review of the literature

Download Report

Transcript review of the literature

ACADEMY OF ECONOMIC STUDIES, BUCHAREST
DOCTORAL SCHOOL OF FINANCE AND BANKING
Real economic convergence in the CEE countries
MSc Student: Iancu Guda
Supervisor: Professor Moisă Altăr
Real economic convergence in the CEE countries
Contents:
1. Abstract
2. Review of the Literature
3. Data and Methodology
4. Estimation results
5. Conclusions and remarks
6. References
ABSTRACT
This study aims to assess the economic convergence of the European countries, and especially the
convergence of the 12 CEE countries that acceded European Union after 2000 (Czech Republic, Estonia,
Cyprus, Latvia, Lithuania, Hungary, Malta, Poland, Slovenia and Slovakia EU members starting with 2004,
Bulgaria and Romania EU members starting with 2007 ).
Convergence among the group is studied, as well as their convergence with the EU 15. Time series, panel
data and SVAR models are used to assess the economic growth, approximate the period of real
convergence of Romania to the EU, as well as to estimate the σ- and β-convergence , and the main
shortcomings of the last indicator.
The objectives of the present study are :
• To know whether the selected 12 EU countries are converging in absolute beta sense independently and
jointly
• To test for conditional convergence across countries
• To measure the speed of absolute and conditional convergence to assess the required time to fill the gap in
the economic development
• To study the underling factors that support and accelerate real convergence.
REVIEW OF THE LITERATURE
Under the convergence hypothesis, countries starting with a low per capita income should have a higher
growth rate. Thus, an inverse relationship between output growth and initial output is interpreted as
evidence in favor of the convergence hypothesis.
The theoretical literature on economic growth has gone through several phases, and the answers to the above
questions depend on the specification of the respective growth model :
• neoclassical (exogenous) growth model (Solow-Swan 1956)
- the economy converges towards a steady state due to diminishing returns to investment in
physical capital
- Assuming a constant population, the long-run growth rate is solely determined by the rate
of technological change, which is assumed to be exogenous.
- economic policy changes will only have a temporary effect on economic activity
- convergence of per capita output across countries with a similar productivity level, savings
rate, depreciation rate, productivity growth and population growth (*)
• Endogenous growth :
- Determine the long-run growth rate within the model ( endogenous growth models )
-Incorporation of R&D theories and imperfect competition ( Romer 1987,1990 ), Grossman
and Helpman ( 1993 )
-In these new framework the long term growth rate depends on governmental action, such
as taxation, maintenance of law and order, provisions and infrastructure services,
protection of intellectual property rights, regulation of international trade
-The new research also includes models of the diffusion of technology
(*)Viewed in the same simple way, the catch-up process would be self-limiting because as a follower catches up, the possibility
of making large leaps becomes smaller and smaller (Abramovitz, MOSES, (1986))
REVIEW OF THE LITERATURE
Neoclassical theory of economic growth – Models of exogenous growth
Y = F(K,L) production function – is neoclassical if the following three properties are satisfied :
1.
2.
3.
F
 0;
K
2F
and

0
K 2
2F
F
;
0
0
L
L2
F(ʎK,ʎL) = ʎ F(K,L) for all ʎ > 0
Inada Conditions :
lim ( FK )  lim ( FL )  
K 0
L 0
and
lim ( FK )  lim ( FL )  0
K 
L 
Solow and Swan Model (1956)
Y = F(K,L) = L*F(K/L,1) = L*f(k) => y=f(k)
.
K / L  s * f (k )   * k
Differential Equation of the Solow-Swan model :
.
k  s * f (k )  (n   ) * k
.
Exogenous variable : population growth rate,
 k  k/ k
rate of technological progress
Constant variables : savings rate, capital depreciation rate
 s * f (k ) / k  (n   )
REVIEW OF THE LITERATURE
.
 k  k / k  s * f (k ) / k  (n   )
.
k  s * f (k )  (n   ) * k
The above case covers the so-called conditional
convergence, that is the alternative implying that all
economies with differences in the initial stock of
capital per capita have the same saving rates (s),
similar technologies (the same parameters A and
δ), as well as the same population (labour) growth
rates (n). Unless such requirements are met, the
equilibrium points of the rich countries differ from
those of the poor countries, and the convergence
cannot take place.
REVIEW OF THE LITERATURE
New methodological approaches to convergence and its determinants
•The new theory is focused in finding out the real causes and mechanisms of the long-term disparities
(through cross-section analysis or long time series), by correlating the growth rate of production and income
per capita at national or/and regional level with several economic, social and political variables that could be
either the engine or the brake of economic growth.
•Usually, conditioning variables such as education attainment, government spending, political instability, and
the growth rate of the terms of trade are included in such an output regression equation to control for effects of
other growth factors (Barro and Sala-i-Martin, 1995).
•There are authors who conducted empiric research on convergence using the modified and augmented
dynamic neoclassical model that involved the human capital and technological progress besides the physical
capital. For example, Mankiw, Romer and Weil (1992), and Islam (1995) revealed, by the new variants of
models, that the economies with an initially low level of the income tended to increase faster than those with
initially high level of the income after they had introduced in the model the saving rate and the population
growth rate, as control variables.
