Introduction

This blog post explores spatial autocorrellation for median gross rent and median household income for census tracts in Virginia. Data comes from Social Explorer (https://www.socialexplorer.com/explore-tables) 5-year estimate survey data for both income and rent.

We expect to see high levels of income and rent in areas outside of the cities, particularly suburbs in Northern Virginia, Richmond, and Hampton Roads (https://dmz1.dhcd.virginia.gov/HB854/part-1-markets.html), and smaller levels of both in small market areas (Southwestern Virginia particularly). We also expect to find the areas of clear positive autocorrellation to be in North Virginia and the Southwestern/Western region.

We use a Global Moran’s I statistic to assess how clustered Virginia itself is. The closer the number is to 1, the more clustered income and rent census tracts tend to be (how high income/rent areas tend to be attracted to other areas of high income/rent). Significance in every spatial weight and cluster analysis is set at the 0.05 level.

Median household income gauges the general area. Rent is more representative for below-median income and smaller-sized households, in a way. This is because, in most cases, renters are more likely to not be able to afford to own property. This is particularly true in smaller markets, and especially true in rural markets (https://dmz1.dhcd.virginia.gov/HB854/part-3-rental.html), as apposed to large market areas, where there are more luxury rentals available. It should also be noted that people are more likely to rent in large market areas (the populations are also much larger). Additionally,renters tend to be made up of small household sizes in all three market sizes. From 2010 to 1019, about 40% of renters lived in one-person household sized spaces, and just under 30% in two-person sized space.

Global Moran I Values for Income, Rent, Bivariate

Over the 11-year period from 2011 to 2021, the Global Moran’s I values for income, rent, and their bivariate relationship remained consistently high, indicating strong spatial clustering patterns across the area. Income demonstrated a stable Moran’s I just above 0.7, suggesting that neighborhoods with similar income levels tended to cluster together and this pattern remained largely unchanged over time. Rent followed a nearly identical trend to income, with slightly higher values in the earlier years, possibly due to some data irregularities. The bivariate Moran’s I, which captures the spatial correlation between income and rent jointly, hovered just below 0.7 throughout the period. This indicates that areas with high (or low) income also tended to have high (or low) rent, and that this joint clustering pattern has been consistent over time. These findings reinforce the spatial linkage between economic status and housing costs in the region.

Univariate Local Moran’s I: Income - 2021

Univariate Local Moran’s I: Income - 2016

Univariate Local Moran’s I: Income - 2011

Cluster’s are colored at the 0.05 significance level. Data for the years 2011 and 2013 are not as rich in data as 2021, giving some error. These three plots and the Global Moran’s I values indicate that Virginia’s demographics hasn’t changed much in the past 11 years. Red can be found in suburban regions (especially North), while the blue exists more in the Southwest region. For the most part, high income census tracts are attracted to/attract other high income census tracts, while low income census tracts are attracted to/attract other low income census tracts.

Rent

Univariate Local Moran’s I: Rent - 2021

Univariate Local Moran’s I: Rent - 2016

Univariate Local Moran’s I: Rent - 2011

Results may vary because of NA values. Trends here are once again similar to those for income. Low to low rent clusters to extend more outwards, however, while high to high slightly more inwards.

For the most part, this bivariate map shows a similar trend to what has been observed in income and rent spatial clusters. This graph, however, shows more instances of high income census tracts being attracted to lower income census tracts, particularly near smaller cities. Red locations are once again dominated by suburban areas, while the majority of low market areas/Southwestern Virginia are shaded blue.

The bivariate map should be taken with a grain of salt. Using this map ignores in-situ correlation between the two variables (https://geodacenter.github.io/workbook/6c_local_multi/lab6c.html#bivariate-local-moran).

Takeaways

Virginia is highly autocorrelated in terms of rent and income - almost all Global Moran I values are found near 0.7. In almost every cluster, whether it be income, rent, or both, high valued census tracts were attracted to other high values while low valued census tracts were attracted to other low values. High valued regions were predominantly found in suburban regions, especially Northern Virginia, while low values regions existed across the map, especially in the Southwestern area (and within cites as well). To summarize, wealth and rent affordability vary significantly by location, and regional disparity is very prominent in Virginia.