disclaimer: the views expressed are those of the author and not necessarily those of dnb did fiscal...
TRANSCRIPT
Disclaimer: The views expressed are those of the author and not necessarily those of DNB
Did fiscal policymakers know what they were doing?
Re-assessing fiscal policy with real time data
Kerstin Bernoth, De Nederlandsche BankAndrew Hughes Hallett, De Nederlandsche Bank
John Lewis, De Nederlandsche Bank
CIRANO Workshop on Data Revision in Forecasting and Macroeconomic Policy
Montreal, 10 October 2008
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 2John Lewis
De Nederlandsche Bank
Introduction
“But as is well known, the actual variables required for implementation of such a rule -potential output, nominal output, and real output-are not known with any accuracy until much later. That is, the rule does not describe a policy that the Federal Reserve could have actually followed” – Orphanides 2001
Growing literature on monetary policy rules which use real time data.
Much less attention to fiscal policy- although Orphanides’ point is just as valid– Any attempt to say what fiscal policymakers were “trying to do”
needs to take account of the information they had at the time they made the decision
– Monetary literature shows us that real time and ex post data can give quite different characterisations of policymakers’ behaviour
Literature on fiscal policy finds that discretionary policy is often a- (or even) procyclical. Is what policymakers intended?– Are fiscal policymakers misintentioned or malinformed?
If misintentioned that acyclicality should also show up in RT data
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 3John Lewis
De Nederlandsche Bank
Overview
Paper makes two key contributions1. Compares fiscal policy reaction based on real time data
with their ex post counterparts2. Develops new technique using real time data to
separate out automatic from discretionary fiscal policy
Overview of Presentation– How data problems affect fiscal policymaking in real time– A simple model of fiscal policymaking in real time
The econometric problem of ex post reaction functions New methodology to split up automatic and discretionary
policy– Dataset– Estimation results– Conclusions
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 4John Lewis
De Nederlandsche Bank
The Cyclicality of Fiscal Policy: A simple example
Cyclically adjusted budget balance usually used to capture discretionary policy– Strip out cyclical effect of automatic stabilisers to leave
discretionary component
Mismeasuring output gap not only the control variable but also the instrument (cyclically adjusted budget balance)
Simple example– Real time output gap +1%, ex post it turns out to be 2%– Automatic stabilisers are 0.5– Budget balance = 1% (both real time and ex post)
– Real time: CAB=+0.5%, output gap=+1% countercyclical fiscal policy
– Ex Post: CAB= 1%, output gap= 2% pro-cyclical fiscal policy
Errors reinforce each other
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 5John Lewis
De Nederlandsche Bank
A first look at the data
Output gap and cyclically adjusted primary balance are prone to quite large revisions
Mean revisions to output gap and capb are approximately mean zero(Hughes Hallet et al, 2007)
Cyclically Adjusted Primary Balance (Germany), % potential GDP
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Real Time
Ex post
Output Gap (Germany), % potential GDP
-4
-3
-2
-1
0
1
2
3
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Real Time
Ex Post
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 6John Lewis
De Nederlandsche Bank
A simple model of Fiscal Policymaking
Fiscal policy is set at time t+s (s=0 or -1)– First subscript is conventional one: year in question– Second subscript is the vintage of the data (f is ex post data)
Fiscal policymaker cannot directly set the cyclically adjusted balance– Instead they enact a set of spending and revenue plans, which
under certain assumptions about the economy, yield a given (cyc adj) budget balance
– Spending and revenue split up into three components
Plan:
Outturn
Expenditures fixed in cash terms(For convenience expressed as ratios to estimated pot. Output at time s)
Do not change as estimates of potential output change
Discretionary expenditures which respond to the real time output gap
Do not change as estimates of output gap revised
Automatic expenditures which respond to the true (“ex post”) level of output
Do change in response to the true output level
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 7John Lewis
De Nederlandsche Bank
A simple model of Fiscal Policymaking (2)
