Transcript Diapositiva 1

Bridging the Gap:
the Role of Trade and
FDI in the Mediterranean
June 8-9, 2006
Convergence and FDI
in the Mediterranean basin:
an empirical evidence
Giuseppe Notarstefano - Raffaele Scuderi
Dipartimento di Contabilità Nazionale ed Analisi dei Processi Sociali
Università degli Studi di Palermo, Italy
Structure
1 – Purpose
2 – Statistical approaches to test convergence hypothesis
i. Review of literature: choice of the methodology
ii. Methodology: the Stochastic Kernel
iii. Empirical findings
3 – Relationship between FDI and growth
i. Review of literature
ii. Methodology: the Stochastic Kernel in a deterministic
conditioning scheme
iii. Empirical findings
4 – Conclusions
1. Purpose
To assess growth dynamics of countries agreeing
to the Euro-Mediterranean Partnership (EMP)
In particular:
Per capita income convergence test
Role of Foreign Direct Investments (FDI) for economic growth
2.i – Convergence: Review (1)
Implication of neoclassical theory of growth: long term reduction of per capita
income gaps (Solow, 1956; Cass, 1965; Koopmans, 1965)
Decreasing marginal productivity of capital
CONVERGENCE HYPOTHESIS (CH)
2.i – Convergence: Review (2)
CONVERGENCE HYPOTHESIS TESTS
Contributions are classified into two categories
Absolute convergence
Per capita incomes of a given set of countries converge to the unique steady state
Conditional convergence
Each economy converges towards its own steady state, proxied by variables
Per capita income needs to be conditioned to a set of proxies
Test the influence of proxy variables to economic growth (endogenous growth theory)
No strong empirical evidence is found about factors influencing economic growth
Durlauf and Quah (1998) - the choice of steady state’s proxies depends on the interest of the
researcher
2.i – Convergence: Review (3)
CONVERGENCE HYPOTHESIS TESTS
Absolute convergence
Conditional convergence
2.i – Convergence: Review (3)
CONVERGENCE HYPOTHESIS TESTS
Absolute convergence
Conditional convergence
“[…] there are reasons other than the testing of economic growth
theories for the empirical study of economic convergence. We,
as economists, are interested in knowing whether the
distribution of income changes over time” (Sala-i-Martin,
1996)
“[…] the new research no longer makes production function
accounting a central part of the analysis. Instead, attention
shifts more directly to questions like, Why do some countries
grow faster than others? It is this changed focus that, in our
view, has motivated going beyond the neoclassical growth
model” (Durlauf and Quah, 1998)
Classification of the approaches, according to the methodology employed
Cross-section
Panel data
Time series
2.i – Convergence: Review (4)
Cross-section approach
Sigma convergence: variability of per capita income, as measured by the coefficient of
variation, reduces over time
Beta convergence:
absolute: no Xi,t
world countries
 i ,t ,t T =  -  log  yi ,t  + ψXi,t +  i ,t ,t -T
 i ,t ,t T  log  yi ,t T yi ,t  T
country i’s per capita income at time t
→ reject CH on a set of
Baumol (1986)
conditional
→ accept CH on a set of
world countries
Barro and Sala-i-Martin (1992)
Mankiw et al. (1992)
2.i – Convergence: Review (4)
Cross-section approach
Sigma convergence: variability of per capita income, as measured by the coefficient of
variation, reduces over time
Beta convergence:
absolute: no Xi,t
world countries
 i ,t ,t T =  -  log  yi ,t  + ψXi,t +  i ,t ,t -T
 i ,t ,t T  log  yi ,t T yi ,t  T
country i’s per capita income at time t
Quah’s criticisms:
→ reject CH on a set of
Baumol (1986)
conditional
→ accept CH on a set of
world countries
Barro and Sala-i-Martin (1992)
Mankiw et al. (1992)
- parametric tests generally refer to the behaviour of a representative unit, and they
are then unsuitable to catch the more real situations of polarization and club
convergence
- sigma and beta convergence approaches are shown to be uninformative about
convergence in some cases
- beta convergence approach does not properly consider dynamics
2.i – Convergence: Review (5)
Panel data approach
Islam (1995): each steady state is better proxied by employing a fixed effects
→ accept CH
panel estimator, otherwise convergence effect would be underestimated.
on a set of world
countries
Quah’s criticism:
“individual heterogeneities“ conditioned out by panel data regression contain those
characteristics which are treated as something not consistently estimable (Quah, 1999)
Time series approach
Bernard and Durlauf (1995) test the presence of common long-run trends
between per capita GDP series
Hobjin and Franses (2000) propose an algorithm to detect clubs of
converging countries
→ reject CH
Evans and Karras (1996) perform a unit root test on panel data
→ accept CH
Quah’s criticism:
these tests do not put in account cross-sectional information
2.ii – Convergence: Methodology (0)
“parametric tests generally refer to the behaviour of a representative unit, and
they are then unsuitable to catch the more real situations of polarization and
club convergence”
2.ii – Convergence: Methodology (0)
“parametric tests generally refer to the behaviour of a representative unit, and
they are then unsuitable to catch the more real situations of polarization and
club convergence”
Univariate density estimates of per capita income
1982
All
(1960-2003)
26 countries: Algeria, Austria, Belgium, Cyprus, Denmark, Egypt, Finland, France, Germany,
Greece, Ireland, Israel, Italy, Jordan, Lebanon, Luxembourg, Malta, Morocco, Netherlands,
Portugal, Spain, Sweden, Syria, Tunisia, Turkey, United Kingdom
1971
1993
2.ii – Convergence: Methodology (1)
Quah’s distribution dynamics approach: the stochastic kernel (SK)
- Time series methodology which also considers sectional information
- Each period’s observation is not a scalar or a finite-dimensional vector, but a distribution
- Nonparametric estimate of the law of motion of the evolving income distribution; based on some
properties of Markov chains
SK is a transition probability matrix in the continuum.
It tracks income distribution’s evolution over time
2.ii – Convergence: Methodology (1)
Quah’s distribution dynamics approach: the stochastic kernel (SK)
- Time series methodology which also considers sectional information
- Each period’s observation is not a scalar or a finite-dimensional vector, but a distribution
- Nonparametric estimate of the law of motion of the evolving income distribution; based on some
properties of Markov chains
SK is a transition probability matrix in the continuum.
It tracks income distribution’s evolution over time
(1)
M
λt
is a measure corresponding to Ft (the cross-country distribution of per
capita income at time t)
λt+1= M * λt
the relation that links λt+1 with λt is analogous to a standard time-series
first-order vector autoregression
maps then how yt evolves at t+1; it also contains information about distribution dynamics and shape.
Iteration of (1) estimates future density distribution:
λt+s= (M * M * … * M) * λt = Ms * λt
2.ii – Convergence: Methodology (2)
Stochastic kernel: estimation (Quah, 1995; 1997)
We consider relative to the mean income
- Epanechnikov kernel nonparametrically estimates the joint density of relative
income at dates t and t+5:
fˆ  Yi ,t ,Yi ,t 1 
n
x  Xi 
ˆf  x   1
K 


