Import Competition and Household Debt Barrot (MIT) Plosser (NY Fed) Loualiche (MIT) Sauvagnat (Bocconi) USC Spring 2017 The views expressed in this paper are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System.
Research question Implications of product market competition for financial markets Focus on import competition Earlier work: implications for the cost of capital Firms in industries more exposed to globalization earn higher expected returns This paper: implications for household debt Workers may adjust balance sheet in response to displacement shocks
Motivation Rise of Chinese import competition in the US Net Chinese imports to the U.S. (billion dollars) 0 100 200 300 Chinese net imports to the U.S. 1985 1990 1995 2000 2005 2010 Year
Motivation Rise of US household leverage in the 2000s U.S. aggregate household debt-to-income ratio Household Debt over Income.8 1 1.2 1.4 1985 1990 1995 2000 2005 2010 Year
This paper Studies the effect of import competition on US household debt growth 2000-07 Measures regional exposure to import competition Uses shipping costs and historical industry composition Shipping cost predict Chinese import penetration Estimates household debt response in 2000-07 Considers mortgage, auto, credit card debt Differentiates new purchase versus refinancing loans Analyzes delinquencies and foreclosures in 2007-11 Not today: net aggregate effects Gains from variety, lower prices (Broda Weinstein 2006)
Main findings Areas more exposed to import competition experienced higher debt growth 2000-07 Total debt growth relative to 2000 Benchmarked to 2000 0.2.4.6.8 1 1.2 High exposure CZ Low exposure CZ 2000 2002 2004 2006 2008 Year DTI
Main findings Higher exposure to import competition: Lower employment/income growth 14%-20% of the cross-regional variation Higher debt growth 30% of the cross-regional variation Mechanism Driven by mortgage debt Driven by refinancing and home equity extraction Outcomes (2007-11) Higher likelihood of delinquencies Higher likelihood of foreclosures
Roadmap 1. Empirical design 1.1 Data 1.2 Identification strategy 2. Results 2.1 Labor market outcomes 2.2 Household debt 2.3 Equity extraction 2.4 Defaults 3. Discussion
Data Import competition Shipping Costs (SC) Markup of Cost-Insurance-Freight value over Free-on-Board value Structural characteristics well-grounded in the literature Related to weight-to-value ratio Large dispersion and high persistence across industries 30% of variation in trade costs in gravity regressions In the cross-section of industries, SC predict Import penetration in 2000-07 Import penetration Output & employment growth in 2000-07 Growth Summary Statistics
Data Import competition Commuting zone (CZ) level analysis CZs represent local labor markets (700+ in the US) Labor clusters estimated by USDA 4-5 counties per CZ Better coverage than MSA CZ exposure Interaction with historical industry composition SC Z = share J,Z SC J J industries Map
Data Household debt Equifax/FRBNY Consumer Credit Panel (CCP) Five percent of individuals with credit score Debt by type: mortgage, auto loans, credit card Demographic information: age, credit score, zipcode Home Mortgage Disclosure Act (HMDA) 90% of mortgage loans includes both refinancing and new purchase loans Panel Study of Income Dynamics (PSID) longitudinal sample of households includes both occupation and balance sheet information Health and Retirement Study (HRS): similar dataset with expectations Demographic variables and house prices IRS and Census: Income and demographics (zipcode) CoreLogic House Price Indices (most granular available) Summary Statistics
Empirical Strategy Specification Cross-sectional regressions at individual or CZ level: Y i,2000 07 = α + β ShippingCosts i + γ X i + ɛ i Where: ShippingCostsi : weighted average shipping costs for industries in CZ i measured in 1998 Xi : controls CZ controls (1998): Share of tradable employment, log employment, log income, log debt, debt-to-income ratio Prior trends: 91,99 CH import penetration, 91,99 Loan originations Census controls (2000): Vacancy rate, percent white, percent black, share of education<high school, share with high school diploma, unemployment rate, poverty rate, percent urban CZ-level regressions weighted by 2000 adult population.
