strategy, and findings.

DATA AND METHODS

Data Sources

This study uses daily data spanning from January 1, 2015, to December 31, 2019, sourced from various publicly accessible websites, including the NYC Open Data Portal and the NYS Open Data Portal. The comprehensive dataset encompasses information from several key city organizations, notably the New York City Police Department (NYPD), the Metropolitan Transportation Authority (MTA), NYC 311, and The New York City Taxi and Limousine Commission (TLC). This rich combination of daily data from multiple entities provides a broad and detailed perspective on the urban dynamics of New York City during this period, offering valuable insights into aspects of city life ranging from weather conditions to public transportation usage and the incidence of violent and nonviolent crime. Our dataset includes daily borough-level data for this five-year period (5 X 365 days + 1 leap year day (5 borough-days) = 9,130).

This study also utilizes five 1-year American Community Survey (ACS) estimates from 2015 to 2019 to retrieve commuter-adjusted population at the borough level, calculated as "resident population + workers working in the borough – workers living in the borough" (Stults & Hasbrouck, 2015). They are chosen as proxies for ambient populations for several reasons. First, interborough commuting is common in New York City due to the concentration of commercial activities in Manhattan, suggesting substantial shifts in daytime population across boroughs. Second, commuter-adjusted populations better capture at-risk populations during daytime, when most social interactions and outdoor activities occur (Stults & Hasbrouck, 2015). Third, our Activity Index encompasses public transit ridership and noise complaints related to parks, reflecting activities aligned more with daytime population dynamics. Lastly, residential crime rates may overestimate relative risk in areas that draw a large influx of people from outside their boundaries (Stults & Hasbrouck, 2015). As such, commuter-adjusted populations are better suited to investigate the effect of routine activities than resident populations.

Dependent Variables

The current analysis focuses on four types of serious violent crime (homicide, shootings, robbery, and assault) and three types of nonviolent crime (larceny, burglary, and unlawful consumption of alcohol or marijuana) occurring within NYC during this period. Alcohol and marijuana consumption offenses were selected because they typically occur in public spaces and are closely tied to outdoor social activities, providing an additional test of whether routine activities mediate the temperature-crime relationship. Historic NYPD complaint data, which includes all valid felony, misdemeanor and violation crimes reported to the police department, was sourced from the NYC Open Data website.1 From this data, we created daily borough-level rates of homicide, robbery, felony assault, burglary and larceny (both petty and grand larceny) using each borough's commuter-adjusted population as the denominator to account for differences in borough size.

Information regarding shooting incidents in New York City was also sourced from the city's open data portal. Specifically, police-reported shooting incident data was compiled from the "historic" file available via the NYPD.2 This dataset, which is updated every quarter, spans from January 1, 2006, to December 31, 2022. To analyze shootings across New York City's boroughs, individual incident records from January 1, 2015, to December 31, 2019, were aggregated to generate a daily count at the borough level which again was computed to a rate per 100,000 borough commuter-adjusted population.

Finally, information on criminal summons issues for unlawful consumption of both marijuana and alcohol was drawn from the city's open data portal.3 Much like the other outcomes explored, information on criminal summons from January 1, 2015, to December 31, 2019, were aggregated to generate a daily count at the borough level which again was computed to a rate per 100,000 borough commuter-adjusted population.

Weather Data

Daily ambient temperature data for each borough were obtained from VisualCrossing.com, a global weather and forecasting platform. The "feels like" or apparent temperature combines air temperature, humidity, and wind speed to reflect how the temperature is experienced by individuals. This measure integrates two key components: the Heat Index, which accounts for increased heat perception at high temperatures and humidity, and Wind Chill, which captures the cooling effect of wind at low temperatures. Both are expressed in temperature units and provide a parsimonious metric for daily maximum and minimum apparent temperatures. For this study, we use the "feels like max" as our primary indicator of temperatures as actually experienced. Additional variables, including precipitation (in inches) and daylight hours, were also drawn from the Visual Crossing dataset.