Output and output gap measured incorrectly in real time
With some simple re-arrangement we can re-write the “final” primary balance as a function of ex post variables, measurement errors, and the policymakers desired parameters:
ex post primary balance
ex postoutput gap
“error” in real time output gap
=zt|t+s
“error” in real time potential output
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 8John Lewis
De Nederlandsche Bank
Econometric Problem with Ex Post Reaction Functions
Strip out cyclical component of previous equation i.e. subtract
Standard ex post reaction function is estimated as:
– Mis-specified as a measure of policymakers intentions- omitted variables
Can test this formally: is either v or z significant? If policymaker somehow has other information from which
they can work out the “true” output gap, then coefficient on z should be zero
y~)(
“error” in real time output gap
“error” in real time potential output
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 9John Lewis
De Nederlandsche Bank
Discretionary vs Automatic
Estimate the equation:
Recall:
(-) denote purely discretionary policycoefficient on output gap error captures discretionary policy(β-φ) denote purely automatic policycoefficient on ex post output gap captures sum of discretionary and
automatic fiscal policy
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 10John Lewis
De Nederlandsche Bank
Taking the model to the data: The dataset
Variables:
– Cyclically Adjusted Balances
– Actual Deficits
– Primary Balances
– Output Gap
– GDP time series
– Government debt
Source: Successive issues of OECD economic outlook (Dec 94 onwards)
14 Countries: EU15 – Luxembourg
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 11John Lewis
De Nederlandsche Bank
Estimation (1) We estimate the following fiscal reaction functions:
where xt|t-j = -- Debt/GDP– eyear: dummy for parliamentarian elections in year t.– “Maastricht variable” (Momigliano et al. (2004)) -
captures effect of fiscal consolidation in the run-up to EMU: For countries prior 1998 with deficit of more than 3%:
Lagged dep. variable included remove (possible) serial correlation due to persistence in budgetary variables.
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 12John Lewis
De Nederlandsche Bank
Estimation (2)
To increase number of observations: Regression (a)-(d) regressed with a panel estimation.– Necessary condition: Countries are ‘poolable’: countries follow
similar fiscal reaction functions.– Poolability test: Testing whether estimated slope coefficient not
significantly different between countries. H0 cannot be rejected.
Choice of vintage: Vintage t|t rather than t|t-1 data fits better for real time model: Adjustment to fiscal policy during the year (see e.g. Kalckreuth and Wolff, 2007).
Application of Blundell and Bond estimation approach: provides consistent estimates of dynamic panel models, when autoregressive parameter is large.
Output gap instrumented with its own lags to account for possible endogeneity.
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 13John Lewis
De Nederlandsche Bank
Results: Baseline Case
(A) Ex post CAPBgap is not significant: Discretionary policy is acyclical
(B) Ex-post data CAPB plus z and vz and gap jointly significant: Discretionary policy is counter-cyclical
(C) Real time CAPB (a la Cimadomo)gap is significant: Discretionary policy is counter-cyclical
(D) Real time PB (our model)Discretionary policy is counter-cyclicalAutomatic policy is counter cyclical: 0.8 - 0.56 = 0.24.
Smaller compared to previous findings.
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 14John Lewis
De Nederlandsche Bank
Results: Extra Control Variables
Coefficients measuring discrectionary and automatic fiscal policy robust to specification of control variables.
Significant Maastricht-effect: governments made addit. effort to tighten fiscal policy in run-up to EMU, if deficit > 3%.
In an election years, fiscal policy is around 0.35 percentage points looser.
Did Fiscal Policymakers Know What They Were Doing?Montreal, 10 October 2008 15John Lewis
De Nederlandsche Bank
Conclusions
Real time measurement error of output gap complicates fiscal policy
Ex post, fiscal policy looks acyclical, but in real time it looks countercyclical– Policymakers are misinformed rather than malintentioned– Remedies based on malintentioned diagnosis are unlikely
to work Problem doesn’t show up in real time, so watchdog may
not spot it
New methodology suggests automatic stabilisers weaker than commonly thought (0.24 rather than 0.35-0.50)– Traditional methodology may wrongly attribute some
counter-cyclical discretionary policy to automatic side