Presentation on theme: "Presented by Mark D. Partridge Swank Professor in Rural Urban Policy The Ohio State University Presented at LaSapienza University Rome, Italy July 12,"— Presentation transcript:
Presented by Mark D. Partridge Swank Professor in Rural Urban Policy The Ohio State University Presented at LaSapienza University Rome, Italy July 12, 2012 Central Place Theory: New Wine for Old Bottles 1
I will summarize several papers. Mulligan, G.F., M.D. Partridge, and J.I. Carruthers. (2012) “Central Place Theory and Its Reemergence in Regional Science.” Annals of Reg. Sci. (48): Chen, A. and M.D. Partridge. Forthcoming. “When are cities engines of growth? Spread and Backwash Effects across the Chinese Urban Hierarchy.” Regional Studies. Partridge, M.D., D.S. Rickman, K. Ali and M.R. Olfert. (2010) “The Spatial Dynamics of Factor Price Differentials: Productivity or Consumer Amenity Driven?” Reg. Sci. and Urban Econ., 40: Partridge, M.D., D.S. Rickman, K. Ali and M.R. Olfert. (2009) “Agglomeration Spillovers and Wage and Housing Cost Gradients Across the Urban Hierarchy.” J. of Int. Econ. 78 (1): Partridge, M.D The Dueling Models: NEG vs Amenity Migration in Explaining U.S. Engines of Growth. Papers in Reg. Sci.. 89: Partridge, M.D., D.S. Rickman, K. Ali and M.R. Olfert. (2008). “Lost in Space: Population Dynamics in the American Hinterlands and Small Cities.” J. of Econ. Geog. 8:
Introduction 4 Economic Geography, Regional Science and Regional/Urban economics have two key models to explain economic geography and the spatial distribution of cities. First is the New Economic Geography (Brakman et al., 2009). Monopolistic Competition with falling long-run average costs and positive transportation costs create a situation in which endogenous growth/decline takes place due to proximity to markets and inputs. We can analytically solve NEG models and develop an urban system. Paul Krugman won the 2008 Nobel Prize in part for the NEG. The JRS’s 50 th Anniversary Issue Points to the NEG as a key reason for the revitalization of Regional Science after World Bank report used NEG for policy advice.
Introduction 5 Second is Central Place Theory of Christaller (1933). CPT is a tiering of urban areas from the hinterlands, to small cities, all the way up to the largest cities based on the order of service/occupation and the market thresholds needed to sustain that service. Larger cities have the fullest range of services and smaller places only have activities with small market thresholds.
6 K=3 Marketing Area Concept. Source Wikipedia.com, CPT’s Hexagons of Central Places
Introduction 7 CPT is criticized for being static and lacking a microfoundation. Nonetheless, Mulligan et al. (2012) report a wealth of empirical evidence suggesting urban systems are organized in this manner. For instance, CPT is consistent with Zipf’s law. One distinction with NEG is that total market potential matters in NEG models, not proximity to large or small cities. NEG is a-spatial in the sense that proximity to different sized cities does not play a role.
Introduction 8 My aim is to show you that CPT does a nice job of predicting where economic activity takes place. CPT does better than NEG in describing location in the US and China. I contend that CPT was abandoned too soon in the 1980s. CPT research was esoteric. Also, because existing GIS was primitive, there were only partial empirical tests of its relevance. Yet, if CPT had held on, GIS would have extended its life. CPT is helpful in guiding policymaking for rural and urban areas in illustrating how economic activity is regional.
