Urban Networks:
Connecting Markets, People, and Ideas
Edward L. Glaeser1
Giacomo A. M. Ponzetto2
1 Harvard
2 CREI,
Yimei Zou3
University
UPF, IPEG and Barcelona GSE
3 UPF
and Barcelona GSE
World Bank Conference
The Rise of the South at a Crossroads
A View from East Asia and Latin America
Monday 16 May 2016
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
1 / 18
Urban Networks
Introduction
Network Formation and Urban Growth
Do megacities outperform smaller but tightly connected cities?
What are they key determinants of their di¤erential performance?
A pressing question for Europe
I
I
We love our historical cities and are invested in their preservation
Are we paying for our love of history?
A pressing question for Asia and Latin America
I
I
How big a megacity is too big, especially in a developing country?
Renewed interest in dispersion and high-speed transit
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
3 / 18
Urban Networks
Introduction
Conceptual Distinction
City: Absence of space between people
1
2
3
Perfect mobility of workers across …rms
Integrated single pool of entrepreneurial ideas
Commuting to work and to man-made amenities
Network: Distinct but connected cities
1
2
3
Mobility of skilled but not unskilled workers
More diverse but less free-‡owing ideas
Separate markets for housing and amenities
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
4 / 18
Urban Networks
Introduction
City Growth
Network Expansion: Add cities to the urban network
Greater …rm creation, higher income, higher welfare
But we don’t model the cost of network infrastructure
Consolidation: Tighten network links to create a megacity
Loss of diversity: innovation may decline
But consumption opportunities improve
Densi…cation: Concentrate people but lose land
Consolidation with scarcity
The elasticity of housing supply is key
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
5 / 18
Urban Networks
Three Building Blocks
Entrepreneurship
Firm creation through recombinant growth
I
Jacobs (1969) and Weitzman (1998)
Skilled workers (Hc ,t , Hr ) and old …rms Nc ,t
µ
Nc ,t = αc ,t Nc ,t
I
I
1
create new …rms Nc ,t
η
ν
1 Hc ,t Hr
City-speci…c productivity αc ,t
Elasticities µ, η and ν with µ + η + ν > 1
Localized entrepreneurial knowledge spillovers
1
2
η
Within city c only: Hc ,t
Within the whole urban netork r : Hrν
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
6 / 18
Urban Networks
Three Building Blocks
Endogenous Consumption Amenities
A resident i of city c has utility
1 γ
ui ,c ,t = hiτ,c ,t qi ,c ,t i
I
I
τ
(gi ,c ,t Gc ,t )γi
Consumption of housing hci ,t , numeraire qci ,t , amenities gci ,t
Positive spillovers in consumption: Gc ,t
Amenities are localized in space and persistent over time
1 φ
Gc ,t = Γc ,t Gc ,t
1
(γ̄Yc ,t )φ
Endogenous amenities are skill-biased
I
High-skill share γH > low-skill share γL
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
7 / 18
Urban Networks
Three Building Blocks
Land Scarcity
The source of congestion in our model
Fixed endowment of land Tc in each city
Endogenous investment in building up
Equilibrium housing supply
hc ,t = Ψc ,t Tcζ (τYc ,t )1
I
I
ζ
Productivity Ψc ,t , income Yc ,t
Congestion ζ 2 (0, 1) ) housing supply elasticity 1/ζ
Giacomo Ponzetto (CREI)
Rising South Conference
1
16 May 2016
8 / 18
Urban Networks
Spatial Equilibrium
Spatial Structure
A continuum of cities partitioned into regions r = 1, 2, ..., R
I
Region r includes measure Cr of cities
Unskilled workers Lc are immobile
Skilled agents Hr are mobile within each region, not across regions
Skilled agents create …rms, employ unskilled workers, produce output
h
Yc ,t = Yt (Ac ,t Lc )σ
I
I
1
Nc ,t
i σ1
Gains from specialization σ > 1 (Ethier 1982)
Productivity Ac ,t , endogenous aggregate output Yt
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
9 / 18
Urban Networks
Spatial Equilibrium
Sorting
Skilled agents sort given fLc , Tc ; Nc ,t
Stability condition
1 , Gc ,t 1 ; αc ,t , Ac ,t , Γc ,t , Ψc ,t g
σ > η (1 + φγH
I
I
ζτ )
High housing share τ and low supply elasticity 1/ζ 1
Limited spillovers in amenities φγH and …rm creation η/σ
Distribution within region r such that
1 η
Hc ,t
1 ζτ +φγH
σ
h
∝
µ
αc ,t Nc ,t 1
Giacomo Ponzetto (CREI)
(Ac ,t Lc )
σ 1
i1
ζτ +φγH
σ
Rising South Conference
Ψc ,t Tcζ
τ
γH
1 φ
Γc ,t Gc ,t
1
16 May 2016
10 / 18
Urban Networks
Full Symmetry
Cost and Bene…ts of Fragmentation
R identical regions, each comprising Cr identical cities
Fragmentation into separate idea networks (R ") is harmful
I
I
I
Reduced entrepreneurship spillovers ) slower …rm creation
Lower incomes across the board, less (and cheaper) housing
Everyone’s utility falls
Fragmentation of idea networks into separate commuting zones (Cr ")
