Consumption Responses to House Price Heterogeneity James Graham - PowerPoint PPT Presentation

Consumption Responses to House Price Heterogeneity James Graham RBNZ Housing Household Debt and Policy Conference 11 December 2017 1

Motivation

Motivation • How can we characterize house price variation faced by households? • Variation in the aggregate, across cities, across neighborhoods, and across individual houses • What is the consumption response to movements in different house price components? • Does consumption respond more to neighborhood price movements than to city price movements? • Are there stronger responses to idiosyncratic price shocks than other shocks? 2

Motivation • House prices may affect consumption through: 1. Substitution effects 2. Wealth effects 3. Collateral effects • Wealth effects may be small if higher house prices today imply higher housing costs in the future (Buiter (2008)) • Older households may experience strong wealth effects since they are likely to downsize housing or sell-up soon (Campbell and Cocco (2007)) • But consider consumption responses to price movements given: • Differential price movements across houses in different locations • Some probability of moving across those locations 3

Motivation 14 . 0 13 . 5 13 . 0 log(median house price) 12 . 5 12 . 0 11 . 5 11 . 0 San Francisco Riverside Fresno 10 . 5 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 Figure 1: Median house prices by zip code (Source: Zillow) 4

Motivation All households Homeowners 0 . 20 0 . 20 0 . 15 0 . 15 Proportion moving Proportion moving 0 . 10 0 . 10 0 . 05 0 . 05 0 . 00 0 . 00 20 30 40 50 60 70 80 20 30 40 50 60 70 80 Age Age Within county Across county, within state Across states Figure 2: Proportion moving across locations (Source: CPS) 5

Overview Motivation House price decomposition Consumption response to house prices Model Elasticity of consumption with respect to house prices Conclusion 6

House price decomposition

Housing data • Zillow Transaction and Assessment (ZTRAX) dataset 1 • Contains over 350 million residential and commerical property transactions, for over 2700 counties in the US, for up to 20 years • Detailed information on deed transfers, mortgages, foreclosures, property characteristics, geographic information, assessor valuations • Want housing market transactions by households: • Non-commerical, single family residences. • Housing sales only (no mortgages or refinancing) • Arm’s-length and non-foreclosure sales • Final data set: • 13 million individual transactions • 38 states + D.C. • 1993–2016 1 Calculated based on data from Zillow (US). The conclusions drawn from the Zillow data are those of the researchers and do not reflect the views of Zillow. Zillow is not responsible for, had no role in, and was not involved in analyzing and preparing the results reported herein. 7

ZTRAX average state prices vs. FRED all-transactions index CA CO CT DC 13 . 2 6 . 6 6 . 2 12 . 5 6 . 2 13 . 2 6 . 8 12 . 6 13 . 0 6 . 4 12 . 4 13 . 0 6 . 6 6 . 0 6 . 0 12 . 3 12 . 8 6 . 4 12 . 8 6 . 2 12 . 4 log-price index 12 . 2 12 . 6 6 . 2 log-price 5 . 8 12 . 6 6 . 0 5 . 8 12 . 2 12 . 1 12 . 4 6 . 0 12 . 4 5 . 8 5 . 6 12 . 0 12 . 2 5 . 8 5 . 6 12 . 0 12 . 2 5 . 6 11 . 9 12 . 0 5 . 6 5 . 4 12 . 0 5 . 4 11 . 8 11 . 8 5 . 4 11 . 8 5 . 4 11 . 8 5 . 2 5 . 2 11 . 7 11 . 6 5 . 2 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 DE FL GA HI 12 . 4 12 . 4 6 . 2 12 . 2 5 . 8 13 . 0 6 . 4 6 . 2 5 . 7 12 . 3 12 . 2 6 . 0 12 . 1 12 . 8 6 . 2 log-price index 5 . 6 12 . 2 6 . 0 12 . 0 5 . 8 12 . 0 log-price 12 . 6 6 . 0 5 . 5 12 . 1 11 . 8 5 . 6 11 . 9 5 . 8 5 . 4 12 . 4 5 . 8 12 . 0 11 . 6 5 . 4 11 . 8 5 . 3 5 . 6 12 . 2 5 . 6 11 . 9 11 . 4 5 . 2 11 . 7 5 . 2 11 . 8 5 . 4 11 . 2 5 . 0 11 . 6 5 . 1 12 . 0 5 . 4 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 IA ID IL IN 12 . 0 5 . 7 11 . 8 5 . 9 12 . 4 6 . 0 5 . 6 11 . 2 5 . 8 5 . 9 11 . 9 5 . 6 11 . 6 12 . 3 5 . 5 5 . 7 5 . 8 11 . 8 5 . 5 11 . 4 log-price index 12 . 2 11 . 0 5 . 6 5 . 7 log-price 5 . 4 11 . 7 5 . 4 11 . 2 5 . 5 12 . 1 5 . 6 10 . 8 11 . 6 5 . 3 11 . 0 5 . 3 5 . 4 5 . 5 12 . 0 11 . 5 5 . 2 10 . 8 5 . 3 5 . 4 10 . 6 5 . 2 11 . 9 11 . 4 5 . 1 10 . 6 5 . 2 5 . 3 11 . 3 5 . 0 10 . 4 5 . 1 11 . 8 5 . 2 10 . 4 5 . 1 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 KS KY LA MA 11 . 6 5 . 6 11 . 9 5 . 8 12 . 2 5 . 6 12 . 8 6 . 6 5 . 5 5 . 5 5 . 7 12 . 1 11 . 4 11 . 8 12 . 6 6 . 4 5 . 4 5 . 4 5 . 6 12 . 0 log-price index 11 . 2 11 . 7 12 . 4 5 . 3 log-price 6 . 2 5 . 3 5 . 5 11 . 9 11 . 0 11 . 6 5 . 2 12 . 2 5 . 2 5 . 4 11 . 8 6 . 0 5 . 1 10 . 8 11 . 5 12 . 0 5 . 1 5 . 3 11 . 7 5 . 0 5 . 8 10 . 6 11 . 4 11 . 8 5 . 0 5 . 2 11 . 6 4 . 9 10 . 4 4 . 9 11 . 3 5 . 1 11 . 5 4 . 8 11 . 6 5 . 6 8 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020 1995 2000 2005 2010 2015 2020

