Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
The views expressed herein are those of the authors and not necessarily those of the Federal Reserve Bank of Minneapolis or the Federal Reserve System WEALTH AND VOLATILITY Jonathan Heathcote Minneapolis Fed Fabrizio Perri University of Minnesota and Minneapolis Fed FRB St Louis, 2011 Economic Policy Conference Source of Business Cycles • Old idea: business cycles can be driven by self-fulfilling wave of optimism or pessimism • Our idea: extent to which non-fundamentals-driven changes in expectations can generate fluctuations depends on the level of household wealth • We will argue that declines in US house prices left the economy fragile and susceptible to a confidence-driven recession Sunspot-driven fluctuations • Rise in expected unemployment → consumers reduce demand → firms reduce hiring → higher unemployment • For a wave of pessimism to be self-fulfilling need high sensitivity of demand to expected unemployment • Sensitivity of demand depends inversely on level of household wealth • High wealth or cheap credit → demand less sensitive to expectations → no sunspot-driven fluctuations • Low wealth and costly credit → demand more sensitive to expectations → confidence-driven recessions possible Outline 1. Micro evidence on the mechanism from the CEX • Wealth-poor reduced consumption more than the rich as unemployment surged during the Great Recession 2. Macro evidence on the importance of sunspot shocks • Historically, US macroeconomic volatility tends to be high in periods when US household wealth is low 3. Model • Committed consumption ⇒ demand sensitive to expectations when wealth is low 4. Great Recession as an equilibrium outcome 5. Policy: Govt spending and unemployment insurance Micro Evidence from the CEX • Households aged 25-60, who report consumption and income in the 2nd and 5th quarterly interviews ⇒ can compute annual consumption and income growth for the same households • Split sample in two based on household net worth to income ratio in the 5th interview • Compare consumption and income growth for the wealth rich versus the wealth poor during the Great Recession Summary statistics (2007.1 -2010.1) Sample Age Family size % Male % College Income after tax Consumption Wealth Wealth Rich 3,522 47 2.9 47 52 $79,659 $62,943 $280,414 Wealth Poor 3,527 41 2.9 48 38 $60,315 $47,394 $2,060 Log Changes in Ave. Total Consumption: Wealth rich v/s wealth poor 6.0% 4.0% Wealth rich (avg. growth =‐2.2%) 2.0% 0.0% ‐4.0% ‐6.0% ‐8.0% ‐10.0% ‐12.0% Wealth poor (avg. growth = ‐4.0%) 2010Q1 2009Q4 2009Q3 2009Q2 2009Q1 2008Q4 2008Q3 2008Q2 2008Q1 2007Q4 ‐2.0% Log Changes in Ave. Disposable Income: Wealth rich v/s wealth poor 10.0% 8.0% 6.0% Wealth poor (avg. growth = 3.6%) 4.0% 2.0% 0.0% 2 2010Q1 Wealth rich (avg. growth = 0.0%) 2 2009Q4 2 2009Q3 2 2009Q2 2 2009Q1 2 2008Q4 2 2008Q3 ‐8.0% 2 2008Q2 ‐6.0% 2 2008Q1 ‐4.0% 2 2007Q4 ‐2.0% ‐0.02 ‐0.04 Wealth poor ‐0.06 ‐0.08 ‐0.1 Wealth rich 2010q4 2010q3 2010q2 2010q1 2009q4 2009q3 2009q2 2009q1 2008q4 2008q3 2008q2 2008q1 2007q4 2007q3 2007q2 2007q1 2006q4 2006q3 2006q2 2006q1 2005q4 Log Changes in Ave. Consumption: 0.1 Extended sample period 0.08 0.06 0.04 0.02 0 Macro Evidence • What is the relation between household-sector net worth and GDP volatility? • Measure volatility as (i) rolling-window standard deviation of GDP growth, (ii) estimate of instantaneous volatility from a Garch (1,1) model • Also investigate relation between net worth and level of GDP Wealth & GDP Volatility: rolling window correlation .014 .012 .010 Volatility .008 Net worth to GDP ratio Correlation = -0.70 4.4 .006 4.2 .004 4.0 3.8 3.6 3.4 W ealth 3.2 3.0 55 60 65 70 75 80 85 90 95 00 05 10 Note: Standard deviation of quarterly GDP growth are computed over 10 years rolling windows. Observations for net worth to GDP ratio are average over the same windows Standard deviation of GDP growth .016 Wealth & GDP Volatility: instantaneous correlation .016 .012 Volatility Net worth to GDP