What drives spatial patterns? First let me dispel conventional thinking. Popular commentators instead focus on new technologies and globalization, which to them makes space much less relevant. advances in ICT maturing and deconcentration of manufacturing globalization improved transportation This implies that agglomeration economies and cities are less important. Economic activity can occur anywhere. There is not much need for economic geography, spatial economics or regional science. 9
Death of Distance The Rural Rebound: Recent Nonmetropolitan Demographic Trends in the United States “Recent improvements in the transportation and ICT infrastructure... thereby diminishing the effect of distance.” “40 Acres and Modem” (Kotkin, 1998) Cairncross “Death of Distance” (1995, 1997) Thomas Friedman World is Flat 10
What drives spatial patterns? Economic Geographers and Urban and Regional Economists believe distance matters more today. Leamer (2007) describes how distance costs are now having a bigger effect on trade. Namely as services rise in importance, distance becomes more important. Face to face contact vs commodity trade (McCann) Small policy differences matter more in global economy if resources are more mobile (Thisse, 2010). “Regional Science is where it is at” (Partridge, SRSA Presidential Address, 2005). 12
Review of CPT 13 Review Christaller assumptions from Wikipedia: an unbounded isotropic (all flat), homogeneous, limitless surfaceisotropic an evenly distributed population all settlements are equidistant and exist in a triangular lattice pattern evenly distributed resources distance decay mechanism perfect competition and all sellers are economic people maximizing profits consumers are of the same income level and same shopping behaviour Consumers have a similar purchasing power and demand for goods & services Consumers Consumers visit the nearest central places that provide the function which they demand. They minimize the distance to be travelled no provider of goods or services is able to earn excess profit(each supplier has a monopoly over a hinterland)
Christaller’s Assumptions—cont. 14 Therefore the trade areas of these central places who provide a particular good or service must all be of equal size there is only one type of transport and this would be equally easy in all directions transport cost is proportional to distance traveled in example, the longer the distance traveled, the higher the transport cost The theory then relied on two concepts: threshold and range. Threshold is the minimum market (population or income) needed to bring about the selling of a particular good or service. Threshold Range is the maximum distance consumers are prepared to travel to acquire goods - at some point the cost or inconvenience will outweigh the need for the good.
What does this mean for China? Many studies of Chinese Growth Processes. Krugman (2010, subsequently published in Regional Studies) argues that NEG applies more to China than (say) US. In these models, market potential (MP) is not affected by its sources. I will stress Ke an Feser (2010); Chen and Partridge (2011, Regional Studies); Chen (2010); and Groenewold et al. (2007). They use CGE models and econometrics. Use CPT, i.e., it matters what city you are near. 15
Chen and Partridge We first use an aggregate market potential (MP) variable from NEG. It is positively linked to GDP growth, but not job growth. We find that China’s urban growth is positivity associated with GDP throughout the nation, without statistically affecting labor migration. This suggests NEG is a good model to understand China’s regional development. 16
Chen and Partridge We split MP into that from the three coastal mega cities, provincial capitals, and prefecture cities. (see graph) We find evidence of considerable heterogeneity. Having greater MP from the nearest provincial capital has the strongest positive link to per-capita GDP growth in smaller county-urban/rural locales. There are also positive and statistically significant association for the prefecture MP variables. 17
18 Figure 1: Illustration of measuring market potential across the city hierarchy Notes: This map illustrates the market potential heterogeneity across city hierarchy. Lai’an Xian is a county in Anhui province. Chuzhou Shi is Lai’an Xian’s nearest prefecture city. Hefei Shi is Lai’an Xian’s own-provincial capital city. Nanjing Shi is the provincial capital city of Jiangsu province, which is also the nearest provincial capital city of Lai’an Xian. Shanghai Shi is the nearest mega city of Lai’an Xian. MPB indicates the market potential in the mega city. MPC indicates the market potential in the county’s own provincial capital city. MPN indicates the market potential in the county’s nearest provincial capital city. MPO indicates the market potential in the prefecture city.
Chen and Partridge MP from the mega-cities is inversely associated with per-capita GDP growth. Our results are more consistent with CPT, not NEG models. Inconsistent with World Bank (2009) view that urbanization is good for all. Illustrates regional growth process in most of China. Gov’t policies that favor the mega cities may be at the expense of growth elsewhere. 19
Chen and Partridge If balanced growth across the entire country is an objective, growth in the three ‘coastal’ mega cities is detracting from the goal (and may be reducing aggregate growth). Fallah et al. (2010) find that MP is positively associated with individual income inequality, creating further social pressures. We conclude the more nuanced view of growth is correct. This view fits into CPT augmented by Spread and Backwash. NEG is too blunt for policy analysis. 20
21 In advanced economies such as the US. I summarize some work I did with my coauthors including Kamar Ali, Rose Olfert and Dan Rickman. Central Place Theory (CPT) and NEG both hypothesize that small urban areas and rural areas support the growth of large cities: Hinterlands Large Cities Residents in the hinterlands purchase services in large cities—forming the market for cities This describes the origination of the urban system through the early 20 th Century
In advanced economies such as the US. But what is better in a maturing urban system? NEG or CPT? New Information-Communication Technology Improved transportation Technological improvements in natural resource industries—labor saving in rural areas Improved importance of manufacturing and then services in urban areas. Decoupling of place of work and place of residence via commuting. The rise of peri-urban/exurban living. 22
In advanced economies such as the US. These have change the urban system to a mature system: Now causation seems to run much more from city to rural areas for economic growth. 23
Do NEG models explain these developments. 24 NEG models generally predict that falling transport costs imply that there should be more urban concentration. The US has had falling transport costs implying US core urban region should have greatly benefited— especially largest cities.