Roadmap 1. Empirical design 1.1 Data 1.2 Identification strategy 2. Results 2.1 Labor market outcomes 2.2 Household debt 2.3 Equity extraction 2.4 Defaults 3. Discussion
Import Competition and Labor Market Outcomes Income growth One std. dev. in SC: 14% of std. dev. in income growth Average annual income growth, 2000-07 Shipping costs 1.449 0.741 0.730 0.694 (0.361) (0.272) (0.310) (0.266) HPI 0.199 (0.024) Census controls No No Yes Yes CZ controls No Yes Yes Yes Observations 715 715 715 715 R 2 0.052 0.312 0.377 0.574 Magnitude SC 0.031 0.016 0.015 0.015 Magnitude HP 0.034
Roadmap 1. Empirical design 1.1 Data 1.2 Identification strategy 2. Results 2.1 Labor market outcomes 2.2 Household debt 2.3 Equity extraction 2.4 Defaults 3. Discussion
Import Competition and Household Debt CZ-level regressions One std. dev. in SC: 30% of std. dev. in debt growth 2000 07 Log total Debt Shipping costs -3.258-2.354-2.670-2.748 (0.784) (0.745) (0.664) (0.638) HPI 0.304 (0.043) Census controls No No Yes Yes CZ controls No Yes Yes Yes Observations 733 733 733 733 R 2 0.051 0.416 0.488 0.574 Magnitude SC -0.069-0.050-0.057-0.058 Magnitude HP 0.052 DTI
Import Competition and Household Debt CZ-level regressions Effects driven by mortgage debt 2000 07 Log debt Total Mortgage Credit mortgage loans HELOC Auto card Shipping costs -3.276-2.469-5.841-0.797-1.327 (0.710) (1.498) (3.005) (0.675) (0.448) CZ controls Yes Yes Yes Yes Yes Census controls Yes Yes Yes Yes Yes Observations 733 730 683 733 733 R 2 0.490 0.477 0.254 0.363 0.327 Magnitude SC -0.069-0.052-0.124-0.017-0.028
Import Competition and Household Debt Individual-level regressions Similar findings at the individual level 2000 07 Log debt +1 Shipping Costs -1.738-2.670-1.854-1.956-2.046 (0.750) (0.664) (0.642) (0.644) (0.625) HPI 0.180 (0.050) Risk Bins No No No Yes Yes Age Bins No No No Yes Yes Census controls No Yes Yes Yes Yes CZ controls No Yes Yes Yes Yes State FE Yes Yes Yes Yes Yes Observations 5,747,023 5,631,162 5,128,389 5,128,389 5,128,389 R-Squared 0.002 0.002 0.076 0.079 0.079 DTI
Import Competition and Household Debt Individual-level data: PSID Shipping cost measured at the worker s industry level Change 1999-2007 Unemployed Log(labor inc.+1) Shipping costs -0.61-0.59 9.36 10.58 (0.33) (0.37) (3.24) (3.54) Individual level controls Yes Yes Yes Yes State fixed effect No Yes No Yes Observations 719 719 719 719 R-Squared 0.110 0.149 0.217 0.257 Log(debt+1) DTI Shipping costs -11.35-12.35-6.88-8.13 (5.16) (5.33) (3.12) (3.58) Individual level controls Yes Yes Yes Yes State fixed effect No Yes No Yes Observations 719 719 719 719 R-Squared 0.240 0.287 0.115 0.155 2SLS
Import Competition and Household Debt Robustness Credit supply? 1. SC positively correlated with denials Denials 2. No effect in high unemployment insurance areas UI 3. No effect on building permits Permits 4. No effect on corporate loans Corporate 5. SC uncorrelated with house prices HPI Alternative specifications? 1. Weight-to-value ratio WVR 2. Weighted by the share of tradable industries Weights 3. Instrumented regressions Instrumented 4. Other proxies for import penetration Alternatives (1/2) 5. Other ways of computing SC Alternatives (2/2)
Roadmap 1. Empirical design 1.1 Data 1.2 Identification strategy 2. Results 2.1 Labor market outcomes 2.2 Household debt 2.3 Equity extraction 2.4 Defaults 3. Discussion
