The determinants of housing demand An analysis for French households 59th North American Meetings of the RSAI 9 November 2012 The determinants of housing demand An analysis for French households Nicolas Coulombel, Clémence Morel Université Paris – Est, LVMT : nicolas.coulombel@enpc.fr
presentation outline Introduction Methodology Comparison of the LES and the QES: a simple model Impact of household size on housing demand Conclusion
context & motivation Housing: increasing weight in French households’ budget yet limited research on housing demand in France Aim of this research determinants of housing demand influence of household size Motivation recover households’ preferences for LUTI modeling design of housing policies
methodology Comparison of two demand systems Step 1: simple models Linear Expenditure System Quadratic Expenditure System Step 1: simple models estimation with expenditure and budget share Step 2: introduction of HH size use of « scaling »
the linear expenditure system (les) Expenditure for good i Derives from the Stone-Geary utility function For two goods : housing h and composite z income elasticity of housing : price elasticity of housing :
the quadratic expenditure system (qes) Expenditure for good i Derives from the indirect utility function For two goods : income elasticity price elasticity
properties of the les and the qes Estimation LES & QES: only 2 periods/regions required QES: non linearity log-likelihood maximization Interpretation of coefficients in the LES ai: marginal budget share bi : subsistence quantity? Elasticities LES: ηy with Y, ηy 1 and ηP with Y, ηP -1 QES: more flexible
data 2006 French Housing Survey Sample Housing price Income private renters excluded: students, outliers (income & net budget share) 5 907 observations Housing price observed mean regional rental price Income current income
housing expenditure Gross monthly rent Preliminary analysis rental charges/utilities excluded subsidies not taken into account ( control variable) Preliminary analysis marked heterogeneity trend: increase with HH income heteroscedasticity
housing (gross) budget share Definition: Gross Housing Expenditure / HH Income > net budget share case of low-income HHs Preliminary analysis budget share decreases with income volatility strongly decreases with income same remarks for net budget share
estimated models Housing outlays LES QES Budget shares: same model divided by Y Heteroscedasticity housing budgets : weight Y-1 budget shares : weight Y0.66
estimation results LES QES Exp Budget Share ah 0.099*** 0.101*** LES QES Exp Budget Share ah 0.099*** 0.101*** 0.124*** aS 0.187*** 0.213*** 0.206*** 0.228*** bz -2043.1*** -1971.5*** -2028*** -1933.3*** bh 12.43*** 11.88*** 6.11*** 6.71*** ch 2.31*** 2.12*** λ -0.157* -0.11652 d 0.0109*** 0.0108*** R² 0.326 0.624 0.337 0.626 B-P (p-value) 0.01 0.24 0.08 0.17
engel curves (budget share)
income elasticity
price elasticity
impact of household size Same models as before (LES & QES) estimated on budget shares Use of scaling (Pollak and Wales, 1981) parameters vary with HH composition Representation of household size 4 dummies: partner/spouse, 1 child, 2 children, 3 children and more test non-linear effects
estimation results LES QES ah 0.101*** 0.158*** ah,spouse -0.011** -0.031*** ah,child1 -0.0074** -0.026*** ah,child2 -0.0041** -0.019*** ah,child3+ -0.0023** -0.022*** as 0.110*** bz -1881*** -1830.1*** bz,spouse -766.6*** -908.7*** bz,child1 -797.9*** -298.4*** bz,child2 -918.7*** -702.2*** bz,child3+ -659.6*** -553.9*** bh 12.25*** 1.67 bh,spouse -4.55*** -4.60** bh,child1 0.43*** 7.13** bh,child2 -0.90*** 2.46** bh,child3+ 3.31*** 6.89** ch 1.106*** ch,spouse 0.692*** ch,child1 0.782*** ch,child2 0.561** ch,child3+ 0.714** λ -0.10033*** d 0.0107*** 0.0105*** Log-likelihood 6324 6373 R² 0,634 0,64 B-P (P-value) 0,19 0,23 QES : significant compared to LES (likelihood ratio test) LES children housing expense increases by a constant spouse: more complex effect QES impact harder to interpret QES > LES: likelihood ratio-test
les – engel curves
qes – engel curves
les – income elasticity
qes – income elasticity
housing budget share and household size in expensive areas Ile de France (Paris area) average rental price = 14,4€/m² ≈ twice the national average Specific feature: couples have a lower housing budget share than single persons plausible interpretation: dual-earner couples might keep the previous home of one of the households’ members
conclusion Comparative analysis of the LES and the QES similar explaining powers LES easier to interpret but elasticities provided by LES are likely biased Household composition and size matters partner/spouse and children do not have the same impact on housing demand Future research investigating the strong heterogeneity of housing budget shares for low-income HHs analysis of the linear and quadratic AIDS