Nobel Symposium Money and Banking https://www.houseoffinance.se/nobel-symposium May 26-28, 2018 Clarion Hotel Sign, Stockholm
Money and Banking: Some DSGE Challenges Nobel Symposium on Money and Banking Harald Uhlig 1 1 University of Chicago Department of Economics huhlig@uchicago.edu Stockholm, May 27th, 2018 Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 1 / 30
Outline 1 Challenges 2 Asset prices and Yield Spreads 3 Financial Frictions 4 Inflation 5 Neo-Fisherian features of New Keynesian models 6 Conclusions Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 2 / 30
Challenges Outline 1 Challenges 2 Asset prices and Yield Spreads 3 Financial Frictions 4 Inflation 5 Neo-Fisherian features of New Keynesian models 6 Conclusions Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 3 / 30
Challenges Main Theme Quantitative DSGE models were meant to rise to the Lucas challenge of constructing general equilibrium models with deep parameters. Now, workhorse models for monetary policy analysis. But: Asset prices and yield spreads. Probably central for monetary policy. Typically ignored or trivialized in QDSGEs. Financial frictions. Much progress has been made. But contracts are often not privately optimal. Perhaps they should be. Inflation. Data: no Phillips-Curve tradeoff. QDSGE: don t account for inflation with monetary policy shocks. Neo-Fisherian features of New Keynesian models. Substantial, but get swept under the rug. The glass is half full. Or half empty. Take your pick. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 4 / 30
Asset prices and Yield Spreads Outline 1 Challenges 2 Asset prices and Yield Spreads 3 Financial Frictions 4 Inflation 5 Neo-Fisherian features of New Keynesian models 6 Conclusions Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 5 / 30
Asset prices and Yield Spreads The skeleton in the closet E.g. log-linearized sol n for cons c t, return R t return. s t : state. log(c t+1 ) = φs t +ǫ t+1 log(r t+1 ) = ξs t +ν t+1 Assume log preferences. Asset pricing equation: ( ) ] ct 1 = E t [β R t+1 c t+1 = βc t e (ξ φ)s t E t [ e ν t+1 ǫ t+1 ] Suppose ν t+1 ǫ t+1 N(0,σt 2 ), conditional on t. Then, [ ] E t e ν t+1 ǫ t+1 = e σt 2/2 Suppose ν t+1 ǫ t+1 t 1000 (0,σt 2 ), conditional on t. Then, [ ] E t e ν t+1 ǫ t+1 = Now what? It gets ignored. I will ignore it too. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 6 / 30
Asset prices and Yield Spreads Risk premia per Epstein-Zin Source: Easy EZ for DSGE (Uhlig, 2010). IES = 1. Risk av = η. Log-linearized: Note: ˆV t = (1 β) ĉ t +β ˆR t ] ˆR t = E t [ˆVt+1 ) ˆM t+1 = ĉ t ĉ t+1 +(η 1) (ˆVt+1 ˆR t E t [ˆM t+1 ] = ĉ t E t [ĉ t+1 ] Thus, EZ has no influence on macro-dynamics (up to first order). Dichotomy between macro and asset pricing. Too easy? If labor is part of utility, it necessarily shows up in asset pricing. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 7 / 30
Asset prices and Yield Spreads Yield Curves: FFR vs 10-year yields 20 FFR vs 10-year yield 15 percent 10 5 0 1940 1960 1980 2000 2020 date Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 8 / 30
Asset prices and Yield Spreads Incorporating Yield Curves in DSGE models Long-term yields important for understanding the effects of monetary policy. Yet, either absent or treated insufficiently in QDSGE models ( expectations theory ). Some promising developments: Piazzesi-Schneider (2007), Equilibrium Yield Curves. Kliem-Meyer-Gohde (2018), (Un)expected Monetary Policy Shocks and Term Premia. Asset prices and yield spreads: bottom line. Probably central for monetary policy. Typically ignored or trivialized in QDSGEs. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 9 / 30