nh i 1  h 
Silverman’s criterion is employed for the choice of the bandwidth h
- The current-period marginal density, implied by that estimated joint density, is
calculated by integration
fˆ  Yi ,t 
- Stochastic kernel is obtained by dividing the joint density by the marginal
ˆ Y Y 
f
i ,t , i ,t 1
fˆ  Yi ,t 1 |Yi ,t  
fˆ  Y 
i ,t
2.ii – Convergence: Methodology (3)
Stochastic kernel: output and interpretation
Ft =
Ft+1 =
λt
λt+1
then $ M | λt+1= M * λt
PROBABILITY MASS
- Located along the 45-degree diagonal: persistence in
the economies’ relative position
- Concentrated along the perpendicular line to the 45degree diagonal: overtaking of economies in their
rankings.
- Parallel to t+k axis: the probability of being in any state
at period t+k is independent of economies’ position at t
- Parallel to the t axis: convergence
2.iii – Convergence: Results
Convergence analysis of relative per capita GDP over 1960-2003, 5 years transition
(EMP)
26 countries: Algeria, Austria, Belgium, Cyprus, Denmark, Egypt, Finland, France,
Germany, Greece, Ireland, Israel, Italy, Jordan, Lebanon, Luxembourg, Malta,
Morocco, Netherlands, Portugal, Spain, Sweden, Syria, Tunisia, Turkey, United
Kingdom
Data source: The World Bank (2005)
3.i – FDI-growth: Review (1)
Foreign debt crisis of developing countries (Eighties)
Attention to no debt creating flows like FDI
Many countries offered tax incentives and subsidies
to attract foreign capital
FDI nowadays accounts for over 60 percent of private capital flows
Increase of FDI towards Mediterranean developing countries (DCs) has been
considerably smaller than the one shown by all the DCs
3.5
3
2.5
LMIC = Low and Middle Income Countries
2
MPC
LMIC
1.5
1
0.5
20
03
20
01
19
99
19
97
19
95
19
93
19
91
19
89
0
19
87
FDI/GDP
ratio
MPC = Mediterranean Partner Countries
3.i – FDI-growth: Review (2)
Theoretical FDI spillover effects:
- stock of human capital
- propension to invest
- per capita income
… but …
Carkovic and Levine (2002): “While
there are sound conceptual reasons
for believing that FDI can ignite
economic growth, the empirical
evidence is divided”
3.i – FDI-growth: Review (3)
Macroeconomic analyses
Prevailing literature: positive relationship between FDI and growth
Borensztein et al. (1995) – set of 69 developing countries receiving flows from industrialized countries
Balasubramanyam et al. (1999) – Asian and South American countries
Alfaro et al. (2000) – three samples of countries, ranging from developing to advanced
Berthélemy and Démurger (2000) – Chinese provinces
Nair-Reichert and Weinhold (2001) – sample of 24 developing countries
Notarstefano and Scuderi (2004) – EMP countries
Carkovic and Levine (2002)
They criticise methodologies employed by macroeconomic approaches, which often
do not control for simultaneity distortions and specific country effects.
Dynamic panel estimation on a set of countries, ranging from developing to advanced.
FDI are found to be not related with growth
3.i – FDI-growth: Review (4)
Microeconomic analyses at firm level
Positive spillover effects of FDI
Liu (2002), – 29 manufacturing Chinese industries
Branstetter (2006) – Japanese firms investing to the US
Absence of spillover effects
Aitken and Harrison (1999) – firms in Venezuela
see also Germidis (1977), Haddad and Harrison (1993), Mansfield and Romeo (1980)
3.ii – FDI-growth: Methodology (1)
Stochastic Kernel: deterministic conditioning scheme (Quah, 1997)
Nonparametric analysis of original
distribution compared to its
conditioned version
The approach aims to catch how a
conditioning factor “alters” the
original distribution
Regression-like
rationale
3.ii – FDI-growth: Methodology (1)
Stochastic Kernel: deterministic conditioning scheme (Quah, 1997)
Nonparametric analysis of original
distribution compared to its
conditioned version
The approach aims to catch how a
conditioning factor “alters” the
original distribution
Conditioning scheme G – collection of the triple
 l t 
an integer indicating the lag in time
Cl  t 
a subset of countries from the set I
l  t 
Yl  t 
Yl  t 
a set of probability weights, which sum
to 1
value of the original distribution Y at time t in region i
conditioned version of Yl  t 
def

