When people talk about evictions, the conversation usually focuses on numbers. We hear about eviction rates, percentages, and housing statistics. But those numbers rarely show where these problems are happening.
That is where Geographic Information Systems (GIS) become useful. GIS helps us move beyond simple statistics and see the geography behind social issues. By mapping eviction rates, we can start to see patterns across neighborhoods and better understand how housing instability spreads across a city.
In this project, I analyzed eviction rates across census tracts in Sacramento, California using ArcGIS Pro. The goal was simple: determine whether eviction rates appear randomly across the city or whether they cluster in certain areas.
Preparing the Data
Before any analysis can begin, the data must be prepared correctly. For this project, eviction data was mapped at the census tract level, which provides a good balance between neighborhood detail and citywide coverage.
The dataset also needed to be projected into a coordinate system suitable for spatial analysis. This step is important because many spatial statistics tools rely on accurate distance calculations between geographic features. If the projection is incorrect, the analysis results can be misleading.
Creating the Spatial Weights Matrix
Spatial analysis tools need to understand how geographic areas relate to each other. In other words, the software needs to know which census tracts are considered neighbors.
To do this, I created a Spatial Weights Matrix using the Queen’s Case method. In this approach, two tracts are considered neighbors if they share a border or even a corner. This method works well for census tracts because neighborhoods often influence each other even when they only touch at a small point.
While setting this up, I encountered a small data issue. The GEOID field that identifies each census tract was stored as text instead of a numeric value. Since the spatial weights tool requires numeric identifiers, I created a new integer field and copied the GEOID values into it so the analysis could run properly.
Measuring Spatial Patterns with Moran’s I
With the spatial relationships defined, I ran a Global Moran’s I analysis to determine whether eviction rates were randomly distributed across Sacramento.
Moran’s I measures spatial autocorrelation, which tells us whether similar values tend to occur near each other. If eviction rates were randomly distributed, the Moran’s I value would be close to zero. If areas with high eviction rates are located near each other, the value becomes positive.
The analysis produced a Moran’s I value of approximately 0.57, along with a very high z-score and a near-zero p-value. These results indicate that the pattern is statistically significant and unlikely to occur by chance.
In simple terms, eviction rates in Sacramento form geographic clusters rather than being randomly distributed.
Global Moran’s I results window
The results show a strong clustered pattern of eviction rates in Sacramento. The positive Moran’s Index (0.567763)
means areas with similar eviction rates are located near each other instead of being randomly distributed. The very high
z-score (21.95) and very small p-value (0.000000) show that this pattern is statistically significant and unlikely to occur by
chance. Since the observed index is much higher than the expected index, the analysis confirms strong spatial clustering.
Overall, eviction rates are geographically concentrated, suggesting that neighboring areas share similar housing and
social conditions influencing eviction patterns.
Identifying Eviction Hotspots
Global Moran’s I tells us that clustering exists, but it does not tell us where the clusters are located. To identify specific neighborhoods experiencing similar eviction patterns, I used Local Moran’s I, also known as Cluster and Outlier Analysis.
This tool compares each census tract with its neighboring tracts and identifies several types of spatial patterns:
- High–High clusters: Areas with high eviction rates surrounded by other high-rate neighborhoods
- Low–Low clusters: Areas where eviction rates are consistently low
- Spatial outliers: Neighborhoods where eviction rates differ from nearby areas
These results reveal the hotspots and coldspots of eviction activity within the city.
Understanding the Moran Scatterplot
Another useful way to interpret spatial autocorrelation is through the Moran’s scatterplot. This chart shows the relationship between a census tract’s eviction rate and the eviction rates of its neighboring tracts.
In this analysis, the scatterplot shows a clear upward trend line. This means that tracts with higher eviction rates tend to be located near other tracts with high eviction rates, confirming the clustering pattern detected by the statistical analysis.
Creating the Final Map Layout
After completing the analysis, the results were compiled into a final cartographic layout. The map highlights areas of high and low eviction clustering across Sacramento and includes the Moran’s scatterplot as a supporting visualization.
This layout helps translate statistical results into a format that is easier to interpret and communicate.
Why This Analysis Matters
GIS allows us to move beyond simple statistics and see the geography of social challenges.
When eviction rates cluster geographically, it often reflects broader social and economic conditions affecting certain neighborhoods. These patterns may be related to housing affordability, economic inequality, job access, or historical development patterns.
By identifying where eviction hotspots occur, policymakers and planners can better target housing assistance programs and community resources.
Spatial analysis does not solve housing instability on its own, but it helps us understand where the problems are concentrated and where solutions may be needed most.
