On the low-frequency relationship between public deficits and - - PowerPoint PPT Presentation

SLIDE 1

On the low-frequency relationship between public deficits and inflation

Martin Kliem1 Alexander Kriwoluzky2 Samad Sarferaz3

1Deutsche Bundesbank 2Universität Bonn 3ETH Zürich

Eltville May 2nd 2014

Measuring the low-frequency relationship Results and conclusion

SLIDE 2

The rediscovery of fiscal policy

◮ fiscal policy as a stabilization tool has been rediscovered in

recent times of crisis

0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 2001‐2007 2008‐2012

Figure: Average primary deficits over debt G7-countries.

⇒ increasing deficits are among the outcomes of recent fiscal policy

Measuring the low-frequency relationship Results and conclusion

SLIDE 3

Are there implications of public deficits for inflation?

economic theory: it depends on the policy regime

◮ Sargent and Wallace: under fiscal dominance seignorage can

be used to finance fiscal deficits and cause inflation

◮ Cochrane, Sims, Leeper: active fiscal policy is unresponsive

to deficits, given passive monetary policy, prices adjust to revalue debt (Fiscal Theory of the Price level)

◮ no long lasting effects under monetary dominance or active

monetary policy pared with passive fiscal policy

Measuring the low-frequency relationship Results and conclusion

SLIDE 4

Are there implications of public deficits for inflation?

empirical evidence

◮ no conclusive evidence from fixed-coefficient time series

models

related literature ◮ classic: King and Plosser (JME, 1985) find no significant

relationship between deficits and seignorage in the US using data from 1953-1982

◮ recent: Catão and Terrones (JME, 2005) as well as Lin and

Chu (JIMF , 2013) find no relationship for advanced economies, but a significant positive relationship in the long run for developing countries

◮ Bianchi/Ilut (2012): regime-switching DSGE model, US data,

1955-2009, show that monetary/fiscal policy mix explains rise and fall of inflation in the US

Measuring the low-frequency relationship Results and conclusion

SLIDE 5

Our paper

◮ we employ a long data set: U.S. data from 1875-2011 ◮ we explicitly account for time-variation

◮ theory suggests policy dependence ◮ long data set calls for a flexible time series model

◮ we consider the low frequency domain:

◮ theory stresses the long run ◮ abstract from business cycle movements

Are fiscal deficits and inflation linked at low frequencies?

Measuring the low-frequency relationship Results and conclusion

SLIDE 6

Outline

Measuring the low-frequency relationship Results and conclusion

Measuring the low-frequency relationship Results and conclusion

SLIDE 7

Measuring fiscal stance

◮ debt growth before interest payments (d) ◮ it measures the change of outstanding liabilities due to fiscal

policy

◮ it is defined as primary deficits relative to debt (Sims (2011,

EER))

Zoom in: fiscal stance Measuring the low-frequency relationship Results and conclusion

SLIDE 8

First pass at the data

Following Lucas (1980):

• 1. filter the data
• 2. run a regression of filtered inflation ˜

π on filtered deficits over debt ˜ d: ˜ πt = const + bf˜ dt + errort (1)

Measuring the low-frequency relationship Results and conclusion

SLIDE 9

Scatter plot

−5 5 10 −5 5 10 Primary deficit over debt Inflation

Figure: 1900 - 2009, dashed line ˜ π on ˜ d

Measuring the low-frequency relationship Results and conclusion

SLIDE 10

Subsample scatter plots

−5 5 10 −5 5 10 Primary deficit over debt Inflation

(a) 1952-1983 (red)

−5 5 10 −5 5 10 Primary deficit over debt Inflation

(b) 1984-2009 (blue)

Figure: Dashed line ˜ π on ˜ d

Measuring the low-frequency relationship Results and conclusion

SLIDE 11

Observations from scatter plots

• 1. relationship is time-varying
• 2. positive relationship between 1952–1983
• 3. almost no relationship between 1984–2009

Measuring the low-frequency relationship Results and conclusion

SLIDE 12

Challenges for the simple approach

• 1. potential endogeneities and omitted variables: estimate a

dynamic system consisting of:

◮ inflation (πt) ◮ money growth (∆mt) ◮ output growth (∆yt) ◮ nominal interest rates (Rt) ◮ primary deficits over debt (dt)

• 2. time variation

⇒ Bayesian time-varying parameter VAR model with stochastic volatility using unfiltered data.

