Discussing ‘Returns to Experience Across and Within Cities’ by Don Davis, Eric Mengus, Tomasz Michalski, Elie Vidal-Naquet

Sciences Po Spatial Workshop 2025

Florian Oswald
florian.oswald@unito.it

UniTo, Collegio Carlo Alberto, Sciences Po Paris

23 June, 2025

Summary

  • Agglomeration effects are a key pillar of urban economics.
  • Some say they are very the reason why cities exist.
  • There are many studies documenting their benefits.
  • There are handful documenting the costs of agglomerations.

This paper

  • Are agglomeration benefits uniformly distributed over space? Does everyone benefit to the same degree?
  • If not, who does?
  • Find that benefits depend on the kind of job a worker has, and where (in the city!) that job is performed.
  • The magnitudes of the effects are striking.
  • So, there is extreme inequality in how those benefits acrue to people.
  • This raises huge issues for urban planning.

Figure 1

Comments

Are we really talking about the same kinds of jobs?

jobs in center are different than outside

Does high density make those jobs as good as they are or does their high productivity merely correlate with high density? Would we recommend a policy that forces jobs to be clustered together very strongly in space?

👉 I don’t think we can tell from this paper, but I think it’s a relevant question.

Comments

The Regression

how long a given individual worked as a nerd in a large city like Lyon before moving to a learning job in the Paris complement.

  • how many such transitions do we actually observe? would be interesting to show.

Comments

Another Regression: Bonhomme, Lamadon, and Manresa (2019, BLM)

BLM Static

BLM Dynamic

Comments

Housing Policy

The authors allude briefly to housing policy. The merits of such policies hinge critically on findings such as those that come out of this study. If the benefits of the big city are tangible only for a very small group of people, well then maybe we should not incentivize so many of them to move there in the first place.

Comments

Hours worked per km2

The most productive jobs may actually be less hours intense (by the very definition of being highly productive). So the measure of density might be a lower bound.

Exclude public sector employment? 21% of total employment.

What’s the role of the public sector in this? Do they not benefit (or generate) agglomeration effects? I think of the important R&D activities provided by public institutions in France.

Comments

Is this a French result?

france - Paris is peculiar. it’s a highly concentrated country: top heavy. Do you think the effect would be smaller in other settings?

The Job characteristics measure is not super precise?

2-digit PCS and 2-digit NACE sector is not overly granular. Could we not do a BLM type of clustering of firm types into classes?

Summing up

  • Striking result.
  • Very careful data work.
  • I would encourage thinking more about the underlying model of wage determination.
  • Looking forward to see the end product!

References

Bonhomme, Stéphane, Thibaut Lamadon, and Elena Manresa. 2019. “A Distributional Framework for Matched Employer Employee Data.” Econometrica 87 (3): 699–739.