Potential Mediators

MTA Subway Entrance Rate. Daily subway ridership information was taken from New York City's open data portal, where hourly MTA subway "entrances" (when a person pays to enter the subway at a turnstile) at each subway station were summed to the daily level for each of the five boroughs. Using this count of subway entrances, a rate per 10 residents was calculated using each borough's commuter-adjusted population (Li & Kim, 2023).

NYC Taxi Pickup Rate. Taxi ridership was estimated using data drawn from the NYC Taxi and Limousine Commission.4 This dataset captures all pick-ups and drop-offs made by NYC taxis as well as the day/time and location. The measure used in the current analysis is the rate of taxi pick-ups that occurred within a given borough per 10 borough commuter-adjusted residents. This was computed by taking a count of each "yellow" and "green" taxi trip that originated within a given borough, regardless of where that trip ended.5 While this does not include other rides completed by ride-sharing companies such as Uber or other private transportation companies, given the persistent use of traditional taxi services within NYC, we believe this data does provide a meaningful measure of day-to-day variation in taxi ridership (Mehrenbod et al., 2022).

311 Complaint Rates. An additional measure was created using data drawn from NYC's 311 system.6 Launched in 2003, 311 serves as a centralized hub for non-emergency government information and services, offering a wide range of assistance related to city agencies, programs, and inquiries. By simply dialing 311 or accessing the 311 website or app, individuals can report issues such as noise complaints, potholes, building violations, or seek information on various topics, including public transportation, health services, and more. From this data we extracted the daily count of 311 complaints associated with noise (specifically noise associated with parks or the street/sidewalk and excluding that labeled as "residential"). Daily complaint rates per 100,000 borough commuter-adjusted population were calculated for each measure.

Activity Index. In constructing the Activity Index, we selected indicators that capture routine activities creating opportunities for crime while avoiding measures that primarily indicate norm violations (which would risk circularity). Our final index focuses on more neutral indicators of urban activity: public transportation usage, taxi pickups, and noise complaints. Though noise complaints represent minor norm violations, they serve primarily as a proxy for outdoor social activity levels. Using the three measures described above we extracted an index (through principal component analysis) believed to represent the volume of routine activities occurring on a given day. An index was chosen due to a high degree of collinearity between the indicators of routine activities (r > .50) and each measure exhibited significant factor loadings (e.g. >.6), indicating a strong association with the extracted factor. The eigenvalue for the primary factor was 1.63, thereby confirming that these measures coalesce into a single latent construct. The resulting standardized measure has a mean of zero and higher values associated with greater levels of activity. The advantage of using PCA over a summative index lies in its ability to reduce dimensionality while capturing the shared variance among variables, thus providing a more accurate representation of the underlying construct without assuming equal weight for each measure.

Additional Controls

The study incorporated several additional variables used in prior research to account for various factors influencing crimes in our study. Variables for federal holidays were included due to their distinct impact on crimes, as changes in routine activities on these days have been observed to affect such crimes (Rudernan & Cohn, 2021). Our analysis also includes weekend indicators that historically report higher overall crime rates (Ceccato, 2005; LeBeau & Corcoran, 1990). We also include a lagged measure of the focal dependent variable as the incidence of crime has been seen to correlate over time (Tennenbaum & Fink, 1994). Including a lagged dependent variable in a model of crime is justified because it accounts for the temporal autocorrelation inherent in crime data, recognizing that past levels of crime can influence current levels. This approach helps to control for unobserved factors that persist over time and provides a more accurate and dynamic understanding of crime patterns and trends. Finally, the models presented include both month- and year-fixed effects to account for seasonality, as well as the potential for annual shocks, to the incidence of crime present across each of the five boroughs included in the analysis.7

