Leads, Lags, and Logs: Asset Prices in Business Cycle Analysis David Backus (NYU), Bryan Routledge (CMU), and Stanley Zin (CMU) Society for Economic Dynamics, July 2006 This version: July 11, 2006 Backus, Routledge, and Zin () Leads, lags, and logs 1 / 20 Overview Leads and lags in business cycles (Almost) the usual equations Loglinear approximation Properties of the model Extensions Backus, Routledge, and Zin () Leads, lags, and logs 1 / 20 Leads and lags Leads and lags Cross-correlation functions of GDP with I I I Stock price indexes Interest rates and spreads Consumption and employment US data, quarterly, 1960 to present Quarterly growth rates (first difference of logs), except I I Interest rates and spreads Occasional year-on-year comparisons (yt+2 − yt−2 ) Backus, Routledge, and Zin () Leads, lags, and logs 2 / 20 Leads and lags Stock prices and GDP Lags GDP 0.50 0.00 −0.50 −10 Backus, Routledge, and Zin () −5 0 Lag Relative to GDP Leads, lags, and logs 5 10 −1.00 −1.00 Cross−Correlation with GDP −0.50 0.00 0.50 Leads GDP 1.00 1.00 S&P 500 3 / 20 Leads and lags Stock prices and GDP (year-on-year) 1.00 0.50 0.00 −0.50 −10 Backus, Routledge, and Zin () −5 0 Lag Relative to GDP Leads, lags, and logs 5 10 −1.00 −1.00 Cross−Correlation with GDP −0.50 0.00 0.50 1.00 S&P 500 (yoy) 4 / 20 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −5 0 5 Lag Relative to GDP −5 0 5 Lag Relative to GDP Backus, Routledge, and Zin () 10 NYSE Composite 10 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −10 Leads, lags, and logs −5 0 5 Lag Relative to GDP −5 0 5 Lag Relative to GDP 10 Nasdaq Composite 10 −1.00−0.50 0.00 0.50 1.00 S&P 500 minus Short Rate −1.00−0.50 0.00 0.50 1.00 −10 Lags GDP −1.00−0.50 0.00 0.50 1.00 S&P 500 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 Leads GDP −1.00−0.50 0.00 0.50 1.00 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 Leads and lags Stock prices and GDP 5 / 20 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −5 0 5 Lag Relative to GDP Backus, Routledge, and Zin () 10 Short Rate (3m) 10 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −10 Leads, lags, and logs −5 0 5 Lag Relative to GDP −5 0 5 Lag Relative to GDP 10 Real Rate 10 −1.00−0.50 0.00 0.50 1.00 Yield Spread (GDP yoy) −1.00−0.50 0.00 0.50 1.00 −5 0 5 Lag Relative to GDP Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −1.00−0.50 0.00 0.50 1.00 Yield Spread (10y−3m) −1.00−0.50 0.00 0.50 1.00 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 Leads and lags Interest rates and GDP 6 / 20 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −5 0 5 Lag Relative to GDP Backus, Routledge, and Zin () 10 Nondurables 10 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −10 Leads, lags, and logs −5 0 5 Lag Relative to GDP −5 0 5 Lag Relative to GDP 10 Durables 10 −1.00−0.50 0.00 0.50 1.00 Services −1.00−0.50 0.00 0.50 1.00 −5 0 5 Lag Relative to GDP Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −1.00−0.50 0.00 0.50 1.00 Consumption −1.00−0.50 0.00 0.50 1.00 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 Leads and lags Consumption and GDP 7 / 20 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −5 0 5 Lag Relative to GDP Backus, Routledge, and Zin () 10 Equipment and Software 10 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −10 Leads, lags, and logs −5 0 5 Lag Relative to GDP −5 0 5 Lag Relative to GDP 10 Residential 10 −1.00−0.50 0.00 0.50 1.00 Structures −1.00−0.50 0.00 0.50 1.00 −5 0 5 Lag Relative to GDP Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −1.00−0.50 0.00 0.50 1.00 Investment −1.00−0.50 0.00 0.50 1.00 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 Leads and lags Investment and