R. L. Bruno, E. Douarin, J. Korosteleva and S. Radosevic
Prepared for the Transition Economies Meets New Structural Economics
SSEES, UCL June 25-26, 2013
• Key objectives.
• Theoretical considerations.
– Development Strategies and Technology Choice: the New
Structural Economics (NSE) paradigm
• Data, Technology Choice Index (TCI) construction and methodology.
• Financial sector distortions and TCI.
• Hypotheses Testing.
– Empirical results I: TCI and growth.
– Empirical results II: Financial sector inefficiencies, TCI, and growth.
• Conclusions.
2
• We test the basic propositions of NSE, i.e. the relationship between TCI and growth:
– We employ a larger sample of countries (164) over a longer time span (1963-2009).
• We further expand this theory to transition economics (TE) by testing whether key propositions of NSE hold for the TE group, and its sub-groups, namely Former Soviet Union (FSU) and Central
European Economies (CEE).
Are transition economies special?
• We examine the association between financial sector distortions and TCI with further implications for growth.
3
• Key propositions of NSE:
– Long term growth is feasible only in contexts where policies and institutions are conducive to the development of sectors consistent with the Comparative Advantage of a country: CAF strategy (following).
– TCI is a valid proxy to capture whether a country follow a CAD strategy, comparative advantage defying (visà-vis a CAF strategy):
• Lin (2012): sample of 122 countries, time period including 1962-1999
(see also Lin 2003)
• On average, TCI is significantly negatively correlated with long-term growth measured by GDP pc.
– TCI is expected to be highly correlated to financial distortions: the higher the ‘structural’ distortion the higher the financial one.
4
5
• What is TCI capturing?
– Higher TCI = more distortion, proxy for CAD strategy
– A context in which policies and institutions favour a capitalintensive manufacturing sector (raising value added in manufacturing AVM, and lowering labour LM)
– This compares with other approaches in the literature:
• Lewis 's two sectors development model (1954) (role of unlimited supply of labour)
• Rostow ’s stages of economic growth (1991) (V: from precondition to take off
…to beyond mass consumption)
• “Big push” idea,
Rosenstein-Rodan (1943) (Eastern and S.E. Europe)
– Key point: the shift from low productivity primary production to higher productivity manufacturing is a pivotal stage in the development process of an economy. This is not always successful, though.
6
• On an intuitive level, one would expect the countries of CEE and FSU to be primary example of the negative relationship between CAD and growth.
• During central planning: heavily distorted economies with strong emphasis on capital intensive manufacturing: surely a CAD!
• With transition, a progressive move towards more liberal market economies (with a lot of variations across countries though): surely a move towards a
CAF!
7
Albania Armenia Azerbaijan Belarus Bosnia and Herzegovina
Bulgaria
Hungary
Macedonia
Croatia
Kazakhstan
Moldova
Czech Republic
Kyrgyzstan
Poland
Estonia
Latvia
Georgia
Lithuania
Romania Russian Federation
Serbia Slovakia Slovenia Tajikistan Ukraine
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 year_num
2020 1960 1980 2000 2020 1960 1980 2000 2020
Graphs by Country 8
• The CAD and CAF strategy cannot be assessed in an institutional vacuum (Lin et al. 2011).
• A country adopting a CAD strategy will require economic distortions introduced through substantial government interventions on the economy
(“forced”).
• The development of financial structure is argued to be endogenous to the government's growth strategy with a CAD strategy being associated with a financial structure deviating farther away from its estimated optimal structure (Lin and Xu, 2012).
9
• Size vs. efficiency
• Financial structure distortions ( size )
Size 1=
Size 2=
• Financial structure gap
(Demirguc-Kunt et al. 2011)
• Estimate financial structure ratio based on sample of
OECD countries.
• Calculate sample-wide country-year residuals based on the estimated regressions, and take the logarithm.
10
• Banking Inefficiency
– the net interest margin is equal to the value of a bank’s net interest revenue as a share of its total earning assets.
