Dissecting Saving Dynamics

Transcription

1 Dissecting Saving Dynamics Measuring Credit, Wealth and Precautionary Effects Christopher Carroll 1 Jiri Slacalek 2 Martin Sommer 3 1 Johns Hopkins University and NBER ccarroll@jhu.edu 2 European Central Bank jiri.slacalek@ecb.int 3 International Monetary Fund msommer@imf.org Presentation at Goethe-Universität Frankfurt, May 2012

2 US Personal Saving Rate (s), Percent of Disposable Income

3 Literature Wealth Effects Modigliani, Klein, MPS model,... s t = 0.05m t + other stuff Precautionary Carroll (1992) Saving rate rises in recessions log C t+1 strongly related to E t(u t+1 u t) Credit Availability Secular Trend: Parker (2000), Dynan and Kohn (2007), Muellbauer (many papers) Cyclical Dynamics: Guerrieri and Lorenzoni (2011), Eggertsson and Krugman (2011), Hall (2011)

4 Great Recession s rises by 4 5 pp Bigger & more persistent increase than any postwar recession But all three indicators also move a lot: Credit conditions tighten Unemployment Expectations rise Wealth falls

5 Personal Saving Rate 2007 Deviation from Start of Recession Value in % Quarters after Start of Recession Historical Range Historical Mean

6 Household Wealth 2007 by 150% of Income Deviation from Start of Recession Value Quarters after Start of Recession Historical Range Historical Mean Recession

7 Sustained Expectations of Rising Unemp Risk Thomson Reuters/University of Michigan E t (u t+4 u t )

8 Tighter HH Credit Supply (Based on Muellbauer)

9 Our Contributions Theory Simple model with transparent role for all 3 channels Qualitative implications of the model Overshooting possible role for fiscal policy Evidence Quantify importance of the 3 channels Two estimated models of s Reduced-form OLS Structural Nonlinear least squares

10 Why Do We Care? Quantify role of credit, wealth and uncertainty Useful for in-sample and out-of-sample analysis Strength of recovery/dynamics of GDP

11 Theory à la Carroll and Toche (2009) CRRA utility, labor supply l, agg wage W, emp status ξ: v(m t ) = max c t [ u(c t ) + βe t v(mt+1 ) ] s.t. m t+1 = (m t c t )R + l t+1 W t+1 ξ t+1 ξ t+1 {ξ u, ξ e } where ξ u < ξ e l and W grow at constant rate Tractability: unemployment shocks are permanent If ξ t = ξ u then ξ t+1 = ξ u Target wealth ˇm exists and is stable: Consumption chosen so that mt ˇm

12 Consumption Function c t e e c t 1 0 e m t 1 0 Steady State c e m Stable Arm m t

13 Target Wealth ˇm Closed-form solution for target wealth depends on unemployment risk and generosity of unemployment insurance ξ u : ˇm = f (, ξ u, preferences,... ) (+) ( )

14 Consumption After a Wealth Shock c e m t 1 0 Target c t c t 1 c m Wealth Shock m t m

15 Permanent Rise in c Sustainable c Target c m c m after unemployment rate increase m m

16 Saving Rate After a Permanent Rise in s s t ' Overshooting Time

17 Overshooting and Fiscal Policy DSGE models: Frictions, frictions everywhere; but missing here If c imposes external costs Sticky prices/wages Capital (or Investment) adjustment costs Other reasons for pecuniary externalities stimulus payments, fiscal policy may reduce cost of cycle Justification for automatic stabilizers?

18 Credit Easing/Financial Innovation & Deregulation c New c m Orig c m e m t 1 0 Orig Target h 0. ˇm is close to linear in credit conditions

19 Data & Sources Quarterly 1966Q2 2011Q1 Saving rate: BEA NIPA Net worth: Flow of Funds Accounts, Fed (Model m corresponds to 1 + ratio of Net worth to disposable income) Credit conditions: Credit Easing Accumulated, CEA Senior Loan Officer Opinion Survey (SLOOS), Fed Banks willingness to provide consumer installment loans Unemployment risk: using Thomson Reuters/UMichigan unemployment expectations

20 Net Worth (Ratio to Quarterly Disp Income)

21 Credit Easing Accumulated (CEA) (à la Muellbauer) Accumulated responses, weighted with debt income ratio, to: Please indicate your bank s willingness to make consumer installment loans now as opposed to three months ago

22 t Implied by Michigan U Expectations Regress: 4u t+4 = α 0 + α 1UExp t U risk: t = u t + 4û t+4 4u t+4 u t+4 u t, 4û t+4 fitted values t tracks but precedes actual U UExp: How about people out of work during the coming 12 months do you think that there will be more unemployment than now, about the same, or less?

