Census data tool proposal

Dustin Carlino

Census data for any boundary polygon

What

• Web tool: Draw a boundary, get demographic breakdown of people living / working there
• and an API: from any language, give a boundary polygon, and get demographic breakdown
• Worldwide scope, “pay as you go” model for data

Why

• How many people live near a bus stop or bike-share station?
  • Between 18-30, employed, not owning a vehicle…
• Who lives inside a proposed low-traffic neighborhood, or nearby where traffic is expected to detour in the short-term?
• Are air-quality sensors equitably distributed? https://github.com/alan-turing-institute/spatial-inequality
• Input into synthetic population pipelines

Many researchers reinventing the same wheel – why?

How (the data)

• One pipeline transforming region-specific data sources into one format
  • US census (census block)
  • UK census (output area)
  • Brazil (enumeration area)
• Use the smallest geography available
  • and intersect with the input boundary area, scale numbers accordingly
• WorldPop fallback

How (the code)

• idea: store in one FlatGeobuf file
  • Proof of concept for US census blocks by Michael Kirk
  • HTTP range requests to cheaply fetch what’s needed (3-4MB over 8-10 requests for one small city)
  • no hosted API to productionize; just a file!
• Rust client library
  • Bindings for any language, cross-platform, WebAssembly, Jupyter notebooks

Alternatives considered

• Commercial: Remix, HASTUS
• Transitland
• Bespoke libraries for certain regions
  • geobr
  • nomisr
• IPUMS