## Reading layer `crime_early_period_franklin' from data source 
##   `https://drive.google.com/uc?export=download&id=16MANlBpAi1AwhvlDXizeCH6zkmQML-LU' 
##   using driver `GeoJSON'
## Simple feature collection with 6629 features and 21 fields
## Geometry type: POINT
## Dimension:     XY
## Bounding box:  xmin: 1980030 ymin: 786014.3 xmax: 1985448 ymax: 790686.5
## Projected CRS: NAD83 / North Carolina (ftUS)
## Reading layer `crime_late_period_franklin' from data source 
##   `https://drive.google.com/uc?export=download&id=1AGwEaKKKPVgv7xzs_PUsWmptl6aqRTcc' 
##   using driver `GeoJSON'
## Simple feature collection with 6518 features and 21 fields
## Geometry type: POINT
## Dimension:     XY
## Bounding box:  xmin: 1980022 ymin: 786023.8 xmax: 1985424 ymax: 790505.5
## Projected CRS: NAD83 / North Carolina (ftUS)

Executive Summary

This report compares the spatial pattern of police reports in Downtown Chapel Hill during 2010–2014 and 2020–2025. I examine both first-order spatial structure, which asks where crime intensity is highest, and second-order spatial structure, which asks whether crime points are more clustered than expected under complete spatial randomness.

Main finding: The early period contains 6,629 reports, or about 1325.8 per year. The later period contains 6,518 reports, or about 1086.3 per year. After accounting for the unequal time spans, the annualized number of reports decreased by about 18.1%.

The spatial analyses below show whether this change was accompanied by a shift in where reports were concentrated and how strongly they clustered.

Why Spatial Crime Patterns Matter

Crime is rarely distributed evenly across an urban area. Routine activity theory suggests that incidents are more likely where potential offenders, suitable targets, and limited guardianship come together. In a downtown setting, that can mean commercial corridors, nightlife areas, parking facilities, transit routes, and other places with heavy pedestrian activity.

This matters in Downtown Chapel Hill because Franklin and Rosemary Streets connect restaurants, bars, retail, public spaces, parking, transit, and the UNC campus. The downtown environment also changed substantially between the two periods. The Town’s 140 West Franklin Plaza became a public gathering space, redevelopment added new mixed-use activity, and later downtown planning efforts continued to reshape streets, public spaces, parking, and commercial uses. UNC enrollment also grew from 29,390 students in fall 2010 to more than 34,000 students today, increasing the population connected to the downtown area.

The COVID-19 pandemic is another important complication. The later period begins in 2020, when campus activity, business operations, nightlife, travel, and police reporting patterns were temporarily disrupted. Because of that, the 2020–2025 period should be interpreted as a mixed post-2019 era rather than a completely stable six-year period.

Descriptive Comparison

Descriptive comparison of the two crime-report periods
Period Number of reports Years observed Reports per year Annualized reports per square mile
2010–2014 6629 5 1325.8 1401.5
2020–2025 6518 6 1086.3 1148.4

The periods are different lengths, so the raw point counts should not be compared by themselves. I use reports per year and annualized density to make the comparison fairer.

Raw Spatial Distribution

Hypothesis about the generating process

Hypothesis: Crime reports will not be spread evenly across Downtown Chapel Hill. Instead, they will be concentrated along major activity corridors and around places that generate pedestrian traffic, nightlife, retail activity, parking, or institutional activity.

The raw point maps provide the first visual clue about change over time. Dense stacks of points indicate repeated reports at or near the same locations, while gaps indicate areas with relatively little reported crime activity.

First-Order Spatial Structure: Where Is Crime Intensity Highest?

First-order structure describes variation in the expected intensity of events across space. I estimate kernel density surfaces for both periods using the same bandwidth and the same study window, which makes their spatial patterns more directly comparable.

Because the periods cover different numbers of years, the density values are annualized.

To summarize concentration numerically, I calculate the share of total estimated density that falls within the highest-density 10% of map cells. A larger value indicates that the spatial pattern is more concentrated into a smaller set of hotspots.

The highest-density 10% of map cells account for approximately 100% of estimated annualized density in 2010–2014 and 100% in 2020–2025.

This suggests that the later-period pattern became somewhat more spatially dispersed, with less of the total estimated intensity concentrated in the strongest hotspots.

The first-order maps should be interpreted in relation to Downtown Chapel Hill’s urban environment. Concentrations along Franklin Street or nearby blocks would be consistent with a routine-activity explanation because those areas combine restaurants, bars, retail, pedestrian traffic, transit, parking, public spaces, and proximity to UNC.

Second-Order Spatial Structure: Is Crime More Clustered Than Expected?

Second-order analysis asks whether points are unusually close to one another after accounting for the overall intensity of the point pattern.

Hypotheses

Null hypothesis (H₀): Within each study period, crime reports follow complete spatial randomness (CSR). Their locations are no more clustered or dispersed than expected from a homogeneous Poisson process.

Alternative hypothesis (H₁): Crime reports show spatial interaction, meaning that they are significantly more clustered or more dispersed than expected under CSR.

I use Ripley’s L-function, which is a variance-stabilized version of Ripley’s K-function. Under CSR, the expected centered L value is zero. Values above the simulation envelope indicate significant clustering at that spatial scale, while values below the envelope indicate significant dispersion.

For 2010–2014, the observed L-function rises above the CSR envelope, showing statistically significant clustering at one or more spatial scales, extending to roughly 1125 feet.