REVIEW OF THE LITERATURE
New econometric testing of the new calculation tools and models, such as :
•
•
•
•
•
•
•
•
•
•
the β and σ indicators (Barro and Sala-i-Martin, 1991)
the augmentent dynamic neoclassical model : Mankiw, Romer, Weil, 1992 ( with human and fix capital ) //
Islam, 1995 // Bassanini, Scarpetta, 2001). Conclude convergence proof after they had introduced in the
model the saving rate and the population growth rate, as control variables.
Barro, Sala-i-Martin, Blanchard and Hall (1991) considered variables like capital mobility, labour migration
Fischer Stanley, Ratna Sahay and Carlos Vegh. 1998, growth regression in panel data with variables like
GDP, inflation, enrolment, investments, government spending ( also Robert J. Barro 1991)
Ross Levine and David Renelt 1992 – Extreme bound analyses to test the robustness, include variables
like trade and investments
Sergio Rebelo 1997 – determinants of economic growth ; Easterly, W., Rebelo, S., 1993 fiscal policy
influence over growth perspective
Robert E. Hall and Charles I. Jones – regression the study the differences of GDP levels among countries
the stochastic convergence model (Lee Kevin, M Hashem Pesaran, Ron Smith ; 1997)
Blanchard, O.J., Perotti, R., 2002. – VAR models testing governments spending and taxation influence
over economic growth
Bart Verspagen 1994 and Ben David : trend calculation for series of per capita income relative to group
average
REVIEW OF THE LITERATURE
The counter-reaction literature
The appropriateness of the cross-country regression approach is challenged by, for example, Quah (1993),
Bernard and Durlauf (1996), and Evans (1997)
•Quah (1993) shows that a negative correlation between output growth and initial output is consistent with a
stable variance in cross-country output.
•Durlauf, 1996; Quah, 1996 – time series models to negate convergence and promote the presence of
convergence clubs and the polarisation of the countries in rich and poor ones. They argue that the initialoutput regression approach tends to reject the null hypothesis of no convergence too often in the presence of
multiple output equilibrium
•Bernard and Durlauf (1995), using the unit root and cointegration techniques, detect the presence of multiple
integrated processes driving the output data of the OECD countries
•Using a panel unit root test, Evans (1998) shows that convergence occurs within a group of developed
countries and different growth patterns are observed among countries with different literacy rates
DATA AND METHODOLOGY
I Time series
I.1 Verspagen (1994)
This test assumes the following relation between per capita income relative to the group average in different
periods t and t - 1.
ln( Yt )   * ln( Yt 1 )
where Yt is defined as (Q/P)/(ΣQ/CP), and Q denotes RGDP, P denotes population (in thousands), and Σ
indicates the sum in some group of countries
The assumed relationship allows for :
•
converging (if ψ < l)
•
diverging (if ψ > l)
•
stable (if ψ= 1) differences in per capita income
Data are comprised from Alan Heston, Robert Summers and Bettina Aten, Penn World Table Version 6.3,
Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania,
August 2009 for the 12 CEE countries and UE15 average values
DATA AND METHODOLOGY
In order to estimate the value of ψ for different periods of time, the next steps are followed:
•
For each ‘case’ (defined as a combination of countries, years, and reference group), a pooled crosscountry time series dataset is set up.
•
In turn, ψ is estimated by OLS for the subset for periods t - 2 to t + 2. This means that the number of
observations in each OLS estimate is five times the number of countries in the analysis.
•
The resulting estimate of ψ is attributed to t, and plotted in a graph together with the estimates for other
periods.
•
The estimated coefficient plus/minus two times the estimated standard error is also plotted, so that a
(reasonably wide) confidence interval is established.
•
Whenever this confidence interval is completely below (above) unity, convergence (divergence) is said to
be observed. Whenever the confidence interval embraces the unit line, no particular trend is found.
The objectives are:
•
•
To test for convergence differences in per capita income within the 12 CEE countries group – to
this purpose the series of RGDP/cap ( Real GDP per capita in 2000 values as base year) are comprised
with annual frequency between 1993-2007. The series are divided in subset of periods t - 2 to t + 2.
Hence, we obtain a final 11 observation for each country. The pooled data comprises 60 observations (12
cross country and 5 time series ) of country individual RGDP/cap difference from groups average.
To test for convergence differences in per capita income between the CEE countries and UE15
average value. The time series is extended to 1970-2007 range, thus including 34 annual observations (
after extracting subsets of 5 years each ). Due to data limitation, only 6 CEE countries are included (
Bulgaria, Cyprus, Hungary, Malta, Poland, Romania ). Hence, the polled data comprises 30 observations (
6 cross section and 5 time series ) of country individual RGDP/cap difference from UE15 average.
DATA AND METHODOLOGY
I.2 . Convergence tests for pairs of countries ( methodology used in M. Hashem Pesaran, 2007)
A time-series based approach to investigate output convergence has been proposed by Bernard and Durlauf
(1996) and Quah (1992). According to Bernard and Durlauf (1996), there is output convergence between two
countries if the long-run forecasts of their real per capita outputs are the same. If pair-wise convergence holds
for all the pairs of countries under consideration, then there is convergence of all the countries simultaneously.
