Tax multipliers: Pitfalls in measurement and identi cation

Transcription

1 Tax multipliers: Pitfalls in measurement and identi cation Daniel Riera-Crichton Bates College Carlos Vegh Univ. of Maryland and NBER Guillermo Vuletin Colby College Indiana University April 25, 2013 Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

2 Fiscal multipliers: The big picture Main objective: Evaluate the impact of changes in government spending and taxes on output G! Y T! Y Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

3 Fiscal multipliers: The big picture Main objective: Evaluate the impact of changes in government spending and taxes on output G! Y T! Y Empirical evidence regarding the size of multipliers varies greatly: G multiplier (increase in $1 of G) : 2.5 to 3.6 T multiplier (increase in $1 of rev.) : 5 to 0.3 Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

4 Fiscal multipliers: The big picture Main objective: Evaluate the impact of changes in government spending and taxes on output G! Y T! Y Empirical evidence regarding the size of multipliers varies greatly: G multiplier (increase in $1 of G) : 2.5 to 3.6 T multiplier (increase in $1 of rev.) : 5 to 0.3 Some variation surely due to country characteristics and/or stance of business cycle (Ilzetzki et al, 2010, Auerbach and Gorodnichenko, 2011) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

5 Fiscal multipliers: The big picture Main objective: Evaluate the impact of changes in government spending and taxes on output G! Y T! Y Empirical evidence regarding the size of multipliers varies greatly: G multiplier (increase in $1 of G) : 2.5 to 3.6 T multiplier (increase in $1 of rev.) : 5 to 0.3 Some variation surely due to country characteristics and/or stance of business cycle (Ilzetzki et al, 2010, Auerbach and Gorodnichenko, 2011) But a lot of variation likely due to identi cation and/or measurement problems Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

6 Sources of endogeneity Endogeneity source # 1: Fiscal variables also respond to output uctuations Y! G Y! T Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

7 Sources of endogeneity Endogeneity source # 1: Fiscal variables also respond to output uctuations Y! G Y! T Endogeneity source # 2: Policy instrument versus policy outcome tax revenue {z } policy outcome = tax {z rate} policy instrument tax base {z } in uenced by Y Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

8 Empirical strategies SVAR (Blanchard and Perotti, 2002): timing assumption Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

9 Empirical strategies SVAR (Blanchard and Perotti, 2002): timing assumption G side: it takes G at least one quarter to respond to Y Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

10 Empirical strategies SVAR (Blanchard and Perotti, 2002): timing assumption G side: it takes G at least one quarter to respond to Y T side: idem + use of revenue-output elasticity to di erentiate discretionary changes in taxes from those driven by changes in output Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

11 Empirical strategies SVAR (Blanchard and Perotti, 2002): timing assumption G side: it takes G at least one quarter to respond to Y T side: idem + use of revenue-output elasticity to di erentiate discretionary changes in taxes from those driven by changes in output Natural experiment approach (mostly for the US): Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

12 Empirical strategies SVAR (Blanchard and Perotti, 2002): timing assumption G side: it takes G at least one quarter to respond to Y T side: idem + use of revenue-output elasticity to di erentiate discretionary changes in taxes from those driven by changes in output Natural experiment approach (mostly for the US): G side (Barro, 1981): focus on military spending (the least likely to respond to output uctuations) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

13 Empirical strategies SVAR (Blanchard and Perotti, 2002): timing assumption G side: it takes G at least one quarter to respond to Y T side: idem + use of revenue-output elasticity to di erentiate discretionary changes in taxes from those driven by changes in output Natural experiment approach (mostly for the US): G side (Barro, 1981): focus on military spending (the least likely to respond to output uctuations) T side (Romer and Romer, 2010): narrative approach of tax changes. Di erentiate exogenous (de cit-driven and long-run growth) from endogenous (to output uctuations) tax changes. Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

14 This paper: Disentangling identi cation from measurement Measurement of tax policy (Finding a tax policy variable under the direct control of the policymaker) Tax rate Cyclically adjusted revenue Identification of exogenous fiscal shocks (Fiscal policy changes that are not the result of policymakers responding to output fluctuations) "Natural experiment" (à la Romer and Romer, 2010) Structural VAR (à la Blanchard and Perotti, 2002) No endogeneity. No measurement error. Endogeneity problems if tax changes are forwardlooking/anticipated. No measurement error. No endogeneity. Possible measurement error, especially due to changes in non policy factors. Endogeneity problems if tax changes are forwardlooking/anticipated. Possible measurement error, especially due changes in to non policy factors. Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