ratio .008 4.8 .004 Correlation = -0.52 4.4 4.0 3.6 3.2 Wealth 2.8 55 60 65 70 75 80 85 90 95 00 05 Note: Standard deviation of quarterly GDP growth are computed using GARCH(1,1). Standard deviation of GDP growth .020 10 Wealth & GDP Growth 4.4 Average GDP growth (per quarter) 4.0 .014 3.8 Wealth Correlation = -0.23 3.6 3.4 .012 3.2 .010 3.0 .008 .006 Growth .004 .002 55 60 65 70 75 80 85 90 95 00 05 Note: Average quarterly GDP growth are computed over 10 years rolling windows. Observations for net worth to GDP ratio are average over the same windows Net woth to GDP ratio 4.2 10 Stylized Model (related to Farmer 2010, Chamley 2011, Guerrieri and Lorenzoni 2009) • Non-durable consumption c, produced by competitive firms using indivisible labor • Durable housing h, in fixed supply with relative price p • Each representative household contains continuum of potential workers • Each representative firm produces with linear technology: y=n where n is mass of workers employed Timing 1. Households co-ordinate expectations on current unemployment, distributions of future unemployment rates 2. Representative household sends out workers with consumption order ct , assets pt ht , reservation wage w∗t 3. Representative firm randomly meets potential workers sequentially, decides whether to hire them 4. Firms pay wages wt = w∗t , workers pay for consumption - must borrow if unemployed and ct > pt ht − d 5. Household regroups, net resources determine ht+1 . Optimal firm strategy: hire worker iff aggregate order ct not yet filled and w∗t ≤ 1 Optimal household strategy: set w∗t = 1 Household Problem max E {ct ,ht+1 } ∞ X β t (log ct + φht ) t=0 s.t. ct + pt (ht+1 − ht ) = (1 − ut )wt − ψ 2 ut min {(pt ht − d − ct ) , 0} + Tt 2 φ : preference weight on housing ψ : cost of credit d : part of home value that cannot be used as collateral ut : fraction of household workers unemployed Tt : lump-sum rebate of credit costs Frictions 1. Labor market friction: No role for labor supply in determining allocations ⇒ output demand-driven 2. Credit friction: Unemployed with low wealth must use expensive credit ⇒ precautionary motive 3. Consumption commitment friction: Consumption chosen before unemployment status known ⇒ precautionary motive sensitive to expected unemployment Equilibrium Conditions 1. wt = w∗t = 1 2. ht = 1 3. Tt = ψut min {(pt − d − ct ) , 0} 2 4. ct = nt = 1 − ut 5. 1 1 pt+1 pt × = βEt φ + ct (1 − ψut min {(pt ht − d − ct ) , 0}) ct+1 Agenda for Theory • Characterize paths for unemployment that satisfy the inter-temporal FOC and the condition ct = 1 − ut • Will mostly abstract from asset price dynamics, focus on how dynamics for unemployment vary with (fixed) price level • Especially interested in expectations-driven multiplicity Role of Asset Prices • Introduce “marginal investor” with same preferences that faces no risk and is measure zero • Assume no housing trade between the two types • Marginal investor establishes a floor p for house prices: pt ≥ p = β φ 1−β • Will see that marginal investor rules out equilibria with very high unemployment Strong housing demand ⇒ full employment If φ ≥ φ̄ = (1 + d) 1−β β then the only steady state is p = p and u = 0 Logic: φ ≥ φ̄ ⇒ p − d ≥ c ... so even the unemployed do not need credit Absent credit constraints, p= β(1 − u) φ≤p 1−β But marginal investor implies p ≥ p, so p = p, u = 0 High wealth ⇒ Strong Demand ⇒ Full Employment Weak housing demand ⇒ positive unemployment If φ < φ̄ and 2 ψ ≥ ψ̄ = (1 − β) (1 − β)(1 + d) − βφ then 1. There is (still) a steady state with p = p and u = 0 2. There is another steady state with p = p and u > 0 • Intuition: p = p & u > 0 ⇒ asset has liquidity value ⇒ c>p−d 3. There are additional steady states with p > p and u > 0. Steady state demand and supply • Demand cd and supply cs defined by FOC and resource constraint: 1 p × cd [1 − ψu (p − d − cd )] cs = = p β φ+ cd 1−u • Closed-form solution with φ = 0: cd = where ρ = 1−β β ρ +p−d ψu is rate of time preference • Consumption elasticity wrt