Specifically, let Yi,t be the (logarithm of the) country i’s real per capita output at time t and Y*,t is the output
variable of the benchmark country. The no-convergence hypothesis is stated as :
H0 : X i ,t  (Yi ,t  Y*t ) = I(1), i = 1,…,N
World Table Version 6.3, Center for International Comparisons of Production, Income and Prices at the
University of Pennsylvania, August 2009 for the 12 CEE countries and UE15 average values. Values for
annual Real GDP /capita (in 2000 values as base year) are found for the following series:
•1970-2007 for Bulgaria, Cyprus, Hungary, Malta, Poland, Romania and UE15
•1987-2007 for Slovak Republic
•1990-2007 for Czech Republic, Estonia, Slovenia
•1993-2007 for Lithuania and Latvia
In testing for:
•Convergence within the 12 CEE countries group, ADF stationary test are applied for N(N-1)/2 possible pairs
of log per-capita output gaps across N economies.
•Convergence between the 12 CEE countries and UE15 average real GDP/capita, stationary is tested for 12
output pairs differentials.
DATA AND METHODOLOGY
II. Panel Data ( methodology also used in Islam 1995 // Crespo-Cuaresma J, Dimitz MA, RitzbergerGrunwald 2002 // Matkowski Z, Prochniak M 2004 // Rajasalu Teet 2003
•
•
•
)
Pooled data
LSDV (Least Square Dummy Variable ) – fixed effects for cross and time series ( intercept / slope
variation)
ECM ( Error Components Model ) or Random Effects
Variables :
real GDP per capita growth rate (difference of consecutive log values )
LN of Real GDP per capita
Trade ( export + import ) % of GDP
Gross Capital Formation % of GDP
Total Government Expenditure % of GDP
Total Government Income % of GDP
Saving rate % of GDP
Enrollment ( secondary + tertiary ISCED Levels )
Data series for 12 EU members acceded after 2000 ( Czech Republic, Estonia, Cyprus, Latvia, Lithuania,
Hungary, Malta, Poland, Slovenia and Slovakia EU members starting with 2004, Bulgaria and Romania
EU members starting with 2007 ), period 1995-2009. Source : Eurostat
Observations : 12 Countries, 15 years ( 1995 – 2009 ), 5 time intervals ( 3 years each ) = 60 panel data
observations (12 Cross countries and 5 interval time series )
DATA AND METHODOLOGY
Argumentation:
•
Why not include 1990-1994 data ? The beginning of the transition of the former socialist countries
caused a massive fall of output which rivaled, and in several countries even surpassed, the dimension of
the United States recession during the Great Depression. After a few years, the fall in output stopped,
and a few countries, mostly Central and Eastern European ones (CEE), started a process of economic
growth. The success of reforms and the sustained economic growth spurred a whole line of research on
the determinants of economic growth and on the long run economic growth perspectives of the CEE
countries.
•
Why 3 years for each period ? The standard approach for reducing the effects of structural changes and
business cycles, is to divide the series and average data on 5 years ( or even 10 if longer data available
) for each interval. Due to data limitation ( total range of data compiled 15 years ), a 5 years interval
would have reduced the degree of freedom
DATA AND METHODOLOGY
II. Panel Data
The above equations are estimated to test for :
•
Absolute Convergence
•
Absolute convergence speed
•
The fixed effects
•
LSDV
•
Random effects
DATA AND METHODOLOGY
III. Structural VAR ( methodology used in Sarah M. Lein, Miguel A. Leo´n-Ledesma, Carolin Nerlich, 2008)
•
-
Variables :
real GDP per capita growth rate (difference of consecutive log values )
Trade ( export + import ) % of GDP
Gross Capital Formation % of GDP
•
empirically assess their relevance using a Structural VAR method for Lithuania and Romania in trying to
find out the real causes and mechanisms of the long-term disparities
Data is compiled from :
•
•
EUROSTAT , with quarterly frequency between 1995 Q1 : 2009 Q4, comprising 60 observations for each
variable for Lithuania
Nations Institute of Statistics ( Romania’s official webpage) with quarterly frequency between 2000 Q1 :
