Soaring economic growth for the sake of environmental degradation? - - PowerPoint PPT Presentation

Soaring economic growth for the sake of environmental degradation? Evidence using second-generation panel methods

Lars Sorge and Anne Neumann Vienna, 05 September 2017

Lars Sorge and Anne Neumann Vienna, 05 September 2017 1 / 27

Motivation

’[...] there is clear evidence, that [...], in the end the best - and probably the only - way to attain a decent environment in most countries is to become rich.’

(Beckerman, 1992)

Lars Sorge and Anne Neumann Vienna, 05 September 2017 2 / 27

Contribution

Inclusion of CO2 emissions in EKC framework Application of recently developed nonstationary panel data methods Consideration of international trade Transparent documentation of data and method

Lars Sorge and Anne Neumann Vienna, 05 September 2017 3 / 27

Agenda

Agenda

1 Literature review 2 Data and empirical strategy

Empirical Specification Data Cross-section dependence test Second-generation panel unit root test Second-generation panel cointegration test Panel long-run estimation Panel causality test

3 Results 4 Conclusions, policy implications, further research Lars Sorge and Anne Neumann Vienna, 05 September 2017 4 / 27

Literature Review

Development of the literature

Threshold income (Environmental Kuznets Curve, EKC) Grossmann and Kruger, 1991 (QJE): ’We find no evidence that environmental quality deteriorates steadily with economic growth’ controversial evidence Causality energy consumption - economic growth higher development requires more energy consumption; more efficient energy use requires more economic development (Kraft and Kraft, 1978) conflicting evidence Energy consumption - economic growth - environmental pollutants impact of energy consumption and economic growth on CO2 emissions (and considering trade)

Lars Sorge and Anne Neumann Vienna, 05 September 2017 5 / 27

Empirical Specification

Empirical Specification Model 1 CO2it = αi + β1iGDPit + β2iGDP2

it + β3iEit + ǫit

CO2: CO2 emissions per capita in metric tons per capita GDP: GDP per capita in constant 2010 USD E: energy consumption in kg of oil equivalents

Lars Sorge and Anne Neumann Vienna, 05 September 2017 6 / 27

Empirical Specification

Empirical Specification Model 2 CO2it = αi + β1iGDPit + β2iGDP2

it + β3iEit + β4iTit + ǫit

T: trade-openness as sum of exports and imports of goods and services measured as share of gross domestic product

Lars Sorge and Anne Neumann Vienna, 05 September 2017 7 / 27

Data

Data

Summary statistics by panel World Bank Development Indicator database WTO countries (N = 70), 1971 - 2013 (t = 43) clustered in three panels (Nhigh = 29, Nmiddle = 17, Nlower = 24)

Lars Sorge and Anne Neumann Vienna, 05 September 2017 8 / 27

Data

CO2, GDP, energy consumption, trade-openness (in logs)

Lars Sorge and Anne Neumann Vienna, 05 September 2017 9 / 27

Empirical Strategy

Empirical strategy

1 detect contemporaneous correlation among countries after controlling

for individual characteristics (i.e. global shocks, local interactions)

2 test for unit roots in the presence of cross-section dependence from a

single common factor

3 check for cointegration 4 dynamic long-run estimation of non-stationary cointegrated panels

and calculation of turning point income

5 identify causal relationship among the variables Lars Sorge and Anne Neumann Vienna, 05 September 2017 10 / 27

Empirical Strategy

Cross-section dependence test

contemporaneous correlation among countries that is left over after controlling for individual characteristics (Moscone and Tosetti, 2009) first-generation panel methods assume cross-sectional independence Pesaran (2004) CD test is robust to the presence of

nonstationary processes, parameter heterogeneity or structural breaks, ... and perfoms well in small samples.

Lars Sorge and Anne Neumann Vienna, 05 September 2017 11 / 27

Empirical Strategy

Second-generation panel unit root test

using nonstationary variables can lead to apparently significant regression results although the data is unrelated Pesaran (2007) CIPS panel unit root test

Cross-sectionally augmented Im-Pesaran-Shin (2003) (IPS) test ∆yit = δ‘

idt + ρiyi,t−1 + ciy t−1 + J

• j=0

dij∆y t−j +

J

• j=1

βij∆yi,t−j + ǫit H0 : ρi = 0 is tested against H1 : ρi < 0 and H1 : ρi = 0

Lars Sorge and Anne Neumann Vienna, 05 September 2017 12 / 27

Empirical Strategy

Second-generation panel cointegration test

Westerlund (2007) error-correction-based panel cointegration test Based on structural rather than residual dynamics ∆yit = δ‘

idt+αi(yi,t−1−β‘ iXi,t−1)+ Ji

• j=1

αij∆yi,t−j +

Ji

• j=0

γij∆Xi,t−j +ǫit H0 : αi = 0: no error-correction implies no cointegration H1 : αi = α < 0 for panel statistics Pτ and Pα H1 : αi < 0 for group mean statistics Gτ and Gα Bootstrapped critical values to account for cross-sectional dependence