– overhead cost is the value of a bank’s overhead costs as share of its total assets.
Higher level of interest margins and overhead costs indicate lower levels of banking efficiency (lower competition).
11
Hypotheses Type of relevance of
NSE propositions
Time relevance • TCI is especially relevant prior to the 1980s but not for later periods in view of decrease incidence of autonomous development strategies
• After 1980s we should observe lower level of TCI as effects of CAF strategies and thus the relationship between TCI and growth should be weaker
Relevance for transition economies
• Transition economies differ from the rest of the sample : a negative relationship between TCI and growth is less likely for them, but we expect some differences across CEE and FSU with the TCI in the former being positively related to growth, whereas TCI in the latter being negatively associated with growth (think at the comparatively more advanced manufacturing sector of CEE vis-a-vis FSU)
Financial distortions • Higher level of TCI is associated with a financial structure deviating farther away from its estimated optimal level , but this effect differs across group of countries.
• The financial sector inefficiencies should be strongly correlated to TCI in highly distorted economies .
• The link between TCI and financial distortions is less pronounced in transition countries , being more robust in FSU countries and less so in Central and East European economies.
• The effect of TCI on growth is further reinforced via larger values of deviation of the optimal financial structure from its actual one.
12
Dependent variable: growth rate of the GDP pc (constant 2000 US$)
Ln TCI2
(1) (2) (3) (4)
Ln_gdp_pc_(start decade)
-0.007***
(0.002)
-0.003***
0.001
-0.008***
(0.002)
-0.003***
0.001
-0.008***
(0.002)
-0.002**
0.001
0.004
(0.003)
-0.040***
0.004
Ln_Population Total _(start decade) ln_axrateLCUperUS
0.001**
(0.001)
0.001***
(0.000)
0.001**
(0.001)
0.001***
(0.000)
0.013*
(0.007)
-0.003
(0.003)
0.001**
(0.001)
0.001***
(0.000)
-0.025***
(0.006)
-.002***
(.001)
Distance to equator
Landlocked
0.015**
(0.007)
-0.003
(0.003)
-
0.011*
(0.007)
-0.003
(0.003)
-
-
Ln TCI2_x_ TE 0.015**
(0.007)
-
-
Ln TCI2_x_ CEE
Ln TCI2_x_ FSU -
-
-
0.017**
(0.007)
-.147*
(0.083)
0.031***
0.032**
(0.014)
-.162**
(0.07)
.566*** Constant 0.035*** 0.034***
Time fixed effects (decade)
Country fixed effects
Observations
(0.013)
Yes
No
439
(0.013)
Yes
No
439
(0.012)
Yes
No
439
(0.088)
Yes
Yes
439
F-st.
15.84*** 15.11***
Source: World Bank Financial Structure Dataset (2012), WB WDI 2012 edition; UNIDO
14.14*** 10.90***
Note: *,**,*** denote significance on the 10, 5 and 1-percent level, respectively. Standard errors reported in parentheses.
13
• Unexpectedly low level of TCI before and at the onset of transition: low level of distortions?
• Positive relationship between growth and TCI for this sub-sample of countries, when TE are pooled together.
• Quite puzzling!
• Variations in the relationship between TCI and growth across CEE and FSU, though.
14
Benin Bolivia Brazil Burkina Faso Burundi Cameroon
Côte d'Ivoire Central African Republic China Ecuador El Salvador Eritrea
Ethiopia
Malawi
Gambia
Nicaragua
Ghana
Niger
Iran
Nigeria
Kyrgyzstan
Papua New Guinea
Lesotho
Paraguay
Peru Rwanda Senegal Somalia Suriname Syria
1960 1980 2000 2020 1960 1980 2000 2020
Zambia Tanzania Thailand Uganda
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 year
1980 2000 2020
Graphs by Country_NUM
15
Table 2: Estimating the effect of TCI2 on growth. Highly-distorted countries: robust regression results
(1) Dependent variable: growth rate of the GDP pc (constant
2000 US$)
Ln TCI2
Ln_gdp_pc_(start decade)
Ln_Population Total _(start decade) ln_xrateLCUperUS
Distance to equator
Landlocked
Ln TCI2_x_ HDD
Constant
Decade time fixed effects
Observations
F-st.