2009 Q4, comprising 40 observation for each variable for Romania.
Observation: All series are seasonally adjusted using the Census X12 seasonal adjustment procedures
ESTIMATION RESULTS
1995 Real GDP per
capita, EUR
Country
Real Per Capita
Average Growth Rate
(1995:2009)
Bulgaria
1,700.00
3.41%
Czech Republic
5,500.00
2.62%
Estonia
3,100.00
5.76%
Cyprus
12,900.00
1.84%
Latvia
2,600.00
5.65%
Lithuania
2,700.00
5.54%
Hungary
4,100.00
3.05%
10,700.00
0.53%
Poland
3,700.00
4.59%
Romania
1,939.48
3.05%
Slovenia
8,700.00
3.31%
Slovakia
3,500.00
4.35%
Malta
Scatter Plot Evidence:
•Romania and Bulgaria, although having the lowest Real
GDP per capita in the group for the starting year, fail to
converge in terms of absolute β convergence ( as in
Barro and Sala-i-Martin, 1991)
•Average growth rate for Romania and Bulgaria in 19952009 bellow group’s average
•Better convergence results ( adjusted R squared, F
statistic ) for the group when excluding Romania and
Bulgaria (Panel Data)
AVERAGE REAL GROWTH RATE - INITIAL GDP
7.00%
2,600.00
6.00%
3,100.00
2,700.00
5.00%
4.00%
1,700.00
3,700.00
3,500.00
4,100.00
3.00%
8,700.00
1,939.48
5,500.00
2.00%
12,900.00
1.00%
10,700.00
0.00%
-
2,000.00
4,000.00
6,000.00
8,000.00
10,000.00
AVERAGE REAL GROWTH RATE - INITIAL GDP
12,000.00
14,000.00
ESTIMATION RESULTS
Year
European
Union (15
countries)
Bulgaria
Czech
Republic
Estonia
Cyprus
Latvia
Lithuania
Hungary
Malta
Poland
Romania
Slovenia
Slovakia
1995
100%
8.33%
26.96%
15.20%
63.24%
12.75%
13.24%
20.10%
52.45%
18.14%
9.51%
42.65%
17.16%
1996
100%
7.25%
28.02%
15.94%
62.32%
13.53%
14.01%
19.81%
52.17%
19.32%
9.73%
43.48%
17.87%
1997
100%
7.08%
26.89%
17.45%
61.32%
14.15%
14.62%
20.28%
50.47%
19.81%
8.96%
44.81%
18.40%
1998
100%
6.88%
26.15%
18.35%
61.93%
14.68%
15.60%
21.10%
49.08%
20.64%
8.31%
44.95%
18.35%
1999
100%
7.14%
25.89%
18.30%
62.50%
14.73%
15.18%
21.43%
48.21%
20.98%
7.92%
46.43%
17.86%
2000
100%
7.33%
25.86%
19.40%
62.50%
15.52%
15.09%
21.55%
46.55%
21.12%
7.76%
46.55%
17.67%
2001
100%
7.66%
26.38%
20.43%
63.40%
16.60%
16.17%
22.13%
45.11%
20.85%
8.09%
47.23%
17.87%
2002
100%
8.02%
26.58%
22.36%
63.71%
17.72%
17.30%
23.21%
45.57%
21.10%
8.86%
48.52%
18.57%
2003
100%
8.40%
27.31%
23.95%
63.45%
18.91%
18.91%
23.95%
44.96%
21.85%
9.24%
49.58%
19.33%
2004
100%
9.09%
28.10%
25.21%
63.64%
20.25%
20.25%
24.79%
44.21%
22.73%
9.92%
50.83%
20.25%
2005
100%
9.39%
29.39%
27.35%
63.67%
22.45%
21.63%
25.71%
44.90%
23.27%
10.20%
52.24%
21.22%
2006
100%
9.96%
30.68%
29.48%
63.35%
24.70%
22.71%
25.90%
45.02%
23.90%
10.76%
53.78%
22.31%
2007
100%
10.16%
31.64%
30.86%
64.45%
26.56%
24.61%
25.78%
45.31%
25.00%
11.33%
56.25%
24.22%
2008
100%
10.94%
32.03%
29.69%
66.02%
25.39%
25.39%
25.78%
46.09%
26.56%
12.16%
58.20%
25.78%
2009
100%
10.90%
32.20%
27.05%
68.08%
22.13%
22.95%
25.41%
47.13%
28.34%
11.85%
55.74%
25.82%
Gap
reduction
1995-2009
2.57%
5.23%
11.85%
4.85%
9.39%
9.72%
5.31%
-5.32%
10.21%
2.34%
13.09%
8.66%
Weights calculated based on Real GDP per capita levels, compared with EU15, SOURCE : calculation based on series from EUROSTAT
Countries bellow the average rhythm of real DGP/capita recovery : Bulgaria, Czech Republic, Cyprus,
Hungary, Malta, Romania, Slovakia
Countries over the average rhythm of real DGP/capita recovery: Estonia, Latvia, Lithuania, Poland,
Slovenia
ESTIMATION RESULTS - VERSPAGEN Test
Convergence among 12 CEE Countries
1.00000
Convergence between 6 CEE Countries and UE15 Average
Convergence among 12 CEE Countries
1.1
Convergence between 6 CEE Countries-UE15
0.90000
1
0.80000
0.70000
0.9
0.60000
1995
1996
ψ
1997
1998
1999
2000
+2se
2001
2002
2003
-2se
2004
2005
19
72
19
74
19
76
19
78
19
80
19
82
19
84
19
86
19
88
19
90
19
92
19
94
19
96
19
98
20
00
20
02
20
04
0.8
0.50000
ψ
+2se
-2se
• For convergence within the 12 CEE group, the solid line is below unity, indicating convergence.
Moreover, for all years considered, the trend is significant ( the confidence interval is completely below unity
).
• For the convergence between the six CEE countries and UE15, a mixed pattern is observed. In most
cases, the solid line is below unity, indicating convergence. However, there are no years for which this trend
is significant. This indicates that although there is a weak trend for convergence, the growth behavior of the
individual countries is so erratic that the overall trend is insignificant.
Convergence tests for pairs of countries - The Augmented Dickey-Fuller test
• Real GDP differences for pairs of countries result in nonstationary series (based on the ADF test),
only for 22.72% they are I(0), hence rejecting convergence
• The result is consistent with the result obtained by M. Hashem Pesaran (2007), who found ~25%
I(0) series when running similar convergence tests to output series in the Penn World Tables over
1950–2000 (“A pair-wise approach to testing for output and growth convergence”, 2006, Journal Of
Econometrics)
Estimation results σ -convergence
Year
Dispersion
1995
0.476395364
1996
0.452627292
0.5
1997
0.429929014
0.45
1998
0.418297471
0.4
1999
0.426234716
0.35
2000
0.413031814
2001
0.394221732
2002
0.367640069
2003
0.336868838
2004
0.309006425
2005
0.285321743
2006
0.261992513
2007
0.250311365
2008
0.250561207
2009
0.272787971
Real GDP per capita dispersion
Barro and Sala-i-Martin
(1992) introduced the
notion of σ -convergence.
