Spatial data (and) science

Fundamentals, applications, and frontiers in env. and resource economics

Ed Rubin

University of Oregon (USA)


Spatial data science


In environmental/resource/energy economics, space really matters.

At the heart of a lot of our empirical work: Who is exposed to what?


  • ind/HH decisions/outcomes
  • schools (test scores)
  • power plants
  • governments
  • crops (in a plot)
  • ecosystems, trees, animals


  • (dis)amenities
    e.g., PM2.5, PFAS, lead, fecal col.
  • weather, climate, disasters
  • govt policy (nat, state, local)
  • information, prices, taxes
  • transportation network

Spatial data science


Spatial data science underpins how we connect

  • outcomes (individuals) to
  • treatments (exposures/instruments).

The process typically involves (1) spatial data and (2) assumptions.

Example Estimate the PM2.5 damage function for perinatal health.

  • Birth data Identified at residence… hospital… city… state?
  • PM2.5 Sparse monitors, coarse AOD satellites, limited emissions data
  • More People move; indoor AQ \(\neq\) outdoor AQ; non-clas. meas. err.

So what do people (we) do?

  • match births to nearest regulatory PM2.5 monitor (or IDW?);
  • use satellite-derived PM2.5 predictions (now with ML!);
  • run a particle transport model for upwind PM2.5 sources (+ ML?);

future: proxy daily movements with phone-based movement data …
… or just hope for the best?

Focusing on (regulatory) monitors…