Interregional Diffusion in the Analytic Climate Econom

Report 2022/24

Andreas Stranden Hoel-Holt

Interregional Diffusion in the Analytic Climate Econom

Category	Reports
Sub-Categories	Modeller og databaserClimate change and the green transformation
Year	2022
Report Number	24
Author(s)	Andreas Stranden Hoel-Holt
	https://vista-analyse.no/no/publikasjoner/interregional-diffusion-in-the-analytic-climate-economy/
Download	file_download (3.8 MB)
Read in browser	PDF

Rapport 2022/	24 | Masteropp	gave

Interregional Diffusion in the Analytic	 Climate Economy

Andreas Stranden Hoel	-Holt

Interregional Diffusion in the Analytic	 Climate Economy

Vista Analyse	 | Rapport 2022/24

Dokumentdetaljer
Tittel	 	Interregional Diffusion in the Analytic	 Climate Economy
Rapportnummer 	 	Rapport 202	2/24
Forfattere	 	Andreas Stranden Hoel	-Holt
ISBN	 	978	-82	-8126	-582	-0
Prosjektleder	 	Orvika Rosnes
Dato for ferdigstilling	 	Mai 2022
Tilgjengelighet	 	Offentlig
Nøkkelord	 	Klima og det grønne skiftet, modeller og databaser

Om Vista Analyse
Vista Analyse AS er et samfunnsfaglig 	analyseselskap med hovedvekt på økonomisk utredning, evaluering, rådgivning
og forskning. Vi utfører oppdrag med høy faglig kvalitet, uavhengighet og integritet. Våre sentrale temaområder er
klima, energi, samferdsel, næringsutvikling, byutvikling og velfe	rd. Vista Analyse er vinner av Evalueringsprisen 2018.
Våre medarbeidere har meget høy akademisk kompetanse og bred erfaring innenfor konsulentvirksomhet. Ved be-
hov benytter vi et velutviklet nettverk med selskaper og ressurspersoner nasjonalt og internasj	onalt. Selskapet er i
sin helhet eiet av medarbeiderne.

Interregional Diffusion in the Analytic
Climate Economy
Andreas Stranden Hoel-Holt
Thesis submitted for the degree of
Master in Economic Theory and Econometrics
30 credits
Department of Economics
Faculty of Social Sciences
UNIVERSITY OF OSLO
Spring 2022

Interregional Di↵usion in the
Analytic Climate Economy
Andreas Stranden Hoel-Holt

©	2022 Andreas Stranden Hoel-Holt
Supervisor: Christian Traeger
Interregional Di↵usion in the Analytic Climate Economy
http://www.duo.uio.no/
Printed: Reprosentralen, University of Oslo

Acknowledgments
Writing this sentence means a lot to me. It symbolizes that I have almost completed my thesis,
thus wrapping up one of the more stimulating periods of my life so far. In navigating through
this process, I have too often felt like the character Clark Griswold in the classic movie Christmas
Vacation, trying to identify the broken bulb after decorating the entire house with lights and
reindeer, only for his wife to discover that a silly mistake caused the whole issue. That is exactly
what it feels like to calibrate a model.
First, I must thank my genius supervisor Christian Traeger for giving me the opportunity to
work closely with him on current research. You have been like a mentor for me the last couple
of years, and I have learned a lot from you. I greatly appreciate all your help and support.
I am happy for all the new friends I have made while studying at the University of Oslo. You
have proved (by contradiction) that there indeed is increasing results to scale in the production
of knowledge. I would also like to thank the participants of the very ﬁrst Master Forum meeting
at Blindern for helping my formulate my research question. A special shout out goes to Tor Odin
who has been a great moral support even though he had his ﬁrst-born child while writing his
own thesis, and to Oliver who lent me his Acemoglu book. I might keep it.
I am grateful to the INTPART programme for the scholarship they have given me. The same
goes for the good people at Vista Analyse for providing me with a scholarship, lunch, an o�ce
space and support through long days (and nights). A special thanks to Orvika Rosnes for her
tips and advice and to Insa for helpful guidance and explanations of IEA data.
Like any good story, I am saving the best for last. My girlfriend, wife, partner and baby-mama
Tina. Thank you for pushing me to go after my goals and dreams, even if it means changing
careers in my 30s and living o↵of L˚anekassen. You inspire me to be my best and I know for
a fact that I could not have completed this thesis without your help and support. Thank you.
EDSM.
i

Abstract
In this thesis, I calibrate a North-South multi-sector integrated assessment model with ex-
plicit energy production and analyze the consequences of technology di↵usion between regions.
Using data from public sources, I sort countries into two regions based on income and degree
of electriﬁcation. Then I calibrate ﬁve separate energy sectors and electricity that goes into the
production of three ﬁnal good sectors for each region. I introduce a simple exogenous inter-
regional di↵usion mechanism to the base model. Then I apply this model to analyze the e↵ects
of inter-regional di↵usion on the welfare in each region, emissions and temperature increase, the
composition of energy use within the economies.
I ﬁnd the welfare e↵ect of di↵usion to be between 40 billion and 370 billion US dollars for
the richer region. These e↵ects are only indirect, through lower global temperature and lower
climate-related damages. The low-income region has a welfare e↵ect between 190 billion and
1800 billion US dollars. The energy composition and emissions of the richer region are not
changing in di↵usion, so the welfare e↵ects on the low-income region are strictly local and direct.
These results depend crucially on the potential for inter-fuel substitution within the production
electricity, represented by the elasticity of substitution.
This thesis contributes to the literature on speciﬁcally analytic types of integrated assessment
models. Analytical models are easily interpretable and permit a large state-space with fast
solution mechanism, which enables quick computation of model results. By calibrating the
present model, I look to narrow the gap between the simplicity and transparency of analytical
models, and the accuracy and richness of more complex numerical models. I hope that this work
lays a foundation for further explorations of the possibilities within numerical use of analytical
models.

Contents
1 Introduction	1
2 Background	4
2.1 Integrated assessment models	.............................	4
2.2 Numerical IAMs analyzing the energy transition in developing countries	.....	5
2.3 Technological di↵usion models	.............................	6
2.4 Analytical IAMs	....................................	7
3 The Model	9
3.1 The Economy	......................................	9
3.2 The Climate System	..................................	12
3.3 Technological Di↵usion	.................................	12
3.4 Sketch of model solution	................................	13
4 Calibration and Data	15
4.1 Calibration procedure	.................................	15
4.2 Data sources	......................................	15
4.2.1 Region classiﬁcation	..............................	16
4.2.2 Volume data	..................................	16
4.2.3 Energy prices	..................................	19
4.2.4 Emissions and Carbon Prices	.........................	20
4.3 On Elasticities of Substitution	.............................	21
5Results	24
5.1 Calibration results	...................................	24
5.2 Model results	......................................	25
5.2.1 Baseline and optimal tax	...........................	25
5.2.2 Di↵usion	....................................	26
5.2.3 Sensitivity analysis	...............................	35
6 Conclusion	38
A Mathematical derivations	45
A.1 Final goods aggregation	................................	45
A.2 Final consumption good production	.........................	46
A.3 Intermediate energy and electricity producers	....................	47
A.4 Reﬁned energy production	...............................	48
ii

B A more general di↵usion model	50
C Supplementary ﬁgures	52
D Supplementary tables	55
E Wrangling script	58
F Calibration script	91
iii

List of Figures
5.1 Output, emissions and temperature.	.........................	25
5.2 Energy by source	....................................	26
5.3 Energy by sector in region	L	.............................	27
5.4 Energy by sector in region	H	.............................	28
5.5 Output, emissions and temperature – with growth in renewable energy.	.....	28
5.6 Energy by source (w/ growth)	.............................	29
5.7 Energy by sector in region	L	(w/ growth)	......................	30
5.8 Energy by sector in region	H	(w/ growth)	......................	31
5.9 Growth rates of renewable energy technology by region (low di↵usion).	......	31
5.10 Growth rates of renewable energy technology by region (high di↵usion).	.....	32
5.11 Output, emissions and temperature. No vs. high di↵usion.	.............	32
5.12 Energy by source (high di↵.)	.............................	33
5.13 Energy by sector in region	L	(high di↵.)	.......................	34
5.14 Energy by sector in region	L	(low substitutability)	.................	34
5.15 Output, emissions and temperature. No vs. high di↵usion. Low substitutability.	35
5.16 Energy by sector in region	L	(low substitutability)	.................	36
5.17 Output, emissions and temperature. No vs. high di↵usion. Low substitutability.	37
C.1 Output, emissions and temperature. No vs. low di↵usion.	.............	53
C.2 Energy by source (w/ growth)	.............................	53
C.3 Energy by sector in region	L	(w/ growth)	......................	54
iv

List of Tables
4.1 Parameters to calibrate – one set for each region	..................	16
4.2 Macroeconomic data	..................................	17
4.3 Energy units input-output table	............................	19
4.4 Energy units conversion factors	............................	20
4.5 CO	2coe�cients	.....................................	21
5.1 Calibrated parameters in base year.	..........................	24
D.1 List of countries in each region	............................	55
v

Chapter 1
Introduction
Energy is the link between climate change and our welfare. For much of human history, we used
a modest amount of energy. This changed during and after the industrial revolution. Anthro-
progenic emissions increased as we discovered how to e↵ectively use coal instead of charcoal for
fuel. With the later additions of oil products and natural gas, fossil fuels gradually replaced
pre-industrial energy sources in all sectors in the industrialized world. However, that large-scale
burning of relatively cheap fossil fuels does not come without consequences.
The Intergovernmental Panel on Climate Change (IPCC) has since the early nineties pub-
lished reports on the state of climate change. They are currently in their sixth assessment cycle
(AR6) and published their latest reports not long before I am writing this sentence. The IPCC
(Masson-Delmotte et al.,	2021	) ﬁnds that human activities already have caused climate change
in all areas on the planet. The average global temperature is estimated to already be more than
1 degree Celcius above pre-industrial levels, and it is only increasing as we continue to break
yearly emission records. The IPCC warns that we might cause irreversible changes to our planet,
which can adversely a↵ect millions, if not billions of people (and other living creatures). Thus,
there are huge potential welfare costs from climate change.
The most straight-forward way to emit less is to consume and produce fewer carbon-intensive
goods and services. However, reducing consumption by a marginal unit has di↵erent consequences
depending on your existing level of consumption. For a rich person, skipping a vacation abroad or
trading in a large SUV with an internal combustion engine (ICE) for a similar but electric vehicle
(EV) may not matter much. This is especially true if ICEs and EVs become more similar means
of transportation and their substitutability increases. For people with very low consumption,
however, reducing consumption could mean to go hungry or worse. Countries have recognized
this potential injustice and have agreed to ﬁnancial transfers from wealthy to poor countries to
help with emission mitigation and climate change adaptation (Harvey,	2021	).
Developing countries without access to modern infrastructure do not have to follow path
of industrialized nations. They can potentially skip some of the fossil-fuel dependence that
characterized the modern energy history of the industrialized world. Such a path of technology
adoption is sometimes called technology leapfrogging. The argument is that falling prices for
renewable sources, abundant local potential for renewable energy supply and political preference
for avoidance of local pollution and foreign fossil fuel imports all point to a future in which
growth in important sectors will be supplied emission-free in developing countries (Bond et al.,
2020	, Bond et al.,	2021	). As a contrasting scenario, developing countries may travel along a path
leading to so-called carbon lock-in. This term refers to the self-perpetuating path-dependency
of a fossil-fuel based energy system. Network externalities, sunk investments into infrastructure,
1

capital vintages, habits, institutions and so on, ‘locks in’ fossil fuels as the main energy source,
similar to what have been seen in the industrialized world (Unruh,	2000	). There are many other
potential futures, of course, but these to constitute two interesting future paths on each end of
the spectrum of possibilities for a country transitioning from using traditional energy sources to
modern ones.
It is not energy use per se that causes emission, but the carbon intensity of the energy use.
Therefore, if developing countries could transition from traditional energy sources to modern
renewables, they could enjoy higher consumption without paying the cost in terms of carbon
taxes or climate-related damages	1. Modern renewable energy systems require a lot of investment,
political will, technology and knowledge that does not come for free, even in richer countries.
But richer countries can make the energy transition in developing countries smoother by sharing
technological knowledge. On the other side of the coin is the political economy of developing
countries. It may (Bond et al.,	2020	) or may not (Newell and Pizer,	2003	)beinthebestinterest
of incumbent leaders to invest in renewable energy systems, to import foreign resources or to
allocate resources e�ciently. All these economical, political and technological factors are part of
a technology- di↵usion mechanism, where knowledge, policy and ideas ﬂow between countries.
In this thesis, I calibrate a North-South multi-sector integrated assessment model (IAM) with
explicit energy production and analyze the consequences of technology di↵usion between regions.
Researchers use IAMs to analyze the interaction between the economy and the climate system in
order to inform climate policy. Such models are used to estimate or calculate various important
ﬁgures, such as the social cost of carbon (SCC), likely future emission paths, the cost of climate
change adaption, the beneﬁt of mitigation and so on. They are called integrated because they
use models of the economy, the climate system, energy and sometimes agriculture and land use
in a combined framework to analyze the complex connections between these areas, and because
they combine insights from several di↵erent ﬁelds (IAM Consortium,	2022	).
Using data from several di↵erent public sources, I sort countries into two regions based on
income and degree of electriﬁcation. Then I calibrate ﬁve separate energy sectors plus electricity
that goes into the production of three ﬁnal good sectors for each region. My motivation for
calibrating this model is to analyze and quantify the sector- and fuel- dependent e↵ects of di↵erent
policy scenarios. To do this, I introduce an exogenous inter-regional di↵usion mechanism to the
base model, in which one region can import technology from another and improve their own
technology levels. Then I apply this model to analyze the e↵ects of inter-regional di↵usion on
the welfare in each region, emissions and temperature increase, the composition of energy use
within the economies.
This thesis contributes to the literature on speciﬁcally analytic types of integrated assessment
models. These models have closed-form solutions which usually limits the functional form choice
and descriptive accuracy compared to more complex numerical models. Analytical models are
easily interpretable and permit a large state-space with fast solution mechanism, which is espe-
cially important when analyzing issues like uncertainty – which I don’t go into in this thesis.
By calibrating the present model, I look to slightly narrow the gap between the simplicity and
transparency of analytical models, and the accuracy and richness of more detailed models. I hope
that this work lays a foundation for further explorations of the possibilities within numerical use
of analytical models.
The rest of the thesis proceeds as follows: In chapter 2, I present the background and review
the relevant literature on integrated assessment models, technological transitions and di↵usion
and on analytical IAMs. In chapter 3, I explain the model, the theoretical assumptions and
the motivation behind the model structure. I also introduce the di↵usion model. In chapter
4, I go through the calibration process. First, I explain the calibration process and list the
1Given that consumption goods also are produced with renewable or low-emission sources.
2

parameters I will calibrate. Second, I present the numerous data sources I use for calibration,
and the assumptions and interpolations I make in cases of missing or incomplete data. Finally, I
discuss how I pick the important elasticity of substitution parameters. In chapter 5, I discuss the
calibration results and simplifying assumption I make in the calibration process. Then I discuss
the numerical results of three main cases: (i) business-as-usual vs. optimal tax implementation;
(ii) the e↵ects of technological growth in the renewable sector and; (iii) the welfare e↵ects of high
and low technological di↵usion. I end the chapter with a brief sensitivity analysis. I conclude in
chapter 6.
I have collected detailed mathematical derivations, supplementary tables and ﬁgures and the
programming codes in the Appendix. I use Matlab for all programming purposes: wrangling
and compiling raw data for further analysis, missing data interpolations, sorting countries into
regions, calculating the calibrated parameters, solving the model and running simulations.
3

Chapter 2
Background
In this chapter I will ﬁrst give a brief history and description of di↵erent kinds of IAMs. Second,
I will give a short literature review on IAMs that analyze the energy transition in developing
countries and those that use technological di↵usion models. Finally, I will give background and
a literature review on analytical IAMs and place my contribution in context.
2.1 Integrated assessment models
Integrated assessment modeling grew out of the 1970’s energy modeling community, as described
in a nice ﬁeld history by William Nordhaus (	2011	). The most well-known IAM is perhaps the
DICE model for which the same Nordhaus (	1993	) was awarded the Nobel Memorial Prize in
Economic Sciences in 2018. IAMs have not become less relevant since then: The IPCC (IPCC,
2001	) uses several di↵erent IAMs in their most recent report on the mitigation of climate change,
such as WITCH (	2006	), AIM (Fujimori et al.,	2017	) and MESSAGE-GLOBIOM (Krey et al.,
2020	). Some countries use IAMs to inform their local carbon tax policy, e.g., the Interagency
Working Group on the Social Cost of Carbon in the US (	2021	) uses DICE, PAGE (Hope,	2006	)
and FUND (Waldho↵et al.,	2014	).
IAMs come in many shapes and forms. There is no single framework that is best suited in
all cases and for all uses. Which model type, solution method or assumptions to use all depend
on the problem on hand. The IPCC (IPCC,	2001	) gives a comprehensive and readable overview
of IAM taxonomies	1. I will give a brief summary of the most relevant general distinctions they
use.
First, models have di↵erent solution methods. Simulation models specify equations and
parameters to represent likely future paths of the economy and corresponding emissions, and
analyze the di↵erent outcomes given expert speciﬁcation of these parameters. Optimization
models selects some control variables optimally given the constraints of the system in order
to maximize utility (for cost-beneﬁt analyses) or minimize system costs (for cost-e↵ectiveness
analyses). The model on which I base this thesis, ACE (Traeger,	2022a	), is a cost-beneﬁt based
optimization model. Such models uses a damage function to close the feedback loop between
the economy and the climate system, where economic activities require energy that generates
emissions which increase temperatures, which in turn harms the economy and human welfare.
But I will also evaluate potential future paths similar to simulation models, where I specify
1See also Farmer et al. (	2015	), Bosetti (	2021	)andN.Stern(	2022	)forcomprehensivereviewsofthehistory,	taxonomy and common critiques of IAMs. For detailed and pointed critiques, see Pindyck (	2013	,2017	)and	Pindyck (	2019	)
4

di↵erent levels of di↵usion and inter-fuel substitutability. The beneﬁt of using an optimization
model as the basis for this analysis is that I can calculate the welfare e↵ects as the di↵erence
between an optimal base-line path and a optimized variation on that path.
Second, dynamics are handled di↵erently within the optimization models category. Inter-
temporal dynamic models ﬁnd the optimal path to follow to maximize the present utility value,
or minimize the total discounted system costs. Such models can assume either assume perfect
foresight, or allow for uncertain futures in which case agents maximize expected present values
(see e.g. Traeger (	2018	)). In contrast, recursive-dynamic models optimize within one period,
but do so sequentially. The state of the world changes over time but the agents do not take
into account the consequences their actions have on future states of the world and there is no
trade-o↵between future and present. There are hybrid models that assume varying degrees of
imperfect foresight. I use a inter-temporal dynamic model in this thesis.
A ﬁnal main distinction is whether the model relies on a general equilibrium framework
where all prices in the economy adjust to equate supply and demand in every market, or if the
researcher analyses a single or a few markets in a partial equilibrium while keeping the e↵ects
on other markets constant. I use a general equilibrium framework in this thesis.
In addition to the more technical di↵erences mentioned previously, models also vary in their
degree of detail and underlying assumptions. The ﬁrst version of DICE (	1993	), for example,
was a global model, but it was later extended to the regional RICE (Nordhaus and Yang,	1996	)
model, increasing the level of detail. The current version of DICE has a single-good economy
produced in each region with capital, labor and two di↵erent energy types. REMIND (Baumstark
et al.,	2021	), however, uses the same basic macro-economic variables, but has a detailed bottom-
up speciﬁcation of the energy sector with more than 50 di↵erent technologies. AIM also has a
detailed energy sector, but also has several di↵erent industrial and agricultural end-use sectors.
Another regional model with a single-good economy is WITCH (Bosetti et al.,	2006	), but that
model allow for endogenous investments in technology which captures strategic game-theoretic
e↵ects between regions or groups of regions.
2.2 Numerical IAMs analyzing the energy transition in de-
veloping countries
There is a rich literature analyzing energy transition in the developing world using IAMs. Lucas
et al. (	2015	) use the models in the LIMITS pro ject	2to analyze African emission scenarios under
various assumptions. Using the same LIMITS framework McCOLLUM et al. (	2013	) ﬁnd that
most supply-side capital investments should ﬂow to developing countries in the two-degree sce-
nario. In particular, investments should shift from upstream fossil fuels to downstream electricity
generation. There exist several other LIMITS pro ject papers (e.g. Van Der Zwaan et al. (	2013	)).
These papers are thorough and very useful for detailed insights in the necessary future energy
system mix in developing countries, how to obtain that mix, and the corresponding emission
paths. They also show through detailed modelling that the energy transition in developing coun-
tries indeed is feasible. In the same spirit, Leimbach et al. (	2018	) quantify the costs and beneﬁt
of climate policy in Sub-Saharan Africa, using REMIND. Their context is the “favorable condi-
tions” for renewable energy on the continent and they include international fossil fuels markets
and technology di↵usion. They ﬁnd emission pathways Sub-Saharan Africa consistent with two
degrees warming at a net zero cost, given equitable burden sharing by international transfers.
Clarke et al. (	2012	) explore how di↵erent models in the Asian Modeling Exercise (AME)
vary in their energy system assumptions. Many of the models use production functions that are
2LIMITS is inter-model study of policies consistent with a maximum warming of maximum 2 degrees Celsius.
5