σ -convergence is said to
be present if the
dispersion of income per
capita across countries
display a tendency to
decline through time
0.3
0.25
0.2
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Real GDP per capita dispersion
Hodrick-Prescott Filter TREND
12
11
10
9
8
Bulgaria
Lithuania
Slovenia
Finland
Italy
Sweden
Czech Republic
Hungary
Slovak Republic
France
Luxembourg
United Kingdom
Estonia
Malta
Austria
Germany
Netherlands
Cyprus
Poland
Belgium
Greece
Portugal
20
06
20
04
20
02
20
00
19
98
19
96
19
94
19
92
19
90
19
88
19
86
19
84
19
82
19
80
19
78
19
76
19
74
19
72
19
70
7
Latvia
Romania
Denmark
Ireland
Spain
Panel Data – Absolute β Convergence – All 12 countries
Pooled Data : All Coefficients Constant across
Time and Individuals
LSDV: Fixed Effects The Time Effect
For all 3 models, sign of Beta Convergence ( as in Barro and
Sala-i-Martin (1992) ), negative relation between Real GDP
per capita growth rates and initial level of Real GDP / capita at
the beginning of each subperiod.
LSDV: Fixed Effects The Individual Effect
• Judged after the value of adjusted R squared, the time variant model seems to be the most appropriate,
although some problems arise with the statistical significance of the estimated coefficients for DY2 and DY3.
•Restricted F test are also employed and favor for The Time Effects model
• This could mean that the production function shifts over time because of factors such as technological
changes, changes in government regulatory and/or tax policies, and external effects such as structural
changes in transitional economies.
• Individual country effects are also statistically significant, with one exception ( country code 10 meaning
Romania )
Panel Data – Absolute β Convergence – All 12 countries
Hausman Test interpretation
The null is that the two estimation methods are both validated and that
therefore they should yield coefficients that are "similar". The alternative
hypothesis is that the fixed effects estimation is validated and the random
effects estimation is not. If this is the case, then we would expect to see
differences between the two sets of coefficients.
This is because the random effects estimator makes an assumption (the
random effects are orthogonal to the regressors) that the fixed effects
estimator does not. If this assumption is wrong, the random effects
estimator will be inconsistent, but the fixed effects estimator is unaffected.
Hence, if the assumption is wrong, this will be reflected in a difference
between the two set of coefficients. The bigger the difference (the less
similar are the two sets of coefficients), the bigger the Hausman statistic.
A large and significant Hausman statistic means a large and significant
difference, and so the rejection of the null that the two methods are valid in
favor of the alternative hypothesis that one is valid (fixed effects) and one
isn't (random effects).
Panel Data – Absolute β Convergence – 10 countries (without RO & BG)
Pooled Data : All Coefficients Constant
across Time and Individuals
LSDV: Fixed Effects The Time Effect
The elimination of RO and BG series :
LSDV: Fixed Effects The Individual Effect
•strengthen the estimator for LN_GDP0, which is statistically significant for all models = > stronger Beta
Convergence process among the group of 10 countries
•Increased the value of adjusted R-squared
•λ, speed of convergence, increases for all the corresponding models ( pooled data to 2.87% , time effect
model to 2.05% and the LSDV to 9.61% )
•For all four models (pooled, LSDV, time effect and random effect), sign of Beta Convergence ( as in Barro
and Sala-i-Martin (1992) ), negative relation between Real GDP per capita growth rates and initial level of
Real GDP / capita at the beginning of each sub period.
Panel Data – Absolute β Convergence – 10 countries (without RO & BG)
The Random Effects
The Hausman test employed also indicates the fixed effects model
against the random effects model.
In line with the arguments of Hsiao (2003) and Baltagi (2005) the
choice of fixed effects option is more appropriate when the research
focuses on a specific set of N countries, which are not drawn
randomly from a large population, and the outcomes of the study are
viewed as conditional on this set of countries. This situation precisely
reflects our convergence study where the choice of country matters for
answers we search for. This choice was confirmed during estimations
with Hausman’s test favoring the fixed effects specification
Panel Data –Conditional β Convergence – All 12 countries
Pooled Data : All Coefficients Constant across
Time and Individuals
LSDV: Fixed Effects The Time Effect
For all 3 models, sign of Conditional Beta Convergence ( as in
Barro and Sala-i-Martin (1992) ), negative relation between
Real GDP per capita growth rates and initial level of Real GDP
LSDV: Fixed Effects The Individual Effect
/ capita at the beginning of each sub period.