Analysis

To examine whether our index of routine activities mediates the association between temperature and crime, a mediation analysis was conducted using a bootstrapped causal step analysis (following a mediation analysis approach similar to prior studies like Akram and Abrar Ul Haq (2022)). The mediation analysis involved a causal step analysis with bootstrapped standard errors in calculating three different tests of mediation (Sobel, Aroian, and Goodman) using the "product of coefficients" approach (MacKinnon et al. 2002; Mize, n.d.). This approach allowed us to estimate the direct effect of the focal independent variable (i.e., daily temperature) after parsing out the effect of the mediator (i.e., the indirect effect of the Activity Index), and the proportion of the total effect that the mediator can explain. The proportion mediated is calculated by dividing the indirect effect by the total effect; moreover, it captures the effect of temperature on crime that can be explained by the indirect effect of routine activities. In line with recommendations from Preacher and Hayes (2004), a bootstrapping procedure (i.e., a non-parametric resampling procedure) with 500 repetitions was used to obtain bias-corrected and percentile 95% confidence intervals (CIs) for the direct and indirect effects.8 This is recommended because the sampling distribution of a*b tends to be non-normal (Preacher & Hayes, 2004). In the current study, the estimated indirect effects were considered statistically significant if the confidence intervals did not contain zero. Finally, in order to make comparisons across each of the models estimated, we standardized each outcome and our focal independent variable (e.g., temperature). This allows us to compare the relative association of our focal variables between the outcomes explored without worrying about the differences in their base rate/prevalence.

RESULTS

Table 1 presents the descriptive statistics associated with our analysis. In addition to the daily rates of crime and information on daily temperature and other weather-related indicators, we include descriptive statistics for each component of our Activity Index. That measure has a mean of zero, with higher values indicating a greater volume of activity on a given day.

We used causal step analysis with bootstrapped standard errors to estimate the proportion of temperature's effect on crime that is mediated by observed routine activities. Seven separate models were estimated (which each include the estimation of three distinct regression models, not shown in tabular form). Importantly, although not shown in tabular form, each model includes all previously described control variables. As discussed above, a bootstrapping procedure with 500 repetitions was used to obtain the direct and indirect effects. Table 2 displays the direct effect of temperature on crime, the indirect effect of temperature through the mediating measure of routine activities, along with the total effect (and corresponding confidence intervals). Using this information, we calculated a percentage of the total effect that is indirect (or explained by the association between temperature and routine activities).

In each case, we see evidence of a significant indirect effect of temperature on crime through our measure of observed routine activities. For all crime types in this study, the direct effect of temperature remains positive and significant. In the majority of the models estimated, a portion of temperature's effect on the daily rate of crime was mediated by our Activity Index, although the magnitude of this indirect relationship varied across the crime types considered. The proportion of the total effect that was indirect was largest in the case of homicide (11.46%) while it was more modest for shootings (6.87%) assault (7.92%), larceny (8.85%) and marijuana and alcohol summonses (7.56%). The indirect effect for both robbery and burglary failed to reach the point of significance. Overall, the results suggest that our measure of routine activities represents a meaningful, yet modest mediator in the relationship between temperature and crime. These results are discussed in greater detail in the next section.9

DISCUSSION

Most previous studies on the relationship between temperature and crime have either been agnostic about the possible mechanisms through which temperature might shape crime, or they have deduced plausible mechanisms through compelling but circumstantial evidence, such as the creative approaches of Harp and Karnauskas (2018). The current study is the first to explicitly assess a possible mechanism via empirically observed changes in a theoretically relevant set of routine activities: how people move about city streets, public transportation, and urban public spaces differently when temperatures rise. Our findings suggest that these observed changes in peoples' movement through New York City explain at least some of the observed impact of temperature on multiple types of crime.

Specifically, our stepwise products of coefficients analysis (with bootstrapped standard errors) yielded strong evidence of at least partial mediation of the association of warmer temperatures with several types of violent and nonviolent crime in New York City from 2015 to 2019. The Activity Index mediated approximately 11.5% of the association of temperature with homicide, 8% with assaults, 7% with shootings, and 9% with larceny. The mediation proportion for marijuana and alcohol summonses was approximately 8%. However, the indirect effects for robbery and burglary were not statistically significant. Our findings are consistent with prior research linking warmer temperatures to increases in these crimes (Ceccato, 2005; Lyons et al., 2022; Corcoran & Zahnow, 2022). In each of those and similar studies, the effects of temperature were deduced or hypothesized to be a function of concomitant changes in routine activities, but our study is the first in this line of research to offer directly observed evidence of this mediating mechanism.