-0.004
(0.002)
-0.001**
0.001
0.001
(0.001)
0.003***
(0.000)
0.016**
(0.007)
-0.001
(0.003)
-0.004**
(0.002)
0.036***
(0.013)
Yes
439
14.75***
(2) Excluding potential outliers
-0.004
(0.003)
-0.001***
0.001
0.001
(0.001)
0.001***
(0.000)
0.015**
(0.007)
-0.002
(0.003)
-0.004*
(0.002)
0.036***
(0.013)
Yes
429
14.15***
Source: World Bank Financial Structure Dataset (2012), WB WDI 2012 edition; UNIDO
Note: *,**,*** denote significance on the 10, 5 and 1-percent level, respectively. Standard errors reported in parentheses.
16
• Highly distorted economies seem to show a behaviour fully predicted by the NSE.
• Negative relationship between growth and TCI for this sub-sample of countries, when are pooled together.
• Variations in the relationship between TCI and growth across CEE and FSU, HDD.
• A non-monotonous relationship?
17
• The relationship between TCI and financial structure gap differs across countries depending on the level of
TCI .
• Our results suggest that deviation from optimal financial structure is only significant for relatively high values of TCI2 (75 th quantile) for our sample of countries as a whole, and for the group of FSU countries specifically .
• The positive relationship between TCI2 and distorted financial structure is strongly and positively significant for the group of Highly-distorted economies regardless of the range of values for TCI2.
18
R-sq 50th R-sq 75th Rsq Quantile/ Dependent variable: TCI2
25th Obs
Whole sample
Size1
Size2
-0.02**
(0.009)
-0.023***
0.08
0.08
-0.003
(0.011)
-0.007
(0.01)
.20
.20
0.018**
(0.01)
0.015*
(0.008)
0.31
0.31
1233
1229
Transition economies
Size1
Size2
0.04
(0.04)
0.04
(0.05)
CEE
Size1
Size2
-0.01
(0.027)
0.001
(0.05)
FSU
Size1 0.035
(0.05)
0.001
Size2
(.5)
Highly-distorted economies
Size1 .20***
(0.068)
0.08
0.08
0.003
0.002
0.22
0.22
-0.01
(0.03)
-0.01
(0.03)
-0.026
(0.039)
0.014
(0.05)
00.07
(0.055)
0.029
(0.064)
0.03
.20
0.03
0.02
0.31
0.30
0.01
(0.017)
0.018
(0.02)
0.006
(0.034)
-0.01
(0.03)
0.16***
(0.04)
0.09
(0.06)
0.27
0.27
0.12
.10
0.51
.48
200
198
159
157
41
41
0.09
0.25***
(0.07)
0.19
0.16***
(0.03)
0.22
124
Size2 .19** 0.06
.18** .17
0.14*** .22
124
(0.07) (0.03)
19 regression controls for the level of economic development proxied by ln of GDP pc at const 2000 US dollars. For reproducibility of the results we run 2000 replications with seed set at 1001.
• There is positive association between bank inefficiency and high values of TCI in the whole sample.
• Decrease in net interest margin and respectively increase in the degree of competition in banking sector has positive effect in Central and Eastern European countries for reducing TCI, we fail to find any significant results for FSU economies, though.
• Similarly, to financial structure results we get robust results for the group of Highly-distorted economies, suggesting that reduction in net interest margin, and consequently increase in bank efficiency will facilitate a move towards CAF strategy.