Measuring the low-frequency relationship Results and conclusion

SLIDE 13

From a VAR model with unfiltered data to bf

• 1. Estimate the VAR model.
• 2. Compute the spectral density at frequency zero.
• 3. Whiteman (1984): Approximate the slope coefficient bf as the

cross-spectral density Sπd and the spectral density Sd at frequency zero: bf ≈ Sπd(0) Sd(0) (2)

Measuring the low-frequency relationship Results and conclusion

SLIDE 14

Low-frequency relationship

1900 1920 1940 1960 1980 2000 −0.2 0.2 0.4 0.6 0.8 1 1.2 1.4

Figure: Long-run relationship between inflation and primary deficits over

• debt. 16% and 84% probability intervals. Grey bars correspond to bf

from OLS regressions.

Measuring the low-frequency relationship Results and conclusion

SLIDE 15

Empirical results

◮ Positive and mostly significant low-frequency relationship up

to 1980s.

◮ The relationship is time-varying. ◮ Remarkable:

◮ Strongest relationship between 1970 and 1980 – neither in

times of crisis nor of high deficits.

◮ Sharp drop after Paul Volcker became chairman of the Federal

reserve.

Additional estimation results Robustness Measuring the low-frequency relationship Results and conclusion

SLIDE 16

Policy implications

Can the time-variation in the low-frequency relationship be attributed to a change in the monetary/fiscal policy regime?

◮ We identify a monetary policy shock using a recursive

identification scheme.

◮ We compute the contribution of the monetary policy shock to

the low-frequency relationship.

Details on structural decomposition Measuring the low-frequency relationship Results and conclusion

SLIDE 17

Why a monetary policy shock?

Fiscal Theory of the Price level:

◮ Active monetary / passive fiscal policy: monetary policy

shocks have no lasting effects

◮ Passive monetary / active fiscal policy: monetary policy

shocks have persistent effects

Measuring the low-frequency relationship Results and conclusion

SLIDE 18

Structural decomposition

1900 1920 1940 1960 1980 2000 −0.2 0.2 0.4 0.6 0.8 1 1.2 Non−Monetary policy shocks Monetary policy shock unconditional

Figure: Structural decomposition of the low-frequency relationship.

Measuring the low-frequency relationship Results and conclusion

SLIDE 19

Counterfactuals

Our VAR model consists of: yt = ct +

p

• j=1

Aj,tyt−j + Btεt εt ∼ (0,Ht) (3)

◮ coefficient matrices At , Bt (systematic response of the

economy)

◮ variances of the error term Ht

⇒ What would have been the estimate of the low-frequency relationship if the systematic response of the economy had been the same as in year XX in all years?

Measuring the low-frequency relationship Results and conclusion

SLIDE 20

Structural decomposition: counterfactual I

1900 1920 1940 1960 1980 2000 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Non−Monetary policy shocks Monetary policy shock unconditional

Figure: Structural decomposition of the low-frequency relationship. Counterfactual A = A1995,B = B1995.

Measuring the low-frequency relationship Results and conclusion

SLIDE 21

Structural decomposition: counterfactual II

1900 1920 1940 1960 1980 2000 0.2 0.4 0.6 0.8 1 1.2 1.4 Non−Monetary policy shocks Monetary policy shock unconditional

Figure: Structural decomposition of the low-frequency relationship. Counterfactual A = A1976,B = B1976.

