Good Data Is Hard to Find

Why this matters and what to do about it

Rachel Franklin

Center for Geographic Analysis
Harvard University

Dani Arribas-Bel

Geographic Data Science Lab
University of Liverpool

2026-03-18

The second biggest urban problem

Our contemporary urban data ecosystem isn’t designed to address wicked urban challenges. (in fact, it’s not designed at all)

This talk

  • Some definitions
  • A list of urban data grievances
  • An illustrative vignette (12 minutes!!)
  • Identifying the solution space
  • Takeaways

So many choices of “big urban problems”

Why focus on the data?



Old methods, new methods, big questions, little questions, causal, descriptive…


…it all depends on data

What do we mean by urban data?

  • Social media
  • Mobile phone
  • Streetview
  • Satellite imagery
  • Sensor
  • Micromoblity and public transit
  • Volunteered
  • Municipal civic
  • Traditional survey and census

The airing of urban data grievances

1. Accidental and incidental

2. Resolution-challenged (temporal and spatial)

3. Access and ownership

4. Policy poor (causal mechanisms, evaluation)

5. Inference fetishization

6. The “Demographic Constraint Problem”

7. Expensive opportunity cost

The solution space

A big-picture observation

This situation isn’t really anything new; we’ve hit other data-infrastructure impasses/inflection points

  • Creation of the American Community Survey (ACS)
  • Laidlaw and neighborhood-level data
  • Panel or longitudinal surveys (like PSID)

Two kinds of solutions

1. Data ecosystem

  • Pause. Deep breath. What data do we need?

  • Data observatories

  • A new survey advent

  • They go big (data); we go small (data)

  • Just because we can, doesn’t mean we should

2. Research ecosystem

  • Fund the data

  • Slow our (research) roll

  • Hold ourselves to a higher theory and policy standard

  • Articulate our data expectations

Takeaways

  • Data infrastructure is its own exciting urban challenge.
  • We should be designing our data to meet our needs–not designing research questions around available or new data.
  • We have some agency.
  • F**k data inference (can I say that?)

The End.

Thank you!!