Preferences of Higher Educated Households for Location Characteristics and Housing TypesLocation Characteristics and Housing Types
Jan Möhlmann
Based on joint work withj
Jasper Dekkers, Mark van Duijn, Or Levkovich, Jan Rouwendal
UvA –VU – PBL seminar, 18 March 2014, The Hague
Research strategyResearch strategy
Estimating household preferences based on revealed preferences
Differentiating between household types
Using estimating results to predict effects of scenarios and policy
Structure of presentationStructure of presentation
S d l Sorting model
Data descriptionData description
Estimation results
Scenario analysis
Conclusions
Sorting modelSorting model
Input of the model: current housing supply and current h h ld lhousehold population
Households choose a region and a housing type based - Sorting
model Households choose a region and a housing type based on regional characteristics and household preferences
model
- Data
Results
Which household preferences will lead to the current equilibrium?
- Results
- Scenario
analysis
- Conclusions
Sorting modelSorting model
Core is a multinomial logit model
Number of alternatives: 472 - Sorting
model(118 regions x 4 housing types)
model
- Data
Results- Results
- Scenario
analysis
- Conclusions
Sorting modelSorting model
Core is a multinomial logit model
Number of alternatives: 472 - Sorting
model(118 regions x 4 housing types)
Utility of household i in alternative n:
model
- Data
Results Utility of household i in alternative n:
in i n i n n inu P X - Results
- Scenario
analysis
- Conclusions
Sorting modelSorting model
Core is a multinomial logit model
Number of alternatives: 472 - Sorting
model(118 regions x 4 housing types)
Utility of household i in alternative n:
model
- Data
Results Utility of household i in alternative n:
in i n i n n inu P X - Results
- Scenario
analysis
Probability that household i chooses alternative n:inue
- Conclusions
in
in u
ee
Endogeneity problemEndogeneity problem
Unobserved characteristics influence utility and ho sehold priceshousehold prices
◦ Housing prices- Sorting
model Housing prices◦ Accessibility◦ Urban amenities
N tUtility
model
- Data
Results ◦ Nature◦ Unobserved characteristics
- Results
- Scenario
analysis
- Conclusions
Endogeneity problemEndogeneity problem
Unobserved characteristics influence utility and ho sehold priceshousehold prices
◦ Housing prices- Sorting
model Housing prices◦ Accessibility◦ Urban amenities
N tUtility
model
- Data
Results ◦ Nature◦ Unobserved characteristics
- Results
- Scenario
analysis
- Conclusions
Estimation strategyEstimation strategy
Solution: estimation in two steps
in i n i n n inu P X - Sorting
modelmodel
- Data
Results- Results
- Scenario
analysis
- Conclusions
Estimation strategyEstimation strategy
Solution: estimation in two steps
in i n i n n inu P X - Sorting
model
1( )i iedu edu 1( )i iedu edu model
- Data
Results- Results
- Scenario
analysis
- Conclusions
Estimation strategyEstimation strategy
Solution: estimation in two steps
in i n i n n inu P X - Sorting
model
( ) ( )u P X edu edu P edu edu X
1( )i iedu edu 1( )i iedu edu model
- Data
Results 1 1( ) ( )in n n n i n i n inu P X edu edu P edu edu X - Results
- Scenario
analysis
- Conclusions
Estimation strategyEstimation strategy
Solution: estimation in two steps
in i n i n n inu P X - Sorting
model
( ) ( )u P X edu edu P edu edu X
1( )i iedu edu 1( )i iedu edu model
- Data
Results
Step 1: estimate and and an alternative specific
1 1( ) ( )in n n n i n i n inu P X edu edu P edu edu X
1 1
- Results
- Scenario
analysis Step 1: estimate and and an alternative specific constant (asc = )n n nP X
1 1- Conclusions
Step 2: explain the asc’s based on characteristics of alternatives using 2SLS
Structure of presentationStructure of presentation
S d l Sorting model
Data descriptionData description
Estimation results
Scenario analysis
Conclusions
Data (households)Data (households)
Data are obtained from Woon Onderzoek Nederland (W ON) 2012(WoON) 2012
57 276 households- Sorting
model 57,276 households
Household characteristics
model
- Data
Results Household characteristics- Results
- Scenario
analysis
Mean Min. Max.
Couple 0.63 0 1
- Conclusions
Children in household 0.35 0 1Higher education 0.30 0 1Age 51.7 17 100
Data (regions)Data (regions)
118 regions based on 415 adjacent municipalities
- Sorting
modelmodel
- Data
Results- Results
- Scenario
analysis
- Conclusions
Data (regions)Data (regions)
Every region provides four alternatives (rentel houses d h f d h )and three types of owner-occupied houses)
Regional characteristics- Sorting
model Regional characteristics
Mean Min. Max.
