Old, New, Borrowed and BLUE?

Rachel Franklin (@rsfrankl)

Center for Geographic Analysis | Harvard University

2025-11-13

What do we talk about when we talk about “spatial economic analysis”?

An age-old question:
“how do we know what we know?”



What we mean by Spatial

  • Role of location or place
  • Geographical variation or disparities (across cities, regions, or countries)
  • Context
  • Spillovers
  • Interactions
  • Networks


…and as statistical artifact

What we mean by Economic

  • Obvious topics like economic growth and development
  • Roles of economic sectors, firms, innovation, and entrepreneurship


…but also well-being, quality of life, and human capital

What we mean by Analysis

“Analysis” comes in lots of flavours

  • Empirical or case studies
  • Descriptive
  • Spatial analytical or geocomputational
  • Machine learning and artificial intelligence


…or a marriage of spatial economics and economic geography
(Spatial Economic Analysis flavour) – so, fairly econometrics orientated

Another way of thinking about analysis



If we’re talking about the SEA marriage, how about old, new, borrowed, and BLUE

Old (your mother’s spatial economic analysis tools)

  • Input-Output
  • Computable General Equilibrium Models (CGE)
  • Shift-share analysis and location quotients
  • Spatial interaction/gravity models
  • Spatial autocorrelation/regression

New(er)

  • Bayesian approaches
  • Instrumental Variables (IV)
  • Geographically Weighted Regression (GWR)
  • Advanced Spatial Econometrics
  • Genetic algorithms
  • Neural networks
  • Clustering (e.g., k means)
  • Microsimulation

Borrowed (your PhD student’s spatial economic analysis tools)

  • Machine learning/Artificial Intelligence (AI)
    • Natural Language Models
    • Large Language Models (LLMs)
    • Computer vision
    • Foundation models

BLUE (Best Linear Unbiased Estimator)

  • Ordinary Least Squares (OLS)



Old school (not much else to report on this)

A little bit of revealed preference

(for a treat)

Where better to look than our journal, Spatial Economic Analysis

  • Founded in 2006

  • With hundreds of papers now published (428)

  • Good proxy for what we consider to be spatial economic analysis

Bibliometric analysis and topic modelling

  • Similar to qualitative analysis but with an entire corpus

  • And automated

  • We use a Large Language Model (LLM–Gemma-3-27B-Instruct)

  • Our corpus is all SEA papers with the exception of virtual special issue papers and issue editorials

So what does a machine think spatial economic analysis is?

And how has this changed over the past 20 years?

Closing thoughts: left behind methods?

How it started

How it’s going

Union versus intersection

What do we think about academic fashions in methods (and topics)?



It’s one thing not to re-invent the wheel. What about a vehicle that never really changes its wheels?



Not about methodological substitution, but rather complementarity

Where are the new forms of data?



Like web-scraping, crowdsourcing, GPS, and satellite imagery

Evolution as survival mechanism

The only thing constant is change.

The End.