We found a particularly pronounced mediation of the association between temperature and homicide. This finding resonates with prior research on homicides being strongly influenced by changes in the opportunity structure of homicidal encounters (Berg & Schreck, 2022; Berg & Felson, 2020; Weisburd et al., 2012). In other words, homicides are often sensitive to shifts in public behavior and mobility patterns that change the likelihood of encounters (Corsaro et al., 2017; Groff & McEwan, 2007; Tita & Griffiths, 2005). Homicides often start as conflicts that escalate violently (Felson & Steadman, 1983; Griffiths et al., 2011), so our findings are consistent with at least some of the association between warmer temperatures and homicide patterns being a function of increasing opportunities for people to interact and find themselves in conflict. Future research could fruitfully explore these behavioral mechanisms at an even more granular or situational level to confirm this interpretation of our findings.

Although the statistical significance of our results is clear, it is important to contextualize the magnitude of the relationships. The direct effects of temperature on crime, as measured by standardized coefficients, range from 0.071 (burglary) to 0.174 (larceny), indicating modest but consistent associations. The indirect effects through routine activities are smaller, ranging from non-significant for robbery and burglary to approximately 0.008 to 0.017 for other crime types. These coefficients suggest that while statistically significant, the relationship between temperature and crime is complex, with the routine activities we observed in the Activity Index playing a meaningful but not dominant role in mediating these associations.

Interestingly, the consistency of these effects across violent and nonviolent crime types warrants further consideration. Contrary to expectations, the mediated proportions for nonviolent crimes such as larceny (8.85%) were not markedly different from those for violent crimes such as homicide (11.46%). The similar mediation effects found across both violent crimes (like homicide and assault) and nonviolent crimes (like larceny and public consumption violations) provides compelling evidence for routine activities as a general mechanism. The fact that our Activity Index explained comparable proportions of the temperature effect for crimes with very different motivational structures suggests that changes in mobility patterns and public activity levels play a consistent role in creating criminal opportunities, regardless of crime type. Future research could examine other crime types, such as pick-pocketing or other drug use in public spaces, or alternative outcomes like hospital admissions to further isolate the contributions of routine activities and heat-induced aggression mechanisms.

Limitations of this study include its focus on empirically observing changes in routine activities without concurrently observing citywide changes in levels of aggression. As such, the two prevailing theories cannot be simultaneously compared in this analysis. This study cannot speak to whether the heat and aggression mechanism might also explain some of the relationship between temperature and violence in this city in this period. The substantial residuals in our models do suggest something not in the models would explain the observed variance in violence (though some of those residuals are surely from routine activities that are not observed via our Activity Index). Overall, this study can only offer the first direct empirical evidence of changes in these routine activities (as measured in our Activity Index) as at least one of the relevant mechanisms. We look forward to future researchers creatively leveraging new data sources to develop research designs that enable simultaneously comparisons of the two prevailing mechanisms to see whether both mechanisms are at play or routine activities alone account for the observed phenomena. Other researchers could also fruitfully develop alternative indicators of routine activities beyond the Activity Index indicator we developed by incorporating additional measures of activity. For example, the proportion of footfall attributed to a borough’s residential population would be instructive in future studies. Further disentangling the explanatory mechanisms would go far in finally explaining Quetelet's (1842) thermic law of delinquency.

Additionally, there are limitations to the crime dataset from the NYPD. Such administrative data cannot include unreported crime, and variance may be a function of changes in police practices rather than a function of actual changes in crime (Baumer & Lauritsen, 2010). This limitation applies to every criminological study using official police statistics, but future research could replicate our study using victimization self-reports and/or other measures of crime such as the use of hospital intake data. An additional potential limitation is that this research was conducted at the borough level rather than a more granular level. Future research examining lower levels of analysis could identify other spatial patterns in both criminal activity as well as routine activities..

This study’s limitations further include its inability to analyze diurnal variation in routine activities or crime patterns, as our data were not collected at a sufficiently granular temporal level. Future research leveraging hourly mobility or crime data could provide valuable insights into how time-of-day interactions with temperature affect crime dynamics. Additionally, while this study focuses on public mobility, it does not account for residential population dynamics or private settings, such as cohabitation-related violence, which may follow different mechanisms. Incorporating data on domestic violence or distinguishing between local residents and visitors could further refine the understanding of temperature’s effects on crime.