calibrated to base-year shares. Some models in AME vary parameters over time, allowing them
to represent predicted changes in technology, infrastructure and so on. The paper includes a
comprehensive overview of the implied constraints on and costs of key technologies and if (and
how) base-year shares are calibrated in the models.
All these models are much more complex and rich than the present model. However, they
are solved numerically and in that respect di↵er from the present thesis.
2.3 Technological di↵usion models
Zhang et al. (	2020	) augment REMIND to include regional technological improvement and dif-
fusion. They ask how multi-level learning a↵ect the regional mitigation costs and technology
di↵usion. This paper is an extension from a earlier paper by Zhang et al. (	2014	) that looked
at di↵erent regional investment costs, but did not include multi-scale learning. Four di↵erent
approaches are investigated: full international spillover, no spillovers and two intermediate cases
with varying degree of market e�ciencies. The crucial exogenous parameter in their model is
the learning rate.
Gu et al. (	2021	) develop a numerical multi-sectoral and regional IAM that models bottom-up
low-carbon technology di↵usion between regions and (exogenous) R&D investments. They use
this model to identify the e↵ects endogenous technology transfer can have on optimal regional
carbon emissions. The use only Cobb-Douglas or Leontief production functions.
Cai et al. (	2015	) combines a top-down dynamic computable general equilibrium model with
a bottom-up model of energy production and consumption. The result is a hybrid IAM called
GTEM-C	3. They want to keep the general equilibrium framework, while permitting more realistic
technological development. To do this they use constant ratios of elasticities of substitution for
homothetic (CRESH) functions for a set of disaggregated energy technologies and household
energy preferences. A CRESH function allows for heterogeneous substitutability between input
factors in the same production functions. I use the simpler but more constraining constant
elasticity of substitution (CES) production function. An interesting extension to the model in
this thesis is try to introduce CRESH production functions. -
Gu et al. (	2021	) analyze the possible carbon mitigation caused by technology di↵usion be-
tween countries. The base model, CIECIA (Wang et al.,	2016	), is a multi-sector, multi-country
general equilibrium model with a damage function, endogenous technology and free trade in
capital and commodities. Gu et al. (	2021	) add a bottom-up technological di↵usion block to
CIECIA in which every sector in each country search for, select, learn and imitate technology
from corresponding sectors in other countries. First, sectors ﬁnd the feasible imitable technolo-
gies and then select among those to ﬁnd imitation targets. The higher the intellectual property
protection, the narrowed the imitation range becomes. Each imitation target is represented their
“attractiviness”, or the magnitutude of the technology gap and the economic gap plus the severe-
ness of path dependency. The selection is governed by a probability distribution where higher
probability of selection are given to countries with higher attractiveness. After having selected
a supplier country, the imitating country gets accelerated energy-saving R%D in the process
technology for the given sector. Progress in “processing” is modeled by a looping stochastic
logarithmic shock mechanism that a↵ect the intermediate input coe�cients in the production
functions. The technology is self-selected by the sector if the shocks have caused lower unit
costs. Analytical models are suited for analyzing uncertainty. Another interesting extension is
to model inter-regional di↵usion more in line with their approach an incorporate uncertainty in
3GTEM-C is built on top of the Global Trade and Environment Model , which is the Global Trade Analysis	Pro ject’s (GTAP) IAM
6

the di↵usion model. This is beyond the scope of this thesis, however.
Furthermore, Gu et al. (	2021	) ﬁnd that technology transfer has signiﬁcant mitigation e↵ects.
They analyze four di↵erent intellectual property regimes. The ideal regime is where low-carbon
technologies are freely shared to the public. Decreasing the technology transfer threshold, so that
more technology becomes imitable, will signiﬁcantly increase di↵usion and mitigation. While this
e↵ect is strongest for technology transfers from developed to developing countries, the authors
point out the importance of technology sharing between developed countries as well. Finally,
Gu et al. (	2021	) interact di↵usion with various exogenous R%D investment shares and ﬁnd that
the combination gives the highest mitigation, particularly in developing countries. Even with
no patent barriers, a relatively low knowledge stock creates a drag on low-carbon technology
progression in developing countries.
2.4 Analytical IAMs
The complexity of all the models mentioned in earlier in this chapter require numerical solutions.
Solving such large and detailed models became feasible only after the ma jor improvements in
computing power during the ﬁnal decades of the previous century. Analytic and semi-analytic
models, in contrast, permit closed-form model solutions. Thus, they allow transparent analyses
of the crucial trade-o↵s and economic mechanisms inherent in the problem of climate change.
While the qualitative closed-forms expressions are often quantiﬁed to inform policy, the necessary
analytic tractability requires closed-from solution, thereby limiting functional form choise and
descriptive accuracy.
Analytic models are especially useful when handling uncertainty	4. They permit large state-
spaces that break the “Bellman curse-of-dimensionality”, referring to the exponentially increasing
computing power needed when adding states in numerically solved models (see e.g. Traeger
(2018	) for a state-of-the-art analytical model of closely related structure, with uncertainty).
An important landmark in the ﬁeld was when Golosov et al. (	2014	), building on the analytic
macro-model of Brock and Mirman (	1972	), included a heterogeneous energy sector with emission
feedback into production damages	5. This ﬂexible but robust framework has supported several
extensions, such as a closed-form general equilibrium version of RICE (Hassler and Krusell,
2012	), the pricing of uncertain catastrophic events (Gerlagh and Liski,	2018b	), game-theoretic
approaches to time-inconsistent preferences (Gerlagh and Liski,	2018a	; Iverson and Karp,	2021	)
and distortionary ﬁscal policy (Barrage,	2019	).
Traeger (	2022a	) made further progress by generalizing the economy, including the energy
sector, while explicitly modeling and improving climate dynamics and thus enabling a clearer
analysis of the economics of climate change. In the appendix of the same paper, Traeger proposes
a production structure in which several sectors produce ﬁnal goods in Cobb-Douglas function of
capital, labor and an CES function of intermediate energy. Traeger (	2022b	) further develops the
multi-sector speciﬁcation.
The present thesis relies on the analytic structure in Traeger (	2022a	) and production functions
in Traeger (	2022b	) to solve the inter-temporal optimization problem characterizing regionally
optimal carbon prices	6. My main contributions are (i) to calibrate the production function
4Several early analytic contributions indeed focus on tackling uncertainty in climate-economic models (Pizer,	1999	;Schauer,	1995	). Later papers make use of linear-quadratic functions to obtain relevant closed-form expres-	sions (Hoel and Karp,	2002	;KarpandTraeger,	2021	;KarpandZhang,	2006	,2012	). These are stylized beneﬁt-cost	models without production or energy sectors.	5Van Der Ploeg and Withagen (	2014	)developedasimilarbutslightlymoregeneralmodelaroundthesame	time. Other generalizations are found in Rezai and Van der Ploeg (	2016	)andvandenBijgaartetal.(	2016	)	6Iassumethatregionstakesintoaccountonlytheirownfuturedamages,notthedamagestootherregions.
7

proposed in Traeger (	2022b	) (ii) to ﬁnd, gather and clean the data used in this calibration
exercise and (iii) to analyze the e↵ects on welfare and energy composition from technological
di↵usion in a two-region version of the model. To my knowledge, this is the ﬁrst analytical
integrated assessment model calibrated to this level of detail.
8

Chapter 3
The Model
The model is based on the ACE-model by Traeger (	2022a	) with the multi-sector production
function speciﬁcation in Traeger (	2022b	). In this chapter, I will ﬁrst present the economy with
the accompanying nested production function. Second, I give a brief explanation of the climate
system. Third, I introduce a simple but general exogenous technological di↵usion model for the
energy sectors across regions. Finally, I brieﬂy explain how to solve the model.
3.1 The Economy
There are two social planners in this model, one for each region	j2	J	=	{L, H	}. Region	L
is low-income and region	H	is high-income. There is no trade between the regions. They are
connected (i) through the carbon-stock in the atmosphere, into which the fossil fuel energy sectors
in both regions emit CO	2and (ii) by the exogenous di↵usion of technical knowledge from region
H	to regions	L. Each social planner cares only about her own region’s consumption, but they
a↵ect the other region’s welfare indirectly through the future damages inﬂicted on each other by
climate-change related damages. If the social planners had a way to credibly commit, they could
agree to lower their own emissions to the beneﬁt of the other social planner, and to the potential
net beneﬁt for both social planners	1. I assume, however, that there are no such commitment
vehicles, and that both social planners therefore have an incentive to free-ride on the other’s
mitigation. Region	H, however will be endowed with a positive but small price on emissions, to
reﬂect the current state of carbon pricing in the World. It is in this setting I analyze di↵erent
technological di↵usion scenarios and their welfare e↵ects.
The social planners maximize their discounted inﬁnite stream of (their own) utility from the
consumption of an aggregated good
max
1X
t=0
�tlog	Ct,	(3.1)
with the discount factor	�=	11+	�⇡	0.87, where the pure rate of time preference	�determines
how much the social planner discounts future utility, and is I crucial parameter in integrated
assessment models (see e.g. Drupp et al. (	2018	) for an expert survey and discussion). I sup-
press regional indexing	jfor brevity, but will include it wherever necessary to avoid confusion.
1Even if the social planners could credibly commit, it is not given that commitment is in the best interest of	their populations.
9

The single good ACE-model has simple logarithmic utility, which implies a unitary intertem-
poral elasticity of substitution. The present model, as explained in the appendix to Traeger
(2022a	), weakens this assumption by disaggregating consumption using the constant elasticity of
substitution (CES) composite
Yt=	At
⇣X
l2Nc
al,t	(cl,t)s⌘1s,	(3.2)
where	P
lal,t	= 1 and the substitutability index	s	1 indicating that all ﬁnal consumption goods
cl,t	are necessary in consumption. The ﬁnal goods are transportation, industry and other, with
l2	Nc=	{tr, in, ot	}respectively. I choose these sectors because they approximately consume
a third each of the total World energy consumption (International Energy Agency,	2019	), and
because they have di↵erent dominating energy inputs with varying degrees of potential interfuel
substitutability. The constant	Atmakes sure	Ctis measured in calibration-period money metric
utility units.
Each ﬁnal good sector have Cobb-Douglas production functions of technology	Al,t, capital,
Kl,t, labor	Nl,t	and the intermediate energy input	dl,t
cl,t	=	Al,tK↵ll,tN1�↵l�⌫l	l,t	d⌫ll,t,	(3.3)
where	⌫lis sector dependent. The overall production function must be homogeneous in capital
to get a closed-form solution (Traeger,	2022a	), which means	↵⌘	↵lmust be sector-independent.
The intermediate energy good	dl,t	represents the energy composite used in production of
ﬁnal good	l. It can be interpreted as a generalized mean of various reﬁned energy inputs	ei,l,t
combined according to the CES production function
dl,t	=	˜Al,t
⇣X
i2Nd
˜al,i,t	e˜sll,i,t
⌘1˜sl	8l2Nd,	(3.4)
where ˜	al,i,t	are CES share coe�cients with	P
i˜al,i,t	= 1 for each sector, and where	˜Al,t	is a
constant turning the function money-metric in calibration-period USD. The notation	ei,l,t	reads
‘reﬁned energy input	iin sector	lin period	t’. The reﬁned energy sources available to sector
lis a sector-speciﬁc subset of	Nd=	{c, o, g, b, el	}, the elements of which corresponds to coal,
oil products, natural gas, bioenergy and electricity, respectively. Note that the substitutability
parameter ˜	slis sector-dependent	2. The model can therefore distinguish between the arguably
harder-to-substitute inputs to the transportation sector, where oil products especially in heavy
transport currently dominates electricity, and the more substitutable other-sector, where fuels
for e.g. heating are easier to switch out.
Electricity production, indexed with both	i= and	l= and where ˜	a0,0,t= 0, happens in a
special in-between transformation sector that uses reﬁned energy from the set	Ne=	{c, o, g, b, r	},
where	ris renewable energy:
e0,t=	˜A0,t
⇣X
i2Ne
˜a0,i,t	e0,i,t
⌘	1˜s0,t,	(3.5)
2An earlier version of the thesis had time-dependent substitutability as well. This feature allows the degree of	substitution to increase over time. This is particularly interesting in the transportation sector, where we are inthe middle of a technological development that makes non-emission vehicles, ships and airplanes more similar totheir combustion engine counterparts. Kaya et al. (	2017	)pointoutthatmanyIAMsine↵ectforcesapreservation	of the technology mix. The reason is the use of constant elasticity of substitution (CES) production functions.This, argues Kaya et al. (	2017	), does not match with empirical data on historical di↵usive transitions. They	suggest several methods to alleviate this issue, one of which is to allow the constant elasticity of substituion varyover time. I removed this feature, however, to simplify the present analysis. In general, substitutability can alsova r y b e twe e n r e g i o n s a n d t h i s m o d e l p e r m i t s i t . B u t I h ave l e f t t h i s f e a t u r e o u t f o r t h e s a m e r e a s o n .
10

with the CES coe�cients and the constant	˜A0,thave the same roles as the production function
in (	3.4	). Electricity has the same notation as other reﬁned energy inputs, but di↵er in inputs
and functional form.
The reﬁned energy products coal, oil products, natural gas, bioenergy and renewable energy
are all produced in the Cobb-Douglas-Leontief-hybrid production function of technology,	¯Ai,t,
capital,	¯K, labor,	¯N	and emissions	Ei,t:
ei,t	=	¯Ai,t	¯K¯↵ii,t	min	�¯Ni,t,¯ai,tEi,t 1�¯↵i	8i2Ne,	(3.6)
where I deﬁne	Ei,t	⌘	¯Ni,t¯ai,t	and where ¯	ai,t	is a calibrated parameter making the units of labor and
emissions comparable. The motivation for using this functional form merits some explanation.
Traeger (	2022b	) introduces the general production function	ei,t	=	gi,t(¯Ai,t,¯Ki,t,¯Ni,t,E	i,t), in-
creasing in all inputs. If there is no saturation mechanism, cost or other constraint on	Ei,t	it will
be a free factor in the absence of carbon pricing, which implies an inﬁnite demand in optimum.
This is not particularly realistic, however. Even with low global average carbon prices, we are
not currently emitting inﬁnite amounts of CO	2, nor have we ever done so. This warrants some
sort of saturation in emissions. Traeger (	2022b	) have several suggestions to ensure ﬁnite optimal
factor demand, among them the functional form in equation (	3.6). This speciﬁcation ensures
that labor and emissions are perfect complements in production of reﬁned energy. Then, to get
inﬁnite emissions, one would need inﬁnite labor in the fossil fuel sectors. The perfect comple-
mentarity introduces an implicit price on emissions, which helps the model in reproducing a
more realistic ‘business-as-usual’ scenario. Finally, note that the capital share of output, in this
case ¯	↵⌘	¯↵i, must yet again be constant across reﬁned energy, for the same reasons as explained
above.
Finally, to ensure that total supply equals total demand in each of the energy sectors, let
ei,t	=	X
l2Nc
el,i,t	8i2Ne[Nd.
The aggregate capital stock,	Kt, does not depreciate fully over the decadal timestep but
is allowed to persist as a stock in the economy, as introduced to analytical IAMs by Traeger
(2022a	). Capital’s equation of motion is
Kt+1	=	It
1+	gk,t
�k+gk,t

were	It=	Yt[1�D(T1,t)]�Ct,
where	�kis the depreciation factor and	gk,t	is an exogenous approximation of the capital growth
rate. The social planner distributes the accumulated capital optimally in each period.
A fraction of GDP is destroyed each year by climate change related damages
D(T1,t)=1	�exp [	�⇠0exp(	⇠1T1,t)+	⇠0],⇠	02R,	(3.7)
where	T1,tis the increase in atmospheric temperature in period	t, compared to year 1900.	⇠0
and	⇠1are damage parameters calibrated to Howard and Sterner (	2017	). While country per-
capita income levels show some geographical correlation, I assume for simplicity that increasing
temperature damages both regions in the same way.
3.2 The Climate System
The climate system in this model is taken from Traeger (	2022a	). Fossil-based energy production
emits CO	2into the atmosphere where it accumulates and causes increased radiative forcing (or
greenhouse e↵ect). This in turn increases atmospheric temperatures which causes damages.
11

Let the global CO	2-emissions in each period be	Et⌘	EL,t	+EH,t	,where	Ej,t	=	P
i2IdEj,i,t	,
and where I have collected fossil-based energy sources in the set	Id. These emissions ﬂow into
the atmosphere, from where they can ﬂow between di↵erent reservoirs in the carbon stock vector
M	t+1	=	�M	t+e1(Et+Eexot	),	(3.8)
where	Eexot	are exogenous global non-energy related emissions.	e1is the unit vector reﬂecting
that emissions initially only go into the atmosphere, and the transition matrix	�	governs carbon
ﬂow between reservoirs.
Increasing levels of carbon in the atmosphere causes a greenhouse e↵ect, also called radiative
forcing
Ft=	⌘log	M1,t+Gt	Mpre
log 2	,
where	M1,tis the level of CO	2in the atmosphere,	Gtare exogenous non-CO	2emissions,	Mpre
is the pre-industrial level of atmospheric CO	2and	⌘is a parameter that captures the radiative
forcing intensity.
There are three temperature layers in the model: atmosphere and the upper and deep oceans,
corresponding to	i=1	,2,3. The 0	th	layer is deﬁned to be the long-term equilibrium temperature
in the atmosphere, which is increasing in the strength of the greenhouse e↵ect
T0,t⌘	s
⌘Ft,
and where climate sensitivity parameter	scaptures the response of temperature equilibrium to a
doubling of atmospheric CO	2levels compared to pre-industrial times. For the other layers, the
temperature equation of motion is a non-linear mean of own- and adjacent temperatures
Ti,t+1	=	1
⇠ilog
⇣
(1��i,i+1	��i,i�1)exp[	⇠iTi,t]
+�i,i+1	exp[	⇠iw�1i	Ti�1,t]+	�i,i�1exp[	⇠iwi+1	Ti+1	,t]
⌘
(3.9)
with parameters	⇠1=	log 2s	⇡	14and	⇠i+1	=	wi⇠i=	Ti�1eqTieq	⇠i. The weight matrix	�	captures heat
exchange between layers. There are only three layers, so	�3,4= 0. Using this speciﬁcation,
atmospheric temperature is a↵ected by the current temperature, the long-term equilibrium tem-
perature and the upper ocean temperature. The current atmospheric temperature enters the
damage function in equation (	3.7	), which closes the economy-climate feedback loop.
3.3 Technological Di↵usion
Remember equation (	3.6	) which for renewable energy (	i=	ren	) reads
eren,t	=	¯Aren,t	¯K¯↵ren,t	¯N1�¯↵	ren,t	,
noting that renewable energy has no emissions (in energy production) so that min	�¯Ni,t,¯ai,tEi,t =	¯Ni,t, and where	¯Aren,t	is technology in production of renewable energy. Since the production func-
tion is Cobb-Douglas, technological change is Hicks-neutral and will not be factor-augmenting.
Growth in the absolute technology level will mean that more output can be produced for the
same level of capital and labor. Furthermore, I let the technology levels for the other energy
12

sources be constant. This keeps the model as simple as possible while capturing that fast relative
growth in renewable energy versus other sources (International Energy Agency,	2021b	). For a
more general version of the model, see the Appendix.
Deﬁne the growth rate of technology in renewable energy production as
gt⌘	¯Aren,t	+1	�	¯Aren,t	¯Aren,t	.	(3.10)
Then, let each region	j2{	L, H	}have the growth rate	gj,t, capturing that the growth rates of
renewable power technology can be di↵erent in di↵erent regions.
Inspired by Acemoglu (	2009	), I propose the following model of technological di↵usion between
the two-regions. Let region	H	be the world leader. Then, let
gL,t	=	✓LˆAL,t	+�L	and (3.11)
gH,t	=	✓H	ˆAH,t	+�H	(3.12)
where	ˆAj,t	⌘	¯AH,t	�¯Aj,t	¯Aj,t	2[0,1) is the rate of di↵erence in technology levels in renewable energy
production between the regions, expressed as a rate. The technology absorption rate or di↵usion
parameter	✓j2[0,1] captures the varying degrees with which regions are able to import various
technologies given their policies, institutions and other economic and social factors. This rate is
exogenous in the current model, but it is possible to endogenize it as a function of e.g. human
capital (see Acemoglu (	2009	)). The absorption rate captures the delay or imperfectness with
which technology becomes available across the world. For	✓j= 0, there is no di↵usion. For
✓j= 1 the region catch up to the World frontier without delay. For example, if the level of
technology in region	H	is ﬁve times region	L,✓L=0	.01 and	�L= 0, then	gL,t	=0	.05. There is
no di↵usion e↵ect for region	H,since	ˆAH,t	= 0. The local innovation parameter	�jdetermines
the locally sourced growth rate in technology. It captures how fast a region uses it existing
knowledge	Ajto improve technology, remembering that technology here means relative to fossil
fuels.
3.4 Sketch of model solution
Traeger (	2022a	) solves a version of the present model with a general production function. I
will only give an heuristic explanation of his solution method and point the interested reader to
his working paper for details. His closed-form solution relies on homogeneity of capital in the
production functions using capital. Therefore, I will show that this requirement holds for this
multi-sector model but that it restricts the parameter space of the model.
To solve the model, Traeger (	2022a	) makes use of the Bellman equation. Here, the control
variables in each period are chosen by the social planner such that she (i) maximizes the sum of
present-period utility and the discounted value of all future utility streams, and (ii) that the value
of this maximization is equal in all periods. First, Traeger (	2022a	) transforms all endogenous
state variables such that their equations of motions become linear in the transformed states.
This step is important, because it makes an a�ne solution to the Bellman equation possible.
The ﬁrst order conditions will then provide optimal controls that do not depend on the states
but only on coe�cients of parameters and shadow values. Thus, when the optimal controls are
inserted back into the Bellman equation, it is possible to collect terms so that the states are each
a linear function of these coe�cients. Second, to make the Bellman equation hold in each period
the state-coe�cients must all equal 0. The coe�cients are therefore manipulated so that this
requirement holds, which will also provide the shadow values along the optimal path as function
13

of parameters only. Finally, one can solve for the optimal control paths also as functions of
parameters only. It is the shadow value on the carbon stock that will, when transformed from
utils to consumption units, inform the value of the social cost of carbon.
The production function in this model has (for each region	j)
Yt=	F(Al,t,K	l,t,N	l,t,dl,t)=	At
⇣X
l2Nc
al,t	(cl,t)s⌘1s=	At
⇣X
l2Nc
al,t
⇣
Al,tK↵ll,tN1�↵l�⌫l	l,t	d⌫ll,t
⌘s⌘1s
Then, multiply capital by
F(Al,t,�K	l,t,N	l,t,dl,t)=	Al,t	(�K	l,t)↵lN1�↵l�⌫l	l,t	d⌫ll,t
If ¯↵i⌘	¯↵then
F(Al,t,�K	l,t,N	l,t,dl,t)=	�¯↵At
⇣X
l2Nc
al,t
⇣
Al,tK↵ll,tN1�↵l�⌫l	l,t	d⌫ll,t
⌘s⌘1s
=	�¯↵F(Al,t,K	l,t,N	l,t,dl,t),
which shows that the production function is homogeneous in capital if all sectors have the same
↵. One can use the same method to show that the production of the reﬁned energy goods also
are homogeneous in capital.
14