Di 100 000 j b (i k ) 12 6 3 6 32 8
model
- Data
Results Distance to nearest 100,000 jobs (in km) 12.6 3.6 32.8Distance to intercity train station (in km) 7.5 1.5 27.8Distance tot highway onramp (in km) 4.1 1.0 20.3Share of surface is nature (in %) 13.8 0.4 65.8Size of historical city centre (in km2) 0.9 0 13.3
- Results
- Scenario
analysis
Prices of owner-occupied houses differ by type- Conclusions
Data (regions)Data (regions)
Price of a standard house is determined using a hedonic l dprice analysis on transaction data
- Sorting
modelmodel
- Data
Results- Results
- Scenario
analysis
- Conclusions
275000 - 425000250000 - 275000225000 - 250000200000 - 225000175000 - 200000129000 - 175000
Structure of presentationStructure of presentation
S d l Sorting model
Data descriptionData description
Estimation results
Scenario analysis
Conclusions
Willingness to payWillingness to pay
5000
- Sorting
model 2000
3000
4000omodel
- Data
Results
0
1000Euro
- Results
- Scenario
analysis
‐2000
‐1000
jobs (km) train station (km)
highway (km) nature (%) city centre (km2)
- Conclusions
(km) (km2)
Willingness to payWillingness to pay
5000
- Sorting
model 2000
3000
4000omodel
- Data
Results
0
1000Euro
- Results
- Scenario
analysis
‐2000
‐1000
jobs (km) train station (km)
highway (km) nature (%) city centre (km2)
Apartment: reference typeTerraced housing: – 500
- Conclusions
(km) (km2)
Detached housing: 39.000
WTP by household typeWTP by household type
F 1 k h h 100 000 b For 1 km higher proximity to nearest 100,000 jobs
- Sorting
model 4700model
- Data
Resultsyes
60
yes
4300
4400
4500
4600
- Results
- Scenario
analysisno no
30
yes60
4000
4100
4200
4300
Euro
- Conclusions
30no
3700
3800
3900
couple children age higher educationcouple children age higher education
WTP by household typeWTP by household type
F d h d h ( l ) For detached housing (relative to apartments)
- Sorting
model 60000model
- Data
Results
yes
yes
60 yes
40000
50000
- Results
- Scenario
analysisno
no30
no
20000
30000
Euro
- Conclusions
0
10000
couple children age higher educationcouple children age higher education
Structure of presentationStructure of presentation
S d l Sorting model
Data descriptionData description
Estimation results
Scenario analysis
Conclusions
Scenario analysisScenario analysis
Estimated parameters for household preferences allow us to sort a given population of households over the alternatives
Scenario input:- Sorting
model Scenario input: Distribution of household types (e.g. education, age) Regional characteristics(e.g. distance to jobs, nature)
H i l (di ib i b i d
model
- Data
Results Housing supply(distribution between regions and composition of housing types within regions)
- Results
- Scenario
analysis
Scenario output: Housing prices Composition of household types for each region
- Conclusions
Composition of household types for each region
Global economy 2030 scenario Global economy 2030 scenario
Example: housing supply in 2030 based on Ruimtep g pp yScanner XL
Global Economy scenario- Sorting
model
Assumption: number of houses is equal to number of households
model
- Data
Results of households
Household demographics and regional
- Results
- Scenario
analysis
characteristics remain constant- Conclusions
Global economy 2030 scenario Global economy 2030 scenario
Pricechange ofdetached
- Sorting
model detachedhousing
model
- Data
Results- Results
- Scenario
analysis
- Conclusions
Global economy 2030 scenario Global economy 2030 scenario
Change in share of higher educated
- Sorting
model higher educatedhouseholds
model
- Data
Results- Results
- Scenario
analysis
- Conclusions
Structure of presentationStructure of presentation
S d l Sorting model
Data descriptionData description
Estimation results
Scenario analysis
Conclusions
ConclusionsConclusions
Sorting model uses revealed preferences to g pdetermine willingness to pay for regional characteristics- Sorting
model
Can distinguish between household types
model
- Data
Results
We find a positive willingness to pay for proximity to jobs, availability of nature and urban amenities,
d f d h d h i
- Results
- Scenario
analysis
and for detached housing
Estimation results can be used to predict the effects
- Conclusions
Estimation results can be used to predict the effects of scenarios and policy on housing prices and regional household composition
Alternative modelsAlternative models
Estimating the sorting model with different g gcharacteristics of households and regions foreign knowledge workers and students
fi ld f d i f i- Sorting
model field of education or profession
Different level of aggregation (e g neighbourhoods
model
- Data
Results Different level of aggregation (e.g. neighbourhoods instead of municipalities)
- Results
- Scenario
analysis
Estimating costs of moving (using distance to previous region)
- Conclusions
Preferences of Higher Educated Households for Location Characteristics and Housing TypesLocation Characteristics and Housing Types
Jan Möhlmann
Based on joint work withj
Jasper Dekkers, Mark van Duijn, Or Levkovich, Jan Rouwendal
UvA –VU – PBL seminar, 18 March 2014, The Hague