All things considered, this study significantly advances our understanding of the relationship between temperature and crime. In other contexts, prior studies have used empirical indicators to directly assess changes in routine activities (Wilcox et al., 2018). Yet none have incorporated such observations of changing behavioral or mobility patterns into solving the perennial riddle of why and how temperature shapes crime. Interpreting the substantive significance of our findings depends on one's prior assumptions. If the prior assumption is that most of the impact of temperature on crime is due to routine activities, then our evidence of only a modest portion of the impact being mediated by changes in routine activities could be interpreted as undercutting the theory. However, if the prior assumption is that routine activities may not explain the effect of temperature on crime since there has previously only been indirect evidence of such a relationship, then our findings strongly support the theory in the sense they show that at least some of the relationship is explainable by routine activities. Our approach was to test the theory by attempting to falsify it, and our results do not falsify the theory. We thus interpret our findings as meaningfully moving the literature forward by being the first study to directly support this theoretical mechanism. It must be noted that our findings are the floor not the ceiling for how much of the relationship is explicable by routine activities, since there are other routine activities not included in our Activity Index that could also mediate the relationship. Future research could empirically assess the contribution of some of these other activities.

One of the strengths of this study is the use of the 'feels like' temperature measure, which incorporates both heat and humidity to capture a more accurate representation of the environmental conditions people experience. By including this measure, our analysis reflects the combined impact of temperature and humidity on crime dynamics, aligning more closely with real-world perceptions of thermal discomfort. Future research on temperature and crime should consider taking this approach.

This study also introduces a novel approach to measuring routine activities through the Activity Index, offering a methodological contribution to criminological research. While developed specifically for this New York City context, the index provides an initial framework that could inform future research designs. Its application may offer insights into routine activity patterns, though broader utility would require careful validation and potential adaptation to other urban settings.

The policy implications of this study include helping focus crime interventions from law enforcement, social workers, community members, and others. Since this study concludes that at least one relevant mechanism involves changes in routine activities, interventions could target places and situations where people are more likely to interact on warmer days (as opposed to interventions targeted toward people experiencing negative affect from heat-induced aggression). In some ways, these findings reinforce law enforcement's conventional wisdom about deploying more officers when temperatures rise, but these findings help further focus the goals and tactics of the interventions. Moreover, the findings could be applied to other approaches to prevention of crime on warm days, such as increasing prosocial community activities near nodes and pathways where people are more likely to interact on warm days. For example, as Boessen and Hipp (2018) argued, whether urban public parks are protective for communities is determined by how those parks are used and activated by their local communities, so there are real possibilities of intentional, transformative warm-day usage of parks that builds local capacity for social control, social ties, and collective efficacy.

Finally, these results are especially germane and meaningful in a period of history in which climate change is accelerating, leading to new patterns of behavioral responses to and expectations about local weather. In the U.S., the most likely climate change scenarios predict not only overall warming of average surface temperatures, but also increased risk of weather phenomena like unseasonable temperatures, weather whiplash events, extreme heat waves, extreme storms, and what the Intergovernmental Panel on Climate Change dubs "compound climate hazards" (such as conditions combining heatwaves, droughts, and fires) (Portner et al., 2023; Thomas & Wolff, 2023). In a time of rapid climate change, behavioral responses to the weather could be evolving in ways that could have more pronounced impacts on crime than in the recent past. More generally, as a society we must grapple with questions of how macro-level changes to our climate are going to shape changes in everyday behavior. Profound changes to the climate could lead to profound changes in everyday behavioral patterns. How will any such changes in our behavioral patterns reshape patterns of crime and responses to crime? If crime is at least partly explained by how activity patterns determine our interactions, it is urgent that criminologists meaningfully contribute to our collective understanding of how changing activity patterns will lead to people interacting differently in the Anthropocene.

References

Akram, F., & Abrar Ul Haq, M. (2022). Integrating agency and resource dependence theories to examine the impact of corporate governance and innovation on firm performance. Cogent Business & Management, 9(1), 2152538.