Measuring the low-frequency relationship Results and conclusion

SLIDE 22

Relation to other studies

◮ Clarida et.al. (QJE, 2000), Lubik and Schorfheide (AER,

2004), Davig and Leeper (NBER, 2006), Bianchi and Ilut (2012), estimate a change in policy regimes

◮ Bianchi and Ilut (2012), Bianchi and Melosi (2013) show that

the interaction of monetary and fiscal policy explains key characteristic of the data after 1965

◮ Sims (2011) argues that the Fed could not control inflation in

the 1970’s

Measuring the low-frequency relationship Results and conclusion

SLIDE 23

Anecdotal evidence I

Alan Meltzer’s history of the Federal reserve system:

◮ In the 70’s: Federal reserve bank acts as the ’junior partner’

(Alan Meltzer) to the fiscal authority. The fiscal authority was not concerned with inflation.

◮ After Paul Volcker took office: central bank independence and

the fiscal authority is concerned with high inflation rates.

Measuring the low-frequency relationship Results and conclusion

SLIDE 24

Anecdotal evidence II

15 20 25 30 35 40

NumberofMeetingsbetweenU.S.PresidentandFedChairmanattheWhiteHouse

5 10

Figure: Number of meetings between US President and Federal Reserve

• chairman. Source: Martin (2012)

Measuring the low-frequency relationship Results and conclusion

SLIDE 25

Summary of the analysis

◮ Counterfactual: change in the systematic part of the economy

accounts for the time-variation in the low-frequency relationship

◮ Structural analysis: long lasting effects of the monetary policy

shock in 1970s ⇒ Bianchi and Ilut (2012) due to monetary/fiscal policy mix

◮ Theory: findings in line with fiscal theory of the price level

(FTPL)

Measuring the low-frequency relationship Results and conclusion

SLIDE 26

Conclusion

Are fiscal deficits and inflation linked at low frequencies?

◮ Yes, the relationship in the US is positive up to 1980 and it is

time-varying.

◮ The interaction between monetary policy and fiscal policy is

crucial for the behavior of the low-frequency relationship.

Measuring the low-frequency relationship Results and conclusion

SLIDE 27

Robustness

We perform robustness exercises of the results w.r.t:

◮ choice of fiscal stance: debt growth

Details

◮ choice of interpolation method: Chow and Lin (1971) and

Litterman (1983)

Details

◮ choice of interest rate measure:

Details

◮ approximation of the spectrum: DOLS and rolling window

estimation

Details Back Measuring the low-frequency relationship Results and conclusion

SLIDE 28

Related literature: question of interest

◮ no conclusive evidence ◮ classic: King and Plosser (JME, 1985) find no significant

relationship between deficits and seignorage in the US using data from 1953-1982

◮ recent: Catão and Terrones (JME, 2005) as well as Lin and

Chu (JIMF , 2013) find no relationship for advanced economies, but a significant positive relationship in the long run for developing countries

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 29

Fiscal stance

◮ Surplus over debt:

st bt−1 =

• (1 + rt) − bt

bt−1

• (4)

◮ Interpretation: net return on the investment due to interest

and retirement of bonds.

◮ In steady state this is the real interest rate. ◮ A change measures reduction in future obligations. ◮ Deficits are the opposite, i.e. a increase in future obligations.

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 30

Supplementary results: inflation and money

1900 1920 1940 1960 1980 2000 0.5 1 1.5 2 2.5

Figure: Long-run relationship between inflation and money growth. 16% and 84% probability intervals.

The low-frequency relationship between inflation and primary deficits over debt does not cancel the one between money and inflation.

back Measuring the low-frequency relationship Results and conclusion

SLIDE 31

Implied Volatilities

1900 1920 1940 1960 1980 2000 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55

(a) Primary deficits over Debt

1900 1920 1940 1960 1980 2000 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18

(b) Inflation

Figure: Standard deviations of the variables.