unemployment decreasing in home equity More general example with φ > 0 Numerical example: ψ = 1 β = 0.9 φ = 0.05 d = 0.1 1. ψ > ψ̄ = 0.15 (credit expensive) 2. φ < φ̄ = 0.12 (housing demand weak) Multiple steady states p’s Multiple paths to a steady state pair (p, u) • Suppose pt = p > p ⇒ constraint always binding • Difference equation defining equilibrium is p 1 1 × = βφ + βpEt (1 − ut ) (1 − ψut [p − d − (1 − ut )]) 1 − ut+1 • Assume no uncertainty / sunspots / expectational errors: 1 1 = Et 1 − ut+1 1 − ut+1 Unemployment Dynamics, p = 0.6 Sunspots generate fluctuations in ut • Low unemployment steady state is dynamically stable ⇒ possibility of “sunspots” • Define sunspot shock vt+1 vt+1 = 1 1 − Et 1 − ut+1 1 − ut+1 where vt+1 is iid over time with mean zero and a support that ensures we stay in the stable region Intuition for Differential Local Dynamics • Consider a hypothetical rise in unemployment starting from steady state • Low unemployment stable steady state • Each unemployed worker borrows a lot ⇒ high marginal credit cost ⇒ optimal to cut consumption sharply even though recovery expected • Expected consumption growth during recovery offsets stronger precautionary motive ⇒ stable demand for savings • High unemployment unstable steady state • Each unemployed worker borrows little ⇒ low marginal credit cost from rise in unemployment ⇒ not optimal to cut consumption sharply Range of equilibrium u decreasing in p p High p Low p Unemployment ranges Unemployment Review: Asset Prices and Macro Volatility • High asset prices ⇒ credit constraint does not bind ⇒ unique full employment equilibrium • Lower asset prices ⇒ constraint binds ⇒ range of equilibrium unemployment rates larger the lower is the asset price Features of the Great Recession 1. Large fall in asset values 2. Sharp decline in consumer spending, especially durables 3. Sharp rise in unemployment, labor productivity strong 4. Slow recovery Great Recession 1.1 1.05 1 Net Worth Price (real, rel. to trend) Durable Cons (real, pc, rel to trend) Unemp. Rate (‐ve, right axis) ‐3 ‐4 ‐5 0.95 0.9 0.85 0.8 ‐6 ‐7 ‐8 0.75 ‐9 0.7 0.65 0.6 ‐10 ‐11 Asset Prices 1.1 ‐3 House Price (real, rel to 2.1% trend) 1 Stock Price (real, rel to 2.1% trend) 0.9 08 0.8 07 0.7 Unemp. Rate (‐ve, right axis) ‐4 4 ‐5 ‐6 ‐7 7 ‐8 ‐9 0.6 0.5 ‐10 ‐11 Using the model to capture The Great Recession 1. Fall in demand for housing (fall in φ) reduces p so that economy becomes fragile 2. Sunspot (Lehman Brothers?) triggers jump in unemployment 3. Slow recovery to low unemployment steady state Graphically p High p 1 Low p 2 3 4 Unemployment Model Can Produce A Great Recession 1.2 1 Unemployment Stock Prices House Prices 0.8 0.6 0.4 0.2 0 ‐9 ‐7 ‐5 ‐3 ‐1 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 Preference Shock at t=‐6, Sunspot Shock at t=0 Policy 1: Tax and Spend p 0.7 equilibria G=0 0.6 0.5 pbar G=0 0.4 equilibria G=0.3 pbar G=0.3 0.3 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 u Policy 1: Review • Reduces elasticity of aggregate demand to expectations • Also reduces asset values • Narrows range of equilibrium unemployment, but need very high G to rule out equilibria with unemployment • Welfare implications will depend on utility from G Policy 2: Unemployment benefit b financed by proportional tax τ on earnings p 0.7 equilibria b=0 0.6 0.5 equilibria b=0.3 0.4 pbar b=0.3 0.3 0.2 0.0 0.1 0.2 0.3 0.4 0.5 0.6 u Policy 2: Review • Policy reduces need for costly credit ⇒ shrinks range of possible unemployment rates • Unique full employment equilibrium if ψ (d + 1) + b≥ β (β−1) φ + (β − 1) (β − 1) + ψ • ... which implies b ≥ 0.61 in our numerical example Conclusions • Model in which macroeconomic stability threatened by low asset values or tight credit markets • Great Recession: Decline in home values + costly credit left economy vulnerable to wave of pessimism • Macro evidence of a link between level of wealth and aggregate volatility • Micro evidence that low wealth households reduced consumption most sharply