Anderson, C. A., & Anderson, D. C. (1984). Ambient temperature and violent crime: Tests of the linear and curvilinear hypotheses. Journal of Personality and Social Psychology, 46(1), 91.

Anderson, C. A., & Bushman, B. J. (2002). Human aggression. Annual Review of Psychology, 53, 27-51.

Anderson, C. A., Deuser, W. E., & DeNeve, K. M. (1995). Hot temperatures, hostile affect, hostile cognition, and arousal. Personality and social psychology bulletin, 21(5), 434-448.

Baryshnikova, N., Davidson, S., & Wesselbaum, D. (2021). Do you feel the heat around the corner? The effect of weather on crime. Empirical Economics, 1-21.

Baumer, E. P., & Lauritsen, J. L. (2010). Reporting crime to the police, 1973–2005: a multivariate analysis of long‐term trends in the National Crime Survey (NCS) and National Crime Victimization Survey (NCVS). Criminology, 48(1), 131-185.

Baumer, E., & Wright, R. (1996). Crime seasonality and serious scholarship: A comment on Farrell and Pease. The British Journal of Criminology, 36(4), 579-581.

Berman, J. D., Bayham, J., & Burkhardt, J. (2020). Hot under the collar: A 14-year association between temperature and violent behavior across 436 US counties. Environmental research, 191, 110181.

Berg, M. T., & Felson, R. (2020). A social interactionist approach to the victim-offender overlap. Journal of Quantitative Criminology, 36, 153-181.

Berg, M. T., & Schreck, C. J. (2022). The meaning of the victim–offender overlap for criminological theory and crime prevention policy. Annual Review of Criminology, 5, 277-297.

Bernasco, W., & Block, R. (2011). Robberies in Chicago: A block-level analysis of the influence of crime generators, crime attractors, and offender anchor points. Journal of Research in Crime and Delinquency, 48(1), 33-57.

Böcker, L., Dijst, M., & Prillwitz, J. (2013). Impact of everyday weather on individual daily travel behaviours in perspective: a literature review. Transport reviews, 33(1), 71-91.

Boessen, A., & Hipp, J. R. (2018). Parks as crime inhibitors or generators: Examining parks and the role of their nearby context. Social Science Research, 76, 186-201.

Brantingham, P.L., and Brantingham, P.J. (1993). Environment, routine and situation: Toward a pattern theory of crime. Advances in Criminological Theory, 5, 259–294.

Browning, C. R., Pinchak, N. P., & Calder, C. A. (2021). Human mobility and crime: Theoretical approaches and novel data collection strategies. Annual Review of Criminology, 4, 99-123.

Ceccato, V. (2005). Homicide in Sao Paulo, Brazil: Assessing spatial-temporal and weather variations. Journal of Environmental Psychology, 25 (3), 307-321.

Ceccato, V., Gaudelet, N., & Graf, G. (2022). Crime and safety in transit environments: a systematic review of the English and the French literature, 1970–2020. Public Transport, 14(1), 105-153.

Cohn, E. G. (1990). Weather and crime. The British Journal of Criminology, 30(1), 51-64.

Cohn, E. G., Rotton, J., Peterson, A. G., & Tarr, D. B. (2004). Temperature, City Size, and the Southern Subculture of Violence: Support for Social Escape/Avoidance (SEA) Theory 1. Journal of Applied Social Psychology, 34(8), 1652-1674.

Cohen, L. E., & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44, 588-608.

Cools, M., Moons, E., Creemers, L., & Wets, G. (2010). Changes in travel behavior in response to weather conditions: do type of weather and trip purpose matter?. Transportation Research Record, 2157(1), 22-28.

Corcoran, J., & Zahnow, R. (2022). Weather and crime: A systematic review of the empirical literature. Crime Science, 11(1).

Corsaro, N., Pizarro, J. M., & Shafer, J. (2017). The influence of planned aggression on the journey to homicide: An examination across typology classifications. Homicide Studies, 21(3), 179-198.

Cruz, E., D’Alessio, S. J., & Stolzenberg, L. (2023). The effect of maximum daily temperature on outdoor violence. Crime & Delinquency, 69(6-7), 1161-1182.

