Housing Market Trends Lessons from International Comparisons Grace Wong The Wharton School, Univ of Penn wongg@wharton.upenn.edu May 10, 2007
International comparisons The impact of macroeconomic variables and policies on housing markets Two worries Country-specific institutions panel data analysis with country fixed effects (non-time-varying component) Country-specific interactions country-specific models How much does the market response to the interest rate in one country tell us about that in another E.g., completeness of mortgage market How useful the estimates of the average impacts are
Average 24 52% 2.50% 0.67 Standard Deviation 13 30% 1.30% 0.48 Source: van den Noord (06)
Methodology Some Difficulties van den Noord (2006, #488) Arbitrary dating techniques Smoothness of housing cycles increases with the size/ cross-regional heterogeneity of the country Variation in best-fit models for each country; significant countryspecific risk identified in Ling & Naranjo (2002) p.11: in the final regressions most variables were dropped because their coefficients were insignificant Short i/r, inflation, unemployment, current a/c balance, residential investment as a share of GDP, equity price index Statistical precision vs. Economic importance Omission changes coeff estimates which feed to the simulation proxy effect might lead to overestimate of the response to i/r How much? Sensitivity tests.
Another international comparison International Comparison of the Residential Risk Premia Joint with Chris Julliard (LSE) Changing risk premia Direct estimates of the risk premia and their link to country-time specific institutional characteristics Qualitative instead of quantitative; Long-run OECD countries (quarterly data; 1966-2004) Campbell & Shiller (1988) and Campbell (1991)
Estimating RE Risk Premium Campbell & Shiller (1988) Start with accounting identity: R=gross return; P=price; L=rental income Gallin (2004): stationarity of price-rent ratio Log Linearization around steady state
Returns and return innovations (4) (5)
Reduced-form VAR To estimate the risk premium and its innovations, we use a reduced form VAR. (Campbell (1991)) i.e. we proxy rational expectations via linear projection z=(rent growth rate, GDP growth rate, inflation, risk-free interest rate, log housing price-rent ratio, and a business cycle proxy)
Panel Analysis Dependent var: Estimated risk premia, demeaned Institutional and regulatory characteristics: Debt market Borrowing constraints Equity market Tax Zoning OLS; Newey-West std errors (Myer & Webb 1994)
R-sq with country, year and quarter fixed effects only: 0.07 Country, year and quarter fixed effects Robust and stable relationships reported
Credit constraints & equity markets Log (liquid liabilities to GDP) 0.224*** 0.272*** 0.141 0.202** 0.275** 0.133 (0.071) (0.084) (0.089) (0.091) (0.094) (0.084) LogGDP 0.002 0.030 0.041 0.015 0.017 0.260*** (0.047) (0.058) (0.061) (0.059) (0.101) (0.088) Log (private bank assets to -0.039*** -0.043*** -0.021-0.036** -0.273-0.121*** central bank assets) (0.015) (0.016) (0.017) (0.017) (0.022) (0.023) Max loan-to-value ratio 0.939* 0.307 0.912-1.355** -1.239*** *GDP growth rate (0.556) (0.549) (0.585) (0.530) (0.464) Max loan-to-value ratio 0.283 1.111*** 0.634** -0.495-0.115 (0.293) (0.312) (0.318) (0.346) (0.295) Gdp growth rate -0.348 0.109-0.329 1.287*** 1.225*** (0.414) (0.408) (0.432) (0.393) (0.340) Log (stockmarket -0.177*** -0.302*** capitalization to GDP) (0.024) (0.036) Log (stockmarket turnover -0.029*** -0.016 ratio to GDP) (0.010) (0.022) Log (equity issues to GDP) 0.039*** 0.088*** (0.009) (0.010) Country dummies Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Quarter dummies Yes Yes Yes Yes Yes Yes R2 0.500 0.535 0.526 0.488 0.783 0.851 No. of obs 678 604 570 577 340 336
Taxes Log (All individual tax -0.190** -0.093 0.219** revenues to GDP) (0.078) (0.078) (0.101) Log (Individual property -0.169*** -0.207*** -0.227*** tax revenues to GDP) (0.019) (0.020) (0.020) Lagged log (All individual 0.011-0.135 tax revenues to GDP) (0.126) (0.136) Lagged log (Individual 0.100*** 0.085*** property tax rev. to GDP) (0.022) (0.024) Federal to local all tax -0.032*** revenue ratio (0.006) Country dummies Yes Yes Yes Year dummies Yes Yes Yes Quarter dummies Yes Yes Yes R2 0.774 0.799 0.812 No. of obs 238 234 198
Zoning Increase in "Decentralization" -0.007 0.114*** zoning legislation (0.019) (0.036) Decrease in "Decentralization" -0.028-0.246*** zoning legislation (0.030) (0.058) Increase in "Public Participation" -0.093** zoning legislation (0.045) Increase in "Simplification" 0.014-0.265** zoning legislation (0.022) (0.117) Decrease in "Incentives" -0.026 0.069 zoning legislation (0.040) (0.081) Increase in "Incentives" 0.039 zoning legislation (0.049) Decrease in "Urban Planning" 0.041 zoning legislation (0.083) Increase in "Urban Planning" -0.011 0.402*** zoning legislation (0.019) (0.148) Increase in "Environmental" 0.027* 0.018 zoning legislation (0.015) (0.037) Increase in "Historical -0.026-0.419** Preservation" zoning legislation (0.028) (0.171) R2 0.137 0.808 No. of obs 1387 238
Consolidated zoning categories Decrease in "Obstacles" -0.053* -0.054* 0.068 zoning legislation (0.031) (0.031) (0.050) Increase in "Obstacles" 0.027-0.026 0.075*** zoning legislation (0.035) (0.018) (0.028) No change in "Simplication" -0.025 zoning legislation (0.048) Increase in "Simplication" -0.010 0.014-0.021 zoning legislation (0.051) (0.020) (0.030) No change in "Guidelines" -0.036 zoning legislation (0.083) Increase in "Guidelines" -0.041-0.004-0.077*** zoning legislation (0.084) (0.013) (0.018) R2 0.134 0.133 0.806 No. of obs 1387 1387 340
Challenges Data Bubbles?
Log Linearization (1) Assume expectations are rational and trasversality condition holds, (2) subscript e denotes excess returns (3)
International comparisons The impact of macroeconomic variables and policies on housing markets Two worries Country-specific institutions panel data analysis with country fixed effects (non-time-varying component) Country-specific interactions country-specific models How much does the market response to the interest rate in one country tell us about that in another E.g., completeness of mortgage market How useful the estimates of the average impacts are