Determinants of US Household Debt

Transcription

1 Determinants of US Household Debt New Evidence from the SCF Rafael Wildauer The Challenge of Inequality Societies in Crisis AK Young Economists Conference 1 October 2015, Vienna

2 Determinants of Household Debt increasing interest in household debt since financial crisis most literature about effects of (household) debt, much less on drivers of debt aim of this paper: compare two explanations of rising debt expenditure cascades: due to status driven consumption housing-driven-debt: due to property price rallies

3 Investigating Household Debt A simple theoretical framework and a summary of the existing literature

4 Simple Accounting The household budget constraint: D it = α (C it Y it ) + β 1 REP it + FAP it + β 2 I it consumption related asset related REP it value of real estate purchases FAP it value of debt-financed purchases of financial assets I it investment in non-incorporated businesses β s debt ratio in asset transactions

5 Two Popular Hypotheses expenditure cascades consumption related Frank et al. 2014, Behringer and van Treeck 2013 housing-driven-debt asset related Mian and Sufi 2011, Dynan and Kohn 2007

6 Empirical Literature: HH Debt authors Specification data findings Cynamon and Fazzari 2015 descriptive, top 5% vs bottom 95%, total household debt US, triannual , NIPA and SCF positive relation between inequality (slow income growth for 95%) and debt of 95% Carr and Jayadev Δln(D/Y)=f(Y, Y~, X): total US (PSID), positive effect of Y~ 2014 household debt Behringer and van Treeck 2013 HFB=f(Y, top5, NFA, fiscal) HFB hh financial balance NFA net foreign assets G7 countries, Income inequality leads to lower household net lending and current account deficits Mian and Sufi 2011 growth(d/y)=f(pp, X, ), instrument PP with housing US, proprietary micro data, Important effects of property prices on borrowing. supply elasticity Boushey and Weller 2008 D/Y=f(Y, Atkinineq, equity, r): credit card debt US quarterly, positive effect of Atkin inequality measure Christen and Morgan 2005 D/Y=f(D/Y(-1), C, ΔY,GINI): total household debt US quarterly, Positive effect of Gini

7 Empirical Literature: Summing Up evidence for expenditure cascades income inequality: positive effects on C and debt evidence for housing-driven-debt hypothesis property prices: positive effects on C and debt rarely investigated simultaneously most surveys don t capture tail of income distribution

8 Data: Survey of Consumer Finances (SCF)

9 Survey of Consumer Finances household survey on behalf of US Fed focus on: assets, debt and income oversampling of rich households using tax data 4,300 to 6,500 observations 3 year intervals ( )

10 Debt to Income Ratios

11 Survey of Consumer Finances key limitation: no panel structure (repeated cross section) D it = D it D it 1 does not work D it 1 not observed work around use household information to construct D it (and thus also D it 1 )

12 Constructing D it 10 debt categories Example: first mortgage on primary residence 6 core pieces of information: year mortgage was taken out/refinanced (tb) how money was used (use) (refinance, equity extract, no prior mortgage) current regular payments (pd) (implicit) current interest payments (rd) amount initially borrowed (B) amount currently outstanding (D)

13 step 1: tb? year ᴧ use =? case 1: case 2: case 3: case 4: tb=year ᴧ use=ref tb=year ᴧ use=ex tb=year ᴧ use=0 tb<year implies: ΔD=D-B implies: ΔD=ex+D-B implies: ΔD=D no conclusion yet about ΔD step 2: P? rd case 4.2: tb<year ᴧ P>rD case 4.1: tb<year ᴧ P rd implies: ΔD=0 step 3: B? D case 4.2.1: B D ᴧ tb<year ᴧ P>rD implies: ΔD=rD-P case 4.2.2: B=D ᴧ tb<year ᴧ P>rD implies: ΔD=0

14 The Econometric Model

15 Specification ln D it = α + β 1 ln Y it + β 2 ln C it + β 3 ln HW it + β 4 ln REP it + β 5 ln D it 1 + β 6 ln FW it + β 7 BUSd it + β 8 CCd it +β 9 normy it + β 10 X it + year t + ε it D it household debt Y it income C it reference group consumption HW it housing wealth REP it value of real estate bought in current year FW it financial wealth BUSd it business founding dummy X it demographic controls (age, education)

16 Definition of C it - Part I

17 Definition of C it - Part II SCF does not provide consumption data expenditure cascades hypothesis holds: debt financed consumption thus C it = Y it + D it

18 Results

19 Results I

20 competing theories / SCF / model / results / conclusion

21 Summing Up positive but statistically insignificant effects of C it positive and statistically significant effects of HW it and REP it housing seems to be binding constraint on borrowing

22 Behringer & van Treeck (2013): Income distribution and current account: A sectoral perspective. IMK Working paper 125. Belabed et al. (2013): Income distribution and current account imbalances. IMK Working Paper 126. Carr & Jayadev (2014): Relative Income and Indebtedness: Evidence from Panel Data. Review of Income and Wealth, online access. Cynamon & Fazzari (2015): Inequality, the Great Recession and slow recovery. Cambridge Journal of Economics. online access. Frank et al. (2014): Expenditure Cascades. Review of Behavioral Economics, 1(1-2), Kapeller & Schütz (2014): Debt, Boom, Bust: A Theory of Minsky-Veblen Cycles. Journal of Post-Keynesian Economics. 36(4), Mian & Sufi (2009): The consequences of Mortgage Credit Expansion: Evidence from the U.S. Mortgage Default Crisis. Quarterly Journal of Economics, 124(4), Mian & Sufi (2011): House prices, home equity-based borrowing, and the US household leverage crisis. American Economic Review, 101(5),

23 Appendix

24

25 Interactions

26

27