Fiscal and Monetary Policy in Australia: an SVAR Model

Mardi Dungey and Renée Fry

University of Tasmania, CFAP University of Cambridge, CAMA Australian NationalUniversity

September 2010

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 1 / 30

Challenges of modelling �scal policy in a VAR

Identi�cation problem

because g and tax are highly correlated its is di¢ cult for VARs todistinguish these empirically

Dynamics of govt expenditure and tax shocks

Extracting automatic versus pure shocks

Blanchard and Perotti (2002) and Perotti (2002)Mountford and Uhlig (2009)

Kirchner et al (2010)

The role of debt

Chung and Leeper (2007), Favero and Giavazzi (2007)

Mixed nature of data: non-stationary and stationary

temporary and permanent shocks

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 2 / 30

Solutions proposed for identifying �scal policy

Ordering using institutional timing

Blanchard and Perotti (2002), Perotti (2002)

Sign restrictions

Canova and de Nicoló (2002), Uhlig (2005), Peersman (2005)

TVP-VAR

Kirchner et al (2010)

Our approach - combine (Dungey and Fry 2009):

traditional restrictions - Dungey and Pagan (2009)sign restrictionslong run restrictions - Pagan and Pesaran (2009)

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 3 / 30

Modelling Australia

Timely issue in the Australian economy

Australian �scal stimulus of $42 billion (about 4% of annual GDP)did it prevent a recession?has it been too stimulatory?

Existing SVAR model to form the framework

Dungey and Pagan (2000, 2009)detailed 11 variable model of the Australian economy+ govt expenditure, govt revenue, debt/GDP

Fiscal policy models

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 4 / 30

A quick primer on the Australian economy

An almost unprecedented period of expansion from 1992-2010.

contributing conditions:

productivity boost in 1990sincreased wage �exibilitylow in�ationimproved terms of trade�nancial market deregulation and expansion

some consequences

housing price boom�scal consolidation

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 5 / 30

Monetary Policy

�oating exchange rate since December 1983

in�ation targeter since 1992ish

0

5

10

15

20

25

80 83 86 89 92 95 98 01 04 07 10

cash rate

10 year bond

Australian Interest Rates%

90

100

110

120

130

140

150

80 83 86 89 92 95 98 01 04 07 10

Trade­Weighted IndexMar95=100

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 6 / 30

Fiscal Policy

latest estimates have net debt/GDP about 8%

source: DiMarco, Piri and Yeung (2009) Australian Treasury

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 7 / 30

Outline

Identi�cation of the benchmark model

traditional restrictionssign restrictionstemporary and permanent shocks

Data

Empirical results

Impulse response functions

Some serious problems

research directions

Conclusions

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 8 / 30

Writing the SVAR

Take account of potential mixed I (1) and I (0) variables with cointegratingrelationships

B(L)Yt = εt , (1)

VECM form:Ψ(L)∆Yt = �ΠYt�1 + et , (2)

Say k variables, n are I (1) with r < n cointegrating vectors thenΠ = α0β is of reduced rankCommon trends representation:

∆Yt = F (L)et = F (L) (B0)�1 εt , (3)

and F (1) = F = β? [α0?Ψ (L) β?] α

�1? ,

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 9 / 30

Permanent and Temporary Shocks

If the �rst (n� r) shocks are permanent then

∆Yt = F (L) (B0)�1�

ε1jtε2jt

�,

for the shocks in ε2jt , to be transitory requires

FB�10

�0(n�r )�rIr+k

�= 0,

equivalently α1 the coe¢ cient on the permanent shocks must equal zero.

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 10 / 30

Sign restrictions

Residualsvt = B�10 εt (4)

De�ne bS as having the estimated standard deviations of the structuralresiduals on the diagonalbvt = bB�10 bSbS�1bεt (5)

= bTbηt (6)

Impact matrix bTEstimated shocks bηtDe�ne a rotation matrix: Q such that Q

0Q = QQ

0= I

bet = TQ 0Qηt (7)

= T �η�t . (8)

rotations are orthogonal but produce alternative impulse responsesDungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 11 / 30

Choose between rotations

Use criteria on sign restrictions to choose rotations which areacceptable

variable/shock absorptiont GDPttax: τt �

govt expenditure: gt +

not enough....multiple shocks problem

can have both shocks in a rotation look like a, say, tax shock. Then weneed to sort them out.If one set of impulses contains both g and tax shock

second set has say only g shockthen assume �rst set refers to a tax shock

Here never the case that both sets contain both shocks

Dungey and Fry 2009 use relative sizes to sort this out.

