What do Chicago, Paris and Los Angeles have in common?
What do Chicago, Paris and Los Angeles have in common? Alex Anas Professor of Economics State University of New York at Buffalo Public Lecture New Economic School Moscow, Russia February 18, 2013
• Chicago, Paris, L.A. have developed differentlybecause of history and initial conditions• But they are shaped by the same economic processes
• The same economic model can be used to study all three places• A Computable General Equilibrium model based on theory and econometrics can be applied to any metropolitan area• To use these models, we just have to recognize that the different areas are shaped by the same processes
The RELU-TRAN Model (Regional Economy Land Use and Transportation)• There is a working version for the Chicago, MSA and the Greater Paris Region• There is a current project underway that will apply the model to the Greater Los Angeles Region.
STARTING POINT p, w, R,V,S G, g RELU-TRAN CYCLE RELU RELU LOOPS CONVERGEDUpdateG and g RELU TRIPSfor next cycle TRAN TRAN ITERATIONS CONVERGED G and g converged? p, w, R, V converged? Excess demands, profits converged? YES RELU-TRAN CYCLES CONVERGED Cyclical linking of the RELU and TRAN algorithms in RELU-TRAN
START POINT p, w, R, V, S, G, g RELU LOOP PRICES, p ( w, R ) p OUTPUTS, X ( p, w, R, S,V ) X WAGES, w ( p, X, R,S,V )w Updatep, w, R, V for next loop RENTS, R (p, X, w, S,V) R VALUES, V RV STOCKS, S VS NO YES p, w, R, V converged? Excess demands converged? Economic profits converged? RELU loops converged The RELU algorithm
RELU TRIPS AUTO MODE CHOICE PROBABILITIES ROUTE CHOICE & NETWORK EQUILIBRIUM FLOWITERATIONS CONVERGED CONGESTED HIGHWAY LINK TRAVEL TIMES ZONE-TO-ZONE EXPECTED TIMES & COSTS G and gTRAN The TRAN Algorithm
DecisionsDecisions are hierarchically linked and involve discrete as well as continuous choices Consumer Workplace- residence Voluntary locations unemployment • Labor supply / leisure All choices on left • Commuting mode choice / vehicle choice apply except those in red • Housing (quantity / type) • Vehicle ownership (quantity / type) • Discretionary travel pattern to obtain goods and services Where to go ? Where not to go ? How many trips per period ? How much to spend ? Mode choice / vehicle choice on each trip
A mix of discrete andcontinuous choices Discrete choice of Working/not workingEnter labor market Stay out of labor market Discrete choice of triplet: i: residence zone j: workplace zone (i,j,k) k: type of housing Discrete choice of mode for commuting Auto Transit Continuous variables•Floor space of type k in residence zone I•Labor hours of work supplied to place of work at j• Number of non-work trips and their destinations and modes• Quantity of goods purchased on non-work trips
FIRMS BUILDING LABOR TYPES TYPES PRODUCTION FUNCTION OUTPUT INTERMEDIATEINPUTS FROM OTHER INDUSTRIES
Congestion and urban development• Many issues can be studied using a model such as RELU-TRAN• Today I will focus on one issue mainly:• How does urban development respond to increases in traffic congestion in Chicago, Paris, Los Angeles
Congestion’s effect on urban development depends strongly on:• 1. How much public transit is available• 2. How spread out geographically are the jobs in the urban arae • 3. How decentralized trips are
1. Use of public transit varies• Los Angeles 2%• USA (average) 4.9%• Chicago 13%• Greater Paris 50%
2. Job concentration varies• Greater Paris 50% in the core (City of Paris & CDTs)• Chicago 30% in 4 job centers• Los Angeles 30% in 30 job centers
3. Trips are decentralized and getting more so• Most travel occurs in the suburbs• Suburban to suburban travel is the most rapidly increasing• This effect is bigger when public transit is less available
US & Canadian Commuting Patterns United States Canada 2001Residence Workplace 2000 Census (%) Census(%)Central Central city city 27.5 46.1Central city Suburb 8.9 7.5 CentralSuburb city 20.2 16.2Suburb Suburb 43.4 30.2 100.0 100.0
Commute time increaseswith city size in the US
Doubling population increases commute time by 10%URBAN AREA WORKERS AVERAGE COMMUTE LOUISVILLE 0.5 million 22.7 minutes PITTSBURG 1.0 million 25.5 minutes HOUSTON 2.0 million 28.8 minutes CHICAGO 4.0 million 31.0 minutes NEW YORK 8.0 million 34.0 minutes• New York has 16 times more workers than Louisville but only 50% higher commute time
Avoidance behavior: workers• Workers avoid congestion by 1) Switching from driving topublic mass transit; 2) Relocating their homes closerto their jobs; 3) Other
Avoidance behavior: firms• Firms respond to congestion by 1) Paying higher wages to attractworkers; 2) Relocate closer to workers andcustomers;
Combined effects of workers and firms• Workers’ response increases housing and job density in the centers of cities• Firms’ response spreads jobs to less congested outlying areas and makes more urban sprawl
Household Firm Access to jobs/shops Household Firm Access to labor and customersLinkages between firms and households
Urban sprawl in the US• One way in which an urban area reduces congestion is by sprawling• Because public transit is not plentiful and easily accessible• US urban areas have adjusted to congestion by jobs moving out to suburbs
How much urban sprawl has happened in the US?• 1972 to 1996: the U.S. urbanized land has sprawled at a rate of 2.48% per annum (2.5 times the 0.98% growth rate of urbanized population)• An example follows about the Buffalo- Niagara Falls area in which I live
Chicago’s congestion1• How does congestion affect public transit, urban sprawl and travel behavior in Chicago?• What would be the effects of a London type or Stockholm type cordon policy in Chicago?
Levels of network and zonal aggregation CITY (5 zone) SUBURBS (9 zone)CHICAGO 14+1 ZONE TEST VERSION Larger Chicago 111+6 zones
Central Business District Lake Rest of City of ChicagoMichigan Inner ring suburbs Outer ring suburbs Exurban areaxban area The Chicago MSA
Real Estate Growth (2000-2030) Single family houses Other buildings Land available for development
Change in Aggregate and Per Capita VMT (Without Highway Capacity Additions )
Change in Aggregate and Per Capita VMT (With Highway Capacity Additions) Aggregate VMT Per capita VMT
The Constancy of Commuting Time by Car Despite Population Growth and Increasing Sprawl Percent commuting by car Minutes of two way commuting
Other results from Chicago• Do not add any road capacity more congestion, more sprawl, less VMT.• Improve transit travel times (5% per decade) less sprawl, less congestion but slightly, more transit ridership, centralization.
• Stable gasoline prices more intra-zonal trips, more non-work trips, more sprawl• Improve car fuel economy (4% per decade) Similar to gas price, but not as strong.
How will rapid rail investments2 affect the Grand Paris Region by 2035?• The RELU-TRAN model was used for the “Ile de France” to model the effect of an estimated 35 billion Euros in planned rapid rail investments.
3 What would be the effects of higher congestion in Los Angeles? • 2% of the trips are by public transit • 30% of the jobs are in 30 sub-centers • It is the 2nd largest metro area in the US but has lower than expected travel times
Commute time increaseswith city size in the US