Measuring the low-frequency relationship Results and conclusion

SLIDE 32

Implied volatilities

1900 1920 1940 1960 1980 2000 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

(a) ∆ GDP

1900 1920 1940 1960 1980 2000 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

(b) ∆ Money

Figure: Standard deviations of the variables.

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 33

Convergence I

1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0.02 0.04 0.06 0.08 0.1

Figure: Running Mean Plot.

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 34

Convergence II

5000 10000 0.5 1 W 5000 10000 0.05 0.1 Q 5000 10000 0.05 0.1 S1

2000 4000 6000 8000 10000 0.05 0.1

S2 5000 10000 0.2 0.4 S3 5000 10000 0.02 0.04 S4

Figure: Trace Plot.

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 35

Convergence III

1000 2000 3000 4000 5000 6000 −0.1 0.1 0.2 0.3 0.4

(a) Autocorrelation at 10th lag.

1000 2000 3000 4000 5000 6000 2000 4000 6000 8000

(b) Minimum Number of Draws.

Figure: Convergence diagnostics.

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 36

Stochastic volatilities I

1900 1920 1940 1960 1980 2000 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04

(a) Primary deficits over Debt

1900 1920 1940 1960 1980 2000 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05

(b) ∆ Money

Figure: Square roots of stochastic volatility.

Measuring the low-frequency relationship Results and conclusion

SLIDE 37

Stochastic volatilities II

1900 1920 1940 1960 1980 2000 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055

(a) ∆ GDP

1900 1920 1940 1960 1980 2000 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045

(b) Inflation

1900 1920 1940 1960 1980 2000 0.005 0.01 0.015 0.02 0.025

(c) 6m Interest Rate

Figure: Square roots of stochastic volatility.

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 38

Parameter Estimates I

−0.1 0.1

cD

−5 5

cY

−1 1

cP

−0.5 0.5

cR

0.5 1

cM

1 2

ADD (1)

−0.5 0.5

AYD (1)

−0.2 0.2

APD (1)

−0.1 0.1

ARD (1)

−0.2 0.2

AMD (1)

−0.01 0.01

ADY (1)

1 2

AYY (1)

0.05 0.1

APY (1)

0.02 0.04

ARY (1)

−0.1 −0.05

AMY (1)

−0.04 −0.02

ADP (1)

−0.4 −0.2

AYP (1)

1 2

APP (1)

0.02 0.04

ARP (1)

−0.05 0.05

AMP (1)

−0.05 0.05

ADR (1)

−0.5 0.5

AYR (1)

−0.2 0.2

APR (1)

1 2

ARR (1)

−0.4 −0.2

AMR (1)

−0.02 0.02

ADM (1)

0.2 0.4

AYM (1)

−0.1 0.1

APM (1)

0.02 0.04

ARM (1)

1 2

AMM (1)

Figure: Time-varying parameter estimates: constants and AR(1) parameter

Measuring the low-frequency relationship Results and conclusion

SLIDE 39

Parameter Estimates II

−1 −0.5

ADD (2)

−0.5 0.5

AYD (2)

−0.2 0.2

APD (2)

−0.1 0.1

ARD (2)

−0.2 0.2

AMD (2)

−0.04 −0.02

ADY (2)

−0.4 −0.2

AYY (2)

−0.1 −0.05

APY (2)

−0.02 0.02

ARY (2)

−0.05 0.05

AMY (2)

0.05 0.1

ADP (2)

−0.2 0.2

AYP (2)

−0.4 −0.2

APP (2)

−0.02 0.02

ARP (2)

−0.1 0.1

AMP (2)

−0.05 0.05

ADR (2)

−0.5 0.5

AYR (2)

−0.2 0.2

APR (2)

−0.4 −0.2

ARR (2)

0.5

AMR (2)

−0.04 −0.02

ADM (2)

−0.2 −0.1

AYM (2)

−0.1 0.1

APM (2)

−0.05 0.05

ARM (2)

−1 −0.5

AMM (2)