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 12 / 30

Impulse responses

In the sign restrictions

Choose the median but retain orthogonality (Fry and Pagan, 2007)

standardize impulses

group into φd

minimize φd 0φd

the corresponding Qd matrix is used to calculate impulses

This ensures that impulses from the same model are selected

In a purely orthogonal system also ensures the system remainsorthogonal

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 13 / 30

Data

Standard for most variables, interesting ones are G,T and debt/GDP,data compiled for us by Australian Treasury.

associated problems:

data frequency: move from annual to quarterlychanging basis of accountsadjustments for large expenditures associated with defense or largeprojectsseasonal adjustment and lack of compatibility between componentseries

Government expenditure: Consumption + Expenditure

Government taxation revenue: Tax - transfers

Debt/GDP ratio: annual ABS data interpolated using the Chow-Lin(1971) using the IFS series for �rst part of the sample and OECDdata for rest.

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 14 / 30

Selection of variables

data cointegratingproperties vectors

Exogenous 1 2 3US GDP I (1) *Terms of trade I (0)Real US interest rate I (0)US Q ratio I (0)Exports I (1) *

EndogenousGovernment Taxation Revenue I (1) * *Government Expenditure I (1) * *Absorption (GNE) I (0) *Debt to GDP ratio I (0)GDP I (1) * *In�ation I (0)Cash rate I (0)Real Exchange Rate I (1) *

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 15 / 30

Restrictions on Australian variables

Dependent variablestax g abs q debt gdp inf short twi

tax +� � �� �� � ��g � +� �� �� � ��q � � �� �� �� ��abs �� +� �� � �� �� �� ��debt � � �� �� ��gdp � � � � �� �� ��inf � � � � � �� �� ��short � � � � � �� ��twi � � � � � � � ��exogenous tot y �, tot y �, xpts tot y �, i�

rus, q� tot px/m+,� sign; � contemporaneous; � lags (p = 3 in levels)

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 16 / 30

Properties of the dataNonstationarity and Cointegration

Testing suggests the following properties of the data

Non-stationary Stationary

y �, xpts tot, (r � � π�), q�,g , tax , gdp, abs, twi q, debt,π, r

Cointegration tests

fy �, xpts, g , tax , yg (9)

fy , abs, twig (10)

Impose for �scal sustainability

fg , tg = [1� a] (11)

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 17 / 30

Transitory and Permanent Shocks

Amongst the 7 nonstationary variables (y �, xpt, g , tax , gdp, abs, twi)

3 cointegrating relationshipsimplies 3 temporary shocks and 4 permanent ones

Choose the permanent shocks

the external sector is permanent: y�, xpt2 domestic shocks need to be permanent

abs permanent - domestic preferences shockgdp permanent - domestic technology shockif either of g or tax is temporary, then the other needs be temporary orthe �scal sustainability condition will be violated ) g and tax aretemporarynot twi

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 18 / 30

Summary of data properties

7 non stationary variables4 permanent shocks y �, xpt, abs, gdp3 temporary shocks g , t, twi

7 stationary variables4 temporary shocks tot, q�, (r � � π�)

q, debt,π, short

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 19 / 30

Government Expenditure Shock

00.10.20.30.40.50.6

1 6 11 16 21 26 31 36 41 46 51 56 61

(b) g shock  on g

00.20.40.60.8

11.2

1 6 11 16 21 26 31 36 41 46 51 56 61

        (a) g shock  on t

­50

­30

­10

10

30

50

1 6 11 16 21 26 31 36 41 46 51 56 61

  (e) g shock  on debt

This shows a 1se shock to government expenditure

temporary shockresults in increased revenue collectioninitial increase in debt/GDPsubsequent pay down of debt

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 20 / 30

Government Expenditure Shock

­0.02

0.02

0.06

0.1

0.14

0.18

1 6 11 16 21 26 31 36 41 46 51 56 61

(f) g shock  on gdp

­2.8­2.0­1.2­0.40.41.22.0

1 6 11 16 21 26 31 36 41 46 51 56 61

     (g) g shock  on inf

­4­202468

1 6 11 16 21 26 31 36 41 46 51 56 61

   (h) g shock  on cash

g shock results in increased GDP and GNE (this is the signrestriction)

cash rate rises and in�ation fallsdomestic currency appreciatesconsistent with g increase being investment rather than consumptionexpenditure

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 21 / 30

Taxation Shock

­0.3

­0.2

­0.1

0

0.1

0.2

1 6 11 16 21 26 31 36 41 46 51 56 61

(m)  t shock  on gne

­0.12­0.08­0.040.000.040.080.120.16

1 6 11 16 21 26 31 36 41 46 51 56 61

     (o) t shock  on gdp

­500

­300

­100

100

300

1 6 11 16 21 26 31 36 41 46 51 56 61

(n) t shock  on debt

This shows a 1se shock to net taxation revenue

temporary shock so there are no long run e¤ectsfall in absorption (this is the sign restriction)gdp falls then rises (also seen in Dungey and Fry for NZ)debt to GDP is reduced

