Monetary and Fiscal Policy Design at the Zero Lower Bound: Evidence from the lab

Transcription

1 Monetary and Fiscal Policy Design at the Zero Lower Bound: Evidence from the lab Cars Hommes, Domenico Massaro & Isabelle Salle CeNDEF, University of Amsterdam Workshop on Socio-Economic Complexity March 2015 Leiden Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

2 Contents Outline 1 Motivation 2 Model 3 Experimental design 4 Experimental results 5 Conclusions Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

3 Motivation 1 Motivation 2 Model 3 Experimental design 4 Experimental results 5 Conclusions Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

4 Motivation Empirical evidence Liquidity trap in the Euro zone (a) Inflation (b) Interest rate Fiscal stimulus at the ZLB? Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

5 Motivation Expectations and fiscal stimulus at the ZLB: theoretical contributions Under rational expectations/perfect foresights (Woodford 1999, Benhabib et al. 2001a,b): ZLB gives rise to a second, deflationary steady state, with the possibility of self-fulfilling decelerating inflation paths. A liquidity trap is a low but stable inflation state. Automatic fiscal stimulus rules out liquidity traps. Under adaptive learning (McCallum 2002, Evans et al. 2008, Benhabib et al. (2012)): The targeted steady state is not globally stable. Deflationary spiral is self-reinforcing. A preventive cut in interest rates is not enough, a fiscal switching rule restores global stability of the targeted state. Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

6 Motivation Main objectives of this experimental paper 1 Characterizing liquidity traps using experimental evidence: Do liquidity traps arise in a learning-to-forecast experiment? Can a liquidity trap be a low inflation steady state or a deflationary spiral? 2 Assessing the empirical relevance of policy recommendations in a learning-to-forecast experiment: Is monetary policy enough to avoid liquidity traps? (with a preventive cut in interest rate) Can a fiscal switching rule prevent a liquidity trap? What are the dynamics along the convergence path towards the targeted steady state? Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

7 Model 1 Motivation 2 Model 3 Experimental design 4 Experimental results 5 Conclusions Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

8 Model An expectation-feedback model: (c, π, R, g) Evans et al. (2008) Aggregate demand (IS curve): Aggregate supply/nkpc: Fiscal policy instrument : g t = ḡ c t = c ( π t = π c e t+1 + (,π e t+1 +,R t c t,π e t+1,g t Monetary policy rule with preventive cut: ( ) R π e t+1 /πt,ct+1 e /c + + R t = ZLB ) ) if π t π if π t < π Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

9 Zero-lower bound and multiple steady states Multiple equilibria with aggressive monetary policy R R(π) π β 0.98 π L π * Figure: Aggressive monetary policy rule (preventive cut to ZLB) and multiple steady states π~ π

10 Model Global dynamics under adaptive learning Evans et al. (2008) : without fiscal policy c e Figure: Phase plot of the economy under adaptive learning Π e Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

11 Experimental design 1 Motivation 2 Model 3 Experimental design 4 Experimental results 5 Conclusions Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

12 Experimental design Learning to forecast experiment General context and information Two forecasts: net output and inflation. Interpretation: private forecasters and representative firm-household. Pay-off as a function of forecast errors. Qualitative information: positive/negative relationships between variables, time series, specific information on policies. A 2 2 treatment design: Monetary policy without or with policy mix. Severe initial pessimistic expectations or expectational shocks. 7 experimental economies in each (28 in total). Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

13 Policy treatments Experimental design MP Monetary policy only Aggressive monetary policy: preventive cut at the ZLB if inflation falls below π 1.6%. Fiscal policy always remains fixed (g = ḡ). MPF Policy mix: Aggressive monetary policy: preventive cut at the ZLB if inflation falls below π 1.6%. "Fiscal switching rule": inflationary-threshold fiscal policy. If inflation falls beyond π despite the ZLB, g is increased so that inflation meets the threshold π. Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

14 Learning to forecast experiment Timing of events (under aggressive monetary policy)... Subjects are Compute asked to the average forecast forecasts c t+1 and π t+1 (c e t+1, πt+1) e t t The CB sets R t based on π e t+1 (and c e t+1 ) t Euler eq. gives c t based on c e t+1, πe t+1, Rt t Fiscal policy g t = ḡ t Phillips curve if π t < π Subjects gives π t The CB sets observe based on πt+1, e R t = ZLB c t, π t, R t, g t c t, g t t + 5 and c t, π t are 8 recomputed t Subjects are asked to forecast c t+2 and π t+2 t t + 6 8