15 What do we do? Identi cation issue (revisited): SVAR vs. Fiscal consolidation (IMF) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

16 What do we do? Identi cation issue (revisited): SVAR vs. Fiscal consolidation (IMF) Measurement of tax policy (less explored): cyclically-adjusted revenue tax rate vs. Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

17 What do we do? Identi cation issue (revisited): SVAR vs. Fiscal consolidation (IMF) Measurement of tax policy (less explored): tax rate vs. cyclically-adjusted revenue Novel VAT quarterly data for 14 industrial countries ( ) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

18 What do we do? Identi cation issue (revisited): SVAR vs. Fiscal consolidation (IMF) Measurement of tax policy (less explored): cyclically-adjusted revenue tax rate vs. Novel VAT quarterly data for 14 industrial countries ( ) Individual income tax for the US (Romer and Romer, 2010; Barro and Redlick, 2011) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

19 Preview of results Punchline # 1 (on measurement): poor proxy for changes in tax rate Cyclically-adjusted revenue is a Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

20 Preview of results Punchline # 1 (on measurement): poor proxy for changes in tax rate Cyclically-adjusted revenue is a Punchline # 2 (on identi cation): Natural experiment does better job in identifying truly exogenous scal shocks Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

21 Data Tax rate data Novel VAT quarterly data for 14 industrial countries ( ) 49 VAT rate changes (36 increases and 13 decreases) Av. marginal individual income for the US (Barro and Redlick, 2011) Cyclically-adjusted revenues Novel VAT quarterly data (cyclically-adjusted revenues) = revenues η Y Individual income for the US (Romer and Romer, 2010) Natural experiment IMF action-based scal consolidation episodes Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

22 IMF scal consolidation episodes World Economic Outlook, October 2010 Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

23 IMF scal consolidation episodes World Economic Outlook, October 2010 Sample: , 15 industrial countries Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

24 IMF scal consolidation episodes World Economic Outlook, October 2010 Sample: , 15 industrial countries Identi ed episodes in which government implemented tax hikes... or spending cuts... to reduce the budget de cit and put public nances on a more sustainable footing Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

25 IMF scal consolidation episodes World Economic Outlook, October 2010 Sample: , 15 industrial countries Identi ed episodes in which government implemented tax hikes... or spending cuts... to reduce the budget de cit and put public nances on a more sustainable footing Roughly corresponds to RR s de cit-driven category of exogenous scal shocks Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

26 IMF scal consolidation episodes World Economic Outlook, October 2010 Sample: , 15 industrial countries Identi ed episodes in which government implemented tax hikes... or spending cuts... to reduce the budget de cit and put public nances on a more sustainable footing Roughly corresponds to RR s de cit-driven category of exogenous scal shocks We use IMF study as our identi cation mechanism: 21 out of 36 increases exogenous Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

27 Tax rate vs cyclically-adjusted: UK?tax rate ?tax rate?cyclically adjusted revenue ?cyclically adjusted revenue :2 1991:1 1992:1 1993:1 1994:1 1995:1 1996:1 1997:1 1998:1 1999:1 2000:1 2001:1 2002:1 2003:1 2004:1 2005:1 2006:1 2007:1 2008:1 2009:1 Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

28 Tax rate vs cyclically-adjusted: Ireland?tax rate ?tax rate?cyclically adjusted revenue ?cyclically adjusted revenue :2 2000:1 2001:1 2002:1 2003:1 2004:1 2005:1 2006:1 2007:1 2008:1 2009:1 Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

29 Correlation between tax rate changes and cyclically-adjusted changes Non-zero tax changes: correlation is 0.26 (and insigni cant) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

30 Correlation between tax rate changes and cyclically-adjusted changes Non-zero tax changes: correlation is 0.26 (and insigni cant) Fiscal consolidation non-zero tax rate changes: correlation is 0.09 (and insigni cant) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

31 Basic speci cation y i,t = α + J β j t i,t j + I 1 β i d i + µ i,t, j=0 i=1 y i,t = real GDP growth t i,t j = scal consolidation tax rate change d i = country dummy By construction, changes in tax rates uncorrelated with contemporaneous and future output Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

32 Cumulative tax multiplier: shock Fiscal consolidation tax rate Quarter Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

33 Cumulative tax multiplier: Robustness checks We control for government spending (since it likely falls in episodes) by running a three variable SVAR Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