Figure: Time-varying parameter estimates: AR(2) parameter

Measuring the low-frequency relationship Results and conclusion

SLIDE 40

Parameter Estimates III

−1.3682 2.6848

B21

−0.5141 1.4001

B31

−0.4058 0.276

B32

−0.2489 0.5926

B41

−0.3816 0.1276

B42

−0.8724 0.1667

B43

−1.4126 0.7659

B51

−0.6142 0.2612

B52

−0.8475 0.3966

B53

−0.2189 1.8025

B54

Figure: Time-varying parameter estimates B

Back Measuring the low-frequency relationship Results and conclusion

SLIDE 41

Debt growth as fiscal stance

1900 1920 1940 1960 1980 2000 −0.5 0.5 1

Figure: ˆ bf: Median and 68% central posterior bands for the time-varying regression coefficient inflation on debt growth. Robustness check with real debt growth instead of primary deficits over debt.

back Measuring the low-frequency relationship Results and conclusion

SLIDE 42

Comparison interpolation methods

1880 1900 1920 1940 1960 1980 2000 −0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 cubic−spline Chow−Lin Littermann

Figure: Interpolated time series for primary deficits over debt using different interpolation methods.

back Measuring the low-frequency relationship Results and conclusion

SLIDE 43

Time-varying VAR and subsample OLS

1900 1920 1940 1960 1980 2000 −0.2 0.2 0.4 0.6 0.8 1 1.2 1.4

Figure: ˆ bf: Median and 68% central posterior bands for the time-varying regression coefficient inflation on primary deficits over debt. Grey lines correspond to the heteroscedastic-serial consistent OLS regression coefficient of the filtered data.

back Measuring the low-frequency relationship Results and conclusion

SLIDE 44

Rolling window OLS and DOLS

1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

(a) OLS estimate

1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 0.2 0.4 0.6 0.8 1 1.2 1.4

(b) DOLS estimate

Figure: Rolling sample (fixed window) regression coefficients.

back Measuring the low-frequency relationship Results and conclusion

SLIDE 45

Nominal interest rates

1900 1920 1940 1960 1980 2000 0.5 1 1.5

Figure: ˆ bf: Median and 68% central posterior bands for the time-varying regression coefficient inflation on primary deficits over debt. Robustness check with 3m nominal interest rates instead of 6m interest rates.

back Measuring the low-frequency relationship Results and conclusion

SLIDE 46

3 month real interest rate

1900 1920 1940 1960 1980 2000 −0.2 0.2 0.4 0.6 0.8 1 1.2 1.4

Figure: ˆ bf: Median and 68% central posterior bands for the time-varying regression coefficient inflation on primary deficits over debt. Robustness check with 3m real interest rates instead of 6m interest rates.

back Measuring the low-frequency relationship Results and conclusion

SLIDE 47

Structural decomposition I

◮ Spectrum Si Y,t|T (ω) associated with i − th column of the

Cholesky decomposition ˜ Bi

t|T:

Si

Y,t|T (ω) = ˆ

Ct|T

• I − ˆ

At|Te−iω−1 ˜ Bi

t|T(˜

Bi)′

t|T

• I − ˆ

A′

t|Teiω−1 ˆ

C′

t|T ◮ Spectrum is decomposed into spectra of structural shocks:

ˆ bf,t|T = Sπd,t|T(0) Sd,t|T(0) = 5

i=1 Si πd,t|T(0)

5

i=1 Si d,t|T(0)

(5)

Measuring the low-frequency relationship Results and conclusion

SLIDE 48

Structural decomposition II

◮ Contribution of the monetary policy shock to the

low-frequency relationship: ˆ bf,t|T = Sm

d,t|T(0)

Sd,t|T(0) ˆ bm

f,t|T + 4

• i=1

Si

d,t|T(0)

Sd,t|T(0) ˆ bi

f,t|T

(6)

Back Measuring the low-frequency relationship Results and conclusion