Import Competition and Equity Extraction HMDA - refinancing loans Effect is going through refinancing loans Home Purchase 2000 07 Log Applications Refinancing Number (#) Value ($) Number (#) Value ($) Shipping costs -1.078-1.190-2.979-4.183 (0.985) (1.018) (1.235) (1.281) CZ controls Yes Yes Yes Yes Census controls Yes Yes Yes Yes Observations 733 733 733 733 R 2 0.458 0.572 0.638 0.779 SC Magnitude -0.023-0.025-0.063-0.089 Graph Originations
Import Competition and Equity Extraction Bhutta and Keys 2016 Direct evidence of equity extraction Log(Home equity extraction 2000-07) Instrumented House price appreciation 2000-07 All High Low Shipping Costs -6.422-18.986-2.189 (1.709) (3.069) (1.477) Risk Bins Yes Yes Yes Age Bins Yes Yes Yes Census controls Yes Yes Yes CZ controls Yes Yes Yes State FE Yes Yes Yes Observations 3,069,768 1,605,283 1,464,485 R-Squared 0.070 0.055 0.065 Instrument Magnitude 0.063 0.132 0.025
Roadmap 1. Empirical design 1.1 Data 1.2 Identification strategy 2. Results 2.1 Labor market outcomes 2.2 Household debt 2.3 Equity extraction 2.4 Defaults 3. Discussion
Import Competition and Defaults Mortgage Delinquencies and Foreclosures Higher probability of delinquency and foreclosure Delinquency Foreclosure 2001-2007 2008-2011 2001-2007 2008-2011 Shipping Costs -0.017-0.206-0.083-0.257 (0.050) (0.073) (0.030) (0.068) Risk Bins Yes Yes Yes Yes Age Bins Yes Yes Yes Yes Census controls Yes Yes Yes Yes CZ controls Yes Yes Yes Yes State FE Yes Yes Yes Yes Observations 5,128,389 5,044,524 5,128,389 5,047,699 R-squared 0.133 0.092 0.124 0.091 Graph Interaction with HPI
Roadmap 1. Empirical design 1.1 Data 1.2 Identification strategy 2. Results 2.1 Labor market outcomes 2.2 Household debt 2.3 Equity extraction 2.4 Defaults 3. Discussion
Discussion Interpretation of the results What hypotheses can we reject? Permanent income hypothesis (PIH) Households use debt to smooth transitory income shocks Cross-sectional implications: Effects driven by those for whom shock was transitory Effects driven by those who perceived the shock as transitory Ratchet effects, consumption commitments Households use debt to smooth any income shock Cross-sectional implications: Unclear
Discussion Response by age Effect concentrated among high middle-aged individuals DTI Shipping Cost Coefficient by Age Shipping Cost Coefficient -3-2 -1 0 1 20 30 40 50 60 Median 1999 Age by Bin Log debt
Discussion Response by credit score Effect concentrated among high credit score individuals DTI Shipping Cost Coefficient by Credit Score Shipping Cost Coefficient -4-3 -2-1 0 1 500 600 700 800 Median 1999 Credit Score by Bin Log debt
Discussion Response by zipcode income Effect concentrated among high income zipcodes DTI Shipping Cost Coeff. by 1998 Income Shipping Cost Coefficient -8-6 -4-2 0 20000 40000 60000 80000 Median 1998 Income by Bin Log debt
Discussion Response by zipcode education Effect concentrated among highly educated zipcodes DTI Shipping Cost Coeff. by Share <College Shipping Cost Coefficient -6-4 -2 0.4.5.6.7.8.9 Median Share <College by Bin Log debt
Conclusion Key takeaways Exposure to import competition explains 30% of the cross-regional increase in household debt in 2000-07 Mortgage debt used to smooth displacement shocks A demand-driven channel for the rise in household debt Next steps Explore the mechanism further Measure/study expectations
International Trade Exposure and DTI Debt-to-income ratio Change in debt to income ratio relative to 2000 Benchmarked to 2000 0.1.2.3.4 High exposure CZ Low exposure CZ 2000 2002 2004 2006 2008 Year
Average SC by Commuting Zone USA Trade Exposure Average shipping costs by Commuting Zones Average Shipping Costs 3.5 4.5 5.5 6.5 15.3 Geometries: Census; Data: CBP, Feenstra, Schott