The Great Depression and the Friedman-Schwartz Hypothesis Lawrence J. Christiano, Roberto Motto and Massimo Rostagno
Background Want to Construct a Dynamic Economic Model Useful for the Analysis of Monetary and Fiscal Policy in a Large Economy Such as the Euro Area or U.S. Would Like a Model That Can be Used to: Identify the Fundamental Shocks Driving Historical Data Identify the Mechanisms Responsible for Propagating the Shocks Serve as a Laboratory For Evaluating Alternative Strategies for Responding to Shocks. We Expand a Standard Monetary Business Cycle Model to Incorporate Multiple Shocks, Financial Frictions and a Banking Sector. We Use our Model to Address one of the Biggest Policy Questions: What Shocks Caused the US Great Depression? What Mechanisms Propagated the Shocks? Could the Monetary Authorities have Mitigated the Severity of the Great Depression By Reacting in a Different Way to the Shocks? 0
Great Depression is a Perfect Testing Ground for Our Model It Was a Big Deal! 6 Log, US GNP 5.5 5 4.5 4 3.5 900 90 920 930 940 950 960 970 980 Years 3
Great Depression is a Perfect Testing Ground for Our Model Big Question: What Shocks? Exogenous Monetary Contraction in 920s (Bernanke, Hamilton) Increased Money Demand in Late 920s (Alexander Field). Deposit Withdrawals (Fear of Bank Closures and Dollar Devaluation). Stock Market Collapse. Institutional Changes, Especially New Deal (Cole and Ohanian). What Propagation Mechanisms? Stock Market Collapse (Mishkin, Romer). Nominal Wage and Debt Rigidities (Irving Fisher, Bordo et al). Banking System (Friedman and Schwartz). Policy Mistakes at Fed (Friedman and Schwartz, Cecchetti). 24
What We Do, and What We Find Incorporate Various Shocks and Propagation Mechanisms into a Single Dynamic General Equilibrium Model. Combine the Model with Data from 920s and 930s. Determine Which Shocks and Propagation Mechanisms Were Crucial. Results: A Liquidity Preference Shock is Important in Contraction Phase (Surprise, consistent with Alex Field?) Financial Frictions Somewhat Important for Investment Not Important Enough to Have a Major Impact on Aggregate Output and Employment (Surprising!) Increased Market Power of Workers Responsible for Duration of Great Depression. A More Responsive Monetary Policy Could Have Substantially Reduced the Severity of the Great Depression. 35
The Great Depression and the Zero Lower Bound on the Interest Rate In Recent Years: Interest In Understanding Monetary Policy in a Low Interest Rate Environment: Can Monetary Policy Resist Deflation and Output Collapse? What Sort of Monetary Policy Can do This? Interest Rates Near Zero In Most of the 930s Analysis of Monetary Policy in 930s Must Confront Zero Lower Bound Constraint Policy We Consider Avoids Zero Lower Bound By Committing to Temporarily High Future Money Growth. Credibility Issues (Eggertsson-Woodford, Krugman). 42