House price decomposition • Estimates of idiosyncratic price variation within a housing market/city: • σ ≈ 8 − 15% (Landvoigt et al. (2015), Giacoletti (2016)) • But prices fluctuate across individual houses, neighborhoods, and cities • Consider an unobserved error components model: log p m , z , i , t = β m , t X i , t + u t + v m , t + w z , t + ε i , t • X i , t : observable house characteristics (hedonic component) • u t : aggregate component • v m , t : city component (CBSA/MSA) • w z , t : neighborhood component (zip code) • ε i , t : idiosyncratic component 9

Cross-sectional variance of house price components 4 . 0 0 . 22 0 . 20 3 . 5 0 . 18 Zip code, idiosyncratic variance 0 . 16 3 . 0 City variance 0 . 14 0 . 12 2 . 5 0 . 10 0 . 08 2 . 0 0 . 06 1 . 5 0 . 04 2000 2002 2004 2006 2008 2010 2012 2014 2016 City (LHS) Neighborhood Idiosyncratic Figure 3: Variance of house price components (Source: ZTRAX) 10

Consumption response to house prices

Consumption response to house prices • Consumption data: Nielsen Consumer Panel 2 • 40,000-60,000 households, 2004-2015 • Households report all purchases tracked by Nielsen ( ∼ 30% of consumption categories) • Demographically balanced panel across: age, race, education, occupation, income, household size, presence of children • Observe the zip code and CBSA of a household • Merge with ZTRAX house price data 2 Calculated based on data from The Nielsen Company (US), LLC and marketing databases provided by the Kilts Center for Marketing Data Center at The University of Chicago Booth School of Business. The conclusions drawn from the Nielsen data are those of the researchers and do not reflect the views of Nielsen. Nielsen is not responsible for, had no role in, and was not involved in analyzing and preparing the results reported herein. 11

Consumption response to house prices log c m , z , i , t = α i + β x i , t + δ q m , t + γ a log p t + γ m log p m , t + γ z log p z , t + ε m , z , i , t • α i : household fixed effects • x i , t : observable household characteristics: • Age, education, race, marital status, household size • q m , t : city-level observables; • Real personal income per capita (BEA), unemployment rate (BLS). • p t : aggregate price • p m , t city price • p z , t : zip code price 12

Consumption response to house prices Dependent variable: log(real consumption expenditure) (1) (2) (3) (4) 0 . 079 ∗∗∗ 0 . 081 ∗∗∗ log( p t ) (0 . 008) (0 . 008) log( p m , t ) − 0 . 040 0 . 043 (0 . 033) (0 . 034) 0 . 077 ∗∗∗ 0 . 056 ∗∗∗ log( p z , t ) (0 . 019) (0 . 015) log( Y m , t ) − 0 . 285 ∗∗∗ − 0 . 343 ∗∗∗ − 0 . 381 ∗∗∗ − 0 . 321 ∗∗∗ (0 . 056) (0 . 052) (0 . 056) (0 . 052) − 2 . 443 ∗∗∗ − 2 . 417 ∗∗∗ log( U m , t ) 0 . 365 0 . 442 (0 . 223) (0 . 293) (0 . 270) (0 . 285) � � � � Household controls Household fixed effects � � � � Observations 966,997 966,997 966,997 966,997 R 2 0.705 0.704 0.704 0.705 Within R 2 0.04 0.04 0.04 0.04 13

Consumption response to house prices Dependent variable: log(real consumption expenditure) (1) (2) (3) (4) 0 . 079 ∗∗∗ 0 . 081 ∗∗∗ log( p t ) (0 . 008) (0 . 008) log( p m , t ) − 0 . 040 0 . 043 (0 . 033) (0 . 034) 0 . 077 ∗∗∗ 0 . 056 ∗∗∗ log( p z , t ) (0 . 019) (0 . 015) log( Y m , t ) − 0 . 285 ∗∗∗ − 0 . 343 ∗∗∗ − 0 . 381 ∗∗∗ − 0 . 321 ∗∗∗ (0 . 056) (0 . 052) (0 . 056) (0 . 052) − 2 . 443 ∗∗∗ − 2 . 417 ∗∗∗ log( U m , t ) 0 . 365 0 . 442 (0 . 223) (0 . 293) (0 . 270) (0 . 285) � � � � Household controls Household fixed effects � � � � Observations 966,997 966,997 966,997 966,997 R 2 0.705 0.704 0.704 0.705 Within R 2 0.04 0.04 0.04 0.04 14

Consumption response to house prices • Result: Neighborhood-level prices have a stronger association with consumption than city-level prices • Preliminary evidence in favor of hypothesis: • Neighborhood level prices have a stronger wealth effect, since households are more likely to move across neighborhoods than across cities • Need a structural model to check this intutition 15

Model