1. A New Test of Ricardian Equivalence Using the Narrative Record on Tax Changes 1 IAAE, Milan, 2016 Alfred A. Haug (University of Otago, New Zealand)

2. Ricardian Equivalence Hypothesis: does not matter whether government expenditures are financed by current taxation or instead by issuing government bonds and hence future taxation (Barro, 1974) Ricardian households are only concerned about the PV of their intertemporal tax liabilities, determined by the PV of the stream of government expenditures and currently outstanding debt Based on restrictive assumptions: – lump-sum taxes – perfect capital markets (no liquidity constraints) – altruistic operative bequests – no uncertainty about the future tax incidence 2

3. Surveys: Seater (1993) and Ricciuti (2003) Recent contributions: – Evans, Honkapohja and Mitra (2012) showed rational expectations are not necessary for Ricardian equivalence and a certain adaptive learning rule instead can produce equivalence – Barro and Mollerus (2014) presented theoretical models How good of an approximation Ricardian equivalence is cannot be determine on theoretical grounds; it is an important benchmark Empirical tests are called for, however, the empirical evidence is not conclusive at all; it crucially matters for tax multipliers Narrative approach opens a new way for estimating the multiplier effects of taxes levied to improve government finances (debt reduction) 3

4. Narrative Methods Romer and Romer (2010): surprise switch from bond financing to taxation; introduced for reasons unrelated to the business cycle or government spending (“exogenous”) Romer and Romer measure: PV of news of a stream of discretionary changes in U.S. tax revenue motivated by concerns about the inherited government budget deficit only They did not analyze this variable in separation (tax changes motivated by log-run economic growth were added in; Romer and Romer (2009a) studied tax cuts driven by long-run economic considerations) Such a switch should have no effects on real GDP and interest rates if Ricardian equivalence is a good approximation 4

5. Construction of the Narrative News Variable Romer and Romer used historical documents to identify types of legislated tax changes from the motivations provided by lawmakers at the time of passage: U.S. Congress, Economic Report of the President, Budget, Senate, Social Security, and Treasury Secretary reports, records or bulletins Fiscal surprise shock (news): is an exogenous "deficit-driven" tax revenue increase; orthogonal by construction to all other information available in a given period Ten (nine) quarters with deficit-driven exogenous tax changes are in the sample Period from 1945:I to 2007:IV (1950:I to 2007:IV) 5

6. 6

7. Estimation OLS (single equation): 7

8. Impact of an exogenous deficit-driven tax increase of 1% of GDP on GDP, with no controls for lagged growth 8

9. Impact of an exogenous deficit-driven tax increase of 1% of GDP on GDP, with controls for lagged growth & bootstrapping 9

10. Structural breaks: CUSUMSQ 10

11. Prior-1980:IV estimated impact of an exogenous deficit- driven tax increase of 1 % of GDP on GDP, with controls & bootstrapping 11

12. Post-1980:IV estimated impact of an exogenous deficit- driven tax increase of 1 % of GDP on GDP, with controls & bootstrapping 12

13. Prior-1980:IV for a VAR with 3 variables (tax shock, real GDP, interest rate) & bootstrapping 13

14. Prior-1980:IV for a VAR with 3 variables (tax shock, real GDP, interest rate) & bootstrapping 14

15. Post-1980:IV for a VAR with 3 variables (tax shock, real GDP, interest rate) & bootstrapping 15

16. Post-1980:IV for a VAR with 3 variables (tax shock, real GDP, interest rate) & bootstrapping 16

17. Summary and Conclusion Ricardian equivalence questions that aggregate demand in the economy can be stimulated by tax cuts, if taxpayers take into account the implications of today's tax cuts, financed by borrowing, on taxes in the future and vice versa for tax increases Empirical results do not support the Ricardian equivalence hypothesis: effects of a tax increase, keeping government spending fixed, have a statistically significant and positive influence on real GDP and nominal interest rates, but only post 1980:IV, not prior to 1980:IV Fiscal foresight explains the statistically significant positive effects of tax increases on real GDP when implementation dates are ignored 17