Previous Quantitative Anlyses of Great Depression Bordo-Choudhri-Schwartz, Could Stable Money Have Averted the Great Contraction?, EJ, 995. Bordo, Erceg and Evans, Money, Sticky Wages, and the Great Depression, AER, 2000. McCallum Could a Monetary Base Rule Have Prevented the Great Depression?, JME, 990. Sims, Christopher, The Role of Interest Rate Policy in the Generation and Propagation of Business Cycles: What Has Changed Since the 30s? 999. 43
A Quick Look at the Data. The Model. Outline Model Estimation Calibration of Some Model Parameters Maximum Likelihood Estimation of Other Parameters Model Fit. Evaluating Alternative Monetary Policies. 44
real gnp 0.25 investment 0.65 consumption 0.9 0.8 0.7 Total manhours (incl. military) 0.2 0.5 0. 0.05 GNP deflator 0.6 0.55 0.5 real M 0.9 0.8 0.9 0.8.4.3.2. 4 3 2 Short term rate 0.8 0.6 0.4 real DOW 0.4 0.35 0.3 0.25 0.2 currency to deposits 0.4 0.3 reserves to deposits 5 4 3 Spread, BAA and AAA 0.55 0.5 0.45 wage 0.2 2 0.4
Traditional Spending Transmission Real Wage Mechanism Rise in Real Wage Labor Demand Curve Drop in Employment Employment
Real Wage Modern Spending Transmission Mechanism No Change in Real Wage Labor Demand Curve (Markup,Capacity utilization,.) Drop in Employment Employment
GNP Falls Over 30%, 929 to 933 The Data Investment: Falls 80% I/Y Goes From 0.25 (P) to 0.06 (T) Consumption: Falls 25% C/Y initially rises from 0.68 in 929IV to 0.77 in 93IV and then falls back to 0.68 in 933IV Employment: Drops Less than GNP (i.e., Productivity Falls) and Never Fully Recovers Price Deflator: Drops and Never Fully Recovers 45
The Data, cont d What Started It? Was it the Fed? Probably Not (Note: M/P Roughly Constant, R Drops) Whatever it Was, it was Something that Hit Investment Hard Stock Prices Collapsed: Was that What Hurt Investment? Flight to Quality : Banks Accumulate Reserves, Households Accumulate Currency General Economic Uncertainty: Rises Big Time Starting 93, With Bank Panics. Real Wage Rose Continually: A Reason for the Long Duration of the Depression? 46
Model Features Suggested by Examination of Data Need a Good Model of: Investment, Employment, Market Power, Especially for Labor Suppliers. Need a Model that Allows for Interaction Between Financial Factors and Real Activity. Need Model That Can Come to Terms with Bank Reserves, Currency, Bank Deposits, etc. Model is a Marriage of Three: Christiano, Eichenbaum and Evans, Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy, forthcoming, JPE. Bernanke-Gertler-Gilchrist, Handbook of Macroeconomics, 999, Credit Market Frictions. Chari-Christiano-Eichenbaum JMCB, 995, Banking Structure. 50
Private Agents in the Model Goods-Producing Firms Intermediate Goods, Final Goods and Capital Goods Entrepreneurs: Own and Rent Out Capital Banks Households Monetary and Fiscal Authorities. 5
Final Goods-Producing Firms: Private Agents in the Model Y t = Z 0 Y jt λ f,tdj λf,t 52
Final Goods-Producing Firms: Private Agents in the Model Y t = Z 0 Y jt λ f,tdj λf,t Intermediate ½ Goods-Producing Firms: t K Y jt = jt α (z tl jt ) α Φz t if t Kjt α (z tl jt ) α > Φz t, 0 <α<, 0, otherwise z t = µ z z t. Profits: P jt Y jt Prt k K jt ( + R t ) W t l jt. Price-setting: with probability ξ p : P jt chosen optimally with probability ξ p : P jt = π t P j,t. 53