Doleac, J. L., & Sanders, N. J. (2015). Under the cover of darkness: How ambient light influences criminal activity. Review of Economics and Statistics, 97(5), 1093-1103.

Felson, M., & Boba, R. L. (2010). Crime and everyday life. Sage.

Felson, R. B., & Steadman, H. J. (1983). Situational factors in disputes leading to criminal violence. Criminology, 21(1), 59-74.

Field, S. (1992). The effect of temperature on crime. The British journal of Criminology, 32(3), 340-351.

Gamble, J. L., & Hess, J. J. (2012). Temperature and violent crime in Dallas, Texas: relationships and implications of climate change. Western Journal of Emergency Medicine, 13(3), 239.

Griffiths, E., Yule, C., & Gartner, R. (2011). Fighting over trivial things: Explaining the issue of contention in violent altercations. Criminology, 49(1), 61-94.

Groff, E. R., & McEwen, T. (2007). Integrating distance into mobility triangle typologies. Social Science Computer Review, 25(2), 210-238.

Guo, Z., Wilson, N. H., & Rahbee, A. (2007). Impact of weather on transit ridership in Chicago, Illinois. Transportation Research Record, 2034(1), 3-10.

Harp, R. D., & Karnauskas, K. B. (2018). The influence of interannual climate variability on regional violent crime rates in the United States. GeoHealth, 2(11), 356-369.

Hipp, J. R., Curran, P. J., Bollen, K. A., & Bauer, D. J. (2004). Crimes of opportunity or crimes of emotion? Testing two explanations of seasonal change in crime. Social Forces, 82(4), 1333-1372.

Irvin-Erickson, Y., & La Vigne, N. (2015). A spatio-temporal analysis of crime at Washington, DC metro rail: Stations' crime-generating and crime-attracting characteristics as transportation nodes and places. Crime science, 4, 1-13.

Kim, Y. A., & Hipp, J. R. (2018). Physical boundaries and city boundaries: Consequences for crime patterns on street segments?. Crime & Delinquency, 64(2), 227-254.

Kim, Y. A., & Hipp, J. R. (2020). Pathways: examining street network configurations, structural characteristics and spatial crime patterns in street segments. Journal of Quantitative Criminology, 36, 725-752.

Kim, Y. A., & Wo, J. C. (2022). Seasonality and crime in Orlando neighbourhoods. The British Journal of Criminology, 62(1), 124-144.

Kim, Y. A., & Hipp, J. R. (2023). Does Street Social Activity Impact Crime? An Analysis in New York City. Crime & Delinquency, 00111287231211262.

LeBeau, J. L., & Corcoran, W. T. (1990). Changes in calls for police service with changes in routine activities and the arrival and passage of weather fronts. Journal of Quantitative Criminology, 6, 269–291.

Li, N., & Kim, Y. A. (2023). Subway station and neighborhood crime: An egohood analysis using subway ridership and crime data in new york city. Crime & Delinquency, 69(11), 2303-2328.

Linning, S. J., Andresen, M. A., Ghaseminejad, A. H., & Brantingham, P. J. (2017). Crime seasonality across multiple jurisdictions in British Columbia, Canada. Canadian journal of criminology and criminal justice, 59(2), 251-280.

Liu, Q., Ding, C., & Chen, P. (2020). A panel analysis of the effect of the urban environment on the spatiotemporal pattern of taxi demand. Travel Behaviour and Society, 18, 29-36.

Lyons, V. H., Gause, E. L., Spangler, K. R., Wellenius, G. A., & Jay, J. (2022). Analysis of daily ambient temperature and firearm violence in 100 US cities. JAMA Network Open, 5(12), e2247207.

MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7(1), 83.

McCord, E. S., & Houser, K. A. (2017). Neighborhood parks, evidence of guardianship, and crime in two diverse US cities. Security Journal, 30, 807-824.

Mehranbod, C. A., Gobaud, A. N., & Morrison, C. N. (2022). Ridesharing and alcohol-related assaults in NYC: A spatial ecological case-crossover study. Drug and alcohol dependence, 232, 109321.

Mize, T. (n.d.). Sobel-Goodman test of mediation in Stata. Retrieved March 13, 2024, from https://www.trentonmize.com/software/sgmediation2.

