Transcript Slide 1

Comments on Antonio Fatás
Luis Servén
The World Bank
Workshop on fiscal policy
IMF, June 2009
Automatic stabilizers
Three components of fiscal policy:
1. Automatic
2. Systematic discretionary
3. Purely discretionary (i.e., unsystematic)
(1) and (2) hard to separate: formal explicit rules vs implicit
ones -- routine responses to economic conditions.
Most research has focused on identifying the effects of (3).
But it likely accounts for a relatively small fraction of the
overall variation in fiscal variables – as implied by high
R2 from projecting them on cyclical and other factors.
Nice to see (1) get some attention too.
Automatic stabilizers: measurement
• Typical ingredients:
– Direct taxes (+ SS contributions)
– Indirect taxes
– Unemployment benefits
• Measurement
– Using tax codes, unemployment rules etc (hard)
– Regression of fiscal outcomes on cycle (easier)
• Reverse causality – with no obvious instruments
• Mixes up all systematic policies – not only automatic
Automatic stabilizers: measurement
Different in poor countries: smaller government,
conventional stabilizers weak on the revenue side, and
virtually absent on the expenditure side.
Country
group
(income
tercile)
Total
Direct taxes
Exp/GDP (+SS) / Total Transfers
(%)
revenue (%) / GDP (%)
Low income
19.5
26.0
6.5
Middle
income
27.8
35.6
11.1
High
income
32.9
53.6
18.4
Source: WDI
Automatic stabilizers: measurement
Latin America vs industrial countries
Source: Suescún 2007
Automatic stabilizers: measurement
Regressions of fiscal aggregates on cyclical indicators.
• Lump together all systematic policy (automatic +
dicretionary)
• Hard to interpret due to reverse causality (Rigobón 2004)
• Typical finding is (more) pro-cyclical policy in poor
countries, acyclical / counter-cyclical in rich ones –
survives a variety of robustness checks (Ilzetzki and
Végh 2008)
• Why?
– Financial frictions: procyclical access to borrowing (Gavin-Perotti
1997; Kaminsky-Reinhart-Végh 2004)
– Institutional failures (Tornell-Lane 1999, Talvi-Végh 2005,
Alesina-Tabellini 2005, Ilzetzki 2007…)
Some empirical support for both explanations – although based on
crude measures of institutional quality and access to finance.
Automatic stabilizers: measurement
Calderón & Schmidt-Hebbel 2008: most of the action seems
to come from institutional differences.
Dependent variable: public consumption growth
IV estimates, 121 countries, annual data (WDI dataset)
GDP growth
0.417***
0.886***
0.469**
0.829**
GDP growth*LDC
0.149**
-0.073
0.085
-0.102
GDP growth * ICRG
GDP growth*credit depth
-0.039**
-0.043**
-0.001
A closer look at LDC fiscal institutions could be illuminating.
0.001
Automatic stabilizers: measurement
But also: shocks in developing countries may be different.
Aguiar-Gopinath 2007:
• Industrial countries: transitory shocks and stable trend
• Developing countries: trend shocks (“the trend is the
cycle”): a current shock signals an even bigger future
change in the same direction
If optimal policy is forward-looking, its response to current
output changes (the focus of conventional regressions)
may well be different in industrial and developing
countries.
Automatic stabilizers: effectiveness
Effectiveness usually stated in terms of output stability –
but consumption stability at least as important.
Methodologically, assessing effectiveness of automatic
stabilizers is different from evaluating unsystematic
policies: it involves evaluating rules – Lucas critique.
In principle one would need structural models with policyinvariant behavioral relations (like McCallum 1999 on
systematic monetary policy).
Not many such exercises (Andrés, Doménech & Fatás 2008 is one)
Instead, reduced-form regressions of aggregate volatility on
measures of cyclical response of budget – or even
government size.
Automatic stabilizers: effectiveness
The relation between volatility
and government size is different
outside rich countries
Source: Perry, Servén and Suescún 2007
Automatic stabilizers: effectiveness
• Why does government size work differently in
developing countries?
-- Nonlinearities: poor countries are below a “minimum
government size” for the stabilizing effect to occur
(the reverse of Buti et al 2003 for rich countries)
-- Composition effect: automatic stabilization is
overwhelmed by procyclical discretionary policy
-- Shocks are different (as before)
Worth a closer look.
Automatic stabilizers: effectiveness
Automatic stabilizers vs discretionary policy
Some hints that the latter may be partly replacing the
former since the 1990s (e.g., Debrun et al 2008,
Auerbach 2009) – is that a good idea?
Not so clear that “automatic stabilizers are more
effective” – what is the metric and evidence?
They do offer the advantages of speed, predictability
and reversibility – especially valuable when fiscal
institutions are weak
Aside from efficiency issues, the political economy of
automatic stabilizers can be hard too: income
taxation, tax enforcement etc – still big hurdles for
many developing countries.
End
Source: Auerbach 2009