Lars Sorge and Anne Neumann Vienna, 05 September 2017 13 / 27

Empirical Strategy

Panel long-run estimation

validation of the EKC hypothesis and turning point income determination of the corresponding long-run elasticities Pedroni (2001) dynamic OLS (DOLS) estimator

applicable to non-stationary cointegrated panels yit = αi + βiXit +

Ji

• j=−ji

γij∆Xi,t−j + ǫit slope coefficients βi are permitted to vary across panel members DOLS estimation technique adds lags and leads of differenced explanatory variables to control for endogeneity consistent and efficient estimates of the long-run relationship

Lars Sorge and Anne Neumann Vienna, 05 September 2017 14 / 27

Empirical Strategy

Panel causality test

Pesaran et al. (1999) pooled mean group (PMG) estimator ∆CO2it = α1i +

J

• j=1

β11ij∆CO2i,t−j +

J

• j=1

β12ij∆Ei,t−j +

J

• j=1

β13ij∆GDPi,t−j +

J

• j=1

β14ij∆GDP2

i,t−j + λ1iECTi,t−1 + ǫ1it

i = countries, t = years, ∆ = first difference operator, j = lags, λki = speed of adjustment ECTt−1 = CO2t−1 − β1GDPt−1 − β2GDP2

t−1 − β3Et−1

Short-run causality: significance of the lagged dynamic terms Long-run causality: significance of the ECT

Lars Sorge and Anne Neumann Vienna, 05 September 2017 15 / 27

Empirical Results

Empirical Results I

Peasaran (2004) CD test all series are highly dependent across all income groups Pesaran (2007) panel unit root test: strong evidence that a panel unit root in the variables exists Westerlund (2007) panel cointegration tests: test statistics with smallest size distortios when cross-sectional dependence is present indicate a long-run equilibrium relationship Implications First generation panel data methods are inappropriate Non-stationary variables that are found to be cointegrated are involved in EKC regression

Lars Sorge and Anne Neumann Vienna, 05 September 2017 16 / 27

Empirical Results

Empirical Results II: Pedroni (2001) DOLS estimates

Empirical evidence of an inverted U-shape pattern

Lars Sorge and Anne Neumann Vienna, 05 September 2017 17 / 27

Empirical Results

Empirical Results III: Summary

Lars Sorge and Anne Neumann Vienna, 05 September 2017 18 / 27

Conclusion

Findings

We find cross-section dependence within all panels. All series are I(1). We find evidence for cointegration. There is empirical support of an inverted U-shape pattern within all panels for model 1 and model 2. positive coefficient for energy consumption negative relation between trade-openness and per capita CO2 emissions only for high- and middle-income panel

Lars Sorge and Anne Neumann Vienna, 05 September 2017 19 / 27

Conclusion

Conclusion

Economic growth accompanied by environmental degradation in early stages of economic development. Policies supporting sustainable economic growth improve environmental quality. The simple existence of the EKC relationship does not guarantee that CO2-emissions will decrease automatically with economic growth. for high-income panel results support conversion hypothesis (economic growth leads energy consumption) for middle- and lower-income panels support neutrality hypothesis (no causal relationship between income and energy consumption) → a reduction in energy consumption might not affect economic growth negatively

Lars Sorge and Anne Neumann Vienna, 05 September 2017 20 / 27

Conclusion

Thank you.

aneumann[at]diw.de

Lars Sorge and Anne Neumann Vienna, 05 September 2017 21 / 27

Conclusion

Causality analysis middle-income panel

Lars Sorge and Anne Neumann Vienna, 05 September 2017 22 / 27

Conclusion

Causality analysis lower-income panel

Lars Sorge and Anne Neumann Vienna, 05 September 2017 23 / 27

Conclusion

Results Pesaran (2004) CD test

Lars Sorge and Anne Neumann Vienna, 05 September 2017 24 / 27

Conclusion

Results Pesaran (2007) unit root test

Lars Sorge and Anne Neumann Vienna, 05 September 2017 25 / 27

Conclusion

Results Westerlund (2007) panel cointegration test

Lars Sorge and Anne Neumann Vienna, 05 September 2017 26 / 27

Conclusion

Empirical Results II: Causality analysis high-income panel

unidirectional causality CO2 → E and T → E bidriectional causality CO2 ↔ T and GDP ↔ T

Lars Sorge and Anne Neumann Vienna, 05 September 2017 27 / 27