Transcript View Conference Presentation

Energy Intensity, Technology and
Institutions
Carmen Álvarez Arbelo
Gustavo A. Marrero
Francisco Javier Ramos Real
Dpto. Análisis Económico, University of La Laguna, Spain
33RD USAEE, Pittsburgh october 25-28, 2015
1
INDEX
1.- MOTIVATION AND AIMS
2.- DATA AND METHODOLOGY
3.- RESULTS, CONCLUSIONS AND FURTHER
RESEARCH
2
I. MOTIVATION AND AIMS

Improved Energy Efficiency (EE) is a critical responde to the pressing
climate change, economic development and energy security challenges
facing many countries (IEA, 2010)

Energy Intensity (EI) is the relationship between energy consumption and
real GDP. EI has commonly been used as an indicator of EE

EI Literature. Country and cross-country comparisons [Mulder and De Groot.
(2012)], decomposition analysis [Ang (2004), Metcalf (2008), Marrero and
Ramos-Real (2013)]…etc.

Main factors at the core of this challenge: technological change, GDP
structure by country and policy in a broad sense
3
I. MOTIVATION AND AIMS

Institutions have became a central subject in economic development.
Empirical analyses have found that progrowth institutions and governance
constitute preconditions to achieve TFP improvements.

A set of indicators reflecting a wide range of institutional features is required
to perform the analysis. (e.g. Beck, Demirgüç-Kunt and Levine, 2010;
Acemoglu, Gallego and Robinson, 2014; Acemoglu, Naidu, Restrepo and
Robinson, 2014).

IEA states that the impacts of market forces and public policy on EE have
been profusely studied, but much less in known about institutional factors
related to Energy Efficiency Governance (EEG)

IEA broader concept o EEG: is the combination of legislative frameworks
and funding mechanisms, institutional arrangements and co-ordination
mechanisms, which work together to support the implementation of EE
strategies, policies and programs
4
I. MOTIVATION AND AIMS

Impossibility to unequivocally define the mechanisms leading to good EEG.
The best way to evaluate EEG is not through the type of involved
mechanisms but through the inspection of outcomes

Researchers in the field of energy economics lack this type of indicators,
which has impeded a systematic empirical study on EE governance.
Available indicators only permit to analyze the linkage between general
institutions and EE.

BUT, successful EEG structure is expected to hinge upon country specific
features like energy import dependance, economy size or GDP structure.

CONTRIBUTION OF THIS RESEARCH:
QUANTIFY THE IMPORTANCE OF
INSTITUTIONS IN EXPLAINING ENERGY INTENSITY

PRELIMINARY EMPIRICAL WORK TO

INCLUSION OF CONTROL
SECTORAL STRUCTURE

PRELIMINARY STEP TO CONSTRUCT EEG
VARIABLES,
INCLUDING TECH. CHANGE AND
COMPOSITE INDICATOR
5
II. DATA AND METHODOLOGY

Dynamic Panel Data Model (DPD) used to analyze the relationship between
institutions and EI at the worldwide level (first step).

Based on widely used macro dynamic regressions (e.g. Forbes, 2000; Metcalf,
2008; Marrero, 2010), we propose and estimate a dynamic panel model (DPD).
We follow Acemoglu et al. (2015, 2015) to present a reduced-form energy model:
EIit  i   'Tt   EIit s   Instit s   ' X it   it
(1)
EIit is final energy intensity of country i at period t (Indust, transp, residential, servic, agriculture)
The lagged terms included the right hand side (5 years):
- A set of fixed, time invariant effects (alphai) and time dummies (TT)
- EIIt-s to control for initial pollluting technology and conditional convergence
- Time and country specific controls (Xit) affecting heterogeneity in EI (TFP, human
capital level , sectoral structure…)
- Preliminary proxy (Inst) of general economic and policy institutions (Polity2
variable) using
indicators of democracy and autocracy, authority characteristics
and polity regime transition (potentially correlated with more specific energy
institutions).
6
II. DATA AND METHODOLOGY

ESTIMATION PROCEDURE. The importance of considering an appropriate
estimation approach when estimating a Dynamic Panel Data Model (DPD)

Traditional OLS-POOL or Within estimator (WG). Endogeneity problem:
biased estimates because the explicative variables and the residual are
correlated (Hsiao, 1986)

Solution: Remove the fixed effect an apply instrumental variables (external
instrument). 2SLS estimator. Anderson and Hsiao (1982). Very difficult to find
with macro-variables

Alternative. Using lagged variables (internal instrument) GMM-DIF. Arellano
and Bond (1991). Bias problem. Weak instrument problem

Arellano and Bover (1995). System of equations (GMM-SYS) using first
differences and level equations to find an additional set of feasible
instruments
7
II. DATA AND METHODOLOGY

DATA. Unbalanced panel of 731 observations taking intervals of every 5
years, 1960-2005 for 116 countries.