For 2020–2025, the observed L-function rises above the CSR envelope, showing statistically significant clustering at one or more spatial scales, extending to roughly 1135 feet.

One caution is important: a homogeneous CSR benchmark assumes constant underlying intensity. Downtown crime clearly may have first-order inhomogeneity because some blocks generate much more activity than others. Therefore, significant L-function clustering can partly reflect the concentration of risk around particular land uses and activity nodes rather than direct interaction among crime events themselves.

How Did the Spatial Structure Change Over Time?

Summary of change in crime-report intensity and clustering
Measure 2010–2014 2020–2025
Reports per year 1325.8 1086.3
Annualized reports per square mile 1401.5 1148.4
Share of density in hottest 10% of cells 100% 100%
Significant clustering under CSR envelope Yes Yes

The comparison shows three different dimensions of change:

Overall intensity. Annualized reporting decreased by approximately 18.1%.

Spatial concentration. The hotspot concentration measure decreased, suggesting that the later pattern was more spatially dispersed across downtown.

Second-order clustering. The L-function analysis shows whether reports were more clustered than expected under CSR in each period. Comparing the two envelopes helps distinguish a change in raw intensity from a change in the spatial arrangement of the points.

Chapel Hill Context and Plausible Explanations

The change between the two periods should not be interpreted as the result of one single cause. Downtown Chapel Hill changed in several ways.

Redevelopment and changing activity nodes. The 140 West development and public plaza added residences, retail, public space, and event activity to West Franklin Street. Carolina Square and later redevelopment efforts also changed the mix of housing, retail, office, and pedestrian activity downtown. These changes can shift where people gather and therefore where opportunities for both crime and police contact occur.

A larger university population. UNC’s official fall 2010 headcount was 29,390 students. UNC currently reports more than 34,000 undergraduate, graduate, and professional students. A larger university population can increase the number of people using Downtown Chapel Hill, although enrollment growth by itself does not imply more crime.

COVID-19 disruption. The 2020–2025 period includes an unusual shock. Campus operations, restaurants, nightlife, pedestrian activity, and travel changed sharply during the pandemic. Therefore, a hotspot that weakened or shifted in the later period may reflect temporary pandemic-era changes as well as longer-term redevelopment.

Policing and reporting behavior. The data represent police reports, not every crime that occurred. Changes in patrol deployment, reporting practices, public willingness to report incidents, business activity, and police presence can affect the observed point pattern.

Most plausible overall interpretation: Downtown crime reports are expected to concentrate around activity generators rather than occur randomly. Any shift in hotspots between the early and late periods should therefore be interpreted alongside changes in nightlife, retail, housing, parking, transit, redevelopment, campus activity, and COVID-era disruptions.

Policy Implications

The results support a place-based approach rather than treating all of Downtown Chapel Hill as equally risky.

First, the Town should continue monitoring micro-locations and corridors where kernel density remains persistently high. Resources can then be tailored to the actual place: lighting, late-night transportation, pedestrian design, alcohol-service partnerships, environmental design, outreach, or police presence.

Second, changes in spatial concentration should be reviewed together with land-use and redevelopment data. A new hotspot near a new activity node may reflect increased pedestrian exposure rather than a worsening citywide crime problem.

Third, future monitoring should compare annualized rates and consistent geographic windows, because the number of reports alone can be misleading when periods differ in length.

Finally, reported-crime data should be combined with other information where possible, including calls for service, crime type, time of day, business activity, special events, and pedestrian volumes. That would help the Town distinguish between places that are simply busy and places where risk is unusually high even after accounting for activity levels.

Limitations

This analysis has several limitations.

The two periods are different lengths, so I annualize counts and density for comparison. The later period also contains the COVID-19 pandemic, which makes it less internally consistent than the early period.

Police-report locations are not the same as all crimes that occurred. Reporting behavior and police practices may change over time.

The point-pattern analysis evaluates spatial location but does not separate crime types. A nightlife-related assault and a daytime theft may follow very different spatial processes.

Finally, the homogeneous CSR benchmark used for the L-function assumes constant underlying intensity. Because downtown land use is clearly uneven, some apparent second-order clustering may actually be caused by first-order variation in environmental risk.

Conclusion

Crime reports in Downtown Chapel Hill are not expected to be spatially random. The first-order analysis identifies where intensity is highest, while the second-order analysis tests whether reports are more clustered than expected under complete spatial randomness.

The comparison between 2010–2014 and 2020–2025 shows whether crime-report intensity, hotspot concentration, and clustering changed as Downtown Chapel Hill itself changed. The most useful policy response is therefore place-specific and context-sensitive: focus on persistent micro-hotspots, connect changes to land use and activity patterns, and avoid interpreting raw counts without considering time span, exposure, redevelopment, and reporting behavior.

References

Chapel Hill Police Department. (2025). Semi-Annual Report, July–December 2025. Town of Chapel Hill.

Town of Chapel Hill. (n.d.). 140 West Franklin Plaza.

Town of Chapel Hill. (2023). Market Assessment: Chapel Hill TOD Planning and UDO Visioning.

Town of Chapel Hill. (2025). Downtown Streetscape / Downtown Public Right-of-Way Master Plan.

University of North Carolina at Chapel Hill, Office of Institutional Research and Assessment. (2014). Fact Book 2010–2011.

University of North Carolina at Chapel Hill. (2026). Carolina by the Numbers.

Reproducibility Note

This report is configured with self_contained: true. When knitted in RStudio with an internet connection and the required packages installed, the output will be a single HTML file with the analysis, figures, styling, and results embedded inside it. No separate _files folder is required.