Pörtner, H. O., Scholes, R. J., Arneth, A., Barnes, D. K. A., Burrows, M. T., Diamond, S. E., ... & Val, A. L. (2023). Overcoming the coupled climate and biodiversity crises and their societal impacts. Science, 380(6642), eabl4881.

Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior Research Methods, Instruments, & Computers, 36, 717-731.

Quetelet, M. A. (1842). A Treatise on Man and the Development of His Facilities. Edinburgh, UK: W. and R. Chambers.

Ranson, M. (2014). Crime, weather, and climate change. Journal of environmental economics and management, 67(3), 274-302.

Rotton, J., & Cohn, E. G. (2003). Global warming and US crime rates: An application of routine activity theory. Environment and Behavior, 35(6), 802-825.

Ruderman, D., & Cohn, E. G. (2021). Predictive extrinsic factors in multiple victim shootings. The Journal of Primary Prevention, 42(1), 59-75.

Schertz, K. E., Saxon, J., Cardenas-Iniguez, C., Bettencourt, L. M., Ding, Y., Hoffmann, H., & Berman, M. G. (2021). Neighborhood street activity and greenspace usage uniquely contribute to predicting crime. Npj Urban Sustainability, 1(1), 19.

Schinasi, L. H., & Hamra, G. B. (2017). A time series analysis of associations between daily temperature and crime events in Philadelphia, Pennsylvania. Journal of urban health, 94, 892-900.

Shepley, M., Sachs, N., Sadatsafavi, H., Fournier, C., & Peditto, K. (2019). The impact of green space on violent crime in urban environments. International journal of environmental research and public health, 16(24), 5119.

Simister, J., & Cooper, C. (2005). Thermal stress in the USA: effects on violence and on employee behaviour. Stress and Health, 21(1), 3-15.

Stover, V. W., & McCormack, E. D. (2012). The impact of weather on bus ridership in Pierce County, Washington. Journal of Public Transportation, 15(1), 95-110.

Stults, B. J., & Hasbrouck, M. (2015). The Effect of Commuting on City-Level Crime Rates. Journal of Quantitative Criminology, 31(2), 331–350.

Szkola, J., Piza, E., & Drawve, G. (2020). Risk terrain modeling: Seasonality and predictive validity. Justice Quarterly, 38(2), 322-343.

Tennenbaum, A. N., & Fink, E. L. (1994). Temporal regularities in homicide: Cycles, seasons, and autoregression. Journal of Quantitative Criminology, 10, 317-342.

Thomas, C., & Wolff, K. T. (2023). Weird winter weather in the Anthropocene: How volatile temperatures shape violent crime. Journal of Criminal Justice, 87, 102090.

Tita, G., & Griffiths, E. (2005). Traveling to violence: The case for a mobility-based spatial typology of homicide. Journal of Research in Crime and Delinquency, 42(3), 275-308.

Tong, H., & Kang, J. (2021). Characteristics of noise complaints and the associations with urban morphology: A comparison across densities. Environmental Research, 197, 111045.

Weisburd, D., Groff, E. R., & Yang, S. M. (2012). The criminology of place: Street segments and our understanding of the crime problem. Oxford University Press.

Wilcox, P., Cullen, F. T., & Feldmeyer, B. (2018). Communities and crime: An enduring American challenge. Temple University Press.

Willis, G., & Tranos, E. (2021). Using ‘Big Data’to understand the impacts of Uber on taxis in New York City. Travel Behaviour and Society, 22, 94-107.

Wolff, D., & Fitzhugh, E. C. (2011). The relationships between weather-related factors and daily outdoor physical activity counts on an urban greenway. International journal of environmental research and public health, 8(2), 579-589.

Wo, J. C. (2019). Mixed land use and neighborhood crime. Social science research, 78, 170-186.

Wu, Y., & Ridgeway, G. (2021). Effect of public transit on crime: evidence from SEPTA strikes in Philadelphia. Journal of experimental criminology, 17, 267-286.

Zahnow, R. (2023). Examining train stations as crime generators and the protective effect of “regular” riders. Crime & Delinquency, 00111287231160737.