Transcript 슬라이드 1 - Keun Lee : homepage

Both Institutions and Policies Matter
but differently at Different Income Levels:
Determinants of Long-run Economic growth
Keun Lee and Byung-Yeon Kim
Recent Debates on
Long Run Economic Growth
Institution vs. Policies vs. Geography
1) Not policies but Institutions matter:
Reversal of fortunes by colonizers
(Acemoglu, Johnson & Robinson 2001, 2002; Rodrik et al 2004)
Institution as a savior of the Washington Consensus
2) Geography matter: A destiny?
( J. Sachs)
3) Does institution matter? -> Human capital
( Glaeser, La Porta, Lopez-de-Silanes & Shleifer, 2004)
Main Ideas and Arguments
 While the institution supremacy view tends to discard
policies in favor of institutions, this paper takes the
view that both matters but differently at different
stage of economic development.
 New policy variables which do not turn into being
insignificant even when added together with institution
variables.->Technology
 Only with several new policy variables, such as
emphasis on tertiary education and R&D expenditure,
can one explain the recent “reversal of fortune”
between East Asian economies and Latin
American countries
1. Different Engines at Different Stages
 engine of growth should be different for
countries at different stage of economic
development, namely between higher income
and lower income countries.
-hypothesize that while primary and/or
secondary education and institution matter
more for lower income countries, tertiary
education and technology policies (higher
R&D/GDP ratio) matter more higher income
countries or in the transition from upper
middle to high income countries.
2. New Policy Variables
 new variables as determinant of long run growth.
much missing in the debate on the determinant of economic
development is the role of technology.
 Technological innovation is now increasingly
recognized as one of the most serious bottleneck in
many countries, especially in middle income
countries in Latin America.
 The difference between more successful Asian economies and
less successful Latin American economies, or the reversal of
fortune between these two groups of countries, can be
explained in terms of the priority to given policies to enhance
long term growth potentials, technology in particular.
Washington Consensus vs. East Asian Consensus
A. Elements of Consensus
South Korea
Taiwan
China
A1. Macroeconomic Stabilization
1. Fiscal Discipline
Yes, generally
Yes
Yes, generally
2. Pub. Expenditure to Health, Edu, Infra
Yes
Yes
Yes, generally
3. Tax Reform, broadening the Tax Base
Yes, generally
Yes
Yes since 1994
4. Unified & Competitive Exchange Rates
Yes, except early time
Yes
Yes, since 1994
5. Secure Property Rights
Yes, except early prd.
Yes, generally
Mixed
Limited
Limited
Limited
7. Trade Liberalization
Limited until the 80's
Limited until the 80's
Limited until 2002
8. Privatization
many SOEs in 50, 60s
Many SOEs in 50, 60s
No, still SOEs dominant
Heavily restricted
subject to control
some sector restricted
Limited until the 80's
Limited until the 80's
Limited until the 80's
A2. Privatization, Deregulation and Liberalization
6. Deregulation
9. Elimination of Barriers to FDI
10. Financial Liberalization
B. Missing Elements from the Washington Consensus
11. Export Promotion + Tariffs
Yes, very strong
Yes
yes, very strong
12. Technology Policy for Upgrading
Yes, since 1970
Yes, since the 1980s
priority since 90s
Yes, since the 1980s
Yes, generally
Yes, since the mid 1990s
13. Higher Education Revolution
3. Better Empirical Methodologies
 Simple or instrumented cross-section regressions
vs. fixed panel and system-GMM estimations,
(taking care of small sample, omitted variable and endogeneity
problems.)
 Cross-sectional estimation, has an advantage that allows one to
use a time-invariant variable like geography as a regressor. –
 But, country-specific aspect of economic growth that is ignored
in a single cross-section regression may be correlated with
independent variables, causing omitted variable bias.
 a fixed-effect panel estimation reduces omitted variable bias
and time-invariant heterogeneity. However, time-varying country
effects and endogeneity are not controlled in this method.
 Thus, also use a system-GMM method, also known to reduce a
small sample bias.
Estimation models
yit  yi 0  1  1GEO i   1INSTi  1POLi  1 yi 0  1PGROWi  ei
yit   2   2 yi 0   2 INSTit   2 POLit  2 PGROW  it
yit   3  3 yi 0   3INSTit   3POLit  3PGROW  i  it
 yit





log of GDP per capita in country i expressed in constant
US dollars in year t,
yi0 log of GDP per capita in country i expressed in constant
US dollars in initial year,
GEOi geography of country i,
INSTi Institutions of country i,
POLi
policy of country i,
PGROWi growth rate of population of country i.
Data Sources and Usages
 Most of the data for GDP’s in current, constant and PPP terms,
population, trade, capital formation and so on are from the World
Development Indicators 2005 (CD ROM), except for those for Taiwan
for which we relies on the official statistics from the Taiwan
government as well as the Penn World tables available at the web site.
 institution variables, we rely on Jaggers and Marshall (2000), Polity IV
Project: “constraints on the executives” =extent of institutionalized
(shown in laws) constraints on the decision-making power of chief
executives, whether individuals or collectives.
 R&D (research and development) expenditure measured by its ratio to
GDP. from the original sources of the UNESCO Yearbooks.
 Technology policy by the number of US patent applications, from 1965
to 2003; classical technology literature (Schmookler 1996, Griliches
1990) observes that patents, close to innovation inputs and the
absolute majority of them are never used in actual productions.
Summary
1) both institutions & policies (Tech. &
education)matter
2) has identified and confirmed the growth impacts
of new policy variables, such as technology and
tertiary education, which have been often
neglected in the literature and field.
3) secondary education and institution, important
for lower income countries;
technology and higher education,effective in
generating growth for upper middle and high
income countries but not for lower middle and
low income countries
Implications : Policy Sequencing matter
 So far, cross-country studies on long run growth
have not taken into heterogeneity among countries
and have suggested uniform prescriptions.
 if we had not divided the countries into two groups,
we might have also said that we should increase
expenditure on tertiary education or R&D, whereas
this has not been confirmed true for lower income
countries or has to be stated with certain sequence
of priorities in mind.
 Stage-Specific Growth Policies – Sequencing matter:
new saviors of the Washington consensus?
Case for State Activism
(market failure in technology)
Missed opportunity for tech. Development in Late-comers
-> Difficulty with building technological capability.
Late-comer Firms feel no reason to spend for R&D
(outcomes, uncertain; technologies often available from
advanced countries).
No strong-enough incentives or high-enough rate of return
for R&D and tertiary education,
and/or gap between social and private returns to R&D
and education.
While innovation for development does not necessarily mean
development of new products or radical innovation, even
process innovation, adaptations and improvement
requires purposeful efforts with specific target and
certain or high level of human capital.