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 22 / 30

Taxation Shock

­2.0

­1.3

­0.6

0.1

0.8

1.5

1 6 11 16 21 26 31 36 41 46 51 56 61

(k) t shock  on g

­0.7

0.1

0.9

1.7

2.5

1 6 11 16 21 26 31 36 41 46 51 56 61

    (l) t shock  on ausq

­3.5

­2.5

­1.5

­0.5

0.5

1.5

1 6 11 16 21 26 31 36 41 46 51 56 61

(r) t shock  on rtwi

This shows a 1se shock to net taxation revenue

decreased debt, increased GDPreduced government expenditure through automatic stabilisersimproved investment con�dencehigher interest rates and higher in�ation, not higher real interest ratesso depreciation of the currency

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 23 / 30

Monetary Policy shock

­0.0055

­0.0035

­0.0015

0.0005

1 6 11 16 21 26 31 36 41 46 51 56 61

(v) cash shock  on gne

­0.002

­0.001

0

1 6 11 16 21 26 31 36 41 46 51 56 61

(x) cash shock  on gdp

­0.012­0.008­0.004

00.0040.0080.012

1 6 11 16 21 26 31 36 41 46 51 56 61

(y) cash shock  on inf

100bp cash shock

decreases GNE and GDPinitial in�ation response a bit uncertainprice puzzle problem after 18 months.....

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 24 / 30

Interacting responses:

­0.012­0.01

­0.008­0.006­0.004­0.002

0

1 6 11 16 21 26 31 36 41 46 51 56 61

(s) cash shock  on t

­3

­2

­1

0

1 6 11 16 21 26 31 36 41 46 51 56 61

(w) cash shock on debt

­0.024­0.02

­0.016­0.012­0.008­0.004

0

1 6 11 16 21 26 31 36 41 46 51 56 61

(t) cash shock  on g

cash shock slows the economy so reduces taxation revenuebut reduction in tax < reduction in GDP, so debt/GDP fallsfall in government expenditure also, this is not clear?

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 25 / 30

Interacting responses:

00.10.20.30.40.50.6

1 6 11 16 21 26 31 36 41 46 51 56 61

(b) g shock  on g

00.20.40.60.8

11.2

1 6 11 16 21 26 31 36 41 46 51 56 61

        (a) g shock  on t

­4­202468

1 6 11 16 21 26 31 36 41 46 51 56 61

   (h) g shock  on cash

­2.0

­1.3

­0.6

0.1

0.8

1.5

1 6 11 16 21 26 31 36 41 46 51 56 61

(k) t shock  on g

­0.5

0.0

0.5

1.0

1.5

2.0

1 6 11 16 21 26 31 36 41 46 51 56 61

         (j) t shock  on t

­2

0

2

4

6

8

1 6 11 16 21 26 31 36 41 46 51 56 61

(q) t shock  on cash

­0.024­0.02

­0.016­0.012­0.008­0.004

0

1 6 11 16 21 26 31 36 41 46 51 56 61

(t) cash shock  on g

­0.012­0.01

­0.008­0.006­0.004­0.002

0

1 6 11 16 21 26 31 36 41 46 51 56 61

(s) cash shock  on t

0

0.3

0.6

0.9

1.2

1 6 11 16 21 26 31 36 41 46 51 56 61

(z) cash shock  on cash

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 26 / 30

Problems

Already have monetary policy shocks not behaving quite as wethought

and di¤er from source model (Dungey and Pagan 2009)

adding g , tax , debt?additional cointegrating relationshipmaking absorption a permanent rather than transitory shock

Problems in sign restrictions literature (Fry and Pagan 2010)

we dont know εt for the shocks identi�ed with sign restrictionsthus no con�dence intervals can be bootstrappedno variance decompositions and historical decompositionsthus far no breakdowns of contributions by various policy shocks

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 27 / 30

Next steps

Attempt to identify some bounds on εt for the sign restrictionidenti�ed shocks

Sort out why monetary policy shocks are behaving so di¤erently tobaseline model

THEN

project model into the crisis periodexamine the deviation of the projected model from observed dataduring the crisisallow us to examine the contribution of the g and cash shocks to thebetter than anticipated y outcomes in the crisis period

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 28 / 30

Contributions of the Paper

Technical Contributions

Combination of identi�cation methods to include �scal and monetarypolicy in a SVAR for AustraliaMixed I(0) and I(1) dataInclusion of long term relationships: Pagan and Pesaran (2007)Identi�cation of permanent and temporary shocks

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 29 / 30

Contributions of the Paper

Analytical Contributions

So far to �nd direction of monetary, government expenditure andgovernment revenue shocks on output and in�ationAim to be able to �nd contribution of these shocks to observedoutcomesAim to project into the crisis period to �nd the contributions of policyshocks to the better than anticipated outcomes.

Dungey and Fry (University of Tasmania, CFAP University of Cambridge, CAMA Australian National University )Fiscal VAR 09/10 30 / 30