Financial Frictions Outline 1 Challenges 2 Asset prices and Yield Spreads 3 Financial Frictions 4 Inflation 5 Neo-Fisherian features of New Keynesian models 6 Conclusions Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 10 / 30
Financial Frictions Financial Frictions It is hard to think about the effects of monetary policy without thinking about pricing or financial frictions. Financial frictions, recent literature: Agent heterogeneity and idiosynchratic shocks: HANK. Financial intermediaries, banks: Gertler-Karadi-Kiyotaki or GKK. Increasingly used for policy guidance and welfare analysis. But then, contracts should be privately optimal. Otherwise: chicken paper conundrum... Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 11 / 30
Financial Frictions The Chicken Paper Conundrum The classic chicken paper (acc. to Ed Prescott): Assumption 1: households enjoy consuming chicken. Assumption 2: households cannot produce chicken. Assumption 3: government can produce chicken. Conclusion: government should produce chicken. For policy guidance, it is important to argue, why agents cannot address these frictions on their own. Example HANK: if income fluctuations are known, full insurance should be possible. Example GKK: if net worth might get destroyed, write insurance contracts. Needed: DSGE models with privately fully-optimal long-term contracts. Example: Krüger-Uhlig (2018). Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 12 / 30
Financial Frictions Example: Krüger-Uhlig (2018) Think: continuous-time Aiyagari model,...... i.e.: agents are endowed with two-state Markov process of labor units, fluctuating between ζ > 0 and 0. Transition rates: ξ dt = P(ζ 0), ν dt = P(0 ζ). Aggregate production Y = K θ L 1 θ. Cap. depr. rate δ. Preferences: discount log-utility with ρ. Equilibrium interest rate r.... but: long-term insurance contracts, with one-sided commitment: Competitive intermediaries, commit long-term. Agents can walk anytime, sign up with the next one. Full information, though no credit history punishment. Contracts: payments from agent are front-loaded, payments from intermediary are backloaded. Intermediaries invest payments from agent in capital. Steady state comparison only. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 13 / 30
Financial Frictions Optimal contract. Case ρ > r. 1.6 1.4 Optimal contract sample path, g=0.4 productivity consumption 1.2 goods/w 1 0.8 0.6 0.4 0.2 0 2 4 6 8 10 time Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 14 / 30
Financial Frictions Stationary Consumption Distribution for three r s Stationary densities plus two point masses 1.5 1 r (c) 0.5 r=0.039 r=0.020 r=0.000 0 0 0.5 1 1.5 c Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 15 / 30
Financial Frictions Results Closed form solution for everything! 0.2 r as a function of 0.05 r as a function of 0.15 0.04 r 0.1 r 0.03 0.02 0.05 0.01 0 0 0.05 0.1 0.15 0.2 0 0 0.05 0.1 0.15 Recall: ξ dt = P(ζ 0) Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 16 / 30
Financial Frictions Financial frictions: bottom line Much progress has been made. But contracts are often not privately optimal. Perhaps they should be: chicken paper conundrum. Recent research shows they can be. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 17 / 30
Inflation Outline 1 Challenges 2 Asset prices and Yield Spreads 3 Financial Frictions 4 Inflation 5 Neo-Fisherian features of New Keynesian models 6 Conclusions Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 18 / 30
Inflation Classic Phillips Curve: textbook. An artificial Phillips Curve 6 inflation 4 2 0 π = β u + ǫ 2 4 6 8 10 12 unemployment per generatingπ t = 6 0.5u t +ǫ, ǫ N(0, 0.3 2 ) Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 19 / 30
Inflation Classic Phillips Curve: data. 15 Phillips Curve: 1948 to 2016 10 inflation 5 0-5 π = β u + ǫ u = β π + ǫ 2 4 6 8 10 12 unemployment Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 20 / 30