def

Yl  t    Yl  t  ,Yˆl  t 
Yˆl  t  
Regression-like
rationale

 j  t Y j t   l t  
j in Cl  t 
3.ii – FDI-growth: Methodology (2)
Our estimation
yˆ j  t  
  d  d y
i
ij
i
ij
i
dij is the absolute distance of FDI value in region i from the value in region j.
Following Quah (1995, 1997):
mean-relative per capita GDP
Silverman’s criterion for the choice of the bandwidth h
Epanechnikov kernel

is the ratio
3.iii – FDI-growth: Results (1)
Original and conditioned-to-FDI-inflows per capita GDP, over 1990-2003
EMP countries
32 countries: Algeria, Austria, Belgium, Cyprus, Czech Republic, Denmark, Egypt,
Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Israel, Italy, Jordan,
Latvia, Lebanon, Lithuania, Morocco, Netherlands, Poland, Portugal, Slovak
Republic, Slovenia, Spain, Sweden, Syria, Tunisia, Turkey, United Kindgom
Data source: The World Bank (2005)
3.iii – FDI-growth: Results (2)
Original and conditioned-to-FDI-inflows per capita GDP, over 1990-2003
non-EU-in-1992 EMP countries
10 countries: Algeria, Cyprus, Egypt, Israel, Jordan, Lebanon, Morocco, Syria,
Tunisia, Turkey
Data source: The World Bank (2005)
3.iii – FDI-growth: Results (3)
Original and conditioned-to-Italian-FDI-inflows per capita GDP, over 2000-2003
EMP countries
34 economies: Algeria, Austria, Belgium, Cyprus, Czech Republic, Denmark, Egypt,
Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Israel, Jordan,
Latvia, Lebanon, Lithuania, Luxembourg, Malta, Morocco, Netherlands, Poland,
Portugal, Slovak Republic, Slovenia, Spain, Sweden, Syria, Tunisia, Turkey, United
Kindgom, West Bank and Gaza
Data source: The World Bank (2005) and Italian Exchange Office (Ufficio Italiano Cambi, 2006)
3.iii – FDI-growth: Results (4)
Original and conditioned-to-Italian-FDI-inflows per capita GDP, over 2000-2003
non-EU-in-1992 EMP
12 economies: Algeria, Cyprus, Egypt, Israel, Jordan, Lebanon, Malta, Morocco,
Syria, Tunisia, Turkey, West Bank and Gaza.
Data source: The World Bank (2005) and Italian Exchange Office (Ufficio Italiano Cambi, 2006)
4 – Conclusions
- Literature on convergence hypothesis tests has been divided about the methods to
employ, and the presence of convergence in real data
- The choice of the methodology has been driven by statistical consideration about
methodologies previously applied and by characteristics of per capita income density
estimate
- SK analysis suggests that no convergence has occurred between countries agreeing to
the Euro-Mediterranean Partnership, as expected. Income gaps have then persisted over
1960-2003.
- A regression-like rationale has been applied to SK as done in Quah (1997). Evidence
indicates that FDI have a positive influence on income distribution. This confirms the
evidence of the prevailing macroeconomic empirical analyses.
- Cluster persistence within EMP countries, as well as the influence of FDI for growth,
have also been shown in a previous contribution (Notarstefano and Scuderi, 2004)
- Future research may be based on the concept of growth as multidimensional
phenomenon, as pointed out by recent literature on living standards, and as also done in
our previous work