SEOUL CUBE LAND MODEL BUILDING AND ITS APPLICATION: The Effects of Housing Preference for Apartment on Housing Market Myung-Jin Jun Professor, Chung-Ang.

SEOUL CUBE LAND MODEL BUILDING AND ITS APPLICATION: The Effects of Housing Preference for Apartment on

Housing Market

Myung-Jin Jun

Professor, Chung-Ang University, Korea

Background• Households of the Seoul metropolitan area (SMA)

have a strong preference for apartment*, unlike citi-zen of the Western cities who likes to live in single-family detached housing.

• Living in a high-rise apartment is a lifelong dream for many Korean.

• The dominant housing type for Korean has dramati-cally shifted from single family housing to multi-family housing (especially apartment) in the nation and the SMA over the last three decades* Apartment is defined in Korea Housing Law as a housing unit oc-

cupying a 5 or more story multi-family housing. As of 2007, 81.6% of apartments are in 10 or more story apartment buildings in Korea.

Background (cont’)

• The share of apartments to total housing stock has significantly increased from 13.6% in 1980 to 63.6% in 2010 for apartments, and decreased from 77.2% to 15.5% for the single family housings in the SMA over the last three decades.

• According to the National Statistical Office, as of 2010, 71% of residents in the SMA lives in multi-fam-ily housings, and 77.5% of them resides in houses in the high-rise apartment buildings, indicating high con-sumer preference for apartments in the SMA.

Study Purpose

• To investigate the effects of household apartment preference on the housing market in the SMA

• To empirically build SEOUL CUBELAND MODEL, a ran-dom utility-based land use simulation model with bid-rent the-ory, that represents the housing market with endogenous prices and a market clearing mechanism.

• To analyze the effects of dwelling preference on the housing market by comparing two different scenarios: – The baseline scenario taking the difference in housing preference by

income group into account, – A counterfactual scenario in which there is no difference in housing

preference among income groups.

BUILDING SEOUL CUBELAND MODEL

Housing Supply

Housing Supply• 74 Zones• 3 Housing Types

1) Single-Family Housing2) Apartment3) Others

• 220 location Options

Households

Agent Category ID

Household Size (person)

Monthly Income ($)

1 1 1,000 or less2 2 1,000 or less3 3 1,000 or less4 4 or more 1,000 or less5 1 1,000-3,0006 2 1,000-3,0007 3 1,000-3,0008 4 or more 1,000-3,0009 1 3,000-5,000

10 2 3,000-5,00011 3 3,000-5,00012 4 or more 3,000-5,00013 1 5000 or more 14 2 5000 or more 15 3 5000 or more 16 4 or more 5000 or more

• Sixteen household types with four in-come levels and four household size levels covering 24.5 mil-lion inhabitants

Data Sources• The primary data sources for the building of the model are the

2006 Household Travel Survey and the 2010 real estate sales data for the SMR from the MLTM (Ministry of Land, Transport, and Maritime Affairs).

• The Household Travel Survey (HTS) data includes socio-demo-graphic and trip information for individual households and persons such as monthly income, household size, and residential and em-ployment locations.

• The HTS also contains information on travel mode and time by trip purpose.

• The real estate sales data includes the lump-sum deposit amount (Jeonse), monthly rent, and floor space by housing type.

• The lump-sum deposits are converted into monthly rent through amortization in terms of US dollars.

Parameter Estimation and Model Calibration

• Bid Function • Rent Function• Cost Adjustment Parameters

Bid Function

• The MNL model estimates 3 equations for 4 household income categories on the dependent variable, assigning the lowest income group as the reference group.

• The explanatory variables for the bid function includes average zonal income, population and employment densities, and accessibility as a location attribute.

• Also included is an apartment dummy variable in order to capture the difference in housing preference for apartments by income group.

Residential Bid Function ParametersIncome Group 2

Vs.Income Group 1

Income Group 3Vs.

Income Group 1

Income Group 4Vs.

Income Group 1

Constant-1.450 (-0.66)

-15.800 (-6.95)**

-32.500 (-12.30)**

Accessibility2.2.E-05(2.35)*

3.7.E-05(3.32)**

1.8.E-04(4.76)**

Employment Density (employment/Km2)

-2.4.E-05(-2.37)*

-2.6.E-05(-2.49)**

-4.0.E-05(-2.95)**

Household Density (person/Km2)

-1.7.E-05(-0.90)

-4.0.E-05(-2.03)*

-5.2.E-05(-2.19)*

Seoul Dummy (1 if locate in Seoul, 0 otherwise)

-0.251 (-2.43)**

-0.119 (-1.09)

0.103 (0.76)

Apartment (1 if housing is apartment, 0 otherwise)

0.603 (9.07)**

1.400 (20.42)*

1.800 (21.87)**

Log(Average Zonal Income: $)

0.154 (0.56)

1.880 (6.57)**

3.870 (11.67)**

N 20,000

Likelihood ratio test 12859.26

Rho-square 0.232

t-value in parenthesis, ** p<0.01, *p<0.05

Rent Function

• The rent function has two components: the logsum of bids and a hedonic part of the rent.