Private Agents in the Model, cont d Producers of Physical Capital Buy Investment Goods and Old Capital and Make K t+ : K t+ =( δ) K t + F (I t,i t ). 54
Private Agents in the Model, cont d Entrepreneurs Buy K t+ at end of t and Rent it out in t + Borrowing: B t+ = Q K K 0,t t+ N t+ 0. Capital Services: K t+ = u t+ Kt+, Capital Utilization costs: a(u t+ ) K t+,a 0,a 00 > 0, Net worth N t+ = past net worth + earnings from renting capital + market value of physical capital - repayments of past debt to bank. total net worth is kept low because random fraction of entrepreneurial wealth is destroyed entrepreneurs receive a costly state verification contract from the bank 58
Entrepreneur of Type ω, Where Eω=. Bank Households Lend Funds to Banks Irving Fisher Debt-Deflation: Exists Because Households Receive Fixed Nominal Return on Time Deposits. 60
Private Agents in the Model, cont d Banks: Issue Liabilities (Time Deposits) Used to Finance Entrepreneurs Issue Liabilities (Demand Deposits) Used to Finance Working Capital Loans to Firms Households: Supply Labor, Consume, Hold Demand and Time Deposits X β l t {u(c t+l bc t+l ) z(h j,t+l ) E j t l=0 υ t+l l = ³Pt+lC t+l Z 0 M t+l ³ θt+l σq θt+l P t+l C t+l D h t+l σ q } (h j ) λ w,t dj λw,t, λ w,t <. 63
Private Agents in the Model, cont d Monetary and Fiscal Authorities Exogenous Government Spending Requirement. Taxes. Monetary Policy: Money µ Base Growth Feeds Back on Shocks: M b log t+ Mt b = µ X µi,t + i Agnostic About Nature of Monetary Policy Could Be Taylor Rule, McCallum Rule, Something Else... 66
Private Agents in the Model Goods-Producing Firms Intermediate Goods, Final Goods and Capital Goods Entrepreneurs: Own and Rent Out Capital Banks Households Monetary and Fiscal Authorities. 67
Steps in the Analysis Select Parameter Values for the Model.. Parameters that Control Nonstochastic Part of the Model Parameters Governing Exogenous Shocks and Monetary Response. Evaluate Model Empirically. Counterfactual Policy Analysis. 72
Selection of Model Parameters Nonstochastic Parameters Match Various Long-Run Averages Use Evidence on the Shocks in the 920s and 930s to Parameterize Exogenous Shocks. 75
Model Parameters (Time unit of Model: quarterly) Panel A: Household Sector β Discount rate.03 0.25 b Habit persistence parameter 0.63 ξ w Probability of Not Reoptimizing Wage in Given Quarter 0.70 Panel B: Goods Producing Sector µ z Growth Rate of Technology (APR).50 ξ p Probability of Not Reoptimizing Price Within Quarter 0.50 δ Depreciation rate on capital. 0.02 α Power on capital in production function 0.36 Panel C: Entrepreneurs γ Quarterly Entrepreneurial Survival Probability 97.00 µ Fraction of Realized Profits Lost in Bankruptcy 0.20 F ( ω) Percent of Businesses that go into Bankruptcy in a Quarter 0.80 Var(log(ω)) Variance of log of idiosyncratic productivity parameter 0.07 Panel E: Policy τ Bank Reserve Requirement 0.00 τ l Tax Rate on Labor Income 0.04 x Growth Rate of Monetary Base (APR).60 76