Chapter 4
Calibration and Data
In chapter 3, the social planners allocated the resources in an optimal utility-maximizing way. To
match this theoretical construct into a construct that permits the use of real data, I transform
the social planner’s problems into the corresponding decentralized cost-minimization problem
for representative ﬁrms. By doing so, I can derive their respective compensated demand equa-
tions, which, when inverted, provides the equations for the calibration targets. For a detailed
mathematical derivation of these equations, see Appendix	A.
This chapter proceeds as follows: First, I brieﬂy explain the practical calibration process.
Second, I list the parameters to be calibrated. Third, I present and discuss the di↵erent data
I use. And ﬁnally, I discuss the most important but also most elusive production function
parameters: the elasticities of substitution. The calibration results are in the next chapter.
4.1 Calibration procedure
A signiﬁcant part of this thesis is the calibration process. My ﬁrst step in this process is to ﬁnd
and collect relevant data. Second, I clean and compile the data and read it into arrays and tables
in Matlab. See Appendix	E	for the wrangling script. Third, I perform interpolations in cases
where there is missing data. Fourth, I aggregate the data into the two regions I will use. Fifth, I
use the regional arrays as inputs to the calibration equations in order to calculate the numerical
values of the parameters. See Appendix	Ffor the calibration script which is appended with the
di↵usion model. I omit the graph-producing code.
I calibrate a total of 44 parameters, see table	4.1	for an overview. These parameters are then
fed into a numerical optimizer in Matlab (provided by my supervisor) along with an initial guess
of the optimal solution controls, as informed by the data. The optimizer produces a result that
should, if the calibration is done correctly, replicate the structure of the data as they were read
into the calibration script initially. If the model do not reproduce the data, I search for bugs,
typos, calculus-errors or similar until the calibration improves. See chapter	5.1	for a discussion
of the ﬁnal calibration results.
4.2 Data sources
There is no single source that can provide up-to-date data covering all the needs in the calibration
process. Therefore, I must collect data from several di↵erent sources, then pre-process and
combine them to a usable form. All the data sources I use are based on country-level data,
15

Table 4.1: Parameters to calibrate – one set for each region
Parameter Domain Explanation Equation
↵l⌘	↵l	2Nc	Capital share in ﬁnal good sectors (	A.5	)
⌫l	l2Nc	Energy share in ﬁnal good sector	l	(A.4	)
¯↵l⌘	¯↵i	2Ne	Capital share in energy production (	A.12	)
al,t	l2Nc	CES-shares in ﬁnal good aggregation (	A.2	)
˜al,i,t	l2Nc,i	2Nd	CES-shares in intermediate energy good (	A.7	)
˜a0,i,t	i2Ne	CES-shares in electricity production (	A.8	)
At	TFP in aggregation sector (	A.2	)
Al,t	l2Nc	TFP in ﬁnal goods sector (	A.6	)
˜Al,t	l2Nc	TFP intermediate energy sector (	A.7	)
˜A0,t	TFP in electricity sector (	A.8	)	¯Ai,t	i2Ne	TFP in reﬁned energy sector (	A.13	)
¯ai,t	i2Id	Leontief parameter in energy production (	A.11	)
which means that my two-region model can be easily extended. The model is already calibrated
to the RICE-regions and can, with some adjustments and additional data, be used as a country-
level model. However, not all countries provide detailed enough data, so the calibration process
rely on regional interpolations where necessary.
The following section has the following structure: First, I set the income cut-o↵that deﬁnes
the two regions. Second, I present the volume data sources. Third, I present price data sources
and the conversions and assumptions I make. Finally, I calculate emission factors and the regional
emission prices.
4.2.1 Region classiﬁcation
Countries are sorted into two regions: high income and low income, with corresponding region
indices	j=	H	and	j=	L. I would like to sort countries to the degree to which they use electricity,
and use GDP per capita as a proxy. I sort the list of countries by PPP GDP per capita, and set
the cut-o↵such that China with a GDP per capita in 2019 of 14 031 PPP USD is included in
the group of high-income countries. This is contrary the common deﬁnition of	developing’ and
developed’ countries, where the cut-o↵normally is set higher. Excluding China, however, which
has a very high degree of electriﬁcation (International Energy Agency,	2019	) seems unnatural.
India, a large country with a low degree of electriﬁcation, is then included in the low-income
group, as are many African countries and poorer countries in South America and Asia. See table
D.1	for a complete list of countries.
4.2.2 Volume data
The base-year for calibration is 2019 and all gathered data corresponds to that year unless
otherwise speciﬁed. I use the Penn World Tables (Feenstra et al.,	2015	) for data on country-
level real GDP, capital stock and employment for 183 countries	1. All values are denoted in
current purchasing power parities (PPP) in millions of 2017 US dollars, which enable cross-
country comparisons in a given year (OECD and Eurostat,	2012	). There are missing capital
1The Penn World Tables have only sparse data for average hours worked per year but good data for number	of persons employed. I will therefore use the latter measurement of labor throughout.
16

and employment data for some smaller countries in the Penn World Tables. For these countries
I use the regional average ratios of capital/GDP and employment/GDP for interpolation. The
interpolating regions are based on the 12 RICE regions (Nordhaus and Yang,	1996	). I ﬁnd the
average yearly growth rate of GDP between 2010 and 2019 by regressing log GDP on year for
each region. I use this growth rate as total factor productivity (TFP) growth in the model, which
I assume is falling over time. The socio-economic data is collected in table	4.2	.
Table 4.2: Macroeconomic data
Variable Unit Region L Region H
No. Countries . 89 94
GDP Trillion PPP 2017USD 27,6 97,3
Capital stock Trillion PPP 2017USD 110 452
Employed Million 1553 1764
Emissions GtCO	2	6,7 27.5
Share of output Transport 0,07 0,05
Share of output Industry 0,44 0,42
Share of output Other 0,49 0,53
Share of labor Transport 0,06 0,06
Share of labor Industry 0,20 0,25
Share of labor Other 0,74 0,68
Share of capital Transport 0,06 0,06
Share of capital Industry 0,16 0,15
Share of capital Other 0,77 0,79
TFP growth rates . 3,1% 2,2%
The Penn World tables do not have data per economic sector, only country-level GDP. GDP
is a value-added concept, and is meant to measure the value added by all the economic activities
within a country in a given period. The value added is deﬁned as output minus the value of inter-
mediate goods used in production, and GDP thus avoids double-counting economic activities. In
my model, however, the three sectors are kept separate and do not provide inputs for each other
except indirectly through the possibility of capital accumulation. The production of ﬁnal good
cl,t	should therefore be interpreted as sector output less the value of the energy consumption
good. But since the energy sectors do not use any intermediate goods, it holds that output and
value added are equivalent measures in my model. However, I cannot use data for value added
by economic sector to calibrate my model, nor only use output data. The ﬁrst reason is that this
will reduce the relative importance of sectors that in reality may use few intermediate goods but
provide many services for other sectors in the economy. The second reason is that total output
will exceed total GDP, which will double count production and therefore total emissions. My
solution is to ﬁnd output shares of total output by sectors using UN (United Nations Statistics
Division,	2022	) and WIOD data (Timmer et al.,	2015	). Then, I use these output shares to ﬁnd
sector shares of GDP in each region based on macro data from the Penn World Tables. UN and
WIOD data cover in total 102 countries, and I use regional average interpolation for missing sec-
tor share data. Some poorer countries’ shares do not sum to unity. The transport and industry
shares are, however, of the same magnitude as in countries with better data, whereas the ‘other’
sector show signiﬁcantly lower shares, resulting in a less-than-unity sum of shares. Therefore, I
take the ‘other’ share to be the di↵erence between total economy output, transport and industry.
This transformation does not matter for countries with good data, and a good second-best ﬁx
17

for countries with inconsistent data. I have 2019 data for most countries, and less recent data for
a few countries, with the oldest being from 2014. The sector shares are fairly stable over time,
which means that older data shouldn’t be too far o↵the correct 2019 shares.
To ﬁnd labor per sector, I use a similar procedure as for output. I ﬁnd country-level employ-
ment shares using ILOSTAT data (International Labour Organization,	2022	) and apply those
share to total employment from the Penn World Tables to get labor per sector. ILOSTAT data
covers 175 of the countries in the Penn World Tables. I use average regional interpolation for
missing data.
There are very few countries with su�ciently detailed capital stock data. The OECD (	2022	)
has capital data per economic sector, which I use to ﬁnd the shares of total capital stock going
to transport, industry and other, and to calibrate capital’s share of output,	↵. Then, I apply
these shares to the total capital stock from the Penn World Tables to ﬁnd the value of capital
per ﬁnal goods sector. I use RICE regions for interpolation of other missing data categories, but
this dataset only covers 29 OECD countries. Therefore, I use the OECD average capital shares
for all countries with missing data which is obviously unrealistic. Remember, however, that	↵
must be independent of sector to keep analytic tractability. As an implication, the sector capital
values are not really needed directly in the calibration. I use them only to check ex-post if the
model reproduces the state of the World from the data.
It is even more di�cult to ﬁnd capital data for the energy reﬁnement sectors than for the
broader economic sectors. To my knowledge, there is no data for capital in bioenergy or renewable
energy production. Investment data for renewables are abundant, but it is beyond the scope
of this thesis to use the perpetual inventory method needed to convert investment ﬂows to
capital stocks. These capital data goes into the calibration of capital’s share in the production
of the reﬁned energy goods, ¯	↵. Similar to	↵, also ¯	↵	has to coincide across fuels to preserve
tractability. Because of the mentioned lack of data, I use only fossil fuel data to calibrate ¯	↵, and
then extrapolate this parameter to bioenergy and renewables. This is perhaps not drastically
unrealistic for advanced biofuel production in developed countries. Bioenergy in the poorest
countries, however, often consist of agricultural waste, ﬁrewood and charcoal, which is probably
much less capital intensive than fossil fuel and renewable production would be in those same
countries. Furthermore, there are very few countries with detailed capital stock data in energy
production, even when only considering fossil fuels. For the high-income region, I use the weighted
average of US and Russian ¯	↵. I get US capital data from US Bureau of Labor Statistics (	2022	).
I could not ﬁnd o�cial Russian data, and based the calculation on manually collected asset
data from the largest Russian oil, gas, and coal companies using publicly available information
(Investopedia,	2022	). Because of the uncertainty in using this method, I give the Russian ¯	↵only
a 20% weight. For the low-income region, I use the Chinese ¯	↵based on data from of Statistic of
China (	2021	). Since China is not included in the low-income group, this will be an out-of-sample
extrapolation. In conclusion, there are several sources of uncertainty surrounding the calibration
of ¯	↵.
I use International Energy Agency (	2019	) data for energy volume balances. They provide
an input-output table of energy ﬂows between fuel sources and sector for 155 countries, and
regional data for countries without individual entries. Production, imports, exports and supply
changes make up the total energy supply in a country, which by deﬁnition has to equal total
consumption. I divide total consumption into ﬁve groups: transport sector, industry sector,
other sector, electricity sector and energy own use and losses. The latter group is the energy
used in energy production itself plus losses from reﬁnement and transport. I add this group to
the other four sectors according to the relative fuel intensity of each sector. Coal, oil, natural
gas and bioenergy are aggregated as such in the IEA data. I add nuclear energy, wind and solar
to get an aggregated “renewable energy” input. I add electricity and heat to a single category as
18

Table 4.3: Energy units input-output table	12
Sector (region) Regions Coal Oil Natural Gas Bioenergy Renewable Electricity
Transport (L) 64 2 068 19 648 792 1 168 282 323 799 147 864
Industry (L) 64 9 058 048 3 540 280 5 913 090 4 598 762 2 824 6 428 874
Other (L) 64 1 411 864 5 011 278 5 412 452 24 495 134 54 981 9 307 931
Electricity (L) 64 20 116 432 2 802 371 9 977 632 1 270 319 6 925 783 16 477 436
Transport (H) 78 1 342 78 848 891 5 278 595 4 022 266 1 519 620
Industry (H) 78 39 411 437 9 778 205 27 649 690 6 218 485 39 028 39 404 152
Other (H) 78 5 108 055 13 326 519 31 223 891 8 380 346 2 244 821 45 814 532
Electricity (H) 78 82 150 361 4 613 042 42 752 393 7 236 553 49 695 890 91 387 683	3
1International Energy Agency (	2019	).	2Units are tera joules (TJ).2Electricity in electricity production is an output.
well. The processed results are in table	4.3	. There are fewer individual countries covered in the
IEA data than in Penn World Tables, but the data still matches when aggregated into regions
since the IEA as regionally aggregated data as well.
4.2.3 Energy prices
Each energy source in this model is homogeneous and therefore will have a single price. There
are of course many di↵erent reﬁnement methods for oil (and coal and gas) which results in many
di↵erent consumer products. Each of these products has di↵erent prices according to quality,
transportation costs, regulation, market conditions and other factors. I use each variant’s share
of total supply as weights wherever there are several prices in the data.
I assume that both regions purchase fuels at the global weighted average prices, and later
adjust these nominal prices with di↵erent regional purchasing power parity (PPP) price levels
to match units with the Penn World Tables macro data. To create these regional price levels,
I weigh the PPP price levels listed in the Penn World Tables for each country according to
their GDP (in PPP) relative to the total GDP for that region. This adjustment corrects for
the di↵erence in purchasing power between the high-income and low-income regions. I use the
regional share of total demand as weights whenever there are regional or country-level prices
instead of World prices in the data. Ideally, each region in the model would have local nominal
prices but local price data is di�cult to ﬁnd or very costly. It is not obvious that local prices,
e.g. prices of Australian coal, only apply for that country or region since fuels are often traded
on a global or super-regional market. Furthermore, I sort countries into regions based on GDP
which does not necessarily group countries together based on geography or according to which
regional fuel market they mostly trade in.
Coal and natural gas prices are from bp (	2021	) and I get monthly oil product prices from the
International Energy Agency (	n.d.	), converted to a yearly average. Renewable energy prices are
more di�cult to ﬁnd or conceptualize. To properly compare di↵erent energy sources in electricity
production, the lifetime costs and beneﬁts for each type should be included. This is the rationale
for using the so-called levelized costs of electricity (LCOE) measure, which gives the present value
for all costs including investments, operation, maintenance and fuel costs. It is the long-term
break-even cost for an electricity plant. I use IRENA (	2020	) data to get world average LCOE
for each type of renewable energy source. Then I ﬁnd the single calibration renewable price by
weighing each LCOE by share of total renewable production. An undesirable side-e↵ect of using
19

the LCOE only for renewables is that it might skew the relative prices in electricity production
in favor of fossil fuels since the LCOE includes all costs, not only marginal fuel costs.
Bioenergy prices are another source of uncertainty. I use the price of wood pellets in the
US (Anna Simet,	2021	) as a catch-all price. In developing countries, however, bioenergy is not
necessarily commercial-grade but collected and used directly from local sources. This means
that it is di�cult to ﬁnd price data. One could use the price of labor, but it is not clear how
many energy units get collected per work hour. Because of this, I assume for simplicity that
bioenergy prices in the low-income region is half the high-income region price (in addition to the
PPP adjustment between the regions).
I convert the idiosyncratic price units to USD per GJ according to the conversion factors in
table	4.4	.
Table 4.4: Energy units conversion factors	1
Unit GJ
1 tonne of coal equivalent 29.3
1 barrel of oil products (basket	2) 5.34
1 million British thermal units (Btu) 1.06
1 000 gallons of ethanol 84.4
1 t wood pellets 18
1 MWh 3600
1https://www.justinto ols.com/unit-conversion/energy.php2bp (	2021	)
4.2.4 Emissions and Carbon Prices
The energy use of each fuel	eiis measured in Tera joules. I use the factors in table	4.5	to
convert energy to CO	2emissions,	Ei. Note that bioenergy has a non-zero conversion factor. The
reason is that burning of non-renewable waste is included in this category in the IEA data. Waste
incineration is a relatively small part of the total energy use in this category (about 5%), and only
a share of that is non-renewable, but it is at least as carbon intensive as oil (U.S. Environmental
Protection Agency,	2022	). Therefore, I assume that biofuels have 1/20 the carbon intensity of
oil to capture the non-zero emissions in this category.
According to the IEA (	2020	), the total energy-related CO	2emissions were 33	.2 Gigatons in
2019. The energy volume data and emission factors I use gives a few percentage points too low
emissions compared with the IEA estimates. The discrepancy probably comes from imprecise
emissions factors. I increase all emission factors reported in table	4.5	correspondingly in order
to match observed energy-related emissions exactly. This will preserves the within-fossil CO	2
intensities, but may cause a small bias in favor of renewable energy.
The price of emissions vary between countries and regions. According to the World Bank
(2021	), 56 carbon pricing initiatives are implemented globally in 2019, covering about 14.5 % of
global emissions. I weigh the price of each initiative with its covered share of global emissions.
Based on this, I ﬁnd the weighted average CO2 price of covered emissions as 2.75 USD per tonne
of CO	2e (in 2017 USD). Only 14.5 % of total emissions are covered, so the average price for all
emissions would be about 0.40 USD/tCO	2e. No countries in region	Lhave pricing initiatives, so
pEL= 0. Region	H’s share of total emissions is about 80.7 %, which implies	pEH	⇡	2.75/0.807	⇡
0.50 USD/tCO	2e. I correct for PPP price levels in the calibration calculations.
20