• Five independent variables explaining residential rents were included: the average of floor space and single family housing dummy variable for the dwelling factor, the average zonal monthly income, education quality, and the Seoul dummy variable representing location factors.

• We employed the OLS method to calibrate the rent function for the SMA.

Residential Rent Function Parameter Estimates

β t-value p

Intercept -0.78242 -9.15 0.0001

Logsum of Bids 0.20242 3.94 0.0001

Average Zonal Monthly Income ($) 0.00023 4.63 0.0001

Average Size of Floor Space (m2) 0.00614 5.86 0.0001

Single Family Housing Dummy -0.14895 -5.69 0.0001

Education Quality (1 if locate in Kangnam 3 Gus, 0 otherwise) 0.13572 2.28 0.0236

Seoul Dummy (1 if locate in Seoul, 0 otherwise) 0.25207 9.73 0.0001

N=222, R2=0.84

Model Calibration

• To match the number of estimated housing units by housing type to the observed units.

• We estimate supply cost adjustment factors us-ing an iterative method as follows:

,

• Iteration continues until the difference between the estimated and observed real estate units is within the tolerance level.

Actual and Estimated Housing Units Sup-plied

0 50000 100000 150000 200000 2500000.00

50000.00

100000.00

150000.00

200000.00

250000.00

f(x) = 0.978488266222283 x + 2103.89527897166R² = 0.990964220030927

Actual

Estimated

The Effects of Apartment Prefer-ence on Housing Supply and Rent

The Effects of Apartment Preference on Housing Supply by Housing Type

Housing Type Region Baseline (A)No-Preference Scenario (B)

Difference (A-B)

Apartment

Seoul 1,257,761 1,142,080 115,681

Incheon 411,159 377,660 33,499

Kyunggi 1,827,759 1,663,363 164,396

Total 3,496,680 3,183,104 313,576

Single Family Housing

Seoul 1,270,029 1,378,673 - 108,644

Incheon 200,729 215,864 - 15,135

Kyunggi 1,024,687 1,109,931 - 85,244

Total 2,495,445 2,704,468 - 209,023

Other Type Housing

Seoul 576,522 626,729 - 50,207

Incheon 191,801 206,375 - 14,573

Kyunggi 476,918 516,690 - 39,772

Total 1,245,242 1,349,794 - 104,553

Spatial Distribution of Apartment Prefer-ence Impact by Zone Type

Zone Type*

Apartment Single Family Housing Other Types

Units % Share Units% Share Units % Share

Central City(City of Seoul)

CBD 1,597 0.5% - 3,730 1.8% - 1,847 1.8%

SUBCENTER

29,360 9.4% - 19,536 9.3% - 10,066 9.6%

NON-CENTER

84,724

27.0% - 85,378 40.8% - 38,294 36.6%

Suburban Area

INNER RING 140,139

44.7% - 69,103 33.1% - 42,055 40.2%

OUTER RING 57,756

18.4% - 31,275 15.0% - 12,290 11.8%

Total 313,576

100.0% - 209,023 100.0% - 104,553 100.0%

Apartment Rent Impacts of Apartment Pref-erence ($/Month)

Conclusions

• The higher apartment preference of the medium- and high-income group has contributed to the addition of substantial apartment units (140,000 units) in the suburban inner ring zones and to the reduction of single family units (85,000 units) in the residential zones of the central city, leading to population suburbanization with dense suburban develop-ment.

• Higher apartment preference of the medium- and high-in-come group has the largest rent impact in the wealthiest communities in Seoul, including Seocho, Kangnam, and Yongsan, demonstrating the high income group's willingness to pay for apartments in these areas.

Policy Implications

• Seoul’s experience presents significant policy implications for the Smart Growth policy

• Some claim that Americans’ preference for single-family homes is so strong that smart growth strategies supporting higher resi-dential density cannot be implemented successfully

• However, consumer preference for large suburban single-family houses is declining, as demographic and economic factors in the housing market are changing, including the aging population, smaller households, rising fuel prices, etc.

• Seoul’s case study supports that multi-family dwellings such as apartments can be an alternative to suburban single-family hous-ing if they offer accessibility and amenity advantages, leading to a dense suburban development