Inflation Phillips Curve: NK version. NK Phillips Curve from 1948 to today 3 2 π(t) - β π(t+1) 1 0-1 -2-3 π(t) - β π(t+1) = κ u(t) + ǫ(t+1) 2 4 6 8 10 12 unemployment NK: π t = βe t [π t+1 ]+κx t. Rewrite: π t βπ t+1 = κx t +ǫ t+1. Use x t = u t, β = 0.99. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 21 / 30
Accounting for Inflation Inflation Source: Fratto-Uhlig (2018). Approach: take pre-crisis Smets-Wouters (2007) model. Decompose inflation into the shocks driving it. 15 10 5 0 5 contribution of shocks to wage and price markup Variance decomp. Technology 3.90 Price Markup 51.09 Wage Markup 27.04 Preferences 7.65 Inv.Spec.Tech. 3.54 Gov t Exp 0.43 Monetary Policy 6.33 1950 1960 1970 1980 1990 2000 2010 Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 22 / 30
Inflation Inflation: bottom line Data: no Phillips-Curve tradeoff. QDSGE: don t account for inflation with monetary policy shocks. The NK / Phillips-Curve-based NK QDSGE models may thus provide a poor guide for monetary policy. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 23 / 30
Outline Neo-Fisherian features of New Keynesian models 1 Challenges 2 Asset prices and Yield Spreads 3 Financial Frictions 4 Inflation 5 Neo-Fisherian features of New Keynesian models 6 Conclusions Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 24 / 30
Neo-Fisherian features of New Keynesian models The three equation NK model Parameters: IS: x t = E t [x t+1 ] 1 σ (i t E t [π t+1 ] r n t ) Phillips: π t = βe t [π t+1 ]+κx t Taylor: i t = ρ+φπ t +ξx t +ν t Persistence: ν t = ψν t 1 +ǫ t β = 0.99,κ = 0.5,σ = 1. ρ = 0, rt n 0 (for impulse response). ξ = 0.1 (Note: ξ = 0.5 might be nice... but gives even weirder results). φ = 1.5. ψ = 0.4 or ψ = 0.6. Let s check some impulse responses to ǫ 0 = 1. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 25 / 30
Neo-Fisherian features of New Keynesian models Impulse responses to ǫ 0 = 1, if ψ = 0.4. Imp. resp. to mon. shock, φ=1.5, ψ=0.4 0.8 0.6 response 0.4 0.2 0 π x i -0.2 0 0.5 1 1.5 2 time (quarters) Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 26 / 30
Neo-Fisherian features of New Keynesian models Impulse responses to ǫ 0 = 1, if ψ = 0.6. Imp. resp. to mon. shock, φ=1.5, ψ=0.6 0.8 0.6 response 0.4 0.2 π x 0 i 0 0.5 1 1.5 2 time (quarters) Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 27 / 30
Neo-Fisherian features of New Keynesian models Neo-Fisherian features of New Keynesian models. Bottom line. Cochrane, Garin-Lester-Sims. Neo-Fisherian features are substantial, but get swept under the rug. A reliable guide for monetary policy? Perhaps not quite. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 28 / 30
Conclusions Outline 1 Challenges 2 Asset prices and Yield Spreads 3 Financial Frictions 4 Inflation 5 Neo-Fisherian features of New Keynesian models 6 Conclusions Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 29 / 30
Conclusions Overall bottom line. Quantitative DSGE models were meant to rise to the Lucas challenge of constructing general equilibrium models with deep parameters. Now, workhorse models for monetary policy analysis. But: Asset prices and yield spreads. Probably central for monetary policy. Typically ignored or trivialized in QDSGEs. Financial frictions. Much progress has been made. But contracts are often not privately optimal. Perhaps they should be. Inflation. Data: no Phillips-Curve tradeoff. QDSGE: don t account for inflation with monetary policy shocks. Neo-Fisherian features of New Keynesian models. Substantial, but get swept under the rug. The glass is half full. Or half empty. Take your pick. Harald Uhlig (University of Chicago) Money and Banking: DSGE Challenges 2018-05-27 30 / 30