Table 4.5: CO	2coe�cients
Source tCO	2/TJ	1
Coal 91
Oil products 69	2
Natural gas 50
Biofuels 3.5	3
Renewables 0
1Source: EIA (	2022	), con-	verted to units of TJ.2Simple average of diesel,gasoline, jet fuel, keroseneand heating oil.3ncludes non-renewablewaste incineration, whichcauses a positive coe�cient.
For regions with a positive carbon tax, the price of emissions,	pEi,t, should already be included
in the end-use energy prices	pei,t. The prices I ﬁnd in the date are gross emission prices, though.
Therefore, I add emissions prices to the observed price of energy. I calibrate	dl,t	according to
pdl,tdl,t	=	pe0,tel,0,t+P
i2Nd(pei,t	+pEi,t)dl,t. I measure the intermediate energy good,	dl,t, in USD
which means I can normalize its prize,	pdl,t, to 1.
4.3 On Elasticities of Substitution
The CES production functions in the model, and therefore their corresponding calibration equa-
tions, depend on the di↵erent elasticities of substitution between factor inputs, where	2�t⌘	11�˜st.
In production theory, the elasticity of substitution is heuristically deﬁned as the ease with which
a pair of input factors can be substituted for one another. The usual intuition is that	�tmeasures
the curvature of a given isoquant of the production function. This parameter is unambiguously
deﬁned for a two-factor production function. Blackorby and Russell (	1981	,1989	) show that the
proper generalization from two factors to	nfactors is the Morishima elasticity of substitution,
MES	i,j	=	✏j,i	�✏i,i,where	✏i,i	is factor	i’s own price elasticity and	✏j,i	the cross-price elasticity
between the goods	iand	j. This elasticity is generally asymmetric, which means that	MES	i,j	is
not necessarily equal to	MES	j,i. However, as shown by Blackorby and Russell (	1981	,1989	), the
constant elasticity of substitution (CES) production function is equivalent to having symmetric
MES.
In the ﬁrst ever meta-analysis on interfuel substitution estimates, converts the various elastic-
ity concept into so-called shadow elasticity of substitution (SES), which is a symmetric cost-share
weighted average of two asymmetric MES. The converted estimates can then be used directly
in interfuel CES production functions. It is not clear, however, how to translate SES estimates
between pairs of inputs to a	n-good CES production function, since there is no obvious way
to weigh each pair’s contribution to the overall elasticity of substitution. The estimates found
in D. I. Stern (	2012	) are often greater than 1 which imply large potential for interfuel substi-
tutability. However, as D. I. Stern (	2012	) points out, the estimates in the literature vary a lot
depending on several decisions and contexts in the underlying studies: econometric estimation
2With the risk of using confusing notation,	�tmeans elasticity of substitution here even if	�denotes temperature	di↵usion in a previous chapter.
21

methods, assumptions about technology, the amount of underlying data, and so on. Some es-
timates in Stern’s (	2012	) meta-regression are signiﬁcantly di↵erent from neither zero nor one,
meaning that neither the Leontief, the Cobb-Douglas nor the gross substitutes case can be re-
jected. Furthermore, the estimates are based on the local political context, market regulatory
systems and factor availability. Local studies can thus not be easily extrapolated to other regions,
which will be necessary in the aggregated regional model I use. These issues cause some concern
for how precise the elasticities ‘picked from the literature’ will be.
In analyses of climate change the time horizon is very long. Therefore, integrated assessment
models should use long-run estimates of sustainability. Unfortunately, such estimates are scarce
and most estimates in Stern’s (	2012	) meta-analysis are short-run. This poses an important and
and di�cult question of how to extrapolate from short-run elasticities to those of longer time
horizons.
Because of these uncertainties, I will do signiﬁcant sensitivity tests with respect to the di↵erent
elasticities to identify how much the key results are a↵ected by di↵erent assumptions. As a
baseline, I pick the elasticity of substitutions as follows, and I assume equal substitutabilities
across both regions for simplicity:
Transport: 0.2	. This is based on the limited substitution in transport found by Serletis et
al. (	2010a	). Bye et al. (	2021	) use a elasticity of 0.5 in the light-duty vehicle production function
between ICE and EVs in their macroeconomic analysis climate regulations of the transportation
sector. However, transportation in my model includes also aviation and shipping. There is
currently impossible to do long-haul shipping, international ﬂights or heavy-duty road freight
transport using electricity, no matter the relative price change. This implies lower elasticity
compared to the personal vehicle market. There are technological developments in short-haul
good transport, ferries, and electric vehicles that could increase this parameter over time.
Industry: 1	. Serletis et al. (	2010a	) ﬁnd limited substitutability in the industrial sector,
except for some fuels in the UK. D. I. Stern (	2012	) and Papageorgiou et al. (	2017	), however,
both ﬁnd greater than unity substitution between most pairs of fuels. I pick an intermediate value
of 1. An elasticity larger than 1 would imply that neither input is essential in production. This
assumption currently seems unrealistic. There are some subindustries with very-hard-to-abate
products which is currently impossible to produce without any fossil fuel input, such as cement
and steel production (International Energy Agency,	2021a	). However, technological development,
such as green steel (Vetter,	2021	), could change this in the future.
Other: 1.2	. Serletis et al. (	2010b	) and Jadidzadeh and Serletis (	2016	) ﬁnd elasticities in
the US and Canadian residential and commercial sectors around 1, while Serletis et al. (	2010a	)
mostly ﬁnd ﬁgures below 1. Wong et al. (	2019	), however, ﬁnd elasticities mostly above unity.
The energy use in the other sector is mainly heating, cooling, cooking, lighting and running
machines and appliances in the commercial, residential and agricultural sectors. There already
exists electrical alternatives to direct fossil use in these sectors, such as electric stoves instead of
gas stoves, heat pumps instead of oil or gas heating, light bulbs instead of kerosene lamps.
Electricity production: 1.8	. Kumar et al. (	2015	), Serletis et al. (	2010a	,2010b	) and
Pelli (	2012	) all ﬁnd less-than-one estimates. There are several reasons: most plants can only
run one fuel, local availability, intermittency, and complementary since fossil electricity is often
used to produce solar panels and windmills (in the beginning). Papageorgiou et al. (	2017	),
however ﬁnd it at 1.8 using a more macroeconomic approach. Technically, there should be
perfect substitutability between energy sources in the production of electricity, since all the inputs
are identical. But there are deﬁnely idiosyncratic factors such as market regulations, resource
availability, transmission constraints that hinders the full substitutability of factors. Moreover,
di↵erent fuel sources have di↵erent technical and physical constraints. Renewable energy has
variable production based on weather conditions and seasons, causing potential intermittency
22

issues in lack of su�cient storage capacity. And coal plants are expensive and time-consuming
to start up and shut down, so they are best used as a peak-load input. Therefore, I pick an
non-inﬁnite elasticity, but larger than 1 to reﬂect that no sources should alone be necessary in
production.
Empirical estimates must necessarily be based on historical data, which pins down the func-
tional form of the production functions today. But these functions may very well change in the
future. For example, Xie and Hawkes (	2015	) ﬁnd that the elasticity of substitution between
oil and electricity in the Chinese transport sector has increased steadily between the 1980s and
2010. Bye et al. (	2021	) calibrate their model to a low present-day elasticity, but argue for a
large expected increase in the substitutability between EVs and ICEs in Norway in the coming
decades. Wong et al. (	2019	) ﬁnd that the interfuel substitution in several Chinese sector becomes
easier as the country steps up the energy ladder. I omit this feature from the present analysis.
23

Chapter 5
Results
In this chapter I will ﬁrst brieﬂy present some of the calibrated parameters and discuss how I
adjusted parameters that caused calibration issues. Then I will present graphical and numerical
results from di↵erent model runs.
5.1 Calibration results
The	⌫l’s in table	5.1	are calibrated as described in equation (	A.4	). The Penn World Table
has only data for capital, labor and GDP, not value of energy or resource rents. Thus, the
directly calibrated	↵’s probably include some of energy’s value share. Therefore I correct these
by subtracting the mean	⌫in the economy to get the results in table	5.1	. I could not ﬁnd reliable
capital data for reﬁned energy production. The initial calculation gave ¯	↵’s that were very high
and close to 1. I could not get the model calibrated correctly when using the data-based values
for ¯	↵. Therefore, I adjusted them until the model better reproduced the calibration period data.
The ﬁt improved as the values approached zero, potentially signaling that the calibration process
and model do not handle capital in reﬁned energy correctly.
Table 5.1: Calibrated parameters in base year.
Explanation	j=	Lj	=	H
↵	Capital share in ﬁnal goods production 0.209 0.336
⌫tr	Energy share in transport 0.454 0.334
⌫in	Energy share in industry 0.045 0.030
⌫ot	Energy share in other sector 0.066 0.025
¯↵	Capital share in reﬁned energy production	1	0.010 0.005
1Assumption.
I also had to use another solution to calibrate ¯	athan equation (	A.11	). This Leontief parameter
is supposed to convert between labor in energy and emissions to match emissions data. However,
it would convert low energy use to high emissions and could not reproduce emissions. Therefore,
I replace ¯	aby the vector of emissions factors in table	4.5	. This will per deﬁnition convert energy
units to CO	2and solves the calibration issue. Finally, I also had issues with the calibration of
At. Both wages and output in all three ﬁnal goods sectors were o↵by a ﬁxed factor. I ﬁxed it by
multiplying	Atby this factor. It was likely a bug causing this issue, but after these corrections
24

Output	Emissions	Temperature increase
Figure 5.1: Output, emissions and temperature.
the optimized model was able to reproduce the data and the solved tra jectories seems reasonable
and robust. Further bug-testing was outside the scope of the thesis.
5.2 Model results
In this section I will ﬁrst discuss the di↵erence between the business-as-usual scenario and the
optimal-tax scenario. This part will be brief, since it is not my main interest in this thesis.
Second, I will present the results of low and high di↵usion. Finally, I will do a sensitivity
analysis.
5.2.1 Baseline and optimal tax
This section shows the di↵erence between the business-as-usual baseline scenario and the optimal-
tax scenario. In the former, the economy is assumed to be structured similarly to the calibration
period. In the latter, each of the social planners implements an optimal carbon tax. Both have
no growth in energy producing sectors and no inter-regional di↵usion.
The left panel in ﬁgure	5.1	shows that outputs in both regions are increasing exponentially
over time. This is a factor of the positive but decreasing TFP growth rates in both economies
(see table	4.2	). Region H abates relatively much compared to region L as is seen in the middle
panel. One explanation is that the optimal carbon tax is equal to the social cost of carbon within
each economy	1. Still, even as a regional tax, the optimal tax causes the temperature increases
to be moderate compared to the business-as-usual case, as seen in the right panel in ﬁgure	5.1	.
Coal has the largest response to the optimal carbon tax in this model. Figure	5.2	shows that
region	L	cuts coal-use almost in half, while it goes from being the most to the least used fuel
in region	H. The other fossil fuels fall in both regions, with oil falling the least. Waste-burning
is a part of bioenergy here, which ceteris paribus should decrease its optimal usage after tax.
However, there is a slight increase in both regions. One reason could be that it substitutes for
more emission-intensive fuels. Renewables increases in both regions, but electricity production
falls. The increase in renewable energy is not su�cient to replace the decline of fossil fuels.
The di↵erence in magnitudes between the regions could be attributed to the di↵erent SCCs.
1This is not a general result, but a function of how much each region internalizes the cost caused on others by	themselves. I have assumed that they only care about future generations within their regions, and not in otherregions.
25

Region L	Region H
Figure 5.2: Energy by source. Dotted lines are baseline and solid lines for the case with optimal
tax. Note the di↵erence in scales.
Region	H	has a high SCC and therefore abates more. Also, the emissions by the poorer and
lower populated region	L	are relatively small. Therefore, they will not cause much temperature
increase and thus little damages which implies a low SCC.
Figures	5.3	and	5.4	show the energy use by sector in region	L	and	H,respectively. These
ﬁgures make the di↵erence in magnitudes of abatement even clearer. All sectors decrease the use
of fossil fuels. The largest abatement e↵ect is in the electricity sectors.
5.2.2 Di↵usion
Here I present three di↵erent comparisons. First, I compare the baseline scenario from the
previous chapter to the present case with growth in renewable energy technology. Second and
third, I change the baseline to be the renewable growth-case and compare it to a low- and
high-di↵usion scenario, respectively.
Technological growth and no di↵usion:	The baseline scenario is identical to the previous
chapter: no carbon tax, and no growth expect for total TFP growth. I set the inherent growth
rates for renewable technology in region	H	as	gH,t	=2020	=	�H,t	=2020	=0	.009 in the ﬁrst period,
and in region	L	as the constant	�L,t	=0	.001 . These parameters are then aligned with the
average growth rates of wind and solar PV as percentage of the electricity supply found in the
analysis of renewable energy growth rates and S-shaped transitions by Cherp et al. (	2021	). I let
gH,t	decline by 2% each year after the year 2030 to capture that some developed countries could
be past the inﬂection point on an S-shaped transition curve (Cherp et al.,	2021	). This means
that	gH,t	=2050	=0	.006 and	gH,t	=2100	=0	.002. Developing countries in region	L	are currently
further left on the S-curve. I assume that all technological growth in excess of	�L,t	are imported
from region	H. This mechanism is captured by the di↵usion parameter	✓L. In this ﬁrst scenario
Iset	✓L= and compare the scenario with technological growth to the no-growth baseline from
the previous section.
Figures	5.7	and	5.8	reveal that growth in renewable energy technology increases electricity
production, as is to be expected. Furthermore, as seen most clearly in the lower right panel of
26

Region L: Energy use by sector
Transport	Industry
Other	Electricity
Figure 5.3: Energy by sector in region	L. Dotted lines are baseline and solid lines for the case
with optimal tax. Note the di↵erence in scales.
27

Transport	Industry
Other	Electricity
Figure 5.4: Energy by sector in region	H. Dotted lines are baseline and solid lines for the case
with optimal tax. Note the di↵erence in scales.
Output	Emissions	Temperature increase
Figure 5.5: Output, emissions and temperature – with growth in renewable energy.
28

Region L	Region H
Figure 5.6: Energy by source with growth in renewable energy. Dotted lines are baseline and
solid lines for the case with growth. Note the di↵erence in scales.
ﬁgure	5.3	, the increase in renewable energy causes a transition away from the other inputs to
electricity production. It is not clear whether the relatively large decrease of coal use in electricity
production is due to its high emission intensity or its high initial share in production.
Another interesting ﬁnd is that oil use in transport is actually increasing. This is likely due
to the inter-fuel complementarity in the transportation sector. The production functions are not
nested between fuels, and all inputs enter on the same level. Therefore, electricity and oil are
complements for an elasticity of substitution	<	1. An interesting extension could be to introduce
time-changing elasticity in the transportation sector to analyze how this relationship can change
over time. However, total oil use decreases in both regions, as seen in ﬁgure	5.6	.
Output in both regions increases and emissions decreases, causing a reduction in temperature
increase. To get the the welfare in utils for both regions, I utilize the ACE numerical optimization
script in Matlab. I calculate the sum of discounted welfare in each period of the planning horizon,
which is 200 year. This gives me the approximate net present value of the scenario. Then I take
the di↵erent of this net present value in utils and convert it to consumption equivalents in USD.
I calculate the welfare e↵ect of the assumed growth in renewable technology to be	⇡	10 trillion
USD for region	H	and	⇡	0.9 trillion USD for region	L, compared to the no-growth baseline.
Low and high di↵usion:	The new baseline is the previous case with growth in renewables,
but without di↵usion. Now I analyse two cases and compare them to the local-growth-only
baseline. The results are qualitatively similar, so I only present ﬁgures for the high case here since
they give clearer visual e↵ects, and put ﬁgures for the low case in the Appendix for completeness.
Iset	✓L=0	.0001 in the low case, and	✓L=0	.002 in the high case. The rate of di↵erence in
renewable technology levels between	H	and	L,¯AL,t	,is	⇡	3.5 in 2020 for the low case. This
means that	gL,t	=0	.0001	⇤3.5+0	.001 = 0	,00135 in the ﬁrst period.
Figures	5.9	and	5.10	reveal some interesting dynamics. Firstly, growth in region	H	is initially
29

Transport	Industry
Other	Electricity
Figure 5.7: Energy by sector in region	L. Dotted lines are baseline and solid lines for the case
with growth. Note the di↵erence in scales.
30

Transport	Industry
Other	Electricity
Figure 5.8: Energy by sector in region	H. Dotted lines are baseline and solid lines for the case
with growth. Note the di↵erence in scales.
Growth rate in renewable energy
Figure 5.9: Growth rates of renewable energy technology by region (low di↵usion).
31

Growth rate in renewable energy
Figure 5.10: Growth rates of renewable energy technology by region (high di↵usion).
Output	Emissions	Temperature increase
Figure 5.11: Output, emissions and temperature. No vs. high di↵usion.
32

Region L	Region H
Figure 5.12: Energy by source with growth in renewable energy. Dotted lines are with no di↵usion
and solid lines for high di↵usion. Note the di↵erence in scales.
high, whereby it declines gradually at 2% a year in both scenarios. Secondly, for the high-di↵usion
case, growth in region	L	becomes higher than in region	H	and stays so for the remainder of the
time horizon. This is partly an e↵ect of the constant	�L,t	=0	.001, which introduces a minor
inconsistency into the analysis. Without that, the growth rates should both converge to zero.
More importantly, the higher growth rate captures the catching-up e↵ect, where technologically
lagging regions experiences faster growth that developed regions, because they can quickly gain
technological progress by using existing innovations instead.
The right panel of ﬁgure	5.12	shows that there are no e↵ects on energy production in region
H, with the solid lines covering the dashed lines. Region	H’s emissions and output also stay
identical as evident from ﬁgure	5.11	. However, there is a positive welfare e↵ect in region	H
which I calculate to be about 40 billion USD for the low case and around 370 billion USD in the
high case. This comes from the indirect e↵ect on region	H	welfare through lower emissions from
region	L. This in turn gives a lower increase in temperature which increases welfare in region
H, as is seen the last panel in ﬁgure	5.11	. The di↵usion e↵ect increases electricity use in region
Land ampliﬁes the qualitative e↵ects from introducing growth alone (see ﬁgures	5.12	and	5.13	).
The welfare e↵ect to region	L	is about 190 billion USD in the low case and 1800 billion USD in
the high case.
There are signiﬁcant gains to be made for both regions by increasing di↵usion. I ﬁnd the
welfare e↵ect of di↵usion to be between 40 billion and 370 billion USD dollars for the richer
region. These e↵ects are only indirect, through lower global temperature and lower climate-
related damages. The low-income region has a welfare e↵ect between 190 billion and 1800 billion
USD. The energy composition and emissions of the richer region are not decreasing in di↵usion,
so the welfare e↵ects on the low-income region is strictly local and direct. Next is to do a
sensitivity analysis of these results with respect to the elasticities of substitution.
33

Transport	Industry
Other	Electricity
Figure 5.13: Energy by sector in region	L. Dotted lines are with no di↵usion and solid lines for
high di↵usion. Note the di↵erence in scales.
Transport	Industry
Other	Electricity
Figure 5.14: Energy by sector in region	L, with low substitutability in the electricity sector.
Dotted lines are with no di↵usion and solid lines for high di↵usion. Note the di↵erence in scales.
34

Output	Emissions	Temperature increase
Figure 5.15: Output, emissions and temperature. No vs. high di↵usion. Low substitutability.
5.2.3 Sensitivity analysis
In this sensitivity analysis I use the high-di↵usion case as the baseline and then vary the elastici-
ties of substitution. In the ﬁrst scenario I decrease the elasticity of substitution in the electricity
sector from 1.8 to just below unity. The results are striking, but maybe expected. For a below-
unity substitutability, the inputs in the electricity sector are gross complements. Thus, increasing
the use of one factor will increase the use of all other factors to keep output constant. The e↵ects
of this is apparent in the bottom right panel of ﬁgure	5.14	. The implied increase in in fossil fuels
increase region	L’s emissions, thus resulting in a negative welfare e↵ect of di↵usion for region	H.
The net beneﬁt is still positive, but I expect it to decrease in lower inter-fuel substitutability.
This e↵ect also relies on the functional forms imposed by the model. Numerical models often
use nested pairs of CES production functions to have a more ﬂexible hierarchy of inputs, and to
allow varying degrees of substitutability between inputs. The present production function has
all inputs in one level which preserves notational tractability and readability but that comes at
the cost of a more ﬂexible production function.
The second sensitivity analysis I make is to increase the elasticity of substitution in transport
from 0.2 to 0.5, capturing the potential for a fast global transition to electric or hydrogen fuel
cell based transportation	2The results can be seen in ﬁgures	5.16	and	5.17	. The energy mix is
not very sensitive to the transportation elasticity. There is a slight decrease in oil use in the
transportation sector compared to the base case, but not enough to change welfare e↵ects much.
The CES production reproduces value shares given the calibrated share. Since oil has a very
large current share (around 95% globally), I do not expect to see large changes in oil until the
elasticity of substitution increases above unity.
2Green hydrogen is made by electrolysis, a electricity-intensive production process. Thus hydrogen and elec-	tricity are equivalent energy carriers in my model.
35

Transport	Industry
Other	Electricity
Figure 5.16: Energy by sector in region	L, with low substitutability in the electricity sector.
Dotted lines are with no di↵usion and solid lines for high di↵usion. Note the di↵erence in scales.
36

Output	Emissions	Temperature increase
Figure 5.17: Output, emissions and temperature. No vs. high di↵usion. Low substitutability.
37

Chapter 6
Conclusion
In this thesis I have attempted to ﬁnd the e↵ects on the energy composition and welfare of
technological di↵usion between regions. To quantify these e↵ects, I calibrated a state-of-the-
art analytical integrated assessment model from scratch and introduced a simple technological
di↵usion mechanism in renewable energy to the base model.
I found that there are signiﬁcant welfare gains to be made for both regions by increasing
technological di↵usion. I estimate the welfare e↵ects of di↵usion to be between 40 billion and
370 billion US dollars for the richer regionm depending on the degree of di↵usion. These e↵ects
are indirect to the region, and comes only through lower global temperature and lower climate-
related damages. The low-income region has a welfare e↵ect between 190 billion and 1800 billion
USD. The energy composition and emissions of the richer region are not decreasing in di↵usion,
so the welfare e↵ects on the low-income region is strictly local and direct.
These results depend crucially, however, on the inter-fuel elasticity of substitution in the
electricity-producing sector in the receiving economy. If renewable energy is a gross complement
to fossil fuels in electricity production, emission increases and the welfare e↵ect for the richer
region turns negative.
This thesis has made a contribution to the literature on analytical integrated assessment
models and I have narrowed the gap between analytic and numerical models. There are sev-
eral interesting extensions to make in future research. First, one could make the technological
di↵usion model more realistic either by using other functional forms, such as logistic growth
functions, or by endogenizing technology investment. Second, implementing dynamic elastici-
ties of substitution has the potential to capture future transitions in transportation, electricity
production and industry.
38