1
Reduces pecuniary variables i¤ µ + η > 1
F
2
If ν > 1
µ
η > 0 local diversity helps entrepreneurship
Reduces individual i’s utility i¤
γi + (1
F
ζτ + φγi )
µ+η 1
>0
σ 1
Unambiguous loss of access to amenities
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
11 / 18
Urban Networks
Full Symmetry
Cost and Bene…ts of Density
Empirically, all fragmentation is probably harmful
But technology can only go so far in eliminating distance
Concentrating population so that ∂ ln T = ∂ ln Cr < 0
1
2
Increases …rm creation and incomes i¤ µ + η > 1
Increases individual i’s utility i¤
γi
ζτ + (1
ζτ + φγi )
µ+η 1
>0
σ 1
Similar results for concentrating into networks
Housing supply may increase if it is su¢ ciently elastic
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
12 / 18
Urban Networks
Full Symmetry
Long-Run Preference Reversals
Log-linear …rst-order VAR: stationary if µ < 1 ) steady state
If η > 1
µ > 0 fragmentation is more harmful in the long run
Consolidation has larger long- than short-run income e¤ects
If γi < ζτ and σ 2 σid + , σ̄id + there is a preference reversal
∂ ln uti
∂ ln uti
∂ ln ū i
∂ ln ū i
+
>0>
+
∂ ln Cr
∂ ln T
∂ ln Cr
∂ ln T
Drivers of preference reversal
1
Large local spillovers: ∂ σ̄id +
F
2
σid + /∂µ > ∂ σ̄id +
Accumulating income gains
Highly persistent amenities: ∂ σ̄id
F
σid + /∂η > 0
σid
/∂φ < 0
Amenities converge more slowly to their steady-state level
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
13 / 18
Urban Networks
Full Symmetry
Distributive Con‡ict and Preference Reversal
Concentration is always relatively regressive (in welfare terms)
I
I
The skilled care more about greater access to amenities
The unskilled care more about cheap housing
Preference reversal by the unskilled increases consensus
I
I
I
In the short run, they dislike concentration because prices rise
But in the long run, income rises and so does housing supply
They come to agree with the skilled that density is good
Preference reversal by the skilled decreases consensus
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
14 / 18
Urban Networks
Heterogeneity
Regional Heterogeneity and Coordination Failure
Heterogeneous regions, composed of identical cities
I
Region-speci…c productivities and initial endowments
Interregional bene…ts of …rm creation
∂ ln Yr ,t
1
Ys ,t
=
> 0 for all s 6= r
∂ ln Ns ,t
σ ( σ 1 ) Yt
Coordination failure for µ + η > 1 and γi < ζτ: insu¢ cient density
Free-riding is a temptation for small regions: Yr ,t /Yt < ŷdi +
I
I
I
Exacerbated by a large, inelastic housing sector: ∂ŷdi + /∂ (ζτ ) > 0
Alleviated by large, transient amenities: ∂ŷdi + /∂γi < 0, ∂ŷdi + /∂φ < 0
And by large local spillovers: ∂ŷdi + /∂η = ∂ŷdi + /∂µ < 0 < ∂ŷdi + /∂σ
Opposite temptation if γi > ζτ (or µ + η < 1): excess density
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
15 / 18
Urban Networks
Heterogeneity
Urban Heterogeneity and Skill Poaching
Let a …nite set of identical cities collapse into one with its original land
Feedback from land and amenities to …rm creation
The number of skilled entrepreneurs rises i¤
γH
ζτ +
η+µ
σ
1
(1
ζτ + φγH ) > 0
Firm creation, output and wages rise i¤
µ + η (1
I
ζτ + γH ) > 1
More stringent than before i¤ ζτ > γH
Unskilled workers’welfare rises i¤
γL
I
ζτ + (1
ζτ + φγL )
µ + η (1 ζτ + γH ) 1
>0
σ η (1 ζτ + φγH )
Triple cost of rising house prices
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
16 / 18
Urban Networks
Heterogeneity
Gains from Reducing Heterogeneity
Finite set of m cities with identical land endowments
Heterogeneous endowment of factors (Lc , Nc ,t
1
Size heterogeneity: empirical distribution of shares
sk = θ ŝk +
I
2
1 , Gc ,t 1 )
1
θ
m
Combination of a non-degenerate distribution and equal shares
Heterogeneous factor proportions
I
I
Fraction ω have i.i.d. shares of each factor
Fraction 1 ω have perfectly correlated shares of all factors
If consolidation is attractive for (θ, ω ), it is for all θ 0
Giacomo Ponzetto (CREI)
Rising South Conference
θ and ω 0
16 May 2016
ω
17 / 18
Urban Networks
Conclusions
Thinking about Urban Networks
Key di¤erences from a megacity
1
2
3
More diverse but less connected entrepreneurs
Less access to amenities
Probably less congestion
Focus on …rm creation
I
I
Lower in the network if there are local increasing returns
Most important in the long run, undervalued in the short run
Key role for housing supply elasticity
I
I
I
If you cannot build up you’ll hate density
Insu¢ cient density from coordination failure
Insu¢ cient density to attract the skilled
Density exacerbates distributive tensions
Hopefully a useful framing for future empirical work
Giacomo Ponzetto (CREI)
Rising South Conference
16 May 2016
18 / 18