25 US Relative Transportation and Warehousing Costs Compared to the CPI and GDP Deflator, (2000 = 1) Notes: Transportation and Warehousing producer price index relative to the GDP deflator and Consumer Price Index. Source for the Transportation and Warehousing Producer Price Index and the GDP deflator is the U.S. Bureau of Economic Analysis [downloaded from on February 16, 2010] and the source for the Consumer Price Index is the U.S. Bureau of Labor Statistics [downloaded from on February 16, 2010].http://www.bea.gov/industry/gpotables/gpo_action.cfm on February 16 Source: Partridge, 2010.
Growth By Metro Area Size in 1969 (%) Notes: Large MSA is > 3 million population in There are 8 MSAs in this category: New York, Los Angeles, Chicago, Philadelphia, Detroit, Boston, San Francisco and Washington DC. The Large-Medium MSA have a 1969 population of 1 million - 3 million (27 MSAs). The Small-Medium Metro Areas are 250, million 1969 population ( 85 MSAs). Small MSAs have a 1969 population of 50, ,000 (230 MSAs). 17 Metros with less than 50,000 in 1969 were omitted due to a small base. These were generally in UT, NV, and FL and grew very rapidly. Big metro growth is dominated by Washington DC’s growth. We use 2008 MSA definitions, which makes nonmetro growth appear especially small. Source: U.S. Bureau of Economic Analysis: Source: Partridge, 2010.
Growth For Representative Metro Type (%) Notes: The Traditional Core includes New York, Boston, Philadelphia and Chicago. The Rustbelt includes Detroit, Cleveland, Pittsburgh and St Louis. Sunbelt includes Miami, Atlanta, Phoenix, Tampa, Orlando and Las Vegas. Mountain/Landscape includes Seattle, Denver, Portland, and Salt Lake. Source: U.S. Bureau of Economic Analysis: Source: Partridge, 2010.
28 U.S. Population Growth by State: Miles Population Growth from 1960 to 2008 (%) Map Created on November 16, 2009 Mean=89.1 Median=43.4 Source, U.S. Census Bureau. Source: Partridge, 2010.
Population Growth by County Source: Partridge, 2010.
The winner? 30 Amenity led growth is a clear winner. NEG fares poorly, but what about CPT?
Partridge et al. (2008) Empirical Model to capture CPT Regress county population growth between periods 0 − t on initial-period geographic variables at time 0 DepVar: Cross Sectional: , ; , , total population growth (one fixed-effect panel model) This mitigates endogeneity Key distance variables are exogenous/predetermined %ΔPOP ist-0 =α+ φGEOG ist-l + γAMENITY is +σ s + ε ist 31
Defining the Hinterlands 1) Rural, 1,300 rural counties (based on 2003 U.S. Census definitions) that never achieved urban status during the sample period; 2) Non-Metro, restrictive (NM-R), adds the micropolitan area counties to rural counties, yielding nonmetropolitan areas (NM); (adding another 600+ counties 3) Non-Metro, inclusive (NM-I), NM-R counties that were NM in 1950 but were assigned to new and existing MAs between : 2,700+ counties in total 32
Small Urban Areas 1) Small MA-R, includes counties in small MAs (2003 definition) of less than 250,000 people in 1990, they were small through the entire period; 2) Small MA-I, adds 218 counties that were part of a small MA at some time during the sample period (even though they are currently in a >250k MA). 33
Main Geography Variables Distance: Nearest Metropoltian Area Rural/NM counties: distance in kms to nearest MA of any size; (population-weighted centroids) Urban MA counties: distance in kms to center of urban core if multi-county, 0 for single county urban area; (pop.-weighted centroids) We also consider nonlinear distance effects. 34
Geography Variables Incrementally Proximate Higher-Tier Areas the incremental distances to reach MAs of at least 250k, at least 500k, and at least 1.5m. pop. (for all counties) We have also considered a ‘highest’ tier of NY, LA, & Chicago in other settings. include population of the nearest or actual urban center (MA) to the county 35
Garfield County, a rural Utah county, about 4,000 residents (1990). 1. The nearest urban area is Cedar City (MICRO) located 88kms away. The nearest MA is St. George (about 90,000 population), 146kms away, an incremental distance of 58kms ( ). 2. Nearest larger MA > 250K, which is Provo-Orem, UT (pop. of 377,000), is 278kms from Garfield County, incremental distance versus St. George is 132kms ( ). 3. The nearest MA > 500K, the next higher tier, Salt Lake City, UT (969,000 people). Salt Lake is 321kms from Garfield County, incremental distance of 43kms ( ). 4. Nearest MA > 1.5 million people, the next higher tier above Salt Lake, is Phoenix, AZ (3.25 million 1990 pop.). Phoenix is 477kms away from Garfield County, an incremental distance of 156kms ( ). Distance calculations 37
Geography Variables NEG Market Potential—follows Hanson (2005) Personal income in surrounding 0-100, , , , and km rings from the population-weighted center. We use measures that predate the initial sample period by one year to mitigate possible endogeneity 38
Other Variables Several amenity variables, population density, population of the nearest MA. For the cross-sectional models, we use GMM to account for spatial autocorrelation. 39
40 40 GMM Regressions of U.S. County Population Change (%) Variables Rural (preferred) NM-RNM-ISmall MA-R (preferred) Small MA-I Intercept ** (4.50) ** (3.44) ** (3.43) ** (2.58) ** (2.31) Distance to nearest MA ** (-6.39) ** (-5.02) ** (-7.50) n.a. (Distance to nearest MA) 2 1.5E-3** (4.38) 2.1E-3** (3.85) 7.0E-3** (6.27) n.a. Distance to center of own MA n.a (0.26) (-0.79) (Distance to center of own MA) 2 n.a. -3.7E-2 (-0.19) 8.2E-3 (0.14) Inc dist to metro > 250,000 pop ** (-5.95) ** (-4.44) ** (-5.91) * (-1.89) ** (-3.86) Inc dist to metro > 500,000 pop ** (-2.79) ** (-3.09) ** (-4.20) (-1.41) ** (-2.68) Inc dist to metro > 1,500,000 pop ** (-2.26) (-1.29) ** (-3.02) ** (-2.56) * (-1.90) Population density ‘ ** (-4.76) ** (-2.14) ** (-2.22) ** (-2.89) ** (-3.08) Population of nearest/own MA ‘50 8.4E-6 (1.41) 3.1E-6 (0.41) 1.3E-5** (2.27) 1.4E-5** (2.02) 5.7E-6 (0.81) Weather/Amenity a YYYYY State fixed effects (FE) YYYYY Adjusted R No. of counties F-statistic: All dist to MA = 0 Inc distance to MA = ** 13.57** 16.24** 12.99** 67.29** 44.48** 3.85** 6.24** 8.49** 14.13**
41 Variables Distance to nearest MA (Distance to nearest MA) 2 Distance to center of own MA (Distance to center of own MA) 2 Inc dist to metro > 250,000 pop Inc dist to metro > 500,000 pop Inc dist to metro > 1,500,000 pop Personal income within 100 km radius, 1969 Personal income within km ring, 69 Personal income within km ring, Personal income within km ring, Personal income within km ring, 69 Weather/Amenity a State fixed effects (FE) Adjusted R 2 No. of counties F-statistic: All dist to MA = 0 Inc distance to MA = 0 Rural (preferred) NM-RNM-ISmall MA-R (preferred)Small MA-I ** (-4.47) ** (-4.66) ** (-6.56) n.a. 7.8E-4** (3.43) 9.0E-4** (3.51) 2.1E-3** (5.62) n.a (0.64) 1.464* (1.84) n.a. -1.6E-2 (-0.39) -2.6E-2 (-1.41) ** (-5.02) ** (-5.36) ** (-5.76) ** (-4.11) ** (-4.13) ** (-2.72) ** (-3.18) ** (-3.79) (-3.07)** ** (-2.72) (-1.42) (-0.78) ** (-2.20) ** (-2.98) (-1.19) 1.6E-03** (2.25) 9.3E-04 (1.25) 5.0E-04** (1.98) -1.2E-04 (-0.27) 9.2E-04** (3.21) 3.1E-04** (2.45) 1.2E-04 (0.86) -3.6E-04 (-1.54) -2.0E-04 (-0.92) -1.0E-04 (-0.58) 1.6E-05 (0.15) -1.3E-05 (-0.12) -2.3E-04* (-1.75) 3.1E-04** (2.06) 2.6E-04* (1.65) 1.6E-04* (1.91) 8.5E-05 (0.69) -1.7E-05 (-0.12) 2.5E-04 (1.35) 3.9E-05 (0.22) 2.9E-05 (0.38) -1.2E-04 (-1.33) -2.2E-04** (-2.09) -4.3E-05 (-0.19) 2.6E-05 (0.14) YYYYY YYYYY ** 14.32** 14.67** 18.10** 40.57** 30.91** 6.19** 8.75** 8.59** 10.37** GMM Regressions of U.S. County Population Change (%) Rural (preferred) ** (-5.40) 9.1E-4** (3.91) n.a ** (-5.83) ** (-3.22) ** (-2.15) N N N N N Y Y ** 23.74**
Conclusions—continued Key results Distance penalties are even bigger for small urban areas than remote areas Distance penalties are more important than NEG style market potential– or proximity to cities matter more than having an otherwise equal market size. Summary: At the mean distance from the urban tiers, the typical Rural county is expected to have 73% less growth than a Rural county that is adjacent to a MA core (i.e., 2.6% less annual growth, all else equal). Controlling for MP, population density, amenities, state fixed effects Distance penalties are growing in importance—distance is not dead. Mature urban system: cities support hinterlands 42
43 Variables\SamplesNon- metro Small metro Large metro Mean pop growth % (std. dev.)32.20 (122.93) (271.64) (257.38) Jan temp (diff − ) July temp (diff − ) July humidity (diff − ) Sunshine hours (diff Detroit−Orlando) Amenity rank (diff between (3) and (5) on a 1-7 amenity scale Mean ‘distance’ penalty due to remoteness from urban hierarchy NA Regression Results for County Population Growth: Selected Variables Note: Boldface indicates significant at 10% level. “Small metro” is counties located in MSAs with 250,000 population, measured in The difference between Detroit and Orlando uses their actual values. “1 std dev.” represents a one-standard deviation change in the variable. Other amenity variables include percent water area, within 50kms of the Great Lakes, within 50kms of the Pacific Ocean, and within 50kms of the Atlantic Ocean, and a 1 to 24 scale of topography—i.e., from coastal plain to extreme mountainous. The models were then re-estimated with USDA Economic Research Service amenity rank replacing all 9 individual climate/amenity variables to calculate the amenity rank effects (available online at USDA ERS). The amenity scale is 1=lowest; 7=highest. Most of the regression results reported here were not reported in Partridge et al. (2008). For more details of the regression specification, see Partridge et al. (2008b). Source: Partridge, 2010.
US Case—Summary Large cities in China are rapidly growing, creating backwash and widening regional differentials. US Large cities are not necessarily growing rapidly. Nice places are winning—amenity growth. NEG model is not a good predictor and amenity led growth wins in the US—Phil Graves. CPT fares better. (Partridge et. al 2008, 2009). CPT approaches illustrate the need for regional approach to governance. 44
45 Figure 3: Distance Penalties (%) for Median Earnings 1999 Source: Partridge et al. (2009) J. of International Economics
46 Figure 4: Distance Penalties (%) for Housing Costs 2000 Source: Partridge et al. (2009) J. of International Economics
Conclusion Space matters! Distance matters and popular folklore about its death is not true. CPT is very helpful in empirical models of growth in very different places such as China and the US. I believe it is equally helpful elsewhere. In both the US and China, it matters what type of cities/places you are near. US growth driven by weather/landscape. NEG is rigorous and formal but it is not a nuanced enough to be good predictor of where economic activity will occur in both China and the US. –at least for policy purposes. Further mega city growth may be detrimental to Chinese growth and socioeconomic goals. 47