Bibliography
Acemoglu, D. (2009).	Introduction to Modern Economic Growth	. Princeton University Press.
Anna Simet. (2021).	A Resilient Year	. Biomassmagazine.com. Retrieved May 11, 2022, from
http://biomassmagazine.com/articles/17970/a- resilient- year
Barrage, L. (2019). Optimal Dynamic Carbon Taxes in a Climate–Economy Model with Dis-
tortionary Fiscal Policy.	The Review of Economic Studies	, rdz055.	https://doi.org/10.
1093/restud/rdz055
Baumstark, L., Bauer, N., Benke, F., Bertram, C., Bi, S., Gong, C. C., Dietrich, J. P., Dirnaich-
ner, A., Giannousakis, A., Hilaire, J., Klein, D., Koch, J., Leimbach, M., Levesque, A.,
Madeddu, S., Malik, A., Merfort, A., Merfort, L., Odenweller, A., . . . Luderer, G. (2021).
REMIND2.1: Transformation and innovation dynamics of the energy-economic system
within climate and sustainability limits.	Geoscientiﬁc Model Development	,14	(10), 6571–
6603.	https://doi.org/10.5194/gmd- 14- 6571- 2021
Blackorby, C., & Russell, R. R. (1981). The Morishima Elasticity of Substitution; Symmetry,
Constancy, Separability, and its Relationship to the Hicks and Allen Elasticities.	The
Review of Economic Studies	,48	(1), 147.	https://doi.org/10.2307/2297127
Blackorby, C., & Russell, R. R. (1989). Will the Real Elasticity of Substitution Please Stand
Up? (A Comparison of the Allen/Uzawa and Morishima Elasticities).	The American
Economic Review	,79	(4), 882–888.
Bond, K., Arunabha Ghosh, Edward Vaughan, & Harry Benham. (2021).	Reach for the sun: The
emerging market electricity leapfrog	. Carbon Tracker. London.
Bond, K., Edward Vaughan, & Harry Benham. (2020).	Nothing to lose but your chains: The
emerging market transport leapfrog	. Carbon Tracker. London.
Bosetti, V. (2021). Integrated Assessment Models for Climate Change. In	Oxford Research Ency-
clopedia of Economics and Finance	. Oxford University Press.	https://doi.org/10.1093/
acrefore/9780190625979.013.572
Bosetti, V., Carraro, C., Galeotti, M., Massetti, E., & Tavoni, M. (2006). WITCH - A World
Induced Technical Change Hybrid Model.	SSRN Electronic Journal	.https://doi.org/10.
2139/ssrn.948382
bp. (2021).	Statistical Review of World Energy	. bp. Retrieved May 2, 2022, from	https://www.
bp.com/en/global/corporate/energy- economics/statistical- review- of- world- energy.html
Brock, W. A., & Mirman, L. J. (1972). Optimal economic growth and uncertainty: The discounted
case.	Journal of Economic Theory	,4(3), 479–513.	https : / / doi . org / 10 . 1016 / 0022 -
0531(72)90135- 4
Bye, B., Kaushal, K. R., Rosnes, O., Turner, K., & Yonezawa, H. (2021). The road to a low
emission society: Costs of interacting climate regulations, 42.
Cai, Y., Newth, D., Finnigan, J., & Gunasekera, D. (2015). A hybrid energy-economy model for
global integrated assessment of climate change, carbon mitigation and energy transfor-
mation.	Applied Energy	,148	, 381–395.	https://doi.org/10.1016/j.apenergy.2015.03.106
39

Cherp, A., Vinichenko, V., Tosun, J., Gordon, J. A., & Jewell, J. (2021). National growth dynam-
ics of wind and solar power compared to the growth required for global climate targets.
Nature Energy	,6(7), 742–754.	https://doi.org/10.1038/s41560- 021- 00863- 0
Clarke, L., Krey, V., Weyant, J., & Chaturvedi, V. (2012). Regional energy system variation in
global models: Results from the Asian Modeling Exercise scenarios.	Energy Economics	,
34	, S293–S305.	https://doi.org/10.1016/j.eneco.2012.07.018
Drupp, M. A., Freeman, M. C., Groom, B., & Nesje, F. (2018). Discounting Disentangled.	Amer-
ican Economic Journal: Economic Policy	,10	(4), 109–134.	https://doi.org/10.1257/pol.
20160240
EIA - U.S. Energy Information Administration. (2022).	Carbon Dioxide Emissions Coe�cients	.
Retrieved May 2, 2022, from	https://www.eia.gov/environment/emissions/co2	vol	mass.
php
Farmer, J. D., Hepburn, C., Mealy, P., & Teytelboym, A. (2015). A Third Wave in the Economics
of Climate Change.	Environmental and Resource Economics	,62	(2), 329–357.	https :
//doi.org/10.1007/s10640- 015- 9965- 2
Feenstra, R. C., Inklaar, R., & Timmer, M. P. (2015). The Next Generation of the Penn World
Table.	American Economic Review	,105	(10).	www.ggdc.net/pwt
Fujimori, S., Masui, T., & Matsuoka, Y. (2017). AIM/CGE V2.0 Model Formula. In S. Fuji-
mori, M. Kainuma, & T. Masui (Eds.),	Post-2020 Climate Action: Global and Asian
Perspectives	(pp. 201–303). Springer.	https://doi.org/10.1007/978- 981- 10- 3869- 3	12
Gerlagh, R., & Liski, M. (2018a). Consistent climate policies.	Journal of the European Economic
Association	,16	(1), 1–44.	https://doi.org/10.1093/jeea/jvx010
Gerlagh, R., & Liski, M. (2018b). Carbon Prices for The Next Hundred Years.	The Economic
Journal	,128	(609), 728–757.	https://doi.org/10.1111/eco j.12436
Golosov, M., Hassler, J., Krusell, P., & Tsyvinski, A. (2014). Optimal Taxes on Fossil Fuel in
General Equilibrium.	Econometrica	,82	(1), 41–88.
Gu, G., Wang, Z., & Wu, L. (2021). Carbon emission reductions under global low-carbon tech-
nology transfer and its policy mix with R&D improvement.	Energy	,216	, 119300.	https:
//doi.org/10.1016/j.energy.2020.119300
Harvey, F. (2021). What are the key points of the Glasgow climate pact? [newspaper].	The
Guardian: Environment	. Retrieved May 20, 2022, from	https://www.theguardian.com/
environment/2021/nov/14/what- are- the- key- points- of- the- glasgow- climate- pact- cop26
Hassler, J., & Krusell, P. (2012). ECONOMICS AND CLIMATE CHANGE: INTEGRATED
ASSESSMENT IN A MULTI-REGION WORLD.	Journal of the European Economic
Association	,10	(5), 974–1000.	https://doi.org/10.1111/j.1542- 4774.2012.01082.x
Hoel, M., & Karp, L. (2002). Taxes versus quotas for a stock pollutant.	Resource and Energy
Economics	, 18.
Hope, C. (2006). The Marginal Impact of CO2 from PAGE2002: An Integrated Assessment
Model Incorporating the IPCC’s Five Reasons for Concern.	The Integrated Assessment
Journal	,6(1), 19–56.	https://www3.nd.edu/	⇠nmark/Climate/Hope	2006.pdf
Howard, P. H., & Sterner, T. (2017). Few and Not So Far Between: A Meta-analysis of Climate
Damage Estimates.	Environmental and Resource Economics	,68	(1), 197–225.	https : / /
doi.org/10.1007/s10640- 017- 0166- z
IAM Consortium. (2022).	IAMC wiki - IAMC-Documentation	. Retrieved May 21, 2022, from
https://www.iamcdocumentation.eu/index.php/IAMC	wiki
Interagency Working Group on Social Cost of Greenhouse Gases. (2021). Technical Support
Document: Social Cost of Carbon, Methane,	https://www.whitehouse.gov/wp- content/
uploads/2021/02/TechnicalSupportDocument	SocialCostofCarbonMethaneNitrousOxide.
pdf
40

International Energy Agency. (n.d.).	Prices – Key World Energy Statistics 2020 – Analysis	. IEA.
Retrieved May 11, 2022, from	https://www.iea.org/reports/key- world- energy- statistics-
2020/prices
International Energy Agency. (2019).	Data and statistics	. IEA. Retrieved May 9, 2022, from
https://www.iea.org
International Energy Agency. (2020, June 16).	Global Energy Review 2019: The latest trends in
energy and emissions in 2019	.OECD.	https://doi.org/10.1787/90c8c125- en
International Energy Agency. (2021a).	Net Zero by 2050 - A Roadmap for the Global Energy
Sector	.
International Energy Agency. (2021b). Renewables 2021 - Analysis and forecast to 2026, 175.
International Labour Organization. (2022).	ILOSTAT	. Data. Retrieved March 30, 2022, from
https://ilostat.ilo.org/data/
Investopedia. (2022).	Markets Today	. Investopedia. Retrieved May 9, 2022, from	https : / / www .
investopedia.com/markets/
IPCC. (2001).	Working Group III: Mitigation - 7.6.4 Integrated Assessment Models	.IPCC-
Intergovernmental Panel on Climate Change. Retrieved February 15, 2022, from	https:
//archive.ipcc.ch/ipccreports/tar/wg3/index.php?idp=311
IRENA. (2020). Renewable Power Generation Costs 2020, 180.	https://www.irena.org/- /media/
Files / IRENA / Agency / Publication / 2021 / Jun / IRENA	Power	Generation	Costs	2020 .
pdf
Iverson, T., & Karp, L. (2021). Carbon Taxes and Climate Commitment with Non-constant Time
Preference.	The Review of Economic Studies	,88	(2), 764–799.	https://doi.org/10.1093/
restud/rdaa048
Jadidzadeh, A., & Serletis, A. (2016). Sectoral Interfuel Substitution in Canada: An Application
of NQ Flexible Functional Forms.	The Energy Journal	,37	(2).	https://doi.org/10.5547/
01956574.37.2.a jad
Karp, L., & Traeger, C. P. (2021). Smart Cap.	SSRN Electronic Journal	.https : / / doi . org / 10 .
2139/ssrn.3798938
Karp, L., & Zhang, J. (2006). Regulation with anticipated learning about environmental damages.
Journal of Environmental Economics and Management	,51	(3), 259–279.	https://doi.org/
10.1016/j.jeem.2005.09.006
Karp, L., & Zhang, J. (2012). Taxes versus quantities for a stock pollutant with endogenous
abatement costs and asymmetric information.	Economic Theory	,49	(2), 371–409.	https:
//doi.org/10.1007/s00199- 010- 0561- y
Kaya, A., Csala, D., & Sgouridis, S. (2017). Constant elasticity of substitution functions for
energy modeling in general equilibrium integrated assessment models: A critical review
and recommendations.	Climatic Change	,145	(1-2), 27–40.	https : / / doi . org / 10 . 1007 /
s10584- 017- 2077- y
Klump, R., McAdam, P., & Willmann, A. (2011).	The normalized CES production function:
Theory and empirics	(No. 1294). European Central Bank.
Krey, V., Havlik, P., Kishimoto, P., Fricko, O., Zilliacus, J., Gidden, M., Strubegger, M., Kar-
tasasmita, G., Ermolieva, T., Forsell, N., Guo, F., Gusti, M., Huppmann, D., Johnson, N.,
Kikstra, J., Kindermann, G., Kolp, P., Lovat, F., McCollum, D., . . . Riahi, K. (2020).
MESSAGEix-GLOBIOM Documentation - 2020 release.	https : / / doi . org / 10 . 22022 /
IACC/03- 2021.17115
Kumar, S., Fujii, H., & Managi, S. (2015). Substitute or complement? Assessing renewable and
nonrenewable energy in OECD countries.	Applied Economics	,47	(14), 1438–1459.	https:
//doi.org/10.1080/00036846.2014.997922
41

Leimbach, M., Roming, N., Schultes, A., & Schwerho↵, G. (2018). Long-Term Development
Perspectives of Sub-Saharan Africa under Climate Policies.	Ecological Economics	,144	,
148–159.	https://doi.org/10.1016/j.ecolecon.2017.07.033
Lucas, P. L., Nielsen, J., Calvin, K., L. McCollum, D., Marangoni, G., Streﬂer, J., van der Zwaan,
B. C., & van Vuuren, D. P. (2015). Future energy system challenges for Africa: Insights
from Integrated Assessment Models.	Energy Policy	,86	, 705–717.	https : / / doi . org / 10 .
1016/j.enpol.2015.08.017
Masson-Delmotte, V., Zhai, P., Pirani, A., Connors, S. L., P´ean, C., Berger, S., Caud, N., Chen,
Y., Goldfarb, L., Gomis, M. I., Huang, M., Leitzell, K., Lonnoy, E., Matthews, J. B. R.,
Maycock, T. K., Waterﬁeld, T., Yelek¸ci,	¨O., Yu, R., & Zhou, B. (Eds.). (2021). Summary
for policymakers. In	Climate Change 2021: The Physical Science Basis. Contribution of
Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on
Climate Change	. Cambridge University Press.
McCOLLUM, D., Nagai, Y., Riahi, K., Marangoni, G., Calvin, K., Pietzcker, R., Van Vliet, J., &
Van Der Zwaan, B. (2013). ENERGY INVESTMENTS UNDER CLIMATE POLICY: A
COMPARISON OF GLOBAL MODELS.	Climate Change Economics	,04	(04), 1340010.
https://doi.org/10.1142/S2010007813400101
Newell, R. G., & Pizer, W. A. (2003). Regulating stock externalities under uncertainty.	Journal
of Environmental Economics and Management	,45	(2), 416–432.	https : / / doi . org / 10 .
1016/S0095- 0696(02)00016- 5
Nordhaus, W. D. (1993). Optimal Greenhouse-Gas Reductions and Tax Policy in the ”DICE”
Model.	The American Economic Review	,83	(2), 313–317.
Nordhaus, W. D. (2011).	Integrated Economic and Climate Modeling	(No. 1839). Retrieved
February 17, 2022, from	https://papers.ssrn.com/sol3/papers.cfm?abstract	id=1970295
Nordhaus, W. D., & Yang, Z. (1996). A Regional Dynamic General-Equilibrium Model of Alter-
native Climate-Change Strategies.	The American Economic Review	,Vo l . 8 6	, 741–765.
OECD. (2022).	OECD Data	. Retrieved May 19, 2022, from	http://data.oecd.org
OECD & Eurostat. (2012).	Eurostat-OECD Methodological Manual on Purchasing Power Pari-
ties (2012 Edition)	.https://www.oecd- ilibrary.org/content/publication/9789264189232-
en
of Statistic of China, N. B. (2021).	China Statistical Yearbook 2020	. Retrieved May 9, 2022, from
http://www.stats.gov.cn/tjsj/ndsj/2020/indexeh.htm
Papageorgiou, C., Saam, M., & Schulte, P. (2017). Substitution between Clean and Dirty Energy
Inputs: A Macroeconomic Perspective.	The Review of Economics and Statistics	,99	(2),
281–290.	https://doi.org/10.1162/REST	a00592
Pelli, M. (2012). The Elasticity of Substitution between Clean and Dirty Inputs in the Production
of Electricity, 40.
Pindyck, R. S. (2013). Climate Change Policy: What Do the Models Tell Us?	Journal of Economic
Literature	,51	(3), 860–872.	https://doi.org/10.1257/jel.51.3.860
Pindyck, R. S. (2017). The Use and Misuse of Models for Climate Policy.	Review of Environmental
Economics and Policy	,11	(1), 100–114.	https://doi.org/10.1093/reep/rew012
Pindyck, R. S. (2019). The social cost of carbon revisited.	Journal of Environmental Economics
and Management	,94	, 140–160.	https://doi.org/10.1016/j.jeem.2019.02.003
Pizer, W. A. (1999). The optimal choice of climate change policy in the presence of uncertainty,
33.
Rezai, A., & Van der Ploeg, F. (2016). Intergenerational Inequality Aversion, Growth, and the
Role of Damages: Occam’s Rule for the Global Carbon Tax.	Journal of the Association
of Environmental and Resource Economists	,3(2), 493–522.	https : / / doi . org / 10 . 1086 /
686294
42

Rutherford, T. F. (2002). Lecture Notes on Constant Elasticity Functions.
Schauer, M. J. (1995). Estimation of the greenhouse gas externality with uncertainty.	Environ-
mental & Resource Economics	,5(1), 71–82.	https://doi.org/10.1007/BF00691910
Serletis, A., Timilsina, G. R., & Vasetsky, O. (2010a). International Evidence on Sectoral Interfuel
Substitution.	The Energy Journal	,31	(4), 1–29.
Serletis, A., Timilsina, G. R., & Vasetsky, O. (2010b). Interfuel substitution in the United States.
Energy Economics	,32	(3), 737–745.	https://doi.org/10.1016/j.eneco.2010.01.013
Stern, D. I. (2012). Interfuel Substitution: A Meta-Analysis.	Journal of Economic Surveys	,26	(2),
307–331.	https://doi.org/10.1111/j.1467- 6419.2010.00646.x
Stern, N. (2022). A Time for Action on Climate Change and A Time for Change in Economics.
The Economic Journal	, ueac005.	https://doi.org/10.1093/ej/ueac005
the World Bank. (2021).	Carbon Pricing Dashboard	. Retrieved May 12, 2022, from	https : / /
carbonpricingdashboard.worldbank.org/map	data
Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R., & de Vries, G. J. (2015). An Illus-
trated User Guide to the World Input-Output Database: The Case of Global Automo-
tive Production: User Guide to World Input-Output Database.	Review of International
Economics	,23	(3), 575–605.	https://doi.org/10.1111/roie.12178
Traeger, C. (2018).	ACE – Analytic Climate Economy (with Temperature and Uncertainty)
(SSRN Scholarly Paper No. ID 3307622). Social Science Research Network. Rochester,
NY.	https://doi.org/10.2139/ssrn.3307622
Traeger, C. (2022a). ACE – Analytic Climate Economy.	American Economic Journal: Economic
Policy (Forthcoming))	.https://doi.org/10.2139/ssrn.3832722
Traeger, C. (2022b). IAMs and CO2 Emissions – An Analytic Discussion (working paper), 90.
United Nations Statistics Division. (2022).	UNSD	.https://unstats.un.org/home/
Unruh, G. C. (2000). Understanding carbon lock-in.	Energy Policy	,28	(12), 817–830.	https :
//doi.org/10.1016/S0301- 4215(00)00070- 7
U.S. BUREAU OF LABOR STATISTICS. (2022).	Productivity Data By Sector, Industry, and
State	. Productivity Data By Sector, Industry, and State. Retrieved March 30, 2022, from
https://www.bls.gov/lpc/tables	by	sector	and	industry.htm
U.S. Environmental Protection Agency. (2022). Emission Factors for Greenhouse Gas Inventories.
https://www.epa.gov/climateleadership/ghg- emission- factors- hub
van den Bijgaart, I., Gerlagh, R., & Liski, M. (2016). A simple formula for the social cost
of carbon.	Journal of Environmental Economics and Management	,77	, 75–94.	https :
//doi.org/10.1016/j.jeem.2016.01.005
Van Der Ploeg, F., & Withagen, C. (2014). GROWTH, RENEWABLES, AND THE OPTIMAL
CARBON TAX: GROWTH, RENEWABLES, CARBON TAX.	International Economic
Review	,55	(1), 283–311.	https://doi.org/10.1111/iere.12049
Van Der Zwaan, B. C. C., R¨osler, H., Kober, T., Aboumahboub, T., Calvin, K. V., Gernaat,
D. E. H. J., Marangoni, G., & McCOLLUM, D. (2013). A CROSS-MODEL COMPAR-
ISON OF GLOBAL LONG-TERM TECHNOLOGY DIFFUSION UNDER A 2	°C CLI-
MATE CHANGE CONTROL TARGET.	Climate Change Economics	,04	(04), 1340013.
https://doi.org/10.1142/S2010007813400137
Vetter, D. (2021).	How Sweden Delivered The World’s First Fossil Fuel-Free Steel	. Forbes. Re-
trieved May 19, 2022, from	https://www.forbes.com/sites/davidrvetter/2021/08/19/
how- sweden- delivered- the- worlds- ﬁrst- fossil- fuel- free- steel/
Waldho↵, S., Antho↵, D., Rose, S., & Tol, R. S. J. (2014). The Marginal Damage Costs of Dif-
ferent Greenhouse Gases: An Application of FUND.	Economics	,8(1), 20140031.	https:
//doi.org/10.5018/economics- ejournal.ja.2014- 31
43

