Why have big cities pulled away from everywhere else since the 1990s?
IT and Urban Polarization
Jan Eeckhout, Christoph Hedtrich, Roberto Pinheiro (2025) · American Economic Journal: Macroeconomics
Article ↗ · 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
Eeckhout, Hedtrich and Pinheiro identify differential information technology investment as a key driver of job and wage polarisation across cities. Firms in large, expensive cities invest most in IT; that investment automates routine work, so routine employment declines fastest exactly where IT spending is highest. The result is that big cities shed middle-skill jobs and add high-skill ones, while smaller cities do neither — polarisation is not uniform across space, it is concentrated where the technology went.
What our data adds
European regional GDP per head shows the kind of spread such a process produces. The capital and financial regions sit at the top — Luxembourg at 127,000 euro per head, Ireland's Eastern and Midland at 126,600, Denmark's Hovedstaden at 90,600, Hamburg at 87,500 — against a poorest region at 6,000. In purchasing power standards the order shifts and Prague enters the top five, which is a reminder that some of the gap is what a euro buys rather than what a region makes. Whether IT investment drove any of this is not something these figures can say.
| Figure | Value | Indicator |
|---|---|---|
| Luxembourg, GDP per head | 127,000 EUR | NAMA_10R_2GDP.EUR_HAB |
| Ireland, Eastern and Midland | 126,600 EUR | NAMA_10R_2GDP.EUR_HAB |
| Hamburg | 87,500 EUR | NAMA_10R_2GDP.EUR_HAB |
| Poorest region | 6,000 EUR | NAMA_10R_2GDP.EUR_HAB |
Every figure here is computed from observations in this database and can be charted, downloaded or queried.
What this cannot show
There is no IT investment data here, no occupational structure and no wage data, which is all three of the paper's key variables. A snapshot also cannot show polarisation, which is a change in the shape of a distribution over time. Ireland's Eastern and Midland region deserves particular caution: its GDP per head is inflated by multinational profit booking, the same phenomenon the profit-shifting notes describe, so it measures where accounts are filed as much as where output happens.
The data behind it
Million euro
NAMA_10R_2GDP.MIO_EUR
WBResearch and development expenditure (% of GDP)
GB.XPD.RSDV.GD.ZS