Transcript Study on the feasibility of a tool to measure the macroeconomic

Econometric models to study the
macroeconomic
effects of structural reforms
Christian Dreger
German Institute for Economic Research (DIW Berlin)
Rationale of structural reforms
• Overly restrictive regulation can worsen the
economic performance
– Information asymmetries, rent seeking of
special interest groups, management and
incentive problems
• Output and employment are stimulated in
competitive and deregulated environment
• Flexible markets contribute to a smoother
transmission of shocks, shorter periods of
unemployment
Areas of structural reforms
• Reforms affect the general framework on
how the economy operates
• Improve functioning of product, labour
and financal markets
• Financial markets highly deregulated,
while product and labour markets are not
• Reforms act on the supply side of the
economy, try to reinforce market-based
adjustment
Product market reforms
• Increase in competition can foster output,
employment and productivity growth
• Static efficiency gains due to lower mark
ups, internal restructuring of firms etc, lead
to one time productivity change
• Higher incentives to research and innovate
provide dynamic efficiency gains
• Dynamic effects can boost productivity over
longer periods of time
Labour market reforms
• Perspectives of workers improve, notably
by creating new jobs
• Labour reallocates faster in response to
shocks, labour demand more in line with
economic conditions
• Higher incentives of households to work
encourage labour supply
• Unemployment traps for low productivity
workers become less important
Financial market reforms
• Abolition of credit and interest rate controls
• Compared to product and labour markets,
financial markets already deregulated, see
Fraser index
• Private consumption does not depend only
on permanent, but also on current income
• Households faced by liquidity constraints, in
particular in economic downturns
Ambiguous effects of reforms
• Reforms might imply long term gains, but
short term losses, as adjustment to the new
conditions might take a while
• Costly reallocation of resources, decline of
inefficient firms, unemployment can increase
over a transition period
• Reforms affect distribution of income
• Losses concentrated, gains widely dispersed
with only limited benefit for each individual
Appropriate econometric
framework
• Models have to analyse the long run, but also
medium and short run effects
• Models should be suited to investigate single
and comprehensive reform strategies
• Interactions between different areas of
reforms. Sequence of reforms important for
political acceptance
• Past reforms provide insights into the effects of
future reforms, only if agents behaviour is
unchanged
Institutional databases
• OECD, Fraser, LABREF, etc, describe
institutional conditions on product, labour and
financial markets
• Often missing values, variables interpolated to
match frequency of macroeconomic data,
merging required
• Definition of indicators not unique. Indicators
often rank-scaled: only the direction, but not
size of impact can be examined
• Useful for international comparisons,
robustness analysis, inclusion in satellite
models
Typology of econometric methods
• Variety of models available to examine
impact of structural reforms
– Microsimulation models if heterogeneities have
to be taken into account
– Macroeconometric techniques rely on
representative agent framework: single
equations (panels), partial and general
equilibrium models, macroeconometric models
and DSGEs
• Models have their individual pros and cons
– Data requirements
– Use for policy consulting (ex ante vs ex post
analysis)
– Reform packages and adjustment processes
Basic building principles
• Models based on empirical fit
– Single equations (reduced form regressions) and
macroeconometric models are data driven
– Equations estimated, in sample measures of fit
– Might not be fully consistent with economic theory
– Risk of structural breaks in out-of sample period
• Models based on theoretical consistency
– CGEs and DSGEs microfounded, i.e. based on optimizing
households and firms
– Consistent with economic theory
– More liberal on data issues, parameters often calibrated
Ex post vs ex ante analysis
• Past reforms provide insights into effects of future
reforms, only if behaviour of agents is unchanged
• Data driven methods subject to Lucas critique,
structural breaks invalidate policy conclusions, only ex
post analysis
– Single equations, macroeconometric models
• Models based on optimizing agents are more robust to
changing parameters, ex ante analysis
– CGEs and DSGEs
• Check relevance of Lucas critique in advance. Impact
of regulation indicators on crucial elasticities must be 0
Single equations
•
•
•
•
•
•
•
•
Easy and fast to implement, standard software
packages (Eviews)
Direct and indirect effects of reforms, only the latter
are interpretable
Tests for nonlinear effects
Useful as satellite models from the perspective of
more elaborated models
Can provide linkages to regulation indicators
Ad hoc specification of the equation
Due to policy dependent parameters, risk of structural
breaks
Ex ante analysis not recommended
Computable general
equilibrium models (CGEs)
• Rely on optimizing agents, micro-macro consistency
(social accounting matrices)
• Data limitations not important. Distributional effects can
be included (sectors, segments of labour markets etc)
• Lack of empirical underpinning, results driven by model
assumptions
– Calibration to a base year, where long run equilibrium is
achieved by assumption
– Parameters derived conditional on that assumption