• Judged after the value of adjusted R squared, F statistic, Darbin-W stat and the statistical significance of
the estimated coefficients, the individual effect (time invariant) model seems to be the most appropriate
•Restricted F test are also employed and favor for The LSDV model
•Variable sign as expected : real GDP growth rate negatively correlated with initial level of real GDP per
capita and government spending / positively correlated with trade, gross capital formation, government
income, saving rate and enrollment ( although the latter two are statistically insignificant)
•Individual country effects are also statistically significant, with no exception
Structural VAR
The identification problem (the recovering of the structural innovations from the reduced form) is solved
by employing Choleski decomposition. Hence, the system is exactly identified by imposing 3 [n*(n-1)/2,
with n=3, the number of variables] short term restrictions on A matrix. The Choleski ordering is : real
GDP per capita growth rates(GDP), trade openness (Trade) and gross capital formation(GCF)
GDP
TRADE
GCF
GDP
1
a
b
TRADE
0
1
c
GCF
0
0
1
The contemporaneous relation between the variables is
described as follows : the variables on each line are influenced,
during each quarter, by variables on each column. “0” is for lack
of contemporaneous influence ( imposed restrictions ), “1” signify
the existence of contemporaneous influence. The diagonal is one
as each variable is influenced by it’s own innovations
Structural VAR
Hence, the influences between the variables are :
•Real GDP per capita growth is influenced by contemporaneous and lagged values of trade openness
and gross capital formation. The argument for this is the panel data test of gradually inclusion of
variables that finally favor for trade openness and gross capital formation (as proxy for investment rate)
as being the most important determinants for growth enhancing for the 12 CEE countries
•Trade openness is influenced by contemporaneous and lagged values of gross capital formation and
only by lagged values of real per capita GDP growth. The intuition for this is that higher ( lagged and
contemporaneous ) investment rates ( approximated by gross capital formation ) will eventually
increase the productivity and generate higher competitiveness for the domestic products. Empirical
results in this regard are also found by Monica Ioana Pop Silaghi (2009) “Exports-Economic growth
causality: evidence from CEE countries” when testing the causality between trade and economic
growth (with the result that the first variable influences that latter )
•Gross capital formation is influenced by lagged values of trade openness and per capita GDP growth.
Economic growth is seen as revenue for sustaining the investments and trade intensification would
eventually increase the competition between foreign and local companies. The local firms, the hold the
competition and protect their market share, would have to optimize their products quality by increasing
productivity and investments.
For the final results to be relevant the econometric model must pass over the diagnostic tests. The main
tests and results are briefly summarized bellow
Structural VAR Lag Selection
Structural VAR – Stability test
VAR is stable ( stationary ) if innovations over the variables within the system diminishes along the
time. Stability is confirmed if the unit roots have absolute values less than unity (this entails that
matrix An tends to zero as n increases to infinity, meaning that the VAR is not explosive). The unit
roots values are compiled bellow. As observed, all values are less than unity, thus confirming the
stability of VAR
Structural VAR Structural Factorization
Structural VAR - Impulse Response
Structural VAR - Impulse Response
Structural VAR – Variance Decomposition & Granger causality tests
Conclusions, Methodology limitations and future research
Final conclusions can be summarized bellow:
•Compared with the cross-country analysis, the time-series approach yields less convincing findings for
the convergence hypothesis
•Strong evidence of σ convergence, as the dispersion for the 12 CEE countries real per capita GDP
levels tends to decrease over time
•CEE countries converge between themselves in terms of absolute convergence. Although strong
evidence is found in this direction, the estimated speed of absolute convergence is low, being around
1.30% for the 12 CEE countries and 2.87% after eliminating the series for Romania and Bulgaria.
•High speed of convergence is obtained when controlling the steady state. Unfortunately, we are
looking at conditional convergence. That is, groups of countries and separate countries are converging
fast to their respective steady state of income level, and nothing can guarantee that it is a good steady
state.
•Individual country specific growth patterns are found within the 12 CEE group
•Panel data test results indicate that the production function for the 12 CEE countries shifted over time
because of factors such as technological changes, changes in government regulatory and/or tax
policies, and external effects such as structural changes in transitional economies.
•trade openness and gross capital formation (as proxy for investment rate) are found as being the most
important determinants for growth enhancing for the 12 CEE countries.
Generally, the lack of conclusive evidence from the data on the rate of catch-up makes it advisable for
transition countries to opt for growth-enhancing policies rather than concentrate their efforts on nominal
convergence with Maastricht criteria.
Conclusions, Methodology limitations and future research
For research in future, conditional convergence can be tested using cross sectional average data on
pertinent growth factors like corruption perception indices, rule of law index, social capital and trust
variables, formal and informal rules governing the society, among others. It will be interesting to find out
the speed of conditional convergence by including such variables in the per capita growth equation.
Another possible source of the growth bonus are the changes in the institutional framework due to
European integration. Whereas the completion of the internal market or, in other words, the openness
of the countries is covered more or less by the trade variable, there are other developments which
could also play a role. Examples are the legal and the institutional framework of the financial sector, the
scale and the nature of foreign direct investment, transport infrastructure and the efficiency of public
administration.
As to other possible future extensions of this research, one can suggest a complementary study of
factors behind the recent growth performance in transition economies ( especially for 2000-2008 range
). It might determine the sustainability of growth in transition economies. This will help to answer the
major convergence question as to how long it will take for transition countries to close the income gap
with the EU.