Wang, Z., Gu, G., Wu, J., & Liu, C. (2016). CIECIA: A new climate change integrated assessment
model and its assessments of global carbon abatement schemes.	Science China Earth
Sciences	,59	(1), 185–206.	https://doi.org/10.1007/s11430- 015- 5141- 3
Wong, D., Mugera, A., & White, B. (2019). The Interfuel Substitutability in China’s Energy
Transition: A National and Sectoral Level Analyses Using Normalized Quadratic Func-
tion.	SSRN Electronic Journal	.https://doi.org/10.2139/ssrn.3360786
Xie, C., & Hawkes, A. D. (2015). Estimation of inter-fuel substitution possibilities in China’s
transport industry using ridge regression.	Energy	,88	, 260–267.	https://doi.org/10.1016/
j.energy.2015.05.034
Zhang, S., Bauer, N., Luderer, G., & Kriegler, E. (2014). Role of technologies in energy-related
CO2 mitigation in China within a climate-protection world: A scenarios analysis using
REMIND.	Applied Energy	,115	, 445–455.	https://doi.org/10.1016/j.apenergy.2013.10.
039
Zhang, S., Bauer, N., Yin, G., & Xie, X. (2020). Technology learning and di↵usion at the global
and local scales: A modeling exercise in the REMIND model.	Technological Forecasting
and Social Change	,151	, 119765.	https://doi.org/10.1016/j.techfore.2019.119765
44

Appendix A
Mathematical derivations
In this Appendix I solve the cost minimization problems for the representative ﬁrms at each
nesting level to obtain their compensated factor demand functions, whose inversions inform the
calibration equation I will use	1. The equations are identical for both regions, so I omit regional
indexing. I initially planned to let the elasticities of substitutions in the model be time-dependent,
but it turned out to be outside the scope of the thesis as mentioned in the main text. As a result
of this path dependency, however, I calculate the inverted compensated demand functions instead
of using the more practical and straight-forward calibrated share form equations in Rutherford
(2002	). The calibrated share form would have permitted a direct interpretation of the CES-
parameters as the value shares of the inputs in the calibration year, instead of the more abstract
parameters below that depend on the elasticity of substitution	2. But it is not clear that it is
possible to use this form when substitutability itself is changing over time. In the following, I
approximately follow the approach laid out in the appendix to Traeger (	2022b	).
A.1 Final goods aggregation
The representative ﬁrm has the following problem:
min	cl,t,8l2Nc
X
l2Nc
pl,tcl,t	s.t.	At
⇣X
l2Nc
al,tcsl,t
⌘1s=	Yt	and	X
l
al,t	=1	.
The Lagrangian to this problem is
L=	X
l2Nc
pl,tcl,t	+
"
Yt�At
⇣X
l2Nc
al,tcsl,t
⌘1s
#
where	�is the shadow value and where I have omitted the second constraint which holds trivially.
The constraints hold for given right-hand-sides, but I omit special notation. The ﬁrst order
1Iinvokethesecondfundamentaltheoremofwelfareeconomics,whichstatesthatanyPareto-optimalallocation	could be realized as a decentralized equilibrium. This model is nicely behaved with the normal convexity and hasalogarithmicutilityfunction,withinteriorsolutionstothemaximizationproblem. Therearenoexternalitieswithin each	social planner’s domain (although the	global	problem does not have a Pareto optimal decentralized	solution).	2See e.g. Klump et al. (	2011	)forathoroughexplanation.
45

condition for ﬁnal good	lis
pl,t	=	�A	t
⇣X
l2Nc
al,tcsl,t
⌘1s�1altcs�1l,t	,
which solved for	cl,t	gives
cl,t	=
"
�A	tal,t
pl,t
⇣X
l
al,tcsl,t
⌘1s�1#	11�s
,	(A.1)
still a function of the shadow price. Multiply both sides by	pl,t, sum over all goods and divide
by output to get the average ﬁnal goods unit costs
UC	t⌘
P
lpl,tcl,t
Yt	=	1
Yt�A	t
⇣X
l
al,tcsl,t
⌘⇣	X
l
al,tcsl,t
⌘1s�1=	�,
which shows that the average unit costs equal the shadow price. Since the production functions
exhibit constant returns to scale, the marginal cost will equal units costs. And since the repre-
sentative ﬁrm is a price-taker, marginal costs equal marginal prices. Thus,	�=	pt.Iinsertthis
into (	A.1	) and solve for	al,t	to get
al,t	=	A1/st	pl,t
pt
✓cl,t
Yt
◆1�s
.
All goods are measured in calibration-period USD, so I set the ﬁnal good prices to unity. I do
the same for	pt. Furthermore, I want the aggregate good to be measured in USD and not in
‘utils’. Therefore, I make the following transformation and get the ﬁnal calibration expression
a⇤l,t	=
✓cl,t
Yt
◆1�s
with	At=
⇣X
l2Nc
a⇤l,t
⌘1s	and	al,t	=	a⇤l,t	P
l2Nca⇤l,t.	(A.2)
A.2 Final consumption good production
The representative ﬁrm production good	lhas the following problem:
min	Kl,t,Nl,t,dl,t	pKtKl,t	+pNtNl,t	+pdl,tdl,t	s.t.	Al,tK↵ll,tN1�↵l�⌫i	l,t	d⌫il,t	=	cl,t.
Note that the prices of capital and labor are sector-invariant, which is a result of e�cient alloca-
tion of these resources. Setting up the Lagrangian and solving the ﬁrst order equations provides
the ﬁrst order conditions below. Furthermore, since there is constant returns to scale and the
ﬁrms are price takers,	�=	pl,t. Inserting for this gives the compensated demand functions
Kl,t	=	pl,t	↵l
pKtcl,t,N	l,t	=	pl,t1�↵l�⌫l
pNt	cl,t,d	l,t	=	pl,t	⌫l
pdl,t
cl,t.	(A.3)
The parameter	⌫lis sector speciﬁc, and I calibrate it according to
⌫i=	pdl,tdl,t
pl,tcl,t	.	(A.4)
46

As explained in chapter 3,	↵lmust be sector-independent. Therefore, I use	Kt=	P
lKl,t	and	Yt
to calibrate it:
↵l⌘	↵=	pKtKt
ptYt	=	Kt
Yt,	(A.5)
where I have set the prices to unity since I measure capital and output both in calibration-period
consumption units. Finally, I calibrate	Al,t	to match total sector output
Al,t	=	cl,t
K↵l,tN1�↵�⌫i	l,t	d⌫ll,t
,
which is a function of the endogenous control variables	Kl,t,Nl,t	and	dl,t. Since these are
measured in calibration-period USD, I set the prices in (	A.3	) to 1 and insert equation (	A.3	)to
get
Al,t	=	cl,t
c1�⌫l	l,t	↵↵⇣1�↵�⌫i	pNt
⌘1�↵�⌫ld⌫ll,t
=	↵�↵✓	pNt
1�↵�⌫i
◆1�↵�⌫l✓cl,t
dl,t
◆⌫l
=	↵�↵⌫�⌫l	l
✓	pNt
1�↵�⌫i
◆1�↵�⌫l
.	(A.6)
A.3 Intermediate energy and electricity producers
The ﬁrms producing the intermediate energy goods solve the following problem:
min	el,i,t	,i2Nd
X
i2Nd
pei,tel,i,t	s.t.	˜Al,t
⇣X
i2Nd
˜al,i,t	el,i,t	˜sl⌘1˜sl=	dl,t,
and the electricity producing ﬁrm solves
min	e0,i,t	,i2Ne
X
i2Ne
pei,te0,i,t	s.t.	˜A0,t
⇣X
i2Ne
˜a0,i,t	e0,i,t	˜sl⌘1˜sl=	e0,t.
Using the same methods as above, I get compensated demand solved for the CES parameters in
the intermediate energy producing sectors as
˜a⇤l,i,t	=	˜A
1˜sll,t
pei,t
pdl,t
✓el,i,t
dl,t
◆1�˜sl
8l2Nc
with	˜Al,t	=
 	X
i2Nd
˜a⇤l,i,t
!	1˜sl
and ˜	al,i,t	=	˜a⇤l,i,t	P
i2Nd˜a⇤l,i,t	.	(A.7)
For the electricity sector, I get
˜a⇤0,i,t	=	˜A1/˜sl	0,t
pei,t
pe0,t
✓e0,i,t
e0,t
◆1�˜s0,t
with	˜A0,t=
 X
i2Ne
˜a⇤0,i,t
!	1˜s0,t
and ˜	a0,i,t	=	˜a⇤0,i,t	P
i2Ne˜a⇤0,i,t	.	(A.8)
47

A.4 Reﬁned energy production
The producers of reﬁned energy face the following cost-minimization problem:
min	¯Ki,t,¯Ni,t,Ei,t	pKt	¯Ki,t	+pNt	¯Ni,t	+pEi,tEi,t	s.t.	¯Ai,t	¯K	¯↵ii,t	min	{¯Ni,t,¯ai,tEi,t}1�¯↵i=	ei,t.
The optimal factor demand for a Leontief production function has the factors in equal propor-
tions. Using this, I rewrite the problem to
min	¯Ki,t,¯Ni,t,Ei,t	pKt	¯Ki,t	+pNt	¯Ni,t	+pEi,tEi,t	s.t.	¯Ai,t	¯K	¯↵ii,t	¯N1�¯↵i	i,t	=	ei,t	and ¯	ai,tEi,t	=	¯Ni,t.
The Lagrangian is
L=	pKt	¯Ki,t	+pNt	¯Ni,t	+pEi,tEi,t	+��ei,t	�	¯Ai,t	¯K	¯↵ii,t	¯N1�¯↵i	i,t
�+µ�¯Ni,t	�	¯ai,tEi,t�,
which has the ﬁrst order conditions
(i)	pKt	=	�¯↵i¯Ai,t	¯K	¯↵i�1	i,t	¯N1�¯↵i	i,t	)	pKt	¯Ki,t	=	�¯↵iei,t
(ii)	pNt	=	�(1	�	¯↵i)ei,t¯Ni,t	�µ)	(pNt+µ)¯Ni,t	=	�(1	�	¯↵i)ei,t
(iii)	pEi,t	=¯	aµ.
I insert (iii) into (ii) to eliminate	µ	and assume that price equals marginal and average costs
which eliminates	�to get
¯Ki,t	=¯	↵ipei,t
pKtei,t	(A.9)
¯Ni,t	=	(1	�	¯↵i)pei,tei,t
pNt+	pEi,t¯ai,t
and (A.10)
Ei,t	=	¯Ni,t
¯ai,t	=	1
¯ai,t
(1	�	¯↵i)pei,tei,t
pNt+	pEi,t¯ai,t
=	(1	�	¯↵i)pei,t
¯ai,tpNt+pEi,t	ei,t,
where I used equation (	A.10	) to eliminate the endogenous	¯Ni,t, and which solved for ¯	ai,t	gives
the calibration equations
¯ai,t	=	¯Ni,t
Ei,t	=(1	�	¯↵i)pei,t
pNt
ei,t
Ei,t	�	pEi,t
pNt	8i2Id.	(A.11)
Remember that capital’s share ¯	↵imust be equal across the producers at this level in the nested
production function. I therefore calibrate ¯	↵	⌘	¯↵iusing equation (	A.9	),	¯Kt⌘	P
i2Ne¯Ki,t	and
petet⌘	P
i2Nepei,tei,t	to get the ‘mean’ capital share of the reﬁned energy sector
¯↵=	¯K,t
petet,	(A.12)
where I used the normalization	pKt	= 1.
48

Next is to calibrate the	¯Ai,t’s to match output. Rewriting the production function, and using
the sector-independent ¯	↵, I get
¯Ai,t	=	ei,t
¯K¯↵i,t	¯N1�¯↵	i,t
.
I use equations (	A.9	) and (	A.10	) to eliminate the endogenous	¯Ki,t	and	¯Ni,t
¯Ai,t	=
ei,t
✓
pNt+	pEi,t¯ai,t
◆1�¯↵
pKt¯↵
�¯↵p	ei,tei,t�¯↵�(1	�	¯↵)pei,tei,t�1�¯↵
=	pei,t�1

pNt+	pEi,t
¯ai,t
!1�¯↵
,	(A.13)
where I used the normalization	pKt	= 1 and	�⌘	�¯↵¯↵(1	�	¯↵)1�¯↵��1.
49

Appendix B
A more general di↵usion model
Recall equation (	3.4	)
dl,t	=	˜Al,t
⇣X
i2Nd
˜al,i,t	e˜sl,tl,i,t
⌘	1˜sl,t
which for each region gives the production of the intermediate energy input in sector	las a
function of electricity	el,0,tand other reﬁned energy inputs. Let	Bl,i,t	be the exogenous factor
augmenting technology for energy input	iin sector	lin a given region. Then
dl,t	=	˜Al,t
⇣X
i2Nd
˜al,i,t	(Bl,i,t	el,i,t	)˜sl,t⌘	1˜sl,t	(B.1)
makes factor-augmenting technological growth in the energy sector explicit.
Let
gl,i,t	⌘	�Al,i,t
Al,i,t	(B.2)
be the growth rate of technology for input	iin sector	lat time	t,with	Al,i,	0>	0. Collect all
the entries	Al,i,t	for region	jin the	l⇥	imatrix	Aj,t, and let	gj,t	be a	l⇥	imatrix with the
corresponding technology growth rates from (	B.2	) for region	jin period	t. The world technology
frontier is represented by	Atwith growth rates	gt. The technological di↵usion equation	1of
motion for each region	jis
Aj,t+1	=	✓j⇤(At�Aj,t)+(	1+�j)⇤Aj,t,	(B.3)
where	⇤is the element-wise matrix multiplication operator and	1is a matrix of ones. The time-
constant and exogenous di↵usion parameters has elements	✓j,l,i	2[0,1]and	�j,l,i	2[0,gl,i]2.The
technology absorption rate	✓jcaptures the varying degrees with which regions are able to import
various technologies given their policies, institutions and other economic and social factors. This
rate is exogenous in the current model, but it is possible to endogenize it as a function of e.g.
human capital (see Acemoglu (	2009	)). The absorption rate captures the delay or imperfectness
1The following is based on the continuous time single-good model in Acemoglu (	2009	), chapter 18.2.1.	2The original single-good law of motion in Acemoglu (	2009	)has	✓j2(0,1). I think that there is a typo here.	Having	✓j>1impliesthataregioncouldabsorbmorethanthedi↵erencebetweentheworldfrontieranditsown	technology level, which is hard to argue. That is why I have closed the interval at an upper boundary of 1.
50

with which technology becomes available across the world. For	✓j=	0, there is no di↵usion.
For	✓j=	1all sectors catch up to the World frontier without delay. If the country is the World
leader in any technology,	Al,i	=	Aj,l,i	, and there is no di↵usion. The local innovation parameter
�jdetermines the locally sourced growth rate in technology. It captures how fast a region uses
it existing knowledge	Ajto improve technology. At the boundaries of the permitted interval, a
region can either have no local innovation or innovate at the World frontier growth rate.
Equation (	B.3	) can be rearranged to get the technology growth rates directly for each sector
land energy input	i
gj,t	=	✓j˜Aj,t	+�j	(B.4)
where	˜Aj,t	⌘	At�Aj,t	Aj,t	2	[0,1) is the rate of di↵erence in technology levels between the World
frontier and region	j. For a positive di↵usion rate, a region with a high local innovation rate can
have a higher technological growth rate than regions on the World frontier. This captures the
catching-up e↵ect.
The discrete-time model used in this thesis allow discontinuous paths of development, which
encapsulates the potential for leap-frogging, where a region does not necessarily visit all the same
‘stops’ along the technological growth path of the World frontier.
51

Appendix C
Supplementary ﬁgures
52

Output	Emissions	Temperature increase
Figure C.1: Output, emissions and temperature. No vs. low di↵usion.
Region L	Region H
Figure C.2: Energy by source with growth in renewable energy. Dotted lines are with no di↵usion
and solid lines for low di↵usion. Note the di↵erence in scales.
53

Transport	Industry
Other	Electricity
Figure C.3: Energy by sector in region	L. Dotted lines are with no di↵usion and solid lines for
low di↵usion. Note the di↵erence in scales.
54

Appendix D
Supplementary tables
Table D.1: List of countries in each region
1	Region L	Region H
1	Albania	Anguilla
2	Algeria	Antigua and Barbuda
3	Angola	Argentina
4	Bangladesh	Armenia
5	Barbados	Aruba
6	Belize	Australia
7	Benin	Austria
8	Bhutan	Azerbaijan
9	Bolivia (Plurinational State of )	Bahamas
10	Bosnia and Herzegovina	Bahrain
11	Burkina Faso	Belarus
12	Burundi	Belgium
13	Cabo Verde	Bermuda
14	Cambodia	Botswana
15	Cameroon	Brazil
16	Central African Republic	British Virgin Islands
17	Chad	Brunei Darussalam
18	Colombia	Bulgaria
19	Comoros	Canada
20	Congo	Cayman Islands
21	Cˆote d’Ivoire	Chile
22	D.R. of the Congo	China
23	Djibouti	China, Hong Kong SAR
24	Dominica	China, Macao SAR
25	Ecuador	Costa Rica
26	Egypt	Croatia
27	El Salvador	Cura¸cao
28	Eswatini	Cyprus
29	Ethiopia	Czech Republic
30	Fiji	Denmark
55

31	Gambia	Dominican Republic
32	Ghana	Equatorial Guinea
33	Guatemala	Estonia
34	Guinea	Finland
35	Guinea-Bissau	France
36	Guyana	Gabon
37	Haiti	Georgia
38	Honduras	Germany
39	India	Greece
40	Indonesia	Grenada
41	Iran (Islamic Republic of )	Hungary
42	Iraq	Iceland
43	Jamaica	Ireland
44	Jordan	Israel
45	Kenya	Italy
46	Kyrgyzstan	Japan
47	Lao People’s DR	Kazakhstan
48	Lesotho	Kuwait
49	Liberia	Latvia
50	Madagascar	Lebanon
51	Malawi	Lithuania
52	Mali	Luxembourg
53	Mauritania	Malaysia
54	Mongolia	Maldives
55	Morocco	Malta
56	Mozambique	Mauritius
57	Myanmar	Mexico
58	Namibia	Montenegro
59	Nepal	Montserrat
60	Nicaragua	Netherlands
61	Niger	New Zealand
62	Nigeria	North Macedonia
63	Pakistan	Norway
64	Paraguay	Oman
65	Peru	Panama
66	Philippines	Poland
67	Republic of Moldova	Portugal
68	Rwanda	Qatar
69	Sao Tome and Principe	Republic of Korea
70	Senegal	Romania
71	Sierra Leone	Russian Federation
72	South Africa	Saint Kitts and Nevis
73	Sri Lanka	Saint Lucia
74	St. Vincent and the Grenadines	Saudi Arabia
75	State of Palestine	Serbia
76	Sudan	Seychelles
77	Syrian Arab Republic	Singapore
78	Ta jikistan	Sint Maarten (Dutch part)
79	Togo	Slovakia
56

80	Tu n i s i a	Slovenia
81	U.R. of Tanzania: Mainland	Spain
82	Uganda	Suriname
83	Ukraine	Sweden
84	Uzbekistan	Switzerland
85	Venezuela (Bolivarian Republic
of )
Taiwan
86	Viet Nam	Thailand
87	Ye m e n	Trinidad and Tobago
88	Zambia	Tu r k e y
89	Zimbabwe	Turkmenistan
90	Turks and Caicos Islands
91	United Arab Emirates
92	United Kingdom
93	United States
94	Uruguay
57

Appendix E
Wrangling script
58

Table of Contents
Data Wrangling for ACE %	 .................................................................................................................. 1
Initial setup	 ........................................................................................................................................ 1
PART 1: READ IN RAW DATA AND BASIC CLEANING: %%	 .............................................................. 1
RICE REGIONS	 ................................................................................................................................. 2
PWT 2019 (Aggregate macro GDP, capital, labor, growth rates etc. Country level.)	 ........................................ 2
MAKING A TABLE WITH MODEL COUNTRY NAMES, 3-LETTER CODES, RICECODES AND RICE
REGIONS.	 ......................................................................................................................................... 5
UN Data. (FOR SECTOR DATA)	 ......................................................................................................... 5
WORLD INPUT-OUTPUT DATABASE (FOR MISSING COUNTRIES IN UN)	 .......................................... 9
ILOSTAT LABOR DATA	 .................................................................................................................. 12
IEA VOLUME BALANCES	 ............................................................................................................... 16
PWT TRANSPORT CAPITAL DETAIL DATA	 .................................................................................... 23
OECD CAPITAL DETAILS DATA ()	 .................................................................................................. 24
US BLS DETAILED CAPITAL DATA FOR FUEL CAPITAL USE	 ......................................................... 26
CHINA CAPITAL DETAILS	 .............................................................................................................. 27
PART 2: COMBINING PRE-PROCESSED DATA INTO USABLE SPREADSHEETS FOR
CALIBRATION %%	 ......................................................................................................................... 28
MERGING FOR FINAL GOOD SECTORS CALIBRATION	 ................................................................... 28
MERGING AND PREPARING INTERMEDIATE AND REFINED ENERGY SECTORS	 ............................. 31
MERGING INTO ONE BIG SHEET	 .................................................................................................... 32
Functions %%	 ................................................................................................................................... 32
Data Wrangling for ACE %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Initial setup
clear 	all
clc
% Wrangle these:
wrangle.RICE = 1; 	% List of RICE codes and region names
wrangle.PWT = 1; 	% Time series of country-level macro data
wrangle.UN = 1; 	% Sector output data (2019 mostly, some 2018)
wrangle.WIOD = 1; 	% Sector output data missing from UN (2014)
wrangle.ILOSTAT = 1; 	% Sector labor data
wrangle.IEA = 1; 	% Sector energy volumes
wrangle.OECD = 1; 	% Sector capital data
wrangle.USBLS = 1; 	% Fine sector capital data (for fuels)
wrangle.sectors = 1; 	% Dataset merging for export
wrangle.energy = 1; 	% Energy data merging for export
PART 1: READ IN RAW DATA AND BASIC
CLEANING: %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1
59

