Does the Confidence Fairy Exist?

Transcription

1 Does the Confidence Fairy Exist? Evidence from a New Narrative Dataset on Fiscal Austerity Announcements Oana Furtuna 1, Roel Beetsma 2 and Massimo Giuliodori 1 1 University of Amsterdam, Tinbergen Institute 2 MN Chair in Pension Economics, University of Amsterdam, Tinbergen Institute and CEPR June 6, 2016 Fiscal Sustainability, XXI Century, Madrid 0 / 45

2 I firmly believe that in the current circumstances, confidence-inspiring policies will foster and not hamper economic recovery, because confidence is the key factor today. (Jean-Claude Trichet, La Repubblica, 2010 ) The need for fiscal austerity rests on the belief that invisible vigilantes will punish us if we re bad and the confidence fairy will reward us if we re good. (Paul Krugman, The New York Times, 2010) 1 / 45

3 Effects of Fiscal Adjustments Fiscal Adjustments T Contractionary Guajardo et al. (2014) Contractionary (Keynesian) Guajardo et al. (2011) C G Expansionary Alesina, Perotti (1997) Alesina, Ardagna (2010) C 2 / 45

4 G versus T Why? Even when G are contractionary, smaller output drop than for T (Guajardo et al.,2014; Alesina et al., 2015). Labor market effects (Ardagna, 2004) More expansionary monetary policy (Guajardo et al., 2014) A spending-based consolidation may have positive expectation effects (Blanchard, 1990) Lower need for larger future adjustments is good news If the adjustment is a sign that government consumption will be permanently reduced 3 / 45

5 Confidence Reacts to Fiscal Adjustments... Consumer confidence Bachmann and Sims (2012) - structural VAR, U.S.A. Positive reaction of consumer confidence to increases in government consumption, only in a recession Beetsma et al. (2015) - monthly event study, panel of OECD countries Negative reaction of consumer confidence to austerity Government bond yields Ardagna (2009) - annual data, panel of OECD countries Yields of long-term government bonds fall when CAPB improves. Akitoby and Stratmann (2008) - annual data, panel of EMEs Fiscal adjustment reduces sovereign bond spreads. 4 / 45

6 ...Differently for Spending and Revenue Measures Tax-Based Spending-Based Consumer Confidence = Beetsma et al. (2015) Producer Confidence Alesina et al. (2015) Gov. bond yield spreads Akitoby, Stratmann (2008) Government bond yields Ardagna (2009) 5 / 45

7 Research Questions This paper Q1. Does the confidence channel contribute to the transmission of fiscal shocks onto real variables? Q2. Do responses to announcements differ for revenues versus spending based measures? 6 / 45

8 Research Questions This paper Q1. Does the confidence channel contribute to the transmission of fiscal shocks onto real variables? Q2. Do responses to announcements differ for revenues versus spending based measures? 6 / 45

9 Research Questions This paper Aims to isolate the role of the confidence channel by using announcements of future fiscal policy actions. 7 / 45

10 Data Contribution Outline 1 Data Contribution 2 Panel VAR 3 Counterfactual Analysis 4 Announcements of Revenue versus Spending-Based Plans 5 Revenue and Spending Components of All Announcements 6 Difference Between Revenue and Spending Plans 7 Conclusion 8 / 45

11 Data Contribution Fiscal Consolidations: Annual Data Devries et al. (2011): annual dataset with implemented fiscal consolidations (CS i,t ) Narrative approach - official records: Budget Speeches, IMF Reports, OECD Economic Surveys Exogenous to business cycle dynamics , 13 European countries, 114 consolidations Alesina et al. (2015a): anticipated (CS A i,t ) and unanticipated (CS U i,t ) components of implemented measures Alesina et al. (2015b): extend time sample to / 45

12 Data Contribution Limitations of Annual Data 1 Recording measures when implemented ignores anticipation effects (legislative and implementation lags). 2 Alesina et al. (2015) record as unanticipated measures announced in the fall of the year before, so this may generate anticipation. 3 There may be more than one announcement moment within a year. 10 / 45

13 Data Contribution New Dataset of Fiscal Consolidation Announcements Real-time policy announcements based on official documents and OECD reports, checked against newspaper articles. Expand the available data by identifying the quarter when the government released information about the consolidation Extend coverage of Alesina et al. (2015) data to include Finland, the Netherlands, Sweden Compute the value of the consolidation announcement by summing its impact on the primary budget balance in all years concerned. 11 / 45