k Steady State Properties of the Model, Versus US Data Variable Model US, 92-29 US, 964-200 y 8.35 0.8 9.79 i y 0.20 0.24 0.25 c y 0.73 0.67 0.57 g y 0.07 0.07 0.9 N K N ( Equity to Debt ) 0.999 -.252 -.25 2 Percent of Goods Output Lost to Bankruptcy 0.37% Percent of Aggregate Labor and Capital in Banking.00% % 3 2.5% 5 Inflation (APR) 0.% -0.6% 4 4.27% 6 78
Money Variable Model 92-929 964-2002 Monetary Base Velocity 9.77 2 6.6 M Velocity 3.92 3.5 6.5 Currency / Demand Deposits 0.28 0.2 0.3 Currency / Monetary Base 0.70 0.55 0.73 Curr. / Household D. Deposit 2.30 79
Shocks Incorporated Into Model Eight Shocks: Monopoly Power of Firms Monopoly Power of Households Demand for Reserves By Banks Two Household Money Demand Shocks Shock to Riskiness of Entrepreneurs Aggregate Technology Shock Shock to Rate of Destruction of Entrepreneurial Wealth 80
Parameterization of Shocks Each Shock, Say x t, Has 4 Parameters Representation: x t = ρx t + ε x,t,σ εx Monetary Policy Response: µ xt = ρ µ µ x,t + φε x,t There are 8 4=32Parameters to be Estimated Monetary Policy µ M b log t+ Mt b = µ X i µi,t + 8
Matrices, A, B Solution to Model z t = A z t + BΨ t. z t ~Core set of 23 Endogenous Variables Ψ t Exogenous Variables, Stacked
Estimation of Parameters of Exogenous Shock Processes Data Used in Estimation: Net Worth (Measured by Value of the DOW) Inflation log, hours Short Term Interest Rate Output Real Wage Investment Velocity of M Consumption Risk Premium (Baa - Aaa Bond Returns) Currency to Deposit Ratio Bank Reserves to Deposit Ratio 2
Data: X t =(log³ Nt+ P t Y t log ³ Wt P t Y t Stochastic Model log (π t )log(l t ) R b t log(y t ) log( I t Y t )log(v t )log( C t Y t ) P e t log(d c t)log(d r t)) 0 Here, N t+ Net Worth, Rt Short b Term Interest Rate Vt Velocity of M P e Premium d c Currency to Deposit Ratio d r Reserves to Deposit Ratio Representation of X t : X t = α + τz t + τ s Ψ t + τz t + τ s Ψ t. 3
State-Observer System State, ξ t ξ t = z t z t Ψ t Ψ t. Law of Motion of State: z t+ z t Ψ t+ Ψ t = A 0 Bρ 0 I 0 0 0 0 0 ρ 0 0 0 I 0 z t z t Ψ t Ψ t + BD 0 D 0 ˆϕ t+, ξ t+ = Fξ t + v t+. 7
Observer Equation y t = Hξ t + w t, where H = τ τ ˆτ s b τ s. Estimate by Maximum Likelihood, Given Parameters of Nonstochastic Part. 8
Results of Model Fit Overall, Model Fit Seems Reasonable Places Where Model Misses Understates Fall in Labor Productivity Overstates Rise in Real Wage Understates Fall in Consumption. Shocks Seem Reasonable 86
Figure 5: Actual and Fitted Data, Converted to Levels -7-7.5-8 -8.5 4 3 2 -.5-2 -2.5-3 Log, real net worth Policy Rate Log, Investment 5 4 3 2 Log, Money Base -.8-2 -2.2-2.4 Premium (APR) 0-0. -0.2-0.3 0. 0-0. -0.2 -. -.2 -.3 -.4 -.5 - -.2 -.4 -.6 Log, Price Level Log, Output Log, M Log, Currency to Deposit ratio -.8 0.3 0.2 0. -7.6-7.7-7.8-7.9-0.3-0.4-0.5 Log, Hours Worked Log, Real Wage Log, Consumption -0.6 Log, bank reserve to deposit ratio - -.2 -.4 -.6 -.8 Notes: (i) Dotted, Solid Line - Model, Actual Data. (ii) Results Obtained by First Adding Actual Data Sample Mean to Results Displayed in Figure 4 and then Aggregating to Levels. (iii) Currency-Deposit, Reserves-Deposit Ratio, Premium Reproduced from Figure 4.