RICE REGIONS
if	 wrangle.RICE == 1
 	% This is a cleaned list to match the available PWT data.
 	% There are more countries in RICE, but these are the ones for which we
 have PWT data
 raw.RICE_regions = readtable(	"raw_data\our_regions_clean.csv"	);
 RICE_regions = renamevars(raw.RICE_regions,
 [	"country_name"	 "country_code"	], [	"country"	 "countrycode"	]);
 RICE_regions.short_name = [];
 	% Converting variables to categories (more functionality, less
 	% memory required)
 RICE_regions = categorizer(RICE_regions, width(RICE_regions));
 	% Merging RICE list with ISO list. Keeping RICE names.
 RICE_regions =
 outerjoin(RICE_regions,isoCountryList,	"LeftKeys"	,"countrycode"	,"RightKeys"	,"alpha_3"	);
 	% Keeping only RICE countries (because of PWT data limitations)
 RICE_regions(isundefined(RICE_regions.country),:) = [];
 RICE_regions(:,[5 6 7 9 12 13 14 15]) = [];
 	% % Saving
 wranglingFile.RICE_regions = RICE_regions;
 save(	"calibData.mat"	, 	"RICE_regions"	, 	"-append"	);
 	%save("rawData.mat", "raw");
else
 load(	"calibData.mat"	, 	"RICE_regions"	);
 	%load("rawData.mat", "raw");
end
% END RICE REGIONS
Error using readtable
Unable to find or open 'raw_data\our_regions_clean.csv'. Check the path and
filename or file permissions.
Error in wranglingScript (line 31)
 raw.RICE_regions = readtable("raw_data\our_regions_clean.csv");
PWT 2019 (Aggregate macro GDP, capital, la-
bor, growth rates etc. Country level.)
if	 wrangle.PWT == 1
 raw.PWT = readtable(	"raw_data\pwt100.xlsx"	, 	"Sheet"	, 	"Data"	);
2
60

PWT = categorizer(raw.PWT, 2);
 	% Adding developed/developing regional codes
 	% NOTE: PWT data is in millions of people. I just want people.
 	% NOTE: PWT data is in millions of USD. I want just USD.
 PWT.cgdpo = 1000000.*PWT.cgdpo;
 PWT.cn = 1000000.*PWT.cn;
 PWT.emp = 1000000.*PWT.emp;
 PWT.pop = 1000000.*PWT.pop;
 PWT.gdpPerCap = PWT.cgdpo ./ PWT.pop; 	% USD per capita
 PWT.thesisCode = ~(PWT.gdpPerCap < 14031 | isnan(PWT.gdpPerCap)); 	%
 GdpPerEmp lower than China is defined as developing.
 	% Getting labsh as extensive variable
 PWT.labshExt = PWT.labsh .* PWT.cgdpo;
 	% Converting the logical to categorical for future use
 	% Note: The normal way was buggy, so had to use workaround.
 PWT = convertvars(PWT, 	"thesisCode"	, 	'categorical'	); 	% Logical to double
 PWT.thesisCode = renamecats(PWT.thesisCode, [	"true"	 "false"	], [	"2"	 "1"	]);
 	% JOINING PWT WITH RICE REGIONS (easy, both has three letter abbr. code)
 PWT = join(PWT, RICE_regions, Keys=	"countrycode"	);
 PWT(:, 	"country_RICE_regions"	) = [];
 	% Saving dynamic PWT data for growth rate calculations
 PWT2019 = PWT(PWT.year == 2019,:);
 countries2019 = unique(PWT2019.countrycode);
 pwtDynamic = PWT(PWT.year > 2009,:);
 PWT = PWT(PWT.year == 2019,:);
 thesis1 = unique(PWT(PWT.thesisCode ==	"1"	, 	"countrycode"	));
 idl = pwtDynamic.year == 2019 & pwtDynamic.thesisCode == 	"1"	;
 lowCountries = pwtDynamic.countrycode(idl);
 idl = pwtDynamic.year == 2019 & pwtDynamic.thesisCode == 	"2"	;
 highCountries = pwtDynamic.countrycode(idl);
 	for	 k = 1:numel(lowCountries)
 id = pwtDynamic.year ~= 2019 & pwtDynamic.countrycode ==
 lowCountries(k);
 pwtDynamic.thesisCode(id) = 	"1"	;
 	end
 	for	 k = 1:numel(highCountries)
 id = pwtDynamic.year ~= 2019 & pwtDynamic.countrycode ==
 highCountries(k);
3
61

pwtDynamic.thesisCode(id) = 	"2"	;
 	end
 PWTsanity1 = PWT;
 G = findgroups(PWT.thesisCode);
 UWplPPP = splitapply(@mean, PWT.pl_gdpo, G);
 	% Sum of GDP per thesis region
 G = findgroups(PWT.thesisCode);
 sumGDP = splitapply(@sum, PWT.cgdpo, G);
 	% Sum of GDP
 sumGDPWorld = sum(PWT.cgdpo);
 PWT.GDPweight = ones(183,1);
 PWT.GDPweightWorld = ones(183,1);
 	%Low income
 PWT{PWT.thesisCode == 	"1"	, 	"GDPweight"	} = PWT.cgdpo(PWT.thesisCode
 == 	"1"	) / sumGDP(1);
 	%High income
 PWT{PWT.thesisCode == 	"2"	, 	"GDPweight"	} = PWT.cgdpo(PWT.thesisCode
 == 	"2"	) / sumGDP(2);
 	%World
 PWT{:, 	"GDPweightWorld"	} = PWT.cgdpo / sumGDPWorld;
 	% Sanity check, each group should sum to 1
 G = findgroups(PWT.thesisCode);
 sumWeights = splitapply(@sum, PWT.GDPweight, G);
 	%Passed
 PWT.weighedPL = PWT.GDPweight .* PWT.pl_gdpo;
 PWT.weighedPLWorld = PWT.GDPweightWorld .* PWT.pl_gdpo;
 	% Finding the weighted average PPP adjusted price levels.
 G = findgroups(PWT.thesisCode);
 wPriceLevel = splitapply(@sum, PWT.weighedPL, G);
 wPriceLevelWorld = sum(PWT.weighedPLWorld);
 wPriceLevel = [wPriceLevelWorld; wPriceLevel];
 	% Removing year variable for easy merging with non-2019 data later
 PWT(:, 	"year"	) = [];
 save(	"calibData.mat"	, 	"PWT"	, 	"-append"	); 	% A full-data 2019 version for
 calibration code
 save(	"calibData.mat"	, 	"pwtDynamic"	 , 	"-append"	); 	% A full-data 2010-2019
 for growth rates
 save(	"calibData.mat"	, 	"wPriceLevel"	, 	"-append"	);
 	%save("rawData.mat", "raw", "-append");
else
 load(	"calibData.mat"	, 	"PWT"	, 	"pwtDynamic"	, 	"wPriceLevel"	);
4
62

%load("rawData.mat", "raw");
end
% END PWT
MAKING A TABLE WITH MODEL COUNTRY
NAMES, 3-LETTER CODES, RICECODES AND
RICE REGIONS.
Subsetting FROM PWT only the country list and codes
countriesPWTRICE = unique([PWT(:,	"countrycode"	), PWT(:,	"country_PWT"	),
 PWT(:,	"rice_code"	), PWT(:,	"rice_region"	), PWT(:, 	"country_code"	),
 PWT(:,	"thesisCode"	),PWT(:, 	"region"	)]);
countriesPWTRICE = renamevars(countriesPWTRICE, [	"country_PWT"	], [	"country"	]);
% IEA data lacks fine level data for some African, Asian and American
 countries.
% Must add these regions manually to country list and assign regions
tabOthAf ={	"OAF"	, 	"Other Africa"	, 	"9"	, 	"Africa"	, 	"99999"	, 	"1"	, 	"Africa"	};
tabOthAm ={	"OAM"	, 	"Other non-OECD
 Americas"	, 	"10"	, 	"LatAm"	, 	"88888"	, 	"1"	, 	"Americas"	};
tabOthAs ={	"OAS"	, 	"Other non-OECD Asia"	, 	"12"	, 	"OthAs"	, 	"77777"	, 	"1"	, 	"Asia"	};
countriesPWTRICE = [countriesPWTRICE; tabOthAf; tabOthAm; tabOthAs];
% END SPECIAL COUNTRYCODES TABLE
UN Data. (FOR SECTOR DATA)
if	 wrangle.UN == 1
 	% First, I must read in the two tables and separately combine the sectors
 	% whithin each system because ofdifferent naming and grouping of sectors.
 	% (See:
 	% https://ilostat.ilo.org/resources/concepts-and-definitions/
classification-economic-activities/)
 	% Second, I must join the two tables to get all the countries in one
 table.
 	% Third, I must update the country list to match the PWT list.
 	% Fourth, I must decide on which sectors to put in which aggregate. What
 to
 	% do with electricity and mining? They should go "below" my three sectors,
 	% which is an argument of using the output method. But then, however, many
 	% other things go "below" mu three sectors as well (such as vehicle
 	% manufacturing, which messes up the energy intermediate interpretation).
5
63

% Fifth, I must find the proper sectoral shares.
 	% Sixth, I must split the table in three -- one for each sector. These
 	% tables become the foundation on which to add further calibration
 	% variables.
 	% These are the variables I need from the UN data
 unVars =
 [	"country"	 "year"	 "sector"	 "accountingType"	 "currency"	 "series"	 "isic"	 "value"	];
 	% READING IN AND COMBINING TABLES
 raw.UN_ISIC4 = readtable(	"raw_data\UN output gva current prices
 ISIC4.txt"	);
 isic = ones(height(raw.UN_ISIC4),1);
 raw.UN_ISIC4.isic = 4.*isic; 	% Assigning ISIC version number for
 later identification
 raw.UN_ISIC3 = readtable(	"raw_data\UN_output_sectors.txt"	);
 isic = ones(height(raw.UN_ISIC3),1);
 raw.UN_ISIC3.isic = 3.*isic;
 raw.UN = [raw.UN_ISIC4;raw.UN_ISIC3]; 	% Vertical append
 unData = renamevars(raw.UN,
 [	"CountryOrArea"	 "Year"	 "SubItem"	 "Item"	 "Currency"	 "Series"	 "isic"	 "Value"	],
 unVars);
 unData = unData(:, unVars);
 unData = categorizer(unData,5);
 	% Keeping value added (GDP) and output. NOTE: A MISTAKE TO KEEP VA.
 	% BUG.
 	% unData = unData(unData.accountingType == "Equals: VALUE ADDED,
 GROSS, at basic prices" | ...
 	% unData.accountingType == "Output, at basic
 prices",:);
 unData = unData(unData.accountingType == 	"Output, at basic prices"	,:);
 unData(:,	"accountingType"	) = [];
 	% Listing all sectors to see which are redundant
 sectorsAndIsicTab = unData(:,[	"sector"	 "isic"	]);
 [~, sectorsAndIsic] = findgroups(sectorsAndIsicTab);
 	% Removing pre-aggregated sectors and subsectors that could cause issues
 in summation
 unData((unData.sector == 	"Agriculture, hunting and related service
 activities (01)"	 | 	...
 unData.sector == 	"Agriculture, hunting, forestry; fishing (A+B)"	 | 	...
 unData.sector == 	"Crop and animal production, hunting and related
 service activities (01)"	 | 	...
 unData.sector == 	"Education; health and social work; other community,
 social and personal services (M+N+O)"	 | 	...
 unData.sector == 	"Financial intermediation; real estate, renting and
 business activities (J+K)"	 | 	...
 unData.sector == 	"Fishing and aquaculture (03)"	 | 	...
 unData.sector == 	"Forestry and logging (02)"	 | 	...
 unData.sector == 	"Forestry, logging and related service activities
 (02)"	 | 	...
6
64

unData.sector == 	"Manufacturing, mining and quarrying and other
 industrial activities (B+C+D+E)"	 | 	...
 unData.sector == 	"Other service activities (R+S+T)"	 | 	...
 unData.sector == 	"Professional, scientific, technical, administrative
 and support service activities (M+N)"	 | 	...
 unData.sector == 	"Public administration and defence, education, human
 health and social work activities (O+P+Q)"	 | 	...
 unData.sector == 	"Wholesale and retail trade, transportation and
 storage, accommodation and food service activities (G+H+I)"	 | 	...
 unData.sector == 	"Wholesale retail trade, repair of motor vehicles,
 motorcycles, etc.; hotels and restaurants (G+H)"	), 	...
 :) = [];
 	% Finding the remanding list of sectors, for merging
 sectorsAndIsicTab = unData(:,[	"sector"	 "isic"	]);
 [~, sectorsAndIsic] = findgroups(sectorsAndIsicTab);
 	% Defining model sectors (see sectorsAndIsic for list of all sectors)
 electricitySector = [{	'Electricity, gas and water supply (E)'	}
 {	'Electricity, gas, steam and air conditioning supply (D)'	}];
 miningSector = [{	'Mining and quarrying (B)'	 }
 {	'Mining and quarrying (C)'	}];
 transportSector = [{	'Transportation and storage (H)'	}
 {	'Transport, storage and communications (I)'	}];
 industrySector = [{	'Construction (F)'	}
 {	'Manufacturing (C)'	 }
 {	'Manufacturing (D)'	}];
 otherSector = [{	'Accommodation and food service activities (I)'	}
 {	'Administrative and support service activities (N)'	}
 {	'Agriculture, forestry and fishing (A)'	}
 {	'Arts, entertainment and recreation (R)'	}
 {	'Education (P)'	}
 {	'Financial and insurance activities (K)'	}
 {	'Fishing (B)'	}
 {	'Human health and social work activities (Q)'	}
 {	'Information and communication (J)'	}
 {	'Other service activities (S)'	}
 {	'Private households with employed persons (P)'	}
 {	'Private households with employed persons (T)'	}
 {	'Professional, scientific and technical activities (M)'	}
 {	'Public administration and defence; compulsory social security (O)'	}
 {	'Public administration and defense; compulsory social security (L)'	}
 {	'Real estate activities (L)'	}
 {	'Water supply; sewerage, waste management and remediation activities
 (E)'	}
 {	'Wholesale and retail trade; repair of motor vehicles and motorcycles
 (G)'	}];
 industrySector = [industrySector;electricitySector;miningSector];
 	% Merging detailed sectors into model sectors
 unData.sector = mergecats(unData.sector, transportSector, 	"Transport (and
 storage)"	);
7
65

unData.sector = mergecats(unData.sector, industrySector, 	"Industry
 (manufacturing and construction)"	);
 unData.sector = mergecats(unData.sector, otherSector, 	"Other"	);
 	%unData.sector = mergecats(unData.sector, electricitySector,
 "Electricity");
 	%unData.sector = mergecats(unData.sector, miningSector, "Mining");
 	% Fixing name differences between UN and PWT. There are 147 countries in
 the UN data and
 	% 183 in PWT. Sorting out spelling differences.
 countriesUN = unique(unData.country);
 countriesPWT = unique(countriesPWTRICE.country);
 inUNnotinPWT = setdiff(countriesUN, countriesPWT);
 inPWTnotinUN = setdiff(countriesPWT,countriesUN);
 unData.country = renamecats(unData.country, 	...
 [	"China, Hong Kong Special Administrative Region"	 ...
 	"China, Macao Special Administrative Region"	 ...
 	"Czechia"	 ...
 	"Tanzania - Mainland"	 ...
 	"Saint Vincent and the Grenadines"	], 	...
 [string(countriesPWT(countriesPWT == 	"China, Hong Kong SAR"	)) 	...
 string(countriesPWT(countriesPWT == 	"China, Macao SAR"	)) 	...
 string(countriesPWT(countriesPWT == 	"Czech Republic"	)) 	...
 string(countriesPWT(countriesPWT == 	"U.R. of Tanzania: Mainland"	)) 	...
 string(countriesPWT(countriesPWT == 	"St. Vincent and the
 Grenadines"	))]);
 countriesUNcheck = unique(unData.country);
 	% Merging some sub-countries into mother country to better mix with
 	% other datasets
 unData.country = mergecats(unData.country,
 [	"Denmark"	 "Greenland"	], 	"Denmark"	);
 	% Collecting sectors
 unData = varfun(@sum, unData, 	...
 GroupingVariables=[	"country"	 "year"	 "sector"	], 	...
 InputVariables=	"value"	);
 	% Merging territories into mother country
 unData.country = mergecats(unData.country,
 [	"Denmark"	 "Greenland"	], 	"Denmark"	);
 	% ADDING COUNTRY CODES TO UN DATA
 	% Note: The joining key is the 3-letter country code because of spelling
 differences in full country names.
 	% JOINING UNDATA TO THE COUNTRYLISTS AND 3-LETTER CODE FROM RICE AND PWT
 	% Joining tables
 unData = outerjoin(unData,countriesPWTRICE); 	% XXX ONLY 130 COUNTRIES.
 NEED TO FIX NAMING BEFORE THIS.
 unCountriesAfterJoin = unique(unData.country_unData); 	%now 193 countries
 	% If the country is not in UN data, I copy the name from PWT data.
8
66

notInUnData = isundefined(unData.country_unData);
 countriesNotInUnData = unData(notInUnData,
["country_countriesPWTRICE"	 "countrycode"	]);
 unData.country_unData(notInUnData) =
 unData.country_countriesPWTRICE(notInUnData);
 	% Assigning year and sector gategory to missing data vars (PWT
 	% countries not in UN)
 isUndefined = isundefined(unData.year);
 unData.year(isUndefined) = 	"9999"	; 	% Just a missing data marker
 isUndefined = isundefined(unData.sector);
 unData.sector(isUndefined) = 	"Missing Data"	;
 	% For now I am just dropping the 10 small countries with UN data and no
 PWT
 	% % data. (XXX Figure out if some can be used)
 	%unData(isundefined(unData.country_countryLists),:) = [];
 	% Keeping only needed vars
 unData = unData(:,
["country_unData"	 "year"	 "sector"	 "sum_value"	 "countrycode"	 "rice_code"	 "rice_region"	]);
 	% END OF ADDING COUNTRY CODES TO UN DATA
 	%unData = unData(unData.year == "2019", :); % 130 countries for 2019, 144
 for 2018, 148 for 2017
 	%unData = unData((unData.sector == "Total Economy" | unData.sector ==
 "Missing Data"),:);
 save(	"wrangling.mat"	, 	"unData"	, 	"-append"	);
 	%save("rawData.mat", "raw", "-append");
else
 load(	"wrangling.mat"	, 	"unData"	);
 	%load("rawData.mat", "raw");
end
% END UN DATA WRANGLING
WORLD INPUT-OUTPUT DATABASE (FOR
MISSING COUNTRIES IN UN)
WIOD is a smaller but more detailed database. It has data for 4 of the large regions missing in UN data: China,
Indonesia, Russia, Taiwan. Only smaller countries are missing in UN, to whom I extrapolate shares when done.
if	 wrangle.WIOD == 1
 	% Reading in raw data (XXX automate somehow?)
 raw.WIOD_CHN = readtable(	"raw_data\WIOD\CHN_NIOT_nov16.xlsx"	,
 Sheet=	"National IO-tables"	, VariableNamesRange=	"A2"	, DataRange=	"A3"	);
 	%raw.WIOD_CZE = readtable("raw_data\WIOD\CZE_NIOT_nov16.xlsx",
 Sheet="National IO-tables", VariableNamesRange="A2", DataRange="A3");
 raw.WIOD_IDN = readtable(	"raw_data\WIOD\IDN_NIOT_nov16.xlsx"	,
 Sheet=	"National IO-tables"	, VariableNamesRange=	"A2"	, DataRange=	"A3"	);
 raw.WIOD_RUS = readtable(	"raw_data\WIOD\RUS_NIOT_nov16.xlsx"	,
 Sheet=	"National IO-tables"	, VariableNamesRange=	"A2"	, DataRange=	"A3"	);
9
67

raw.WIOD_TWN = readtable(	"raw_data\WIOD\TWN_NIOT_nov16.xlsx"	,
 Sheet=	"National IO-tables"	, VariableNamesRange=	"A2"	, DataRange=	"A3"	);
 	% Manually adding countrycodes (XXX could automate from regexp on
 filename)
 raw.WIOD_CHN.countrycode(:) = 	"CHN"	;
 	%raw_WIOD_CZE.countrycode(:) = "CZE"; Found CZE data in UN under
 	%different name
 raw.WIOD_IDN.countrycode(:) = 	"IDN"	;
 raw.WIOD_RUS.countrycode(:) = 	"RUS"	;
 raw.WIOD_TWN.countrycode(:) = 	"TWN"	;
 	% Vertical concat. of each country into one big table for all five
 	% countries
 wiodData = [raw.WIOD_CHN;raw.WIOD_IDN;raw.WIOD_RUS;raw.WIOD_TWN]; 	% Add
 raw.WIOD_CZE if needed later
 wiodVars = [	"countrycode"	 "year"	 "code"	 "sector"	 "origin"	 "value"	];
 wiodData = renamevars(wiodData,
 [	"countrycode"	 "Var1"	 "Var2"	 "Var3"	 "Var4"	 "TotalOutput"	], wiodVars);
 wiodData(wiodData.year ~= 2014,:) = [];
 wiodData = wiodData(:,wiodVars);
 wiodData(wiodData.origin ~= 	"Domestic"	,:) = [];
 wiodData = categorizer(wiodData, 5);
 electricitySector = {	'Electricity, gas, steam and air conditioning
 supply'	};
 miningSector = {	'Mining and quarrying'	};
 transportSector = [{	'Land transport and transport via pipelines'	}
 {	'Water transport'	}
 {	'Air transport'	}
 {	'Warehousing and support activities for transportation'	}
 {	'Postal and courier activities'	}];
 industrySector = [{	'Construction'	}
 {	'Manufacture of food products, beverages and tobacco products'	}
 {	'Manufacture of textiles, wearing apparel and leather products'	}
 {	'Manufacture of wood and of products of wood and cork, except
 furniture; manufacture of articles of straw and plaiting materials'	}
 {	'Manufacture of paper and paper products'	}
 {	'Printing and reproduction of recorded media'	}
 {	'Manufacture of coke and refined petroleum products'	}
 {	'Manufacture of chemicals and chemical products'	}
 {	'Manufacture of basic pharmaceutical products and pharmaceutical
 preparations'	}
 {	'Manufacture of rubber and plastic products'	}
 {	'Manufacture of other non-metallic mineral products'	}
 {	'Manufacture of basic metals'	}
 {	'Manufacture of fabricated metal products, except machinery and
 equipment'	}
 {	'Manufacture of computer, electronic and optical products'	}
 {	'Manufacture of electrical equipment'	}
 {	'Manufacture of machinery and equipment n.e.c.'	}
10
68