Energy data from IEA database; population and GDP (real and PPP
adjusted) form the PWT (7.1). TFP and human capital index from PWT
(8.0). Sectoral composition data from WB (developing indicators database)
and Democracy is taken from the Polity4 database (-10/+10)

No perfect methodology. Results may depend on the set of controls and the
econometric approach considered

Strategy: A certain number of versions of (1) are estimated to test the
robustness of our results. A constant term and time dummies are always
included.
8
III. RESULTS
Dependent variables: Energy Intensity (final energy consumptio / GDP-ppp adjusted)
2SLS instrumental variables
GMM-SYS
ln(EI), lag
ln(GDP_pc), lag
Polity2, lag
0.857***
(12.96)
-0.0132
(-0.59)
-0.209***
(-2.65)
Agric/GDP, lag
Indust/GDP, lag
Num. obs
Hansen (p-val)
m1-test (p-val)
m2-test (p-val)
Cross section
Num. instruments
678
0.294
0.00571
0.305
116
57
0.919***
(16.49)
0.0938***
(2.93)
-0.412***
(-3.12)
0.00539*
(1.88)
-0.00681*
(-1.81)
494
0.649
0.103
0.936
108
76
0.886***
(23.51)
-0.00883
(-0.45)
-0.134***
(-2.88)
678
0.247
0.0213
0.310
116
57
0.937***
(20.27)
0.0439
(1.12)
-0.276**
(-2.48)
0.00207
(0.71)
-0.00595*
(-1.66)
494
0.375
0.126
0.979
108
74
9
III. RESULTS
Dependent variables: Energy Intensity (final energy consumptio / GDP-ppp adjusted)
2SLS instrumental variables
GMM-SYS
ln(EI), lag
ln(GDP_pc), lag
Polity2, lag
Agric/GDP, lag
Indust/GDP, lag
Human capital, lag
0.941***
(17.88)
0.106**
(2.27)
-0.257***
(-3.15)
0.00587*
(1.89)
-0.00622
(-1.64)
-0.0602
(-0.80)
0.929***
(19.89)
0.0715*
(1.71)
-0.167*
(-1.84)
0.00251
(0.88)
-0.00428
(-1.35)
-0.103
(-1.55)
-0.152***
(-2.87)
TFP growth
Num. obs
Hansen (p-val)
m1-test (p-val)
m2-test (p-val)
Cross section
Num. instruments
0.897***
(23.44)
0.0264
(0.78)
-0.246***
(-3.24)
0.000459
(0.15)
-0.00466*
(-1.65)
461
0.739
0.00649
0.930
99
88
391
0.741
0.0000266
0.0921
85
88
0.908***
(30.56)
0.0168
(0.50)
-0.155*
(-1.91)
0.00156
(0.50)
-0.00306
(-1.14)
-0.0901**
(-2.42)
461
0.543
0.00499
0.939
99
86
391
0.758
0.0000339
0.0812
85
86
10
III. CONCLUSIONS

The most important result is that we obtain robust result that the lagged
degree of democracy is beneficial to reduce EI. More quality of institutions
implies lower levels of EI.

If we control by sectoral composition (which implies that EI is a better of
Energy Eficiency), results are even more significant and higher impact

Human capital (proxy of technical adoption) parameter is negative (as
expected). The lagged human capital (proxy of technical progress) shows a
negative coefficient but is not significant.

TFP growth (proxy of Tech. Progress) is significant and with the expected
negative coefficient for most of the specifications

And, more important, results Institutions effect remains unchanged
11
III. FURTHER RESEARCH
CONSTRUCTION OF EE GOVERNANCE INDICATORS



Needed for a systematic empirical study on EE governance
Background: general governance and institutional indicators-development
economics (e.g. World Governance Index; World Governance Indicators;
Polity IV Project)
Methodology: Handbook on Constructing Composite Indicators (OECD)
Indicators based on EE governance pillars (IEA, 2010)
Pillar
Dimensions
Enabling frameworks
Law, funding, plans
Institutional arrangements
Agencies, resources; providers; co-operation;
international assistance
Co-ordination mechanism
Co-ordination; targets; evaluation
Data sources: WEC (EE Policies and Measures Database); IEA
(Policies and Measures Database); Limaye, Heffner and Sarkar
(2008); IEA (2010)
12
THANK YOU VERY MUCH FOR YOUR ATTENTION
Francisco Javier Ramos Real
Contact for specific questions: [email protected]
13