15 Learning to forecast experiment Timing of events (under policy mix)... Subjects are asked to forecast Compute the average forecasts The CB sets R t based on πt+1 e Euler eq. gives c t based on c t+1 and π t+1 t (c e t+1, πt+1) e t (and c e t+1 ) t c e t+1, πe t+1, Rt t Fiscal policy g t = ḡ t Phillips curve if π t < π Fiscal policy Subjects gives π t The CB sets if π based on t π observe πt+1, e R t = ZLB c c t, g t, π t, R t, g t t g t = ḡ t + 5 and c t, π t are 9 recomputed if π t + 8 t < π 9 Subjects are asked to forecast c t+2 and π t+2 t t sets g t so that π t = π t + 7 9

16 Experimental design Expectations treatments a Initial mild pessimism: [ 5, 8]% (5%) and [60, 80] (74). NEWS Experts predict a decrease in inflation and net output Figure: News shocks at periods 8,9 and 10. b Initial severe pessimism: [ 8, 8]% (5%) and [50, 80] (74). Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

17 Experimental results 1 Motivation 2 Model 3 Experimental design 4 Experimental results 5 Conclusions Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

18 Overview of the results Monetary policy only Mild pessimism (news shock) 4 deflationary spirals after news 3 convergences to steady state (1 with ZLB) Severe pessimism 5 deflationary spirals (initial pessimism) 2 convergences to steady state with ZLB

19 Overview of the results Monetary policy only Policy mix Mild pessimism (news shock) 4 deflationary spirals after news 3 convergences to steady state (1 with ZLB) 1 convergence with standard policy 6 convergences with policy mix: 3 oscillatory convergence paths, 3 liquidity traps Severe pessimism 5 deflationary spirals (initial pessimism) 2 convergences to steady state with ZLB 1 convergence with standard policy 6 convergences with policy mix: 2 oscillatory convergence paths, 4 liquidity traps

20 Treatments without fiscal policy and initial severe pessimism Monetary policy only Policy mix Mild pessimism (news shock) 4 deflationary spirals after news 3 convergences to steady state (1 with ZLB) 1 convergences with standard policy 6 convergences with policy mix: 3 oscillatory convergence paths, 3 liquidity traps Severe pessimism 5 deflationary spirals (initial pessimism) 2 convergences to steady state with ZLB 1 convergence with standard policy 6 convergences with policy mix: 2 oscillatory convergence paths, 4 liquidity traps

21 Example of a deflationary spiral after severe pessimism Treatment MP_b group Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

22 Example of a convergence to the steady state despite severe Treatment MP_b group 7 5 pessimism 0 5 Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

23 80 Phase plot of the experimental economy L T net output stable unstable 50-8% π T +8% π L inflation

24 80 Initial expectations in Tr. MP/severe pessimism (b) L T net output stable unstable 50-8% π T +8% π L inflation

25 80 Fall in a liquidity trap in Tr. MP/severe pessimism (b) L T net output stable unstable 50-8% π T +8% π L inflation

26 80 Convergence in Tr. MP/severe pessimism (b) L T net output stable unstable 50-8% π T +8% π L inflation

27 Treatments without fiscal policy and news shocks Monetary policy only Policy mix Mild pessimism (news shock) 4 deflationary spirals after news 3 convergences to steady state (1 with ZLB) 1 convergences with standard policy 6 convergences with policy mix: 3 oscillatory convergence paths, 3 liquidity traps Severe pessimism 5 deflationary spirals (initial pessimism) 2 convergences to steady state with ZLB 1 convergence with standard policy 6 convergences with policy mix: 2 oscillatory convergence paths, 4 liquidity traps

28 Example of a deflationary spiral after a bad news shock Treatment MP_a group Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

29 Example of a convergence to the steady state despite a bad Treatment MP_a group 7 5 news shock 0 5 Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

30 80 Initial expectations in Tr. MP/mild pessimism (a) L T net output stable unstable 50-8% π T +8% π L inflation

31 Expectations before the news in Tr. MP/mild pessimism (a) 80 L T net output stable unstable 50-8% π T +8% π L inflation

32 80 Expectations after the news in Tr. MP/mild pessimism (a) L T net output stable unstable 50-8% π T +8% π L inflation