23 Reduced-Form Regressions s t = γ 0 +γ m m t +γ CEA CEA t +γ Eu E t u t+4 +γ t t +γ uc (E t u t+4 CEA t )+ε t Model Time Wealth CEA Un Risk All 3 Baseline Interact γ (0.61) (1.73) (0.57) (0.42) (2.56) (2.16) (2.56) γ m (0.32) (0.42) (0.35) (0.46) γ CEA (1.74) (1.94) (0.57) (1.72) γ Eu (0.05) (0.12) (0.08) (0.11) γ t (0.00) (0.00) (0.01) (0.00) (0.01) (0.01) γ uc 0.32 (0.16) R F stat p val DW stat

24 Fit: Baseline vs Time Trend

25 Model Baseline Uncert s t 1 Debt Full Controls Post-80 IV γ m (0.35) (0.36) (0.22) (0.36) (0.31) (1.25) (0.49) γ CEA (0.57) (0.65) (0.53) (0.73) (0.63) (2.00) (1.17) γ Eu (0.08) (0.09) (0.05) (0.07) (0.09) (0.14) (0.13) γ σ 0.26 (0.47) γ s 0.57 (0.07) γ d 1.91 (1.16) γ r 0.13 (0.04) γ GS 0.12 (0.08) γ CS 0.31 (0.14) γ 0post (7.90) γ mpost (1.29) γ CEApost (2.13)

26 Fit: Baseline vs Post

27 Fit: Baseline vs Full Controls

28 Reduced-Form Regressions Summary The three factors explain saving well: 1. Credit conditions 2. Wealth 3. Unemployment risk

29 Structural Estimation Nonlinear Least Squares Minimize distance between model-implied s theor t and actual s meas t : ˆΘ = arg min where T ( t=1 s meas t Θ = {β, θ m, θ CEA, θ, θ u } st (Θ; theor m t ˇm ( m(cea t ), (E t u t+4 ) ))) 2, m t = θ m + θ CEA CEA t t = θ + θ u E t u t+4 β: discount factor

30 Structural Estimation Asymptotics Delta Method where D = E qt(θ) Θ E = var ( q t (Θ) ) Scores q t (Θ) = ( s meas t T 1/2 ( ˆΘ Θ) d N (0, D 1 ED 1 ), s theor t (Θ) Θ (Θ) ) st theor

31 Structural Estimates s theor t = st theor ( Θ; mt ˇm( m t, ) t), m t = θ m + θ CEA CEA t, t = θ + θ ue tu t+4. Parameter Description Value Calibrated Parameters r Interest Rate 0.04/4 W Wage Growth 0.01/4 ρ Relative Risk Aversion 2 Estimated Parameters Θ = {β, θ m, θ CEA, θ, θ u} β Discount Rate (0.0018) θ m Scaling of m t (0.0206) θ CEA Scaling of m t (0.1396) θ Scaling of t ( ) θ u Scaling of t (0.1227) R DW stat 0.950

32 Estimated Extent of Credit Constraints m t (Frac of DI)

33 Estimated Permanent Unemployment Risk t 9.5 x

34 Fit of the Structural Model 12 Actual PSR Fitted PSR

35 Decomposition of Fitted PSR Fix t and CEA t at their sample means, back out the implied s t 12 Fitted PSR Fitted PSR excl. Uncertainty Fitted PSR excl. Uncertainty and CEA

36 Fit: Structural Model vs Reduced-Form Actual Reduced Form Structural

37 Reduced-Form Regressions on Model Data s theor t = γ 0 +γ m m t +γ CEA CEA t +γ Eu E t u t+4 +γ t t+γ uc (E t u t+4 CEA t )+ε t Model Time Wealth CEA Un Risk All 3 Baseline Interact γ (0.50) (1.11) (0.41) (0.16) (0.60) (0.53) (0.55) γ m (0.25) (0.12) (0.10) (0.11) γ CEA (1.12) (0.59) (0.14) (0.47) γ Eu (0.02) (0.04) (0.02) (0.03) γ t (0.00) (0.00) (0.01) (0.00) (0.00) (0.00) γ uc 0.19 (0.04) R F stat p val DW stat