GDP 8 / 20 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −5 0 5 Lag Relative to GDP Backus, Routledge, and Zin () 10 Avg Weekly Hours (All) 10 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −10 Leads, lags, and logs −5 0 5 Lag Relative to GDP −5 0 5 Lag Relative to GDP Employment (Household Survey) Avg Weekly Hours (Manuf) 10 −1.00−0.50 0.00 0.50 1.00 10 −1.00−0.50 0.00 0.50 1.00 −5 0 5 Lag Relative to GDP Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 −10 −1.00−0.50 0.00 0.50 1.00 Employment (Nonfarm Payroll) −1.00−0.50 0.00 0.50 1.00 Cross−Correlation with GDP −1.00−0.50 0.00 0.50 1.00 Leads and lags Employment and GDP 9 / 20 Leads and lags Lead/lag summary Things that lead GDP I I I Stock prices Yield curve and short rate Consumption (a little) Things that lag GDP I Employment Backus, Routledge, and Zin () Leads, lags, and logs 10 / 20 The usual equations (Almost) the usual equations Basic real business cycle model except I Recursive preferences (Kreps-Porteus/Epstein-Zin-Weil) I CES production I More complex shock process (predictable component in productivity) Backus, Routledge, and Zin () Leads, lags, and logs 11 / 20 The usual equations Preferences Equations Ut = V [ut , µt (Ut+1 )] ut = ctγ (1 − nt )1−γ V (a, b) = [(1 − β)aρ + βb ρ ]1/ρ ¡ ¢1/α α µt (Ut+1 ) = Et Ut+1 Interpretation IES = 1/(1 − ρ) CRRA = 1 − α α = ρ Backus, Routledge, and Zin () ⇒ additive preferences Leads, lags, and logs 12 / 20 The usual equations Technology Equations yt = f (kt , zt nt ) = [ωktν + (1 − ω)(zt nt )ν ]1/ν yt = ct + i t kt+1 = (1 − δ)kt + it Interpretation Elast of Subst = 1/(1 − ν) Capital Share = ω(y /k)−ν Backus, Routledge, and Zin () Leads, lags, and logs 13 / 20 The usual equations Productivity process Equation log zt+1 − log zt = ḡ + ∞ X χj εt+1−j j=0 Interpretation I Moving average allows predictable component Backus, Routledge, and Zin () Leads, lags, and logs 14 / 20 Logs Loglinear decision rules We’re looking for decision rules of the form ĉ = hck k̂ + n̂ = hnk k̂ + ∞ X j=0 ∞ X hcj εt−j hnj εt−j j=0 Backus, Routledge, and Zin () Leads, lags, and logs 15 / 20 Logs Loglinear decision rules Traditional methods I I I Dependence of decisions on capital independent of shocks Scale by z for stationarity Decision rules follow from loglinear approximation of derivatives of value function With recursive preferences I I I I I I Dependence of decisions on capital independent of shocks Scale by z for stationarity We need the value function, not just its derivatives Tallarini: logquadratic value function Us: loglinear value function Impact: risk aversion has no impact on quantities Backus, Routledge, and Zin () Leads, lags, and logs 16 / 20 Properties of the model Asset prices Pricing kernel mt+1 = β (ct+1 /ct )ρ−1 [Ut+1 /µt (Ut+1 )]α−ρ Short rate rt Backus, Routledge, and Zin () = − log Et mt+1 Leads, lags, and logs 17 / 20 Properties of the model Impulse response of short rate Backus, Routledge, and Zin () Leads, lags, and logs 18 / 20 Bottom line Bottom line Asset prices contain information about business cycles Specifically: they lead the cycle Not true of traditional business cycle models We add I I recursive preferences predictable component to productivity Lots left to do I I I equity macro-based bond pricing stochastic volatility Backus, Routledge, and Zin () Leads, lags, and logs 19 / 20 Related work Related work Leads and lags I Ang-Piazzesi-Wei, Beaudry-Portier, Jaimovich-Rebelo, Stock-Watson Predictable component I Bansal-Yaron Computation with recursive preferences I Hansen-Sargent, Tallarini, Uhlig Kreps-Porteus pricing kernel I Hansen-Heaton-Li, Weil Backus, Routledge, and Zin () Leads, lags, and logs 20 / 20