Economic complexity analysis: a policy framework - Neffke

Presentation by Frank Neffke, Team leader of the Science of Cities program, Complexity Science Hub, Austria at the 26th OECD Spatial Productivity Lab meeting "Smart moves: How data can help discover what regions could be great at" held online on 23 October 2025.

More info: https://oe.cd/6dd

1. Economic Complexity Analysis
A policy framework
Meeting of the Spatial Productivity Lab
October 23, 2025
Frank Neffke
Complexity Science Hub

2. An economic complexity world
Three central assumptions
1.Many input factors or “capabilities” (Hausmann & Hidalgo, 2011)
2.Strong complementarities between these capabilities (Gomez-
Lievanoet al., 2016)
3.Complexity increases with the number of and complementarities
between capabilities
Metaphor: Scrabble
•Capabilities are like letters
•Products are like words

3. PRODUCTS
Top Hat: By Nikodem Nijaki(Own work) [CC BY-SA
30
COUNTRIES CAPABILITIES

4. The world as we observe it
COUNTRIES PRODUCTS

5. The world as we observe it
COUNTRIES PRODUCTS

6. The product space
COUNTRIES PRODUCTS

7. metallurgy
vegetables
fishing
garments
coffee and cocoa
products
fruits
7

8. Emerging related diversification literature
•Principle of relatedness: Hidalgo et al. (2018)
•Products in export baskets: Hidalgo/Klinger/Barabasi/Hausmann 2007 –Science; Bahar/Hausmann/Hidalgo 2014 –
JoIE
•Industries in regions: Neffke/Henning/Boschma2011 –EG, Delgado/Porter/Stern 2014 – ResPol; Zhu/He/Zhou 2017 –
JoEG, …
•Innovation in cities: Rigby 2015 –RS; Boschma/Balland/Kogler2017 –ICC; …
•Science fields in cities: Boschma/Heimeriks/Balland2014 –ResPol; Guevara/Hartmann/Aristarán/Mendoza/Hidalgo
2016 –Scientometrics
•Methodological papers
•Li & Neffke (2024 - ResPol)
•Beyond RCAs: Van Dam et al. (2023 –JoRS)
•“Genotypic” approach to ECA: inferring the Ppaand Ccamatrices (Diodatoet al., 2024 –work in progress)
•Policy frameworks (e.g., smart specialization)
•Balland/Boschma/Crespo/Rigby 2018 –Regional studies
•Rigby/Roesler/Kogler/Boschma/Balland2022 –Regional studies
•Green product spaces: e.g., Mealy and Teytelboym2022 –Research Policy
8

9. The real resolution of economic activity
Metroverse

10. From product space to industry space
Dun & Bradstreet World Base
•Second approach
•Dun and Bradstreet World Base
•Large-scale global company registry
•Entire economy, not just exported products
•Number of establishments in dataset:
•2011 batch: 124,816,570
•2016 batch: 121,019,655
•2019 batch: 148,315,749
•2020 batch: 170,000,618
•Variables:
•Consistent industry codes: up to 6 NAICS (2019, 2020) or SIC (all batches) codes
•Geocoded
•Estimated number of employees
•Total sales (not fully reliable)
•Domestic and ultimate global owner
Yang Li

11. City-industry level aggregates
City definitions taken from EU’s JRC
•Establishments are merged through spatial joins to city polygons
•Currently, we use the urban core definition, others can be
implemented
Middlesex County
boundary
Boston urban core boundary
Boston functional urban area boundary

12. Industry space
(co-occurrences in establishments)

13. Understanding a city’s specialization
and its “adjacent possible”
New York City Chicago

14. Shenzhen-Guangzhou Hong Kong
Understanding a city’s specialization
and its “adjacent possible”

15. Software: Stack Overflow
Open resource for support in coding
•Established 2008
•Alexa top 100 website at its peak.
•6M users
•Locations (50k+)
•280+ programming languages
•Content
•Questions (23M)
•Answers (33M)
•60k Tags

16. •In the 2022 snapshot, out of ~6 million active users we can
geolocate:
•4.0 million to a country.
•2.8 million to a FUA.

17. Software task space

18. Atlases of economic complexity
Regions’ position in abstract activity spaces
Metroverse
(Vienna)
Software task space
(Data Science in Europe)

19. ECA and policy
If it’s so easy to develop this, why hasn’t the region done so yet?
•Possibility 1:
•The region hasn’t found the time to enter the activity yet
•Possibility 2:
•The region has recently changed a lot and only now has come close by
•Possibility 3:
•There are (observable or unobservable) barriers to entry

20. ECA and policy
Anomaly detection (Li and Neffke, 2024)
•Sometimes,
•opportunities are real opportunities
(e.g., guided missiles in Seattle,
home to Boeing)
•In other cases,
•they are an expression of a missing
crucial capability (e.g., deep sea
transportation in landlocked
Arizona)
•Use anomalies to learn about the
missing capabilities,
•complementing ECA with in-depth
analysis of a region ECA as part
of a Growth Diagnostics framework

21. Conclusion
•Using ECA to inform regional development policy
•Diversification trajectories are predictable
•Study the “adjacent possible”
•Three ways to apply ECA in development policy
•Map a country’s options
•Industry spaces: map capability base
•Complexity: benchmark development against productive portfolio
•Find “low-hanging fruit” of nearby, un(der)developed activities
•Easier: weed out targets based on wishful thinking
•Best combined with smart objectives: good jobs, green, digital, etc.
•Anomaly detection
•Understand what is hindering seemingly easy development trajectories
•Integrate into wider diagnostics framework