33 80 Fall in a liquidity trap by expectational shock L T net output stable unstable 50-8% π T +8% π L inflation

34 80 Fall in a liquidity trap by expectational shock L T net output stable unstable 50-8% π T +8% π L inflation

35 80 Fall in a liquidity trap by expectational shock L T net output stable unstable 50-8% π T +8% π L inflation

36 Treatments with policy mix and news shocks Monetary policy only Policy mix Mild pessimism (news shock) 4 deflationary spirals after news 3 convergences to steady state (1 with ZLB) 1 convergence with standard policy 6 convergences with policy mix: 3 oscillatory convergence paths, 3 liquidity traps Severe pessimism 5 deflationary spirals (initial pessimism) 2 convergences to steady state with ZLB 1 convergence with standard policy 6 convergences with policy mix: 2 oscillatory convergence paths, 4 liquidity traps

37 Example of a oscillatory convergence path with the policy mix Treatment MFP_a group Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

38 Example of a liquidity trap path with the policy mix Treatment MFP_a group Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

39 Effect of policy mix after the news in Tr. MFP/mild pessimism (a) 80 L T net output stable unstable 50-8% π T +8% π L inflation

40 Effect of policy mix after the news in Tr. MFP/mild pessimism (a) 80 L T net output stable unstable 50-8% π T +8% π L inflation

41 Treatments with policy mix and news shocks Monetary policy only Policy mix Mild pessimism (news shock) 4 deflationary spirals after news 3 convergences to steady state (1 with ZLB) 1 convergence with standard policy 6 convergences with policy mix: 3 oscillatory convergence paths, 3 liquidity traps Severe pessimism 5 deflationary spirals (initial pessimism) 2 convergences to steady state with ZLB 1 convergence with standard policy 6 convergences with policy mix: 3 oscillatory convergence paths, 3 liquidity traps

42 Example of a oscillatory convergence path with the policy mix Treatment MFP_b group Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

43 Example of a liquidity trap path with the policy mix Treatment MFP_b group Inflation Interest Rate Consumption (left axis) Gov. Exp. (right axis) Period 15

44 80 Effect of policy mix in Tr. MFP/severe pessimism (b) L T net output stable unstable 50-8% π T +8% π L inflation

45 80 Effect of policy mix in Tr. MFP/severe pessimism (b) L T net output stable unstable 50-8% π T +8% π L inflation

46 Experimental results Estimates of the fiscal multiplier in the experiments Direct estimates using MFP (a & b): fiscal multiplier = y t g = 1 + c t g Treatment MFP/a Group 2 Group 4 Group 5 Group 6 Group 7 Group 8 average number of fiscal shocks Treatment MFP/b Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 average i i number of fiscal shocks Table: Average effects of fiscal shocks on output in Treatments MFP. fiscal multiplier [1, 1.5] Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

47 Experimental results Estimates of the fiscal multiplier in the experiments Estimates based on counter-factual analysis (MP/a vs. MFP/a): fiscal multiplier t = 1 + c MFP,t c MP,t g MFP,t ḡ t = 8,9,10 (following the news shocks) gvt. expenditure MFP MP news shock consumption MFP MP news shock estimated multiplier news shock Figure: Comparison of Treatments MP/a and MPF/a after the news Fiscal policy crowds in private consumption through the expectations channel (drop in expected real interest). Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

48 Conclusions 1 Motivation 2 Model 3 Experimental design 4 Experimental results 5 Conclusions Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

49 Conclusions Sum-up 1 Deflationary spirals can arise as a result of pessimistic expectations even after a preventive cut in the interest rate. 2 A policy mix rules out deflationary spirals. The fiscal multiplier raises above one through the expectations channel. 3 However, an increase in public spending 1 is necessary only if the drop in expectations is strong enough, otherwise aggressive monetary policy alone is enough; 2 may lock the economy in a liquidity trap (low and stable inflation state) for an extended period of time. Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

50 Conclusions Research questions: Behavioural Experiments 1 Expectations are the most important transmission channel of macro policies Learning to Forecast Experiments empirical validation of policy recommendations that rely on specific expectation formation mechanisms e.g. what to target? forward guidance, fiscal policy taking into account sovereign risk channel. 2 Learning to Optimize Experiments: expectations are only one part of the story. 3 Larger scale-experiments. Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51

51 Conclusions Thank you for your attention Hommes, Massaro & Salle (UvA) Liquidity Trap in the Lab Leiden / 51