14 Data Contribution Fiscal Consolidations: Announcement Data quarter 1992:Q1 1992:Q3 1992:Q4 1993:Q2 1993:Q3 month March August October April July spend spend t tax tax t Full announcement Revenue component Spending component Type of announcement TB SB TB TB SB News: Belgium / 45

15 Panel VAR Outline 1 Data Contribution 2 Panel VAR 3 Counterfactual Analysis 4 Announcements of Revenue versus Spending-Based Plans 5 Revenue and Spending Components of All Announcements 6 Difference Between Revenue and Spending Plans 7 Conclusion 13 / 45

16 Panel VAR PVAR Model Y i,t = A(L)Y i,t 1 + U i,t U i,t = zero-mean, stationary reduced-form disturbance Y i,t = [F i,t, CC i,t, LTI i,t, I i,t, C i,t, GDP i,t ] F = announcement shock (% of GDP), CC = log-consumer confidence index, LTI = long-term government bond yield, I = log-private investment, C = log-private consumption, GDP = log-real GDP. Baseline: VAR in levels, 4 lags, country fixed effects, country-specific trends, seasonal dummies 14 / 45

17 Panel VAR Identification Choleski factorization with F i,t shock ordered first Responses invariant to ordering of variables following shock (Christiano et al., 1999) F i,t unanticipated at quarterly frequency F i,t exogenous to business cycle 15 / 45

18 Panel VAR Impulse Responses (CC) 1 Consumer confidence Impulse responses to an austerity announcement of 1% of GDP. 90% confidence bands based on 1000 bootstrap replications. 16 / 45

19 Panel VAR Impulse Responses (LTI ) 30 Long-term interest rate Impulse responses to an austerity announcement of 1% of GDP. 90% confidence bands based on 1000 bootstrap replications. 17 / 45

20 Panel VAR Impulse Responses (I ) 1 Investment Impulse responses to an austerity announcement of 1% of GDP. 90% confidence bands based on 1000 bootstrap replications. 18 / 45

21 Panel VAR Impulse Responses (C) Consumption Impulse responses to an austerity announcement of 1% of GDP. 90% confidence bands based on 1000 bootstrap replications. 19 / 45

22 Panel VAR Impulse Responses (GDP) Real GDP Impulse responses to an austerity announcement of 1% of GDP. 90% confidence bands based on 1000 bootstrap replications. 20 / 45

23 Panel VAR Robustness Different deterministic components Here Sample choice Time sample Country sample Here Here Alternative specifications Here Reduced form local projection IRFs Here 21 / 45

24 Panel VAR Q1. Does confidence contribute to the transmission of fiscal shocks onto real variables? 22 / 45

25 Counterfactual Analysis Outline 1 Data Contribution 2 Panel VAR 3 Counterfactual Analysis 4 Announcements of Revenue versus Spending-Based Plans 5 Revenue and Spending Components of All Announcements 6 Difference Between Revenue and Spending Plans 7 Conclusion 23 / 45

26 Counterfactual Analysis Joint Confidence Channel 1 Investment Switch off both Baseline Both consumer confidence and long-term interest rate channels switched off 24 / 45

27 Counterfactual Analysis Joint Confidence Channel Consumption Swith off both Baseline Both consumer confidence and long-term interest rate channels switched off 25 / 45

28 Counterfactual Analysis Joint Confidence Channel Real GDP Baseline Switch off both Both consumer confidence and long-term interest rate channels switched off 26 / 45

29 Counterfactual Analysis Q2. Do responses to announcements differ for revenues versus spending based measures? 27 / 45

30 Announcements of Revenue versus Spending-Based Plans Outline 1 Data Contribution 2 Panel VAR 3 Counterfactual Analysis 4 Announcements of Revenue versus Spending-Based Plans 5 Revenue and Spending Components of All Announcements 6 Difference Between Revenue and Spending Plans 7 Conclusion 28 / 45

31 Announcements of Revenue versus Spending-Based Plans Revenue and Spending-Based Announcements Consumer confidence Revenue-based announcements Spending-based announcements 29 / 45

32 Announcements of Revenue versus Spending-Based Plans Revenue and Spending-Based Announcements Long-term interest rate Revenue-based announcements Spending-based announcements 30 / 45

33 Announcements of Revenue versus Spending-Based Plans Revenue and Spending-Based Announcements Investment Revenue-based announcements Spending-based announcements 31 / 45

34 Announcements of Revenue versus Spending-Based Plans Revenue and Spending-Based Announcements Consumption Revenue-based announcements Spending-based announcements 32 / 45

35 Announcements of Revenue versus Spending-Based Plans Revenue and Spending-Based Announcements Real GDP Revenue-based announcements Spending-based announcements 33 / 45

