Wealth Effects (Plural) and U.S. Consumer Spending *

Transcription

1 Wealh Effecs (Plural) and U.S. Consumer Spending * John Duca, Federal Reserve Bank of Dallas & Oberlin College John Muellbauer, Oxford Universiy & INET Anhony Murphy, Federal Reserve Bank of Dallas December 2017 * Corresponding auhor: anhony.murphy@dal.frb.org. The views expressed are hose of he auhors, and are no necessarily hose of he Federal Reserve Bank of Dallas or of he Federal Reserve Sysem. 1

2 Inroducion Convenional models, in which long-run consumpion depends on wealh, ineres raes and permanen income, canno accoun for: Secular decline in he saving rae from lae 1970s o 2007 The behavior of consumpion and savings since hen. Need o accoun for he evolving credi marke archiecure of U.S. household finance in order o do so Idenify and quanify wo imporan financial innovaions conribuing o household financial acceleraors: Changing consumer credi sandards Changing liquidiy of housing wealh, he housing wealh ( collaeral) effec Esimaed wealh effecs differ by asse: Approx. 8% for liquid asses minus consumer and morgage deb Approx. 1.5% for socks and oher illiquid financial asses Housing collaeral: Rising from abou 1% in early 1980 s o over 3¾% in he mid 2000 s; Falls o under 2% in 2015; Abou 2¼% in mid

3 Figure 1: Wealh-o-Income Raio Alone Canno Accoun for Saving Rae Trends Percen Personal Savings Rae Ne Wealh / Disposable Personal Income Raio

4 Figure 2: Consumer Credi Availabiliy Conribued o Recovery from Grea Recession CCI Index Tougher Bank Reguaion Deposi Deregulaion & Rise of Credi Scoring Technology Recovery From Grea Recession 0.0 Spread of Credi Card & Insalmen Credi Noes: Consumer Credi Index (CCI) derived from cumulaed exogenous componen of he responses o he Senior Loan Officer Opinion Survey response o willingness o make consumer insallmen loan quesion. 4

5 Sae Space Model of Consumpion and Refinancing We use a credi augmened consumpion funcion We consider forward looking consumers who are aware of heir ineremporal budge consrain We do no assume a represenaive agen wih raional expecaions ec. as in mos Euler equaions In sandard life-cycle, permanen income model consumpion C is p proporional o permanen income Y and wealh W, which implies: P lnc = α0 + lny + γw 1 Y + ln( Y Y) + u where Y is labor and ransfer income Adding expeced income growh and ineremporal subsiuion yields: P lnc = α0 α1r + lny + γw 1 Y + Eln( Y Y) + u Parial adjusmen dynamics arise by adding habis, raional inaenion ec. * ln C = λ(lnc ln C ) + u 1 P = λα { αr + lny lnc + γw Y + E ln( Y Y )} + u

6 Credi Augmened Consumpion Funcion (Con d) Add measure of consumer credi condiions (CCI) Disaggregae wealh ino ne liquid asses (NLA = liquid asses consumer & morgage deb), ne illiquid financial asses (NIFA) and gross housing wealh (HSG) Allow variable liquidiy of housing wealh (HLI) Add uncerainy and oher credi consrain effecs: ln = λα + ln ln + α + α + αln p C Y C r CCI Y Y { ( ) ( ) γ1nla 1 Y + γ2nifa 1 Y + γ3hsg 1 Y + HLI HSG 1 Y + β1 ur + β2 i + + u where ur is he unemploymen rae, i is he nominal ineres rae The unobserved liquidiy of housing wealh sae variable evolves as HLI = HLI 1 + ε, where ε is a whie noise, normal innovaion Use sae space model of permanen income wih 10 year horizon à la Friedman (1956,1963), Deaon (1991, 1992) and Carroll (2001), and a high 5% discoun rae à la Skinner (1988), Zeldes (1989), Kimball (1990) and Carroll (1997, 2001) } 6

7 Morgage Refinancing Equaion HLI also appears in refi equaions which helps us pin down he housing liquidiy effec: refi = φ refi + φ HLI + z δ + HLI z δ + v z δ δ 0 + δ 1 PosGap 1 + δ 2 PosGap 2 + δ 3 Payback + δ 4 Low + δ 5 MorForeclose where refi is share of agency morgages ha are refinanced HLI is posiively relaed o he abiliy (willingness) o refinance, and negaively o he cos of refinancing As HLI, so does refi and he response of refi o ineres rae gaps ec. Iner alia, he z δ erm capures he incenive o refinance: PosGap = ousanding minus new morgage rae gap (if posiive) Low = 30 quarer morgage rae low 0/1 indicaor Payback = capures end of morgage refinancing booms - ineracion of number of morgage rae lows in he 8 prior quarers and an indicaor if prior quarer marked an ineres rae low bu curren quarer is no MorForeclose = Percen of morgages enering foreclosure 1 7

8 Figure 3: Refi Response o Morgage Ineres Rae Differenials Varies Over Time Percen 2.4 Percen Posiive Gap: Exising vs New Morgage Rae Agency Refi Rae

9 Housing Wealh vs Housing Collaeral Effecs In sylized Classical model (perfec capial markes, dynasic Ricardian households), house prices have a small negaive effec on non-housing consumpion Posiive esimaed housing wealh effec in U.S. daa may arise from: Non-raional expecaions Non-dynasic family behavior - lile evidence of sronger housing wealh effec for older households Omied fuure income expecaions, because permanen income no curren income maers Credi consrains (more imporan for young and alered by financial innovaions) HLI allows for a collaeral role for housing o affec consumpion, so HLI HSG 1 Y maers, no HSG 1 Y 9