Secondly, one can resort to a number of case studies related to the accession experiences of countries
such as Greece, Spain and Portugal. These countries started with large income disparities with the EU
average. The study of what happened to them before and after accession may shed light on what may
happen to the present and future candidates
References
Ali M. Kutana, Taner M. Yigit (2007) European integration, productivity growth and real convergenceEUROPEAN ECONOMIC REVIEW
Arbia G, Piras G (2005) Convergence in per-capita GPD across european regions using panel data models extended to spatial autocorellation
effects, Institute for studies and economic analyses, working paper nr 51/2005
Barro, Robert J., and Xavier Sala-i-Martin,
(1992) "Convergence," Journal of Political Economy
Bart Verspagen : (1994) Convergence in the global economy. A broad historical viewpoint, STRUCTURAL CHANGE AND ECONOMIC
DYNAMICS,
Bassanini, A., Scarpetta, S., Hemmings, P., 2001. (2001) Economics Department Working Papers No. 283. Organization for Economic
Cooperation and Development, ParisEconomics Department Working Papers No. 283. OECD
BERNARD _ DURLAUF (1994) INTERPRETING TESTS OF CONVERGENCE HYPOTESIS
Bernard, A. and Durlauf, S. (i 996). (1995) Interpreting tests of the convergence hypothesis.' Journal of Econometrics
Bianchi M. (1995) (1995) "Testing for convergence: evidence from nonparametric multimodality tests", Working Paper Bank of England, London
Blanchard, O.J., Perotti, R., 2002. (2002) An empirical characterization of the dynamic effects of changes in government spending and taxes on
output. Quarterly Journal of Economics 117 (4), 1329–1368.
Carl-Johan Dalgaard, Jacob Vastrup (2000) On the measurement of s-convergenceeconomic Letters
Carmela Martín ; Francisco J. Velázquez (2001) AN ASSESSMENT OF REAL CONVERGENCE OF LESS DEVELOPED EU MEMBERS:
LESSONS FOR THE CEEC CANDIDATESEuropean Economy Group (EEG)
Carree M., and Klomp L (1997) (1997) ‘Testing the convergence hypothesis: A comment’Review of Economics and Statistics,77, 683-686
Cheung,Yin-Wong and Pascual,Antonia Garcia(2004 (2004) "Testing for Output Convergence:A Rexamination",Oxford Economic Papers,Oxford
University Press, 56,45-63
Crafts Nicholas and Kai Kaiser. 2004. (2004) “Long-Term Growth Prospects in Transition Economies: A Reappraisal Structural Change and
Economic Dynamics
Crespo-Cuaresma J, Dimitz MA, Ritzberger-Grunwald D (2002) Growth effects of european integration: implications for EU enlargement,Focus
on transition, VOL 1/2, 2002, pag 87-99
References
Danny Quah (1998) Twin Peaks: Growth and Convergence in Models of Distribution Dynamics', WORKING PAPER 280, CENTRE FOR
ECONOMIC PERFORMANCE
Davide Furceri
() BETA and SIGMA -convergence: A mathematical relation of causality
DURLAUF ST. N, (1996) “The Economic Journal”, vol. 106, No. 437 (June 1996), 1016-1018On the Convergence and Divergence of Growth
Rates,
Durlauf, S.N., Johnson, P.A., 1995. (1995) Multiple regimes and cross-country growth behavior. Journal of Applied Econometrics 10, 365–384
Easterly, W., Rebelo, S., 1993. (1993) Fiscal policy and economic growth: an empirical investigation. NBER Working Paper No. 4499. National
Bureau of Economic Research, Cambridge, MA
Enrique López-Bazo ; Esther Vayá Valcarce (1) Antonio José Mora Corral (1) Jordi Suriñach Caralt (1 () REGIONAL ECONOMIC DYNAMICS
AND CONVERGENCE IN THE EUROPEAN UNION
Evans, P. (1997), (1997) “How Fast Do Countries Converge?”, Review of Economics and Statistics, 79: 219-225.
FARHAD RASSEKH (1998) The Convergence Hypothesis: History, Theory, and Evidence
Fischer Stanley, Ratna Sahay and Carlos Vegh. 1998. (1998) ”From Transition to Market: Evidence and Growth ProspectsIMF WP
Frankel, J.A., Romer, D., 1999. (1999) Does trade cause growth? The American Economic Review 89 (3), 379– 399
FRIEDMAN MILTON, (1992) Do Old Fallacies Ever Die?, “Journal of Economic Literature”, vol. 30, No. 4 (December 1992), 2129-2132Journal
of Economic Literature
G.E. Boyle and T.G. McCarthy, (1999) SIMPLE MEASURES OF CONVERGENCE IN PER CAPITA GDP: A NOTE ON SOME FURTHER
INTERNATIONAL EVIDENCEApplied Economics Letters, vol. 6, pp. 343-347.
Herbert Brticker () European Union Eastward Expansion Accelerates Convergence
Ho-Chuan Huang
() Diverging evidence of convergence hypothesis,Journal of macroeconomics
Iancu, A. (2007a), (2007) “The Question of Economic Convergence - First Part”,Romanian Journal of Economic Forecasting, 8(3): 5-18.
References
Iancu, A. (2007b), (2007) “The Question of Economic Convergence - Second Part”,Romanian Journal of Economic Forecasting, 8(4): 24-48.
Iancu, A. (2008), (2008) “Real Convergence and Integration”, Romanian Journal of Economic Forecasting, 9(1): 27-40.
Islam, Nazrul, (1992) "Small Sample Performance of Dynamic Panel Data Estimators: A Monte Carlo Study," Department of Economics, Harvard
University, 1992
Ismael Sanz Francisco J. Velázquez (2001) The evolution and convergence of the government expenditure composition in the OECD countries:
an analysis of the functional distributionEuropean Economy Group (EEG)
Jones, Larry E. and Manuelli, Rodolfo E. (1994) “The Sources of Growth.” Journal of Economic Dynamics and Control, January 1997, 21(1), pp.
75-114
Juan Cuadrado Roura : Regional convergence in the European Union : (2000) From hypothesis to actual trendThe annals of Regional Science
Kevin Lee, M. Hashem Pesaran, Ron Smith (1998) Growth Empirics: A Panel Data Approach -- A CommentThe Quarterly Journal of Economics
KUTAN ALI M., TANER M. YIGIT, (2003) Convergence of Candidate Countries to the European Union, 2003
Lee Kevin, M Hashem Pesaran, Ron Smith (1995) Growth and Convergence: A multi-country empirical analysis of the solow growh model, DAE
Working Paper No 9531, University of Cambridge, 1995
Lichtenberg, F.R. (1994) (1994) ‘Testing the convergence hypothesis’Review of Economics and Statistics, 76, 576-579
M. Hashem Pesaran (2007) A pair-wise approach to testing for output and growth convergenceJOURNAL OF ECONOMETRICS
Matkowski Z, Prochniak M (2004) Real economic convergence in the EU accesion countries,International journal of applied econometrics and
quatitative studies, VOL 1, NR 3, 2004, pag 5-38
MATKOWSKI ZBIGNIEW, MARIUSZ PROCHNIAK, (2004) Economic Convergence in the EU Accession Countries, Warsaw School of
EconomicsInternational Journal of Applied Econometrics and Quantitative Studies
Michael K. Fung () Financial development and economic growth: Convergence or divergence?
Miguel St. Aubyn* (1999) Convergence across industrialised countries (1890±1989): new results using time series methods EMPIRICAL
ECONOMICS
References
Monica Ioana POP SILAGHI (2009) EXPORTS-ECONOMIC GROWTH CAUSALITY: EVIDENCE FROM CEE COUNTRIESRomanian Journal of
Economic Forecasting
MRW - N. Gregory Mankiw, David Romer, David N. Weil (1992) A Contribution to the Empirics of Economic GrowthThe Quarterly Journal of
Economics
Nazrul Islam (1995) Growth Empirics: A Panel Data ApproachThe Quarterly Journal of Economics
Nazrul Islam (1998) Growth Empirics: A Panel Data Approach -- A ReplyThe Quarterly Journal of Economics
OECD, OECD Economic Studies, No. 33, 2001/II. (2001) THE DRIVING FORCES OF ECONOMIC GROWTH:PANEL DATA EVIDENCE FOR
THE OECD COUNTRIESOECD
Quah D.T. (1996c) (1996) Convergence empirics across economies with (some) capital mobility, Journal of Economic Growth, 1:95-124
Rajasalu Teet (2003) Indicators of Economic Freedom and Economic Structure as Determinants of Growth and Convergence in Enlarging EU
and Priorities for Estonia
Raphael Bergoeinga, Norman Loayzab, Andrea Repetto (2004) Slow recoveriesJournal of Development Economics
Robert E. Hall and Charles I. Jones (1999) Why Do Some Countries Produce So Much More Output Per Worker Than Others?The Quarterly
Journal of Economics
Robert E. Lucas, Jr. (1990) Why Doesn't Capital Flow from Rlch to Poor Countries? American Economic Review
Robert J. Barro (1990) Government Spending in a Simple Model of Endogeneous GrowthThe Journal of Political Economy
Robert J. Barro (1990) Government Spending in a Simple Model of Endogeneous GrowthThe Journal of Political Economy
Robert M. Solow (1956) A Contribution to the Theory of Economic GrowthThe Quarterly Journal of Economics
Rodrik, D. 1999. (1999) "Where Did All the Growth Go? External Shocks, Social Conflict, and Growth Collapses."Journal of Economic Growth,
4:4, pp. 385-412
Rodrik, Dani and Rodriguez, F. b (1999) Trade Policy and Economic Growth: A Sceptic’s Guide to the Cross National Evidence,Q Centre for
Economic Policy Research Discussion Paper: 2143, May 1999.
References
Ross Levine and David Renelt (1992) A Sensitivity Analysis of Cross-Country Growth RegressionsThe American Economic Review
Sala-i-Martin, Xavier. (1996) “The Classical Approach to Convergence Analysis.” The Economic Journal 106 (1996
Sarah M. Lein Miguel A. Leo´n-Ledesma ; Carolin Nerlich : () How is real convergence driving nominal convergence in the new EU Member
States?
Sergio Rebelo (1997) On the determinants of economic growth
Sergio Rebelo (1991) Growth in Open EconomiesThe World Bank
Somesh K. Mathur (2005) Absolute and Conditional Convergence: Its Speed for Selected Countries for 1961--2001
Stanley Fischer 1993 (1993) The role of macroeconomic factors in growth, Journal of monetary economics, 32, 485:512, 1993
William J. Baumol (1986) Productivity Growth, Convergence, and Welfare: What the Long-Run Data ShowThe American Economic Review
Xavier Sala-i-Martin, Gernot Doppelhofer, Ronald I. Miller (2004) Determinants of Long-Term Growth: A Bayesian Averaging of Classical
Estimates (BACE) ApproachThe American Economic Review
Xavier X. Sala-I-Martin (1997) I Just Ran Two Million RegressionsThe American Economic Review
References
Thank you for your attention !