36 Revenue and Spending Components of All Announcements Outline 1 Data Contribution 2 Panel VAR 3 Counterfactual Analysis 4 Announcements of Revenue versus Spending-Based Plans 5 Revenue and Spending Components of All Announcements 6 Difference Between Revenue and Spending Plans 7 Conclusion 34 / 45

37 Revenue and Spending Components of All Announcements Revenue and Spending Components of Announcement Consumer confidence Revenue component Spending component 35 / 45

38 Revenue and Spending Components of All Announcements Revenue and Spending Components of Announcement Long-term interest rate Revenue component Spending component 36 / 45

39 Revenue and Spending Components of All Announcements Revenue and Spending Components of Announcement 2 Investment Revenue component Spending component 37 / 45

40 Revenue and Spending Components of All Announcements Revenue and Spending Components of Announcement 0.5 Consumption Revenue component Spending component 38 / 45

41 Revenue and Spending Components of All Announcements Revenue and Spending Components of Announcement Real GDP Revenue component Spending component 39 / 45

42 Difference Between Revenue and Spending Plans Outline 1 Data Contribution 2 Panel VAR 3 Counterfactual Analysis 4 Announcements of Revenue versus Spending-Based Plans 5 Revenue and Spending Components of All Announcements 6 Difference Between Revenue and Spending Plans 7 Conclusion 40 / 45

43 Difference Between Revenue and Spending Plans What Explains The Difference Between Revenue and Spending? Inflation expectations? We control for ex-post CPI and ex-ante one-year ahead inflation expectations. No significant difference in responses. Monetary policy (Guajardo et al., 2014)? Not significant (Alesina et al, 2015). Also, since 1999 EMU. Credibility? The extent of follow-up of the plan. Deviations of implementations from plans: larger for spending than for tax components of plans. 41 / 45

44 Difference Between Revenue and Spending Plans What Explains The Difference Between Revenue and Spending? Inflation expectations? We control for ex-post CPI and ex-ante one-year ahead inflation expectations. No significant difference in responses. Monetary policy (Guajardo et al., 2014)? Not significant (Alesina et al, 2015). Also, since 1999 EMU. Credibility? The extent of follow-up of the plan. Deviations of implementations from plans: larger for spending than for tax components of plans. 41 / 45

45 Difference Between Revenue and Spending Plans What Explains The Difference Between Revenue and Spending? Inflation expectations? We control for ex-post CPI and ex-ante one-year ahead inflation expectations. No significant difference in responses. Monetary policy (Guajardo et al., 2014)? Not significant (Alesina et al, 2015). Also, since 1999 EMU. Credibility? The extent of follow-up of the plan. Deviations of implementations from plans: larger for spending than for tax components of plans. 41 / 45

46 Difference Between Revenue and Spending Plans What Explains The Difference Between Revenue and Spending? Inflation expectations? We control for ex-post CPI and ex-ante one-year ahead inflation expectations. No significant difference in responses. Monetary policy (Guajardo et al., 2014)? Not significant (Alesina et al, 2015). Also, since 1999 EMU. Credibility? The extent of follow-up of the plan. Deviations of implementations from plans: larger for spending than for tax components of plans. 41 / 45

47 Difference Between Revenue and Spending Plans What Explains The Difference Between Revenue and Spending? Inflation expectations? We control for ex-post CPI and ex-ante one-year ahead inflation expectations. No significant difference in responses. Monetary policy (Guajardo et al., 2014)? Not significant (Alesina et al, 2015). Also, since 1999 EMU. Credibility? The extent of follow-up of the plan. Deviations of implementations from plans: larger for spending than for tax components of plans. 41 / 45

48 Difference Between Revenue and Spending Plans What Explains The Difference Between Revenue and Spending? Inflation expectations? We control for ex-post CPI and ex-ante one-year ahead inflation expectations. No significant difference in responses. Monetary policy (Guajardo et al., 2014)? Not significant (Alesina et al, 2015). Also, since 1999 EMU. Credibility? The extent of follow-up of the plan. Deviations of implementations from plans: larger for spending than for tax components of plans. 41 / 45

49 Conclusion Outline 1 Data Contribution 2 Panel VAR 3 Counterfactual Analysis 4 Announcements of Revenue versus Spending-Based Plans 5 Revenue and Spending Components of All Announcements 6 Difference Between Revenue and Spending Plans 7 Conclusion 42 / 45

50 Conclusion Conclusion Consumer confidence drops and long-term bond yields increase (= expectations ) in response to announcement shock. Output, investment and consumption react to announcements with a lag The joint expectations channel explains most of the movement in real variables. Stronger effect of revenue- than spending-based announcements. 43 / 45