10 Table 1: Sae Space Model of Consumpion and Morgage Refinancing Consumpion Funcion: Refinancing Equaion: ln = λα + ln ln + α + α + αln p C Y C r CCI Y Y { ( ) ( ) γ NLA Y + γ IFA Y + γ HSG Y + HLI HSG Y β ur + β i + β CrConrols + u u N(0, σ ) u 2 refi = φ refi + φ HLI + z δ + HLI z δ + v v N(0, σ ) z δ δ + δ PosGap + δ PosGap + δ Payback + δ Low δ MorForeclose 5 Sae Equaion for Liquidiy ( mpc ) of Housing Wealh: HLI 1 2 = HLI 1 + ε ε NID(0, σ ε ) 4 v } 10

11 Consumpion Funcion: Table 2: Parameer Esimaes (1972 q q2) *** * * ** ln = ln ln ln p C { ( Y C ) r CCI ( Y Y ) (8.97) (1.29) (1.70) (1.80) (2.28) *** ** NLA Y IFA Y + HLI HSG Y (2.93) (2.03) *** *** i *** + σ = *** CrConrols u u (3.97) (4.32) (4.65) (13.25) ur Refinancing Equaion: refi = refi HLI + ˆ δ + z HLI z δ + vˆ vˆ = *** *** *** 1 (18.79) (2.95) (22.66) *** *** *** *** z δ (0.715 PosGap PosGap Payback Low 1 2 ( 3.84) (4.09) (5.40) (7.78) (1.68) * MorForeclose ) / Sae Equaion Housing Liquidiy: HLI = HLI + ε ˆ σ ε = } (5.28) *** ˆ ˆ 11

12 Figure 3: Housing Liquidiy / Wealh Effec Smoohed Sae Esimaes mpc Rise in Cash Ou Refis Subprime Bus Rise of 2nd Morgages Basel I Rise in Capial Raios Higher GSE Cash Ou Refi Fees HLI (Smoohed Sae) ± 2 RMSE

13 Esimaed Wealh Effecs Esimaed wealh mpcs: 7.9% for ne liquid asses 1.4% for socks and illiquid financial asses Esimaed ime varying housing collaeral effec: Close o zero in he early 1970 s Approx. 1% in he lae 1970 s and early 1980 s Jumps in he mid 1990 s, peaking a over 3¾% in he mid 2000 s Falls back o under 2% by 2015 Recovered o abou 2¼% in mid-2107 Overall mpc somewha smaller, more variable, han convenional 3% figure Housing collaeral effec no a radiional housing wealh effec, consisen wih micro sudies, e.g. Hurs and Safford (2004), Cooper (2009), and Browning, Gorz & Leh-Peersen (2008) Esimaed housing mpc smaller han in some recen sudies e.g., long run 9% mpc in Carroll, Osuka & Slacalek (2011). Why? We conrol for permanen income and consumer credi availabiliy 13

14 Table 3: Esimaed Wealh Effecs and Specificaion of Consumpion Funcion (1972 q q2) MPC Adj. SE x Model / Specificaion Speed R NLA NIFA HSG ˆλ Baseline wih W 3.0% Disaggregae W ino NLA,NIFA,HSG 10.2% 2.3% 6.6% Add CCI, HLI and Δur 8.2% 1.4% HLI Convenional ARDL in C, Y, W, r, Δur 2.7% Noes: W = ne wealh, NLA/NIFA/HSG = dis-aggregaion of ne wealh, CCI = credi condiions index, HLI = esimaed housing liquidiy index from 2 equaion sae space model, ur = unemploymen rae. The baseline and auoregressive disribued lag (ARDL) models are: lnc = λα + lny lnc + αr + αcci + αln Y p Y + γw Y + β i { 0 ( 1) ( + 1 ) 1 } 2 + β3crconrols + u, and lnc = β + β lnc + β lnc + β lny + β lny + β ( lnc lny ) + β W Y β ur + β r + β CrConrols + u

15 Table 4: Changing Credi Condiions and Wealh Imporan Drivers of Changes in Savings Rae Esimaed Long Run Effecs on Ln C/Y (pps) Period Housing & Sock Bubbles 1995q q4 Housing & Financial Crises 2006 q q2 Modes Recovery 2009 q q4 Moderae Growh 2013 q q2 Acual Change in CY Raio Esimaed Long Run Credi & Wealh Effecs Consumer Credi Availabiliy (CCI) Conribuion o Esimaed Effecs Liquid Asses (Consumer + Morgage Deb) Sock + Oher Ne Illiquid Asses Housing Asses 4.8% 4.8% 0.9% -4.4% 2.1% 6.5% -5.4% -4.7% -0.1% 2.7% -1.8% -5.4% 1.7% 0.8% 0.6% -0.2% 1.4% -1.4% 3.5% 3.3% 0.6% 1.0% 1.3% 1.4% The esimaed long run effec is: p r CCI ln ( Y+ 1 Y) + (0.079NLA NIFA 1+ HLI HSG 1) Y Muliply enries by minus wo-hirds o ge approx. savings rae effecs 15

16 Summary Financial innovaions and fricions maer, especially in gauging he impac of wealh componen and rends in he personal saving rae. Imporan roles for: Changing composiion of ne wealh Exogenous changes in supply of consumer credi (CCI) Changing liquidiy of housing wealh (sae space HLI esimaes) No single wealh effec : Liquid asses: 8% Consumer and morgage deb: -8% Socks & oher illiquid financial asses: 1½% Housing collaeral: Rising from 1% in early 1980 s o over 3¾% in he mid 2000 s Falls o under 2% in 2015 and recovers o 2¼% in mid-2107 In a housing boom, he collaeral effec of higher housing wealh firs booss consumpion, bu he negaive deb overhang effecs linger afer a bus 16