Being near other firms helps — but how near is near enough?
How Close is Close? The Spatial Reach of Agglomeration Economies
Stuart S. Rosenthal, William C. Strange (2020) · Journal of Economic Perspectives 34(3)
Replication data ↗ (openICPSR)
The archive requires an account we do not hold, so the files were not opened. What follows rests on the published abstract and on this platform's own observations.
What the authors found
Rosenthal and Strange review the evidence on how agglomeration benefits attenuate with distance. The effects operate at several spatial scales at once — neighbourhood, metropolitan and regional — and they fall off sharply: much of the benefit of proximity is exhausted within a few miles, sometimes within a few blocks. Which scale you measure at therefore determines what you find, and a regional average can miss an effect that is real at street level.
What our data adds
The finest geography in this database is NUTS-3, and NUTS-2 for regional GDP — coarse against the neighbourhood scale where the authors say much of the action is. What that view does show is how unequal the outcome of agglomeration is, and how much the answer depends on the measure. Across 276 EU NUTS-2 regions in 2024, GDP per head runs from 127,000 euro in Luxembourg to 6,000 in Şanlıurfa-Diyarbakır, a ratio of 21 to one, with a Gini of 0.301. Convert to purchasing power standards, which strips out the fact that expensive places are expensive, and the ratio falls to 9 to one and the Gini to 0.215. Half the apparent gap between European regions is price levels rather than output.
NUTS-2 regions. Compare with the purchasing-power version below.
Adjusting for price levels roughly halves the measured spread.
| Figure | Value | Indicator |
|---|---|---|
| EU NUTS-2 regions with GDP per head | 276 | NAMA_10R_2GDP.EUR_HAB |
| Luxembourg, GDP per head | 127,000 EUR | NAMA_10R_2GDP.EUR_HAB |
| Şanlıurfa-Diyarbakır | 6,000 EUR | NAMA_10R_2GDP.EUR_HAB |
| Gini across regions, per head in euro | 0.301 | NAMA_10R_2GDP.EUR_HAB |
| Same, in purchasing power standards | 0.215 | NAMA_10R_2GDP.PPS_EU27_2020_HAB |
Every figure here is computed from observations in this database and can be charted, downloaded or queried.
What this cannot show
An earlier version of this note reported a Gini of 0.518 and a 530-fold gap between the largest and smallest region. Those figures were computed on total regional GDP, which measures how big a region is rather than how well off it is, and they have been replaced now that regional population is loaded. Keeping the correction visible seems more useful than quietly restating: a plausible number computed on the wrong denominator is the most common way a regional statistic misleads. Nothing here reaches the neighbourhood scale the paper is actually about.