How Are Derelict and Vacant Land Sites Spatially Related to Deprivation in Glasgow?

Aims

Deindustrialisation in Scotland accelerated during the second half of the twentieth century, marked by the collapse of key industries such as coal mining, shipbuilding, steel, and heavy engineering. This decline left a substantial legacy of neglected land across the nation that coincided with significant economic and social shifts, particularly within the Central Belt, which was once a highly industrialised region of the country.

Focusing on Glasgow City, this project investigates whether derelict and vacant land sites are spatially associated with patterns of socioeconomic deprivation, and whether the regeneration of such sites has the potential to influence deprivation outcomes in nearby areas over time.

Distribution of Derelict and Vacant Land Sites Across Scotland

National Spatial Pattern

Derelict and vacant land sites are unevenly distributed across Scotland, with a clear concentration in the Central Belt, highlighted by the outlined purple boundary.

How Does Glasgow Compare to Other Scottish Local Authorities in Terms of Deprivation?

Glasgow City is one of Scotland's most deprived areas. Based on Overall Scottish Indices of Multiple Deprivation (SIMD) Rankings, it has the second-highest share of its population living in data zones classified within the three most deprived deciles (deciles 1 to 3).

Is Deprivation Spatially Associated with Distance to Derelict and Vacant Land in Glasgow?

Across the four of the five observed SIMD domains, higher levels of deprivation are seen in data zones located closest to derelict and vacant land sites, as shown by the clustering in the bottom-left of each plot. A weak but consistent positive gradient between distance and SIMD rank across those domains suggests that proximity to derelict or vacant land is associated with deprivation in Glasgow.

Given A Data Zone's Deprivation Profile, Can We Predict its Distance to the Nearest Derelict or Vacant Land Site?

Hypothesis: A data zone's deprivation profile can predict its distance to the nearest derelict or vacant land site.

Interpretation: Deprivation indicators predict distance with weak accuracy (R² ≈ 20%), indicating some predictive power but still significant unexplained spatial variation.

While the regression model shows that deprivation profiles only weakly predict a data zone's proximity to derelict or vacant land sites, investment in regeneration is a tool that has been widely used by policymakers to improve social and economic outcomes in Scotland.

Does Regeneration of Derelict and Vacant Land Affect Nearby Area Deprivation Over Time?

Between 2012 and 2014, a previously neglected industrial brownfield site in Glasgow was repurposed through a regeneration project to form the Athletes' Village for the 2014 Commonwealth Games. Following the Games, the site was further developed into permanent housing and green spaces for public use. Using SIMD data from approximately two and six years after the site became publicly accessible, the chart below examines changes in median deprivation levels across data zones at varying distances from the site.

How Do Deprivation Outcomes Compare Across Domains and by Distance from Athletes' Village?

The heatmap below provides a comparative view of changes in deprivation outcomes across SIMD domains by distance band, defined by proximity to the repurposed derelict site, by subtracting 2016 median SIMD ranks from those in 2020. It reveals that changes in deprivation levels are not uniform across domains, and that the relative magnitude of improvements varies across distance bands in ways that differ from one domain to another.

Conclusions

Glasgow is one of Scotland’s most deprived cities, with a high concentration of derelict and vacant land rooted in its industrial history. While deprivation and the presence of such sites frequently coincide, proximity alone does not fully explain Glasgow’s high deprivation levels, as reflected in the weak predictive power of the supervised learning model (R² = 20.2%).

Analysis of the Athletes’ Village regeneration project reveals no clear short-term improvement in deprivation outcomes in surrounding areas. However, socioeconomic change often unfolds over longer time horizons than those observed here, and these findings should not be interpreted as representative of all regeneration or repurposing projects.

Overall, the results suggest that deprivation levels are shaped by a range of interacting factors beyond simple proximity to neglected sites. As such, when seeking to improve economic and social outcomes through regeneration initiatives, policymakers should combine physical redevelopment with nuanced, context-specific interventions to ensure that benefits are effectively targeted and sustained.

Appendix

Challenges

Challenge 1: Data availability and access


One of the main challenges I faced when conducting this project was data availability. Relevant datasets were often difficult to find in the first place: files were inconsistently named, and links to datasets would frequently redirect to error pages. In one case, a government website hosting a dataset I was working with crashed in December, which meant I had to source new data midway through completing the project.

Challenge 2: Defining the Central Belt spatially


When trying to highlight the Central Belt region of Scotland in Chart 1, I could not find any standard or official GeoJSON definition of the Central Belt. As a result, I created the highlighted boundary manually by combining planning authorities commonly identified as part of the Central Belt, based on publicly available regional maps. This boundary should therefore be viewed as an approximate representation rather than a precise regional definition.

Challenge 3: Access to derelict and vacant land data

The derelict and vacant land site data used in this project come from the Scottish Vacant and Derelict Land Survey and were cleaned and processed in Python. I filled out the relevant form to request access to the dataset's API, but unfortunately received no response. My analysis therefore relies on a downloadable CSV file of the 2024 site register. If this project were to be reproduced in the future, newly recorded sites uploaded after this date would need to be added manually in order to reflect the most up-to-date data.

Challenge 4: Changes in Data Zone geographies over time

Another major challenge was the change in Data Zone geographies between SIMD releases. SIMD indicators from 2004 to 2012 use Data Zones based on the 2001 Census, whereas SIMD 2016 and 2020 use Data Zones based on the 2011 Census. Similarly, the TopoJSON file used to map Glasgow’s Data Zones corresponds to the 2011 geography. Although both the SIMD indicators and the boundary files refer to “Data Zones", the fact that they originate from different census years means that they use non-overlapping Data Zone codes and could not be directly joined. I looked extensively for a Glasgow TopoJSON file using 2001 Data Zone boundaries, as I had initially hoped to conduct a complete time series spatial analysis from 2004 to 2020, but I was unable to find one nor any clear guidance on how to map Data Zones across different census geographies.

Data Sources
Colab Notebooks
Academic Integrity Statement

Generative AI tools were used to aid in troubleshooting and assist with CSS styling. All analysis, interpretation, and conclusions are my own.