Transcript La congiuntura e le prospettive

Comments on
“Potential Output from a Euro-Area
Perspective” by R. Anderton et al.
Alberto Locarno
Banca d’Italia
12th Macroeconomic Policy Research Workshop on
Growth, Rebalancing, and Macroeconomic Adjustment after Large Shocks
Objectives and scope of the paper
• The paper, focused on the EZ, provides a wide-ranging analysis
of the issues associated with the estimation of potential output.
It assembles three short surveys and a model-based analysis.
• The three short surveys deal with:
1. the different techniques used to estimate potential output;
2. the assessment (made by international organizations) of
potential output up to 2060 for the euro area;
3. the valuation of the impact of the 2008-? crisis on actual and
potential output growth, comparing post-crisis developments in
the US and the euro area.
• The model-based analysis aims at assessing the growthenhancing impact of structural reforms.
2
Potential output: data and methods (1)
• Regardless of the preferred approach, the first question to
answer involves which variables to use. The authors pay due
attention to capital but somewhat less to employment or output.
- Is it better to use GDP at market prices, basis prices or factor
costs? Including or excluding the public sector?
- Is it better to use hours, FTE workers or employment (ILO
definition)? Possible inconsistency with the unemployment rate.
• The UC and production function approaches are described in
details, being the most-often used filtering techniques. No
attention is instead paid to other approaches, like SVAR models,
univariate time-varying coefficient AR methods or DSGE models.
- SVAR- and DSGE-based approaches allow to identify the shocks
driving potential output, which is vital if one wants to attach an
economic interpretation to changes in potential output estimates.
- TV methods tend on average to generate more stable (i.e. less
subject to revision) estimates.
3
Potential output: data and methods (2)
• The different sources of uncertainty affecting potential output
estimates are listed with care. Most attention is devoted to endof-sample and model uncertainty, the latter proxied by the
variability of the estimates of International Organizations.
- UC methods (as well as SVAR and DSGE, but not PF approaches)
allow to assess estimation uncertainty.
• For unobserved series, the issue of uncertainty is tricky and
cannot be answered objectively; it depends e.g. on how smooth
we want potential output to be.
4
Potential output in the long run
•The three main results highlighted:
1. long-term developments in potential output are
mostly driven by demographics, pointing to a
negative contribution of the labour input;
2. TFP projections are more favourable, but they are
driven by ad-hoc assumptions (about cross-country
convergence in TFP);
3. potential output growth persistently lower (by 0.60.8pp) in the EZ than in the US.
5
Assessing the impact of the crisis (1)
•
Potential growth was in 2012 ≃ 1.5pp lower than in 2007, due
to: (i) labour contribution turned negative; (ii) weaker
support from capital accumulation; (iii) a modest decline in
TFP contribution.
•
Which are the factors driving these changes? NAIRU and
participation rates (old age cohorts excluded) dragged the
trend increase in employment. Capital accumulation suffered
from credit supply shocks, uncertainty and possibly changes
in the scrapping ratio.
•
TFP hardly affected: negative impact of direct effects (drop in
innovations) offset by positive impact of indirect ones
(resource reallocation). Spurious effects (due to the residual
nature of the Solow residual)?
•
Countries displaying huge imbalances (in foreign/public debt
or sectoral composition of GDP) or real rigidities were hit the
most.
6
Assessing the impact of the crisis (2)
So far so good. Problems arise when the authors try to pin down
the “first causes” of the changes in potential output.
1.
How can you infer from Chart 18 that macroeconomic
imbalances are responsible for the decline in potential output
growth? And how can you say that potential growth has
declined more than implied by the loss in the level of p.o.?
2.
Chart 20(A) is viewed as showing that public indebtedness is
not responsible for the slowdown in p.o., essentially because
the relationship between debt and growth is the same before
and after the crisis. This is objectionable. The results can be
simply due to the use of debt (rather than indebtedness) and
the presence in the sample of ES and IE (hidden liabilities,
e.g. the guarantees provided to the banking sector).
3.
The black line in Chart 21 is unreliable (and indeed is quite
flat), due to reverse causality, and cannot be used to support
the claim that pre-crisis current account imbalances led to
lower potential growth.
7
Model-based analysis (1)
• According to the authors, the New MCM predicts that the crisis
heavily affected potential growth in FR, IT, ES and NL, but not in
DE. Moreover, the recovery would be quicker in DE, FR and IT
than in ES and the smaller countries. Quite reasonable, but how
can it be, if N*, the income shares and technical progress are
exogenous in the model?
• Which set of projections have been used to extend the sample
period beyond 2013? What justifies the larger TFP contributions
for most countries but ES?
• The claim at page 60 (i.e. “in DE, FR and IT the estimates suggest
that the drop in potential output growth was limited (and a quick
return to close to previous trend growth rates is more or less
accomplished by 2012”) is certainly far from true for Italy.
8
Model-based analysis (2)
• The authors presents results (obtained by simulating EAGLE)
regarding the implementation – in isolation and simultaneously –
of structural reforms (namely a 10pp decline in the wage
markups) in euro-area countries.
• Assuming that structural reforms boost GDP and potential
output, at least in the medium/long-run, it is not surprising that a
coordinated effort would have a larger impact than an isolated
one.
• But is this a meaningful experiment? Some countries badly need
reforms, other much less so. Country heterogeneity would
persist. Wouldn’t it be preferable to design an experiment with
convergence to the best practices, which, in addition, would be
more informative on the actual measures to adopt?
9