Figure 7: Estimated Economic Shocks Firm Markup, λ f,t Banking Money Demand Shock, ξ t Money Demand, θ t % dev. from steady state 2 0-2 -4 Fraction 0.95 0.9 Fraction 0.9 0.8 0.7 0.6 ratio to steady state Labor Market Power, ζ t.4.2 0.8 Technology Shock, ε t 0.995 0.99 0.985 0.98 % dev. from steady state 400 200 0-200 Liquidity Demand, υ t -400 0.5 0.4 0.3 0.2 Riskiness of Entrepreneurs, σ t Fraction.05 0.95 0.9 Financial Wealth Shock, γ t 0.85
Liquidity Preference Shock Demand Deposits Drop Working Capital Loans Drop Consumption Drops Liquidity Shock Time Deposits Fall Net Worth Falls Entrepreneur Loans Fall Financial Accelerator Investment Falls Asset Price Falls Output Drops Debt Deflation Price Level Drops Rental Rate Of Capital Falls
Response to One-Standard Deviation Innovation to Liquidity Preference, υ t.0 Household Currency Relative to SS 0.995 Household Deposits Relative to SS ratio.005 ratio 0.99 0.985 0.98 Household Time Deposits Relative to SS Consumption ratio 0.995 0.99 0.985 Percent 0-0.5 0.98 - Interbank Loan Rate Investment 5.9 - APR 5.8 5.7 Percent -2-3 -4 0 Output Percent -0.5 - -.5 5 0 5 20 quarters
Figure 9: Model Response with Only υ t Shocks, and Data -7-7.5-8 -8.5 4 3 2 -.5-2 -2.5-3 5 4 3 2 Log, real net worth Policy Rate Log, Investment Premium (APR) Log, Money Base -.8-2 -2.2-2.4-2.6 0-0.2-0.4 0. 0-0. -0.2 -. -.2 -.3 -.4 -.5 - -.2 -.4 -.6 Log, Price Level Log, Output Log, M Log, Currency to Deposit ratio -.8 Notes: Results Correspond to Those in Figure 5, Except Model Simulation Only Includes Estimated υ t Shocks 0.3 0.2 0. -7.6-7.7-7.8-7.9-0.2-0.4-0.6 - -.2 -.4 -.6 -.8 Log, Hours Worked Log, Real Wage Log, Consumption Log, bank reserve to deposit ratio
0.4 0.2 0-0.2-0.4-0.6-0.8 - -.2 consumption 0 5 0 5 20 quarters 0 - -2-3 -4-5 investment 0 5 0 5 20 quarters percent deviation from unshocked path percent deviation from unshocked path 0.5 0-0.5 - -.5-2 Dynamic Response to a v Shock - benchmark (*); no entrepreneur (o) output 0 5 0 5 20 quarters 0.5 0-0.5 - -.5-2 -2.5-3 -3.5 price of installed capital, in units of C 0 5 0 5 20 quarters percent deviation from unshocked path percent deviation from unshocked path
Counterfactual Experiment Identify Alternative Monetary µ Policy Within a Particular Class M b log t+ Mt b = µ X µi,t + i µ it = θ 0 ε i,t + θ ε i,t + θ 2 µ i,t Trade-off: Immediate Monetary Response Delayed Monetary Response Selected Quantities in Baseline and Counterfactual, 929IV - 939IV Counterfactual Growth, APR Baseline Delayed Response Immediate Response Real Output -0.7.0.0 Inflation -2.5 3.5.0 Monetary Base 5.3 2.0 0.7 M 0.3 7.4 5. 400(max R min R) -3.6-3.6-5.8 95
Figure 2: Baseline Estimated Policy (Solid Line) and Counterfactual Policy (Dotted Line) Log, real net worth Log, Price Level Log, Hours Worked Policy Rate.8.6.4.2 0.8 0.6 0.2 0-0.2-2.55-2.6-2.65-2.7-2.75 7 6 5 4 0-0. -0.2 Log, Output 2.2 2.5 2. 2.05 2 Log, Real Wage -.5-2 -2.5 Log, Investment 0.8 0.6 0.4 0.2 0-0.2 Log, M Log, Consumption Premium (APR) Log, Currency to Deposit ratio Log, bank reserve to deposit ratio -0.3-0.35-0.4-0.45 Log, Money Base 3 2 0 - -.2 -.4 930 932 934 936 938 -.4 -.6 -.8-2 -2.2-2.4 930 932 934 936 938-0.2-0.4-0.6-0.8 - -.2 930 932 934 936 938
Concluding Remarks Fit A Model to 920s and 930s data Liquidity Preference Shock Important in Contraction Increased Worker Bargaining Power Important in Delaying Recovery Financial Frictions: Exacerbate Fall in Investment But, Not Enough to Have a Major Impact on Aggregate Output and Employment Question: Could the Great Depression Have Been Substantially Mitigated Under an Alternative Monetary Policy? Answer: Yes. 02