Scale and Spatial Complexity: Agent-based models and geographical economics December 7 th, 2007 Christopher Fowler and Mark Ellis
Scale Body Household City Region Nation Interscalar Relationships Example: Household racial heterogeneity and neighborhood diversity
Spatial Complexity Spatial economic models often treat space as a distance matrix But space can also be understood in terms of differentiation among places and gradients between them. Agent-based modeling is well-suited to representing space in terms of local differentiation Once space is meaningfully introduced into economic models we find that it has significant implications for model results and, by extension, for policy recommendations
Space in Economic Modeling: Geographical Economics Developed initially by Paul Krugman (1991) extended by many others (Journal of Economic Geography) Identifies economic trajectories of cities with workers and firms moving among cities in search of higher utility (for workers) and profits (for firms) Goods are available in different cities at cost plus a transportation fee. This differentiation, combined with increasing returns to scale, drives the model with a home market effect that induces consolidation of economic activity
A critique of geographical economics Equilibrium is enforced exogenously by linking the movement of firms and workers to maintain optimal proportions Because equilibrium is enforced rather than arrived at, space is reduced to an added cost on certain goods Focus is on distance rather than differentiation
Agent-based Reinterpretation Start from simple propositions (Box A), build towards replication of formal deductive model (Box C) Version 1.0 (Very Box A) No movement permitted (equilibrium relationship between workers and firms maintained). Space as distance
Model Results Conclusion: When space is present only as an additional cost for certain goods firms have no difficulty in defining a stable equilibrium
Agent-based models: Further complications (on the way to Box C) Version 2.0 Introduce the possibility of worker migration Version 3.0 Introduce the possibility of firm relocation Space is still represented very simply no cost for migration, no policies that would advantage or disadvantage one city over another but focus is now on points of disequilibrium that act as drivers for the movement of people and capital
Model Results Model does not have a deterministic outcome. Equilibrium is attainable, but not necessarily stable. Disequilibrium is nearly always a possibility
Space and Equilibrium Simulation Simulation Simulation 1: Workers 3: Workers 2: and Workers and Firms Firms Can Fixed Can Move in Space Move Deductive Model Results 100 100 500 100 90 90450 80400 70350 60300 Steps to to stable Stable equilibrium Equilibrium Steps to stable equilibrium 50 250 (10 (10 turn turn minimum) Simulation 2: Workers Can Move Scenario 3: Workers and Firms Move 2500 2000 40200 30150 2500 2000 Steps to stable equilibrium (100 turn minimum) 1500 1000 500 0 10 20 100 20 10 50 10 0 0 0 13 16 19 22 25 28 31 34 37 40 43 46 Starting Wage 99 75 51 27 3 49 10 2 11 10 10 13 13 13 16 16 16 19 19 19 22 22 22 25 25 25 28 28 28 28 31 31 31 31 34 34 34 34 37 37 37 37 40 40 40 40 43 43 43 43 46 46 3 9 46 15 46 49 21 49 27 33 49 49 39 Number of Workers Starting Starting Wage Wage Starting Wage Steps to stable equilibrium (100 turn minimum) 1500 1000 500 0 Number of Workers 93 93 93 93 75 75 75 75 57 57 57 57 39 39 39 Number Number of of of Workers Workers 39 Number of Workers 10 21 21 21 17 21 24 33 3 31 Starting Wage 38 3 45 51 57 63 69 75 81 87 93 99 45
Spatial Complexity: Implications Increasingly complex spatial representations produce increasingly complex economic systems Scale: As we begin to consider the impact of space, we become increasingly aware of the need for more complex spatial representations at different scales Policy: Complex and multiscalar spatial representation can radically alter predictive model outcomes
Questions?
Model starting conditions Three cities located on an isotropic plane. Cities 1 and 2 are equidistant from City 3 10 units distant City 3 10 units distant City 2 6 units distant City 1
The basic model of geographical economics L L W Y r r r r ) 1 ( γ φ γ λ + = λ α γ ρ β = = 1/(1 1 1 1 ) /(1 1 R s s rs s r W T L I ρ α δ ρβ 1/ 1 1 1 1/ 1) ( = = R s s rs s r I Y T W