– No measures of uncertainty
• Only static analysis, adjustment process, sequence of
reforms cannot be discussed
• Even not immune to structural change
Partial equilibrium models
• Advantages and disadvantages similar to CGE
approach
• Specific sectors or markets considered in very high
detail, used in network liberalization
• Spillovers to other parts of the economy ignored
• Policy measures investigated only in directly
affected markets
• Reliable results, only if linkages to other markets
can be neglected. Otherwise, policy conclusions
would be biased
• Easier to maintain than CGEs. Quick adaption to
study new policy relevant questions
Macroeconometric models
• Based on national accounts data
• Long and short run separated, equations specified in
error correction form
• Wage and price rigidities, long run determined by supply,
short run by demand side
• Satellite models bridge the gap between frequency of
regulation indicators and macroeconomic data
• Designed for simulation and forecasting, models might
be subject to Lucas critique
• Long run ex ante analysis questionable, but models are
suited to study short run impact of reforms
Dynamic stochastic general
equilibrium models (DSGEs)
• Microfoundation of macroeconomeric models
• Consistent with economic theory, as for CGEs, Lucas
critique less relevant
• Ex ante analysis of reforms
• Impact of reforms studied in terms of the change in
welfare
• Dynamic specification allows to examine sequence of
reforms and adjustment processes
• Deficits in short run analysis, macroeconometric models
might still be useful
• Even not immune to structural change
Implementation of structural
reforms in CGEs and DSGEs
• Satellite models extend simulation capabilities of the
core model, improve maintainance
• Satellite models constructed for certain transmission
channels of structural reforms
– Identification of transmission variables like TFP, markups,
reservation wage
• Reforms have to be proxied by institutional variables,
preferably on a quantitative scale
• Once institutional impact has been determined, the
change in the transmission variable is passed to core
model
– Change in transmission variable affects core variables.
Effects studied by counterfactual exercise
Product market reforms
• Product market reforms should try to improve the
level of competition and innovation
• Satellite models focus on the link of reforms on the
mark up of prices over marginal costs, the TFP rate
and the degree of the administrative burden
• Transmission variables available on a quantitative
scale, while some institutional regressors like PMR
or MICREF indicators are not
• If possible, quantitative proxies should be used
– Involvement of the state in the economy, number of days
necessary to establish a business, working time needed to
fulfill bureaucratic requirements etc
• Quality of approximation can be examined by rank
correlation analysis
Labour market reforms
• Reforms include employment protection, wage
bargaining, taxes and transfers and active labour
market policies
• Main transmission channels are labour income
taxes, reservation wage and the wage mark up
• Taxes and transfers, bargaining power and active
labour market policies available on a quantitative
scale
• Other indicators rank scaled, proxied by quantitative
indicators
– Long term unemployment for overall degree of regulation
– Employment protection: Severance pay, length of notice
periods, maximum duration allowed for fixed term working
contracts etc
• Common factor of quantitative proxies might be
appropriate
Knowledge based economy
• R&D investments or policies to stimulate innovation
work through an increase in efficiency, translates into
a TFP shock
• Satellites include the link between TFP and R&D
investments and R&D and its basic determinants
– Number of scientists and engineers, patents, level of tertiary
education, research in public sector etc
• Core models can also be modified to consider
directly the role of R&D or different skills of labour
• Augmentation of the production function to include
human capital stock, transformation of unskilled to
skilled labour
Decision tree for model selection
• Models have their individual pros and cons, no
method superior in general
• Most accurate tool depends on specific topic of
research
• Steps in model decision process involve
–
–
–
–
–
Direct or indirect effects of structural reforms
Stability of the relationship: ex post or ex ante analysis
Partial or entire economy analysis
Static or dynamic analysis
Relevance of distributional issues: heterogeneous or
homogeneous agents
Social security in Germany
• Social contributions to health, pensions, unemployment
insurance, nursing care
• Contributions are paid by employees and employers, by
equal amounts
• Funds used to finance extra insurance benefits of public
interest
– East German workers did not contribute, but receive pensions
– Maternity benefits, health insurance for parental leave
• System increases costs of labour, worsens employment
perspectives
– Social contribution rate exceeds 40 percent of the wage bill
• Reforms in the financial structure could improve the
employment record
Social security in Germany (II)
• Effects studied in a macroeconometric model for
German economy
• Benefits of public interest. Financing by the
public, not only by workers
• Contribution rates cut by 1 percentage point
financed by higher government debt or increase
in VAT rate
• Debt financing more expansive. Employment
rises by 70Tsd workers 2 years after the shock,
increase in GDP growth by 0.2 points
• Tax increase might be beneficial in the long run
from a regulatory policy perspective