51 Conclusion Conclusion Consumer confidence drops and long-term bond yields increase (= expectations ) in response to announcement shock. Output, investment and consumption react to announcements with a lag The joint expectations channel explains most of the movement in real variables. Stronger effect of revenue- than spending-based announcements. 43 / 45

52 Conclusion Conclusion Consumer confidence drops and long-term bond yields increase (= expectations ) in response to announcement shock. Output, investment and consumption react to announcements with a lag The joint expectations channel explains most of the movement in real variables. Stronger effect of revenue- than spending-based announcements. 43 / 45

53 Conclusion Further work Difference between revenue and expenditure measures: credibility Difference between revenue and expenditure measures: disaggregate further (VAT, income tax, etc.) Trade-off short-term and medium-term effects Effects dependent on business cycle (projection method) 44 / 45

54 Conclusion Further work Difference between revenue and expenditure measures: credibility Difference between revenue and expenditure measures: disaggregate further (VAT, income tax, etc.) Trade-off short-term and medium-term effects Effects dependent on business cycle (projection method) 44 / 45

55 Conclusion Further work Difference between revenue and expenditure measures: credibility Difference between revenue and expenditure measures: disaggregate further (VAT, income tax, etc.) Trade-off short-term and medium-term effects Effects dependent on business cycle (projection method) 44 / 45

56 Conclusion Further work Difference between revenue and expenditure measures: credibility Difference between revenue and expenditure measures: disaggregate further (VAT, income tax, etc.) Trade-off short-term and medium-term effects Effects dependent on business cycle (projection method) 44 / 45

57 Conclusion Two Fairies. Confidence is not one. 45 / 45

58 Conclusion The sad truth is that we economists don t know very much about what drives the animal spirits of economic participants. Until we figure it out, it is best to be suspicious of any policy whose benefits are supposed to work through the amorphous channel of confidence. (Gregory Mankiw, 2009) Questions, Comments, Suggestions: O.Furtuna@uva.nl 45 / 45

59 Appendix Robustness 1 Consumer confidence Baseline 8 lags Lagged debt Common trend Time effects 45 / 45

60 Appendix Robustness Long-term interest rate Baseline 8 lags Lagged debt Common trend Time effects 45 / 45

61 Appendix Robustness 1 1 Investment Baseline 8 lags Lagged debt Common trend Time effects 45 / 45

62 Appendix Robustness Consumption Baseline 8 lags Lagged debt Common trend Time effects 45 / 45

63 Appendix Robustness Real GDP Baseline 8 lags Lagged debt Common trend Time effects 45 / 45

64 Appendix Robustness 2: Time Consumer confidence Baseline Pre-2000 Post-2000 Pre / 45

65 Appendix Robustness 2: Time Long-term interest rate Baseline Pre-2000 Post-2000 Pre / 45

66 Appendix Robustness 2: Time Investment Baseline Pre-2000 Post-2000 Pre / 45

67 Appendix Robustness 2: Time 0.2 Consumption Baseline Pre-2000 Post-2000 Pre / 45

68 Appendix Robustness 2: Time 0.2 Real GDP Baseline pre-2000 post-2000 pre-2009 Back 45 / 45

69 Appendix Robustness 2: Countries Consumer confidence / 45

70 Appendix Robustness 2: Countries Long-term interest rate / 45

71 Appendix Robustness 2: Countries Investment / 45

72 Appendix Robustness 2: Countries Consumption / 45

73 Appendix Robustness 2: Countries Real GDP Back 45 / 45

74 Appendix Robustness 3: Local Projection IRFs 3 Confidence Linear Projection VAR 45 / 45

75 Appendix Robustness 3: Local Projection IRFs Long-term interest rate Linear Projection VAR 45 / 45

76 Appendix Robustness 3: Local Projection IRFs Investment Linear Projection VAR 45 / 45

77 Appendix Robustness 3: Local Projection IRFs Consumption Linear Projection VAR 45 / 45

78 Appendix Robustness 3: Local Projection IRFs Real GDP Linear Projection VAR 45 / 45

79 Appendix Implementation - Announcement Deviations Higher deviation Italy Spain Ireland Finland Belgium Portugal Lower deviation Germany France Austria Sweden Denmark United Kingdom Netherlands Country groups based on ranking of average deviations Back 45 / 45

80 Appendix Robustness 4: Alternative specifications Back Including government expenditure and tax revenue Including unemployment rate Including exports and imports Including stock price index Including short-term interest rate 45 / 45