34 Cumulative tax multiplier: Robustness checks We control for government spending (since it likely falls in episodes) by running a three variable SVAR We control for changes in other taxes (corporate and income taxes) by running a four variable SVAR Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

35 Fiscal consolidation tax rate shock: GDP components Panel A. Cumulative tax multiplier for consumption GDP Consumption Quarter Panel B. Cumulative tax multiplier for investment GDP Investment Quarter Panel C. Cumulative tax multiplier for exports Panel B. Cumulative tax multiplier for imports 5 GDP Exports Quarter 5 GDP Imports Quarter Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

36 Cyclically-adjusted revenue multiplier Using scal consolidation episodes to identify scal shocks Using fiscal consolidation tax rates Using fiscal consolidation cyclically adjusted revenues Quarter Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

37 Cyclically-adjusted revenue multiplier Using all cyclically-adjusted changes Using fiscal consolidation tax rates Using all cyclically adjusted revenues Quarter Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

38 Response of tax rate to GDP Percent Using fiscal consolidation tax rates Using all cyclically adjusted revenues Quarter Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

39 Summary: Tax multipliers Measurement of tax policy (Finding a tax policy variable under the direct control of the policymaker) Tax rate Cyclically adjusted revenue Identification of exogenous fiscal shocks "Natural experiment" (à la Romer and Romer, 2010) Very Contractionary Short run multiplier = 1.02 Long run multiplier = 2.76 Neutral Multipliers = 0 (Fiscal policy changes that are not the result of policymakers responding to output fluctuations) Structural VAR (à la Blanchard and Perotti, 2002) Contractionary Short run multiplier = 0.37 Long run multiplier = 1.32 Expansionary Short run multiplier = 0.18 Long run multiplier = 0.78 Tax rate (contractionary) Cyclically-adjusted (neutral or expansionary!) Neutral: Perotti (2004), Favero and Giavazzi (2010), Ilzetzki (2011) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

40 United States: Measurement Using scal consolidation episodes to identify scal shocks and US individual income tax data 1. Using cyclically-adjusted revenue Quarter 3. Using Barro-Redlick av. marginal tax rate Quarter 2. Using Romer-Romer revenue change Quarter Corr(CA, tax rate)=0.17 Corr(RR, tax rate)=0.35*** Corr(RR, CA)=0.59*** Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

41 Identi cation problems bias multipliers in di erent directions Response of endogenous tax rate changes to output shock 0.02 Industrial Developing Quarter Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

42 Tax multiplier: Industrial versus developing Tax multiplier (industrial) Tax multiplier (developing) Using exogenous tax changes 5 Using all tax changes Using all tax changes Using exogenous tax changes 5 6 Quarter Quarter Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

43 Conclusions We use novel VAT data for 14 industrial countries and existing US individual income data Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

44 Conclusions We use novel VAT data for 14 industrial countries and existing US individual income data We revisit the discussion on identi cation of scal shocks: natural experiment vs SVAR Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

45 Conclusions We use novel VAT data for 14 industrial countries and existing US individual income data We revisit the discussion on identi cation of scal shocks: natural experiment vs SVAR Finding: natural experiment is a better way of identifying exogenous scal policy changes Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

46 Conclusions We use novel VAT data for 14 industrial countries and existing US individual income data We revisit the discussion on identi cation of scal shocks: natural experiment vs SVAR Finding: natural experiment is a better way of identifying exogenous scal policy changes We analyze the (much less discussed) issue of measurement of tax policy (i.e., measure of policy instrument) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

47 Conclusions We use novel VAT data for 14 industrial countries and existing US individual income data We revisit the discussion on identi cation of scal shocks: natural experiment vs SVAR Finding: natural experiment is a better way of identifying exogenous scal policy changes We analyze the (much less discussed) issue of measurement of tax policy (i.e., measure of policy instrument) Finding: Tax rates are a better measure of tax policy instrument as opposed to widely-used revenue-based measures, such as cyclically-adjusted revenues. Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24

48 Conclusions We use novel VAT data for 14 industrial countries and existing US individual income data We revisit the discussion on identi cation of scal shocks: natural experiment vs SVAR Finding: natural experiment is a better way of identifying exogenous scal policy changes We analyze the (much less discussed) issue of measurement of tax policy (i.e., measure of policy instrument) Finding: Tax rates are a better measure of tax policy instrument as opposed to widely-used revenue-based measures, such as cyclically-adjusted revenues. tax rate (contractionary), cyclically-adjusted (neutral or expansionary!) Riera-Vegh-Vuletin (Indiana University) Tax multipliers April 25, / 24