38 Reduced-Form Regressions on Actual Data s meas t = γ 0 +γ m m t +γ CEA CEA t +γ Eu E t u t+4 +γ t t+γ uc (E t u t+4 CEA t )+ε t Model Time Wealth CEA Un Risk All 3 Baseline Interact γ (0.61) (1.73) (0.57) (0.42) (2.56) (2.16) (2.56) γ m (0.32) (0.42) (0.35) (0.46) γ CEA (1.74) (1.94) (0.57) (1.72) γ Eu (0.05) (0.12) (0.08) (0.11) γ t (0.00) (0.00) (0.01) (0.00) (0.01) (0.01) γ uc 0.32 (0.16) R F stat p val DW stat

39 Structural Estimation Summary Model fits well almost as well as reduced form (Mincer Zarnowitz puts weight 0.45 on structural model) Substantial role for time-varying precautionary saving CEA matters for low frequency, wealth for business-cycle frequency

40 PSR Forecasts In Sample Great Recession Variable Reduced-Form Model Structural Model Actual s t γ m m t = = 1.34 γ CEA CEA t = = 0.67 γ Eu E tu t = = 1.39 Explained s t

41 PSR Forecasts Out of Sample (percent of disposable personal income) Baseline Scenario 2 Upside Risk Scenario Downside Risk Scenario Fitted values of model Scenarios based on SPF and our judgement

42 Conclusions All three effects present Easier borrowing largely explains secular decline s Order of importance in Great Recession: 1. Wealth shock 2. Labor income risk 3. Credit tightening PSR to remain elevated

43 References Carroll, Christopher D. (1992): The Buffer-Stock Theory of Saving: Some Macroeconomic Evidence, Brookings Papers on Economic Activity, 1992(2), , Carroll, Christopher D., and Patrick Toche (2009): A Tractable Model of Buffer Stock Saving, NBER Working Paper Number 15265, Dynan, Karen E., and Donald L. Kohn (2007): The Rise in US Household Indebtedness: Causes and Consequences, in The Structure and Resilience of the Financial System, ed. by Christopher Kent, and Jeremy Lawson, pp Reserve Bank of Australia. Eggertsson, Gauti B., and Paul Krugman (2011): Debt, Deleveraging, and the Liquidity Trap: A Fisher-Minsky-Koo Approach, Manuscript, NBER Summer Institute. Guerrieri, Veronica, and Guido Lorenzoni (2011): Credit Crises, Precautionary Savings and the Liquidity Trap, Manuscript, MIT Department of Economics. Hall, Robert E. (2011): The Long Slump, AEA Presidential Address, ASSA Meetings, Denver. Parker, Jonathan A. (2000): Spendthrift in America? On Two Decades of Decline in the U.S. Saving Rate, in NBER Macroeconomics Annual 1999, ed. by Ben S. Bernanke, and Julio J. Rotemberg, vol. 14, pp NBER.

44 Background Slides

45 Alternative Measures of Credit Availability CEA/Debt Income Ratio Abiad et al. Index of Financial Liberalization

46 Assumptions/Scenarios for Out-of-Sample Forecasts 700 Household net wealth (percent of disposable personal income) Unemployment rate (percent of labor force) Baseline scenario 6 Upside risk scenario Baseline scenario Upside risk scenario Downside risk scenario Credit conditions Downside risk scenario Unemployment expectations Household saving rate Baseline scenario Upside risk scenario Downside risk scenario (percent of disposable personal income) 8

47 6 Upside risk scenario 6 Assumptions/Scenarios for Out-of-Sample Downside risk Forecasts Baseline scenario Upside risk scenario Downside risk scenario Credit conditions 400 Baseline scenario scenario Unemployment expectations Household saving rate Baseline scenario Upside risk scenario Downside risk scenario (percent of disposable personal income) Baseline Scenario Upside Risk Scenario Downside Risk Scenario Fitted values of model 2 Sources: Haver Analytics and authors' estimates

48 Actual and Target Wealth Actual Wealth Target Wealth