{	'Manufacture of motor vehicles, trailers and semi-trailers'	}
 {	'Manufacture of other transport equipment'	}
 {	'Manufacture of furniture; other manufacturing'	}
 {	'Repair and installation of machinery and equipment'	}
 electricitySector
 miningSector
 ];
 otherSector = [{	'Crop and animal production, hunting and related service
 activities'	}
 {	'Forestry and logging'	}
 {	'Fishing and aquaculture'	}
 {	'Water collection, treatment and supply'	}
 {	'Sewerage; waste collection, treatment and disposal activities;
 materials recovery; remediation activities and other waste management
 services'	}
 {	'Wholesale and retail trade and repair of motor vehicles and
 motorcycles'	}
 {	'Wholesale trade, except of motor vehicles and motorcycles'	}
 {	'Retail trade, except of motor vehicles and motorcycles'	}
 {	'Accommodation and food service activities'	}
 {	'Publishing activities'	}
 {	'Motion picture, video and television programme production, sound
 recording and music publishing activities; programming and broadcasting
 activities'	}
 {	'Telecommunications'	}
 {	'Computer programming, consultancy and related activities;
 information service activities'	}
 {	'Financial service activities, except insurance and pension funding'	}
 {	'Insurance, reinsurance and pension funding, except compulsory social
 security'	}
 {	'Activities auxiliary to financial services and insurance
 activities'	}
 {	'Real estate activities'	}
 {	'Legal and accounting activities; activities of head offices;
 management consultancy activities'	}
 {	'Architectural and engineering activities; technical testing and
 analysis'	}
 {	'Scientific research and development'	}
 {	'Advertising and market research'	}
 {	'Other professional, scientific and technical activities; veterinary
 activities'	}
 {	'Administrative and support service activities'	}
 {	'Public administration and defence; compulsory social security'	}
 {	'Education'	}
 {	'Human health and social work activities'	}
 {	'Other service activities'	}
 {	'Activities of households as employers; undifferentiated goods- and
 services-producing activities of households for own use'	}
 {	'Activities of extraterritorial organizations and bodies'	}
 ];
 	% NOTE: Can get electricity and so on as separate sectors. Here they are
 	% merged into other.
11
69

wiodData.sector = mergecats(wiodData.sector, transportSector, 	"Transport
 (and storage)"	);
 wiodData.sector = mergecats(wiodData.sector, industrySector, 	"Industry
 (manufacturing and construction)"	);
 wiodData.sector = mergecats(wiodData.sector, otherSector, 	"Other"	);
 	% Collecting sectors
 wiodData = varfun(@sum, wiodData, 	...
 GroupingVariables=[	"countrycode"	 "year"	 "sector"	], 	...
 InputVariables=	"value"	);
 	% Creating Total Economy sector
 wiodDataTotal = varfun(@sum, wiodData, 	...
 GroupingVariables=[	"countrycode"	 "year"	], 	...
 InputVariables=	"sum_value"	);
 wiodDataTotal = renamevars(wiodDataTotal, [	"GroupCount"	 "sum_sum_value"	],
 [	"sector"	 "sum_value"	]);
 wiodDataTotal =
 convertvars(wiodDataTotal,wiodDataTotal.Properties.VariableNames{	"sector"	},	'categorical'	);
 wiodDataTotal.sector = renamecats(wiodDataTotal.sector, 	"3"	, 	"Total
 Economy"	);
 	% Merging Total Economy category and data into main wiod table
 wiodData(:,	"GroupCount"	) = [];
 wiodData = [wiodData; wiodDataTotal];
 wiodData = sortrows(wiodData, 	"countrycode"	);
 save(	"wrangling.mat"	, 	"wiodData"	, 	"-append"	);
 	%save("rawData.mat", "raw", "-append");
else
 load(	"wrangling.mat"	, 	"wiodData"	);
 	%load("rawData.mat", "raw");
end
% END WIOD WRANGLING
ILOSTAT LABOR DATA
if	 wrangle.ILOSTAT == 1
 	% Basic read-ins and preprocessing
 raw.ILOSTAT = readtable(	"raw_data\ilostat labor data sector shares (189C
 with SECTOR SHARES).csv"	);
 ilostatVars = [	"country"	, 	"sector"	, 	"secLabSh"	];
 ilostatData = renamevars(raw.ILOSTAT,
 [	"ReferenceArea"	 "EconomicActivity"	 "Value"	], ilostatVars);
 ilostatData = ilostatData(:,ilostatVars);
 ilostatData = categorizer(ilostatData, 2);
 electricitySector = {	'Detailed: Utilities ~ISIC rev.4 D; E'	}; 	% Note:
 Utilities contain more than just electricity production
 miningSector = {	'Detailed: Mining and quarrying ~ISIC rev.4 B'	};
12
70

transportSector = [{	'Detailed: Transport; storage and communication ~ISIC
 rev.4 H; J'	}];
 industrySector = [{	'Detailed: Manufacturing ~ISIC rev.4 C'	}
 {	'Detailed: Construction ~ISIC rev.4 F'	}
 electricitySector
 miningSector];
 otherSector = [{	'Detailed: Agriculture; forestry and fishing ~ISIC rev.4
 A'	}
 {	'Detailed: Wholesale and retail trade; repair of motor vehicles and
 motorcycles ~ISIC rev.4 G'	}
 {	'Detailed: Accommodation and food service activities ~ISIC rev.4 I'	}
 {	'Detailed: Financial and insurance activities ~ISIC rev.4 K'	}
 {	'Detailed: Real estate; business and administrative activities ~ISIC
 rev.4 L; M; N'	}
 {	'Detailed: Public administration and defence; compulsory social
 security ~ISIC rev.4 O'	}
 {	'Detailed: Education ~ISIC rev.4 P'	}
 {	'Detailed: Human health and social work activities ~ISIC rev.4 Q'	}
 {	'Detailed: Other services ~ISIC rev.4 R; S; T; U'	}
 ];
 ilostatData.sector = mergecats(ilostatData.sector,
 transportSector, 	"Transport (and storage)"	);
 ilostatData.sector = mergecats(ilostatData.sector,
 industrySector, 	"Industry (manufacturing and construction)"	);
 ilostatData.sector = mergecats(ilostatData.sector, otherSector, 	"Other"	);
 ilostatData = ilostatData( 	...
 (ilostatData.sector == 	"Transport (and storage)"	 | 	...
 ilostatData.sector == 	"Industry (manufacturing and construction)"
 | 	...
 ilostatData.sector == 	"Electricity"	 | 	...
 ilostatData.sector == 	"Mining"	 | 	...
 ilostatData.sector == 	"Other"	),:);
 	% Collecting sectors
 ilostatData = varfun(@sum, ilostatData, 	...
 GroupingVariables=[	"country"	 "sector"	], 	...
 InputVariables=	"secLabSh"	);
 	% Cleaning
 ilostatData = renamevars(ilostatData,	"sum_secLabSh"	, 	"secLabSh"	);
 ilostatData = ilostatData(:,[	"country"	 "sector"	 "secLabSh"	]);
 	% Unstacking to get a variable per sector (a wider table, one country per
 row)
 ilostatData = unstack(ilostatData,	"secLabSh"	,"sector"	,...
 	"VariableNamingRule"	,"preserve"	);
 ilostatData = renamevars(ilostatData, 	...
 [	"Transport (and storage)"	 "Industry (manufacturing and
 construction)"	 "Other"	], 	...
 [	"transportLabSh"	 "industryLabSh"	 "otherLabSh"	]);
13
71

% Removing ILOSTAT regions
 dropIlostat = {	'ASEAN'
 	'Africa'
 	'Africa: High income'
 	'Africa: Low income'
 	'Africa: Lower-middle income'
 	'Africa: Upper-middle income'
 	'Americas'
 	'Americas: High income'
 	'Americas: Low income'
 	'Americas: Lower-middle income'
 	'Americas: Upper-middle income'
 	'Arab League'
 	'Arab States'
 	'Arab States: High income'
 	'Arab States: Low income'
 	'Arab States: Lower-middle income'
 	'Arab States: Upper-middle income'
 	'Asia and the Pacific'
 	'Asia and the Pacific: High income'
 	'Asia and the Pacific: Low income'
 	'Asia and the Pacific: Lower-middle income'
 	'Asia and the Pacific: Upper-middle income'
 	'BRICS'
 	'Caribbean'
 	'CARICOM'
 	'Central Africa'
 	'Central America'
 	'Central Asia'
 	'Central and Western Asia'
 	'Central and Western Asia: High income'
 	'Central and Western Asia: Low income'
 	'Central and Western Asia: Lower-middle income'
 	'Central and Western Asia: Upper-middle income'
 	'Eastern Africa'
 	'Eastern Asia'
 	'Eastern Asia: High income'
 	'Eastern Asia: Low income'
 	'Eastern Asia: Lower-middle income'
 	'Eastern Asia: Upper-middle income'
 	'Eastern Europe'
 	'Eastern Europe: High income'
 	'Eastern Europe: Lower-middle income'
 	'Eastern Europe: Upper-middle income'
 	'Europe and Central Asia'
 	'Europe and Central Asia: High income'
 	'Europe and Central Asia: Low income'
 	'Europe and Central Asia: Lower-middle income'
 	'Europe and Central Asia: Upper-middle income'
 	'European Union 27'
 	'European Union 28'
 	'G20'
 	'G7'
 	'Latin America and the Caribbean'
14
72

'Latin America and the Caribbean: High income'
 	'Latin America and the Caribbean: Low income'
 	'Latin America and the Caribbean: Lower-middle income'
 	'Latin America and the Caribbean: Upper-middle income'
 	'MENA'
 	'Northern Africa'
 	'Northern Africa: Lower-middle income'
 	'Northern Africa: Upper-middle income'
 	'Northern America'
 	'Northern America: High income'
 	'Northern Europe'
 	'Northern, Southern and Western Europe'
 	'Northern, Southern and Western Europe: High income'
 	'Northern, Southern and Western Europe: Upper-middle income'
 	'Pacific Islands'
 	'South-Eastern Asia'
 	'South-Eastern Asia and the Pacific'
 	'South-Eastern Asia and the Pacific: High income'
 	'South-Eastern Asia and the Pacific: Lower-middle income'
 	'South-Eastern Asia and the Pacific: Upper-middle income'
 	'Southern Africa'
 	'South America'
 	'Southern Asia'
 	'Southern Asia: Low income'
 	'Southern Asia: Lower-middle income'
 	'Southern Asia: Upper-middle income'
 	'Southern Europe'
 	'Sub-Saharan Africa'
 	'Sub-Saharan Africa: High income'
 	'Sub-Saharan Africa: Low income'
 	'Sub-Saharan Africa: Lower-middle income'
 	'Sub-Saharan Africa: Upper-middle income'
 	'Western Africa'
 	'Western Asia'
 	'Western Europe'
 	'World'
 	'World excluding BRICS'
 	'World: High income'
 	'World: Low income'
 	'World: Lower-middle income'
 	'World: Upper-middle income'	};
 ilostatData(ismember(ilostatData.country, dropIlostat),:) = [];
 	% Sorting out naming differences.
 countriesILO = unique(ilostatData.country);
 countriesPWT = unique(countriesPWTRICE.country);
 inILOnotPWT = setdiff(countriesILO, countriesPWT);
 inPWTnotinILO = setdiff(countriesPWT,countriesILO);
 ilostatData.country = renamecats(ilostatData.country, 	...
 [	"Bolivia"	 ...
 	"Cape Verde"	 ...
 	"Congo, Democratic Republic of the"	 ...
 	"Czechia"	 ...
15
73

"Hong Kong, China"	 ...
 	"Iran, Islamic Republic of"	 ...
 	"Korea, Republic of"	 ...
 	"Lao People's Democratic Republic"	 ...
 	"Macau, China"	 ...
 	"Moldova, Republic of"	 ...
 	"Occupied Palestinian Territory"	 ...
 	"Saint Vincent and the Grenadines"	 ...
 	"Taiwan, China"	 ...
 	"Tanzania, United Republic of"	 ...
 	"Venezuela, Bolivarian Republic of"	], 	...
 [string(countriesPWT(countriesPWT == 	"Bolivia (Plurinational State
 of)"	)), 	...
 string(countriesPWT(countriesPWT == 	"Cabo Verde"	)), 	...
 string(countriesPWT(countriesPWT == 	"D.R. of the Congo"	)), 	...
 string(countriesPWT(countriesPWT == 	"Czech Republic"	)), 	...
 string(countriesPWT(countriesPWT == 	"China, Hong Kong SAR"	)), 	...
 string(countriesPWT(countriesPWT == 	"Iran (Islamic Republic
 of)"	)), 	...
 string(countriesPWT(countriesPWT == 	"Republic of Korea"	)), 	...
 string(countriesPWT(countriesPWT == 	"Lao People's DR"	)), 	...
 string(countriesPWT(countriesPWT == 	"China, Macao SAR"	)), 	...
 string(countriesPWT(countriesPWT == 	"Republic of Moldova"	)), 	...
 string(countriesPWT(countriesPWT == 	"State of Palestine"	)), 	...
 string(countriesPWT(countriesPWT == 	"St. Vincent and the
 Grenadines"	)), 	...
 string(countriesPWT(countriesPWT == 	"Taiwan"	)), 	...
 string(countriesPWT(countriesPWT == 	"U.R. of Tanzania:
 Mainland"	)), 	...
 string(countriesPWT(countriesPWT == 	"Venezuela (Bolivarian Republic
 of)"	))]);
 countriesILOcheck = unique(ilostatData.country);
 save(	"wrangling.mat"	, 	"ilostatData"	, 	"-append"	);
 	%save("rawData.mat", "raw", "-append");
else
 load(	"wrangling.mat"	, 	"ilostatData"	);
 	%load("rawData.mat", "raw");
end
% END ILOSTAT WRANGLING
IEA VOLUME BALANCES
if	 wrangle.IEA == 1
 raw_IEA = readtable(	"raw_data\IEA energy volumes.xlsx"	, Sheet=	"Data"	,
 VariableNamesRange=	"A4"	, DataRange=	"A6"	);
 raw_IEA_Other = readtable(	"raw_data\WBAL_Extract-Sharing_V2.xlsx"	,
 Sheet=	"Data_Add"	, VariableNamesRange=	"A4"	, DataRange=	"A5"	);
 	% Cleaning to merge the two datasets.
 ieaDataOther = renamevars(raw_IEA_Other, 	"Var2"	, 	"PRODUCT"	);
 ieaDataOther = movevars(ieaDataOther, 	'Coal'	, 	'Before'	, 	'Coal_1'	);
 ieaDataOther = movevars(ieaDataOther, 	'Coal_1'	, 	'Before'	, 	'Coal_2'	);
16
74

ieaDataOther = movevars(ieaDataOther, 	'Coal_2'	, 	'After'	, 	'Total_2'	);
 ieaDataRawMerged = [raw_IEA;ieaDataOther];
 	% We have 2015, 2018, 2019 data. Keeping only 2019 data.
 ieaData = removevars(ieaDataRawMerged,3:24);
 	% Taking absolute values of everything
 	% Negative values were transformation. Not needed.
 	%ieaData = abs(ieaData)
 	% Combining fuels to our aggregated fuels
 ieaData.oil = ieaData.CrudeOil_2 + ieaData.OilProducts_2; 	% Crude oil and
 oil products
 ieaData.renewables = ieaData.Nuclear_2 + ieaData.Hydro_2 +
 ieaData.Solar_Wind_Etc__2;
 ieaData.elAndHeat = ieaData.Electricity_2 + ieaData.Heat_2;
 ieaVars = [	"sector"	 "country"	 "natGas"	 "bio"	 "total"	 "coal"	];
 ieaData = renamevars(ieaData,
 [	"Var1"	 "PRODUCT"	 "NaturalGas_2"	 "BiofuelsAndWaste_2"	 "Total_2"	 "Coal_2"	],
 ieaVars);
 ieaData = ieaData(:,
["country"	 "sector"	 "coal"	 "oil"	 "natGas"	 "bio"	 "renewables"	 "elAndHeat"	]);
 ieaData = categorizer(ieaData, 2);
 	% % Removing IEA regions
 	% ieaData((ieaData.country == "Africa" | ...
 	% ieaData.country == 'Memo: European Union-27' | ...
 	% ieaData.country == 'Memo: European Union-28' | ...
 	% ieaData.country == 'Memo: FSU 15' | ...
 	% ieaData.country == 'Memo: IEA Total' | ...
 	% ieaData.country == 'Memo: Non-OECD Total' | ...
 	% ieaData.country == 'Memo: OECD Total' | ...
 	% ieaData.country == 'Non-OECD Americas' | ...
 	% ieaData.country == 'Non-OECD Asia (excluding China)' | ...
 	% ieaData.country == 'Non-OECD Europe and Eurasia' | ...
 	% ieaData.country == 'OECD Americas' | ...
 	% ieaData.country == 'OECD Asia Oceania' | ...
 	% ieaData.country == 'OECD Europe' | ...
 	% ieaData.country == 'World'), :) = [];
 ieaDataSanity = ieaData;
 	% Sorting out naming differences.
 countriesIEA = unique(ieaData.country);
 countriesPWT = unique(countriesPWTRICE.country);
 inIEAnotPWT = setdiff(countriesIEA, countriesPWT);
 inPWTnotinIEA = setdiff(countriesPWT,countriesIEA);
 ieaData.country = renamecats(ieaData.country, 	...
 [	"Bolivarian Republic of Venezuela"	 ...
 	"Chinese Taipei"	 ...
 	"Curaçao/Netherlands Antilles"	 ...
 	"Democratic Republic of the Congo"	 ...
 	"Hong Kong (China)"	 ...
 	"Islamic Republic of Iran"	 ...
 	"Korea"	 ...
17
75

"Lao People's Democratic Republic"	 ...
 	"People's Republic of China"	 ...
 	"Plurinational State of Bolivia"	 ...
 	"Republic of North Macedonia"	 ...
 	"Republic of the Congo"	 ...
 	"Slovak Republic"	 ...
 	"United Republic of Tanzania"	], 	...
 [string(countriesPWT(countriesPWT == 	"Venezuela (Bolivarian Republic
 of)"	)), 	...
 string(countriesPWT(countriesPWT == 	"Taiwan"	)), 	...
 string(countriesPWT(countriesPWT == 	"Curaçao"	)), 	...
 string(countriesPWT(countriesPWT == 	"D.R. of the Congo"	)), 	...
 string(countriesPWT(countriesPWT == 	"China, Hong Kong SAR"	)), 	...
 string(countriesPWT(countriesPWT == 	"Iran (Islamic Republic
 of)"	)), 	...
 string(countriesPWT(countriesPWT == 	"Republic of Korea"	)), 	...
 string(countriesPWT(countriesPWT == 	"Lao People's DR"	)), 	...
 string(countriesPWT(countriesPWT == 	"China"	)), 	...
 string(countriesPWT(countriesPWT == 	"Bolivia (Plurinational State
 of)"	)), 	...
 string(countriesPWT(countriesPWT == 	"North Macedonia"	)), 	...
 string(countriesPWT(countriesPWT == 	"Congo"	)), 	...
 string(countriesPWT(countriesPWT == 	"Slovakia"	)), 	...
 string(countriesPWT(countriesPWT == 	"U.R. of Tanzania: Mainland"	))]);
 countriesIEAcheck = unique(ieaData.country);
 ieaDataSanity2 = ieaData;
