Housing Demand in Germany and Japan

1
Housing Demand in Germany and Japan

• Borsch-Supan, Heiss, and Seko,
• JHE 10, 229-252 (2001)
• Presented by Mark L. Trueman, 11/25/02

2
Commonalities

• Market economies with essentially private housing
  markets
• Strong government intervention in the form of
  housing subsidies/ market regulations
• Comparable standards of living
• Comparable demographicse.g. high elderly share
• Americanized- similar consumption patterns

3
Differences
4
Problem/Approach

• Problem
• Typically, we draw inferences about components of
  demand by looking at differences and similarities
  across HH. Such cross-sectional data in a single
  country features little price variations.
• Goal Exploit cross-national variation
• Identify determinants of housing demand
• Separate differences in preference parameters
  from HH attributes socio-economic
  characteristics
• Econometric method
• A flexible discrete choice model is used- Mixed
  Multinomial Logit Model (MMLM)

5
Methodology

• Since housing is a durable good, use permanent
  (or normal) income when est. demand.
• Since housing is a heterogeneous good, normalize
  quality and quantity (i.e., find a well-defined,
  standard dwelling). Use classical hedonic
  approach.
• Model housing demand as a multidimensional
  choice- major attributes are tenure, size,
  structure type

6
Prior Econometric Specifications

• The basic MNL models derive choice probabilities
  and a likelihood function.
• Mistaken assumption is that the error components
  which carry all unobserved heterogeneity are
  independent of each other.
• Problem seriously biased parameter estimates of
  housing demand determinants.

7
Enhanced Model Specification

• Model housing demand as a dynamic, decision
  making process using a mixed multinomial logit
  (MMNL) model.
• Model allows for
• a flexible substitution pattern among
  alternatives.
• unobserved heterogeneity in panel data

8
Model Summary

• Housing demand equations joint-dependent
  variables are the probabilities of the 8 housing
  alternatives, explained by housing prices, HH
  income, set of socio-econ characteristics (age of
  HH, HH size both quadratic forms)
• Plus, a linear time trend
• Alternative specific constants-interaction of the
  other variables with the 3 dimensions-tenure,
  type, size.

9
Structure/ Results

• Goodness of Fit- MMNL model provides a much
  better fit than basic MNL model

10
Interpreting Elasticities from Significant
Coefficients

• Coefficients in discrete choice models carry
  little intuitive information.
• They must be converted to be meaningful.
• Effects are roughly symmetrical.

11
More!
12
Conclusions

• Key pooling data from different countries offers
  more variation in explanatory variables.
• Result we get better estimates of the income,
  age, and size effects (elasticities).