20
Table 4: TCI2 and Bank Inefficiency measures: Simultaneous Quantile
Regression results
Quantile/
Dependent variable: TCI2
25th R-sq 50th R-sq 75th Rsq Obs
Whole economy
Overhead costs
NetIntMargin
Transition economies
Overhead costs 0.07
NetIntMargin
(.1)
0.01
(.11)
CEE
Overhead costs
-.13***
(0.02)
-0.03
(0.04)
NetIntMargin
0.07
(0.08)
.18**
(0.08)
FSU
Overhead costs
NetIntmargin
-.43*
(.233)
-.27**
(.108)
Highly-distorted economies
Overhead costs -0.02
(0.06)
.1
0.08
.14
.13
0.01
0.02
.21
.28
-0.08***
(0.03)
0.01
(0.03)
.24***
(0.058)
0.005
(0.074)
.26***
(0.07)
.23***
(0.07)
-.41
(.33)
-.33**
(.13)
.23
.21
0.03
.24
0.09
0.09
.22
.27
0.000
(0.03)
0.09***
(0.032)
0.19
(0.085)
0.01
(0.075)
0.23***
(0.06)
0.23***
(0.065)
-.44
(.32)
.-25
(.20)
0.35
.34
0.31
.30
.16
.13
.28
.23
1246
1269
233
238
172
170
172
68
.19
0.06**
(0.024)
.28
0.03
(0.06)
.27
160
NetIntMargin 0.32*** .21
0.35*** .32
0.28** .29
162
(0.09) (0.06)
Source: World Bank Financial Structure Dataset (2012), WB WDI 2012 edition; UNIDO .
(.12)
21 regression controls for the level of economic development proxied by ln of GDP pc at const 2000 US dollars. For reproducibility of the results we run 2000 replications with seed set at 1001.
• Larger deviations in actual financial structure from its optimal one further reinforce a negative effect of TCI on growth.
22
Explanatory variables
Ln_gdp_pc_start
Ln TCI
FinStr gap
Ln TCI_x_FinStr_gap
Landlocked
Legal origin UK
Ln Population Size
Natural Resources Exports
Private Credit to GDP
Number obs.
F st.
Pr>z AR(1) / Pr>z AR(2)
Hansen test of overid. restrictions, Chi2
(Pr.>chi2)
Coefficient
-0.014**
(0.06)
-0.019**
(0.009)
-0.001
(0.004)
-0.005**
(0.002)
0.003
(0.006)
-0.016
(0.010)
0.007*
(0.002)
0.0001
(0.0001)
0.008
(0.007)
171
14.68
00.016/0.666
.294
Note: SYS GMM regression results
Dependent variable: growth averaged over 5-year non-overlapping periods of time.
23
• Our analysis confirms the negative relationship between TCI and growth found by Lin (2012).
• These results also hold for the group of highly distorted economies .
• However, we find that the NSE propositions cannot be generalized to the overall group of transition economies for which the relationship is positive.
• This positive relationship is due to two different subgroups. For CEE, the relationship between TCI2 and growth is positive while for the FSU is negative.
24
• We also confirm that the positive relationship between larger financial sector distortions and Technology Choice
Index, although this relationship is not homogeneous, and it differs across different groups of countries depending on the aspect of financial structure or financial sector distortions we look at (size vs. efficiency)
• While controlling for potential endogeneity between economic growth, TCI and financial sector distortions our study also reveals that the negative effect of TCI on growth is further reinforced by larger deviation in the actual financial structure from its estimated optimal level.
25
• The Technology Choice index is very important element of the development strategy of a country
• Also the Financial sector Distortion (size and/or efficiency) is a key ingredient for a sustainable development strategy.
26
• We discussed the relevance/adequacy of TCI as an indicator of CAD/CAF strategy and/or distortion.
• In fact, the relationship between TCI/FinDist and growth seems to be characterised by:
1. Non-monotonicity
2.
Groups’ specific effects (CEE, FSU, HDD)
We are developing our research in this direction by constructing a simultaneous estimation of the
TCI=f(FinDist)
Growth=f(TCI) =>
Growth=f(TCI(FinDist))
We are aware of important challenges for endogeneity and selection.
27
28