A Spatial Ecoometric Model of the Korea Ecoomy Doleswar Bhadari BBER, Uiversity of New Mexico Thomas G. Johso Deis P. Robiso Commuity Policy Aalysis Ceter Uiversity of Missouri
Motivatio Applicatio of commuity policy aalysis framework i Korea regios Collaborative effort of CPAC with Korea Rural Ecoomic Istitute Impact of ecoomic developmet expeditures ad employmet Forecastig of depedet variable based o projected exogeous variables
Commuity Policy Aalysis System Framework Use of impact model for aalysis of rural issues Flexible eough to address specific places ad a variety of diverse eeds (e.g., housig market impacts ad demographic chages) Employmet chage is the mai driver of the model
Employmet Model Structure Exteral employmet Sceario developmet Ecoomic developmet expeditures Exteral ecoomic developmet expeditures Labor Market Ecoomically Active Populatio I-commuters Out-commuters Employmet i o-basic sectors Demography Populatio Number of studets Local public fiace Local Reveues Local Expeditures Number of firms Housig market Total housig uits
Spatial Equilibrium of Labor Market A B L S Wage gap L S Wage L D L D Employmet Employmet
Basic model Β X A E (1) _ B X C A U (2) Kelejea ad Prucha (2004) estimatio procedure (istrumet selectio)
Spatial likages Weight/distace Weight/distace squared WG 1 [ w ] ij N Weight/distace squared W [ w ] Uiform G2 ij N [ ] w J 1 X j D X J 1 X j X J D W 1 U ij J 2 ij D ij D 2 ij N ij i
Reduced form solutio Reduced form solutio Reduced form solutio Reduced form solutio U A C X B _ { } m x I C W A I B I I y ) ' ( )] ' ( ) ' [( ) ( 1 0 1 12 β β λ 0 0 1 L 0 0 ' 21 1 12 ) ( m m m B β β β M O M M O M M KK A m m λ λ 0 0 0 0 ' 2 1 M O M 0 1, 1 m m β m β K M O M m λ 0 K 0 K
Model s Equatios 1. Populatiof( f(lag, ecoomically active populatio) labor force-na 2. Ecoomically active populatiof(lag, populatio, EMP_NBAS) 3. Number of studetsf(lag, populatio) p
Model cotiued.. 4. Out-commutigf(lag, ecoomically active populatio, EMP, area, area*emp, CEMP, ecoomic developmet expeditures) 5. I I-commutigf(lag, ecoomically active populatio, EMP, exteral EMP, area, area*emp) 6. No-basic employmetf(lag, EMP, area, area*emp, ecoomic dev. expeditures, area* ecoomic dev. expeditures)
Model cotiued 7. Local public reveuesf(lag, populatio, o- basic EMP, i-commutig commutig) icome-na 8. Local public expedituresf(lag, populatio, i- commutig, o-basic EMP) icome-na 9. Housig uitsf(lag, populatio, i-commutig commutig) house price-na. 10. Firm_tot totf(lag, populatio, i-commutig, ecoomic developmet expeditures, area, area*emp)
Data ad Data Sources Korea regios o 7 metropolita cities o 77 cities o 88 couties Data sources o Local reveue ad expeditures Korea Local Fiacial ear Book 2005 o Employmet, populatio, housig, busiess Korea s Si or Gu s Statistical ear Book 2005
GS3SLS Results Usig Differet Weight Matrices Uiform Weight ad distace Weight ad distace squared Model Local Reveue Variables Sig Sig Sig Itercept W_REV_LOC POP_TOT EMP_NBAS COM_IN - - - Itercept W_EXP_LOC a - - - Local Expediture POP_TOT COM_IN - - - EMP_NBAS Itercept W_HOUS_TOT a Total Housig Uits POP_TOT COM_IN - - - Itercept Populatio W_POP_TOT POP_EAP Itercept - - - Ecoomically Active Populatio W_POP_EAP a - - - POP_TOT EMP_NBAS Studets Total Itercept W_STDT_TOT POP_TOT 1% 5% 10% NS
Model Uiform Weight ad distace Weight ad distace squared Variables Sig estimates estimates Itercept - W_COM_OUT POP_EAP Out-commuters EMP_TOT - - - AREA A_EMP - - - CEMP C_EMP EXP_ED - - - Itercept - - - W_COM_IN POP_EAP - - - I-commuters EMP_TOT C_EMP AREA AEMP _ - - - Itercept W_FIRM_TOT - - - Firms Total POP_TOT AREA A_EMP Itercept - - - W_EMP_NBAS a - - - Employmet i o-basic EMP_TOT sectors AREA EXP_ED A_EXPED - - - 1% 5% 10% NS
Ecoomic Impact Estimated From Spatial Lag Model a. Provice Gu or Si REV_LO C EXP_L OC HOUS_T OT POP_T OT POP_E AP STDT_T OT COM_O UT COM_ IN FIRM_T OT EMP_NB AS Pusa Gijag-Gu 3602.8 3202.5 481.87 2206.21 1660.13 356.00 266.77 319.16 191.57 813.28 ulsa Gyug-Buk Gyug-Buk Gyug- Nam Gyug- Nam ulju-gu Pohag-Si Gyugju-Si Chagwo- Si Gimhae-Si -7-8.14-1.52-0.14-0.09-0.037 2.268 3.584 0.006 0.016-0.01-0.01-0.002 0 0 0 0.004 0.004 0 0 0.18 0.17 0.044 0.01 0 0.002-0.07-0.096 0.001 0 0 0-0.001 0 0 0 0 0.001 0 0 0.23 0.23 0.057 0.03 0.02 0.01-0.008-0.105 0.004 0.001 Gyug- 0 0-0.001 001 0 0 0 0 0.001001 0 0 Nam Milyag-Si Gyug- Nam Pusa ulsa Gyug- Nam agsa-si Pusa ulsa Jihae-Si -6.98-7.85-1.57-0.35-0.24-0.093 2.175 3.539 0.017 0.043-6.6969-5 -2.096-2.83-1.93-0.762 0.189 2.878 0.088088 0.306-6.95-7.59-1.623-0.62-0.42-0.168 1.879 3.454 0.02 0.067 0 0 0 0 0 0 0 0.001 0 0 Total Impact 3575.58 3174.31 475.156 2202.31 1657.47 354.951 273.206 332.42 191.706 813.709 a Effects of 1,000 ew jobs created i Gijag Couty of Pusa Provice, Korea
Itra-Couty Ecoomic Impact Compariso of a Spatial ad No-Spatial Model Impact from Variable spatial model Impact from ospatial model Percetage differece Local reveue (millio wo) 3603 3144-13% Local expeditures (millio wo) 3203 2587-19% Housig uits 482 466-3% Populatio 2206 2253 2% Ecoomically active populatio 1660 1695 2% Number of studets 356 364 2% Out-commuters 267 331 24% I-commuters 319 381 19% Number of firms 192 172-10% Employmet i o-basic sector 813 830 2% Effects of 1,000 ew jobs created i Gijag Couty of Pusa Provice, Korea
Types of Spatial Simultaeous Equatio Models Types of Spatial Simultaeous Equatio Models yp p q yp p q Cosidered Here Cosidered Here (1) E A X B U A C X B _ (1) (2) E A X B U A C X B (2) E C X B U E E Ρ (3) E C X B E A C X B _ (4) U E E Ρ
Mea Absolute Percetage Error as a Measure of Forecastig Accuracy i Differet Models Equatios Spatial lag ad spatial error model Spatial error model Spatial lag model Nospatial model Local reveues 21.8 22.5 21.9 25.1 Local expeditures 15.0 15.6 15.2 19.1 Total housig uits 13.0 13.7 12.7 13.6 Populatio 11.0 11.9 11.1 11.9 Ecoomically active populatio 10.2 9.1 9.1 9.7 Number of studets 48.4 53.0 45.5 53.3 Out-commuters 423.0 544.7 470.2 730.4 I-commuters 173.0 167.4 177.6 183.5 Number of firms 21.3 22.4 23.4 22.6 Employmet i o-basic sectors 72.2 86.0 85.8 87.1 Average 81.0 94.7 87.3 115.8 Coefficiet of variatio 4.0 4.6 4.2 5.1
Coclusios Provided a comprehesive modelig framework for local ecoomies i Korea ad made a uique applicatio of spatial ecoometric aalysis Solved for the spatial reduced form solutio ad performed simulatio aalysis Both the spatial iteractio ti ad cross equatio iteractios are sigificat However, the equatio parameter estimates are sesitive to the structure of the spatial likages used (i.e., weight matrix). This appears to be due to the heterogeeity of sizes of spatial uits i Korea (metro vs rural couties)
Coclusios cotiued.. Other key fidigs Addig spatial compoets icreases the model s explaatory power More importatly, the spatial compoets appears to improve the accuracy of the itra-couty impacts Ay type of spatial model cosidered here was better tha the o-spatial alterative model
Better data Recommedatios Cosider simultaeous spatial models that have differet spatial structures (spatial error, spatial lag, both ad o-spatial) Cosider other spatial structures Istitutioal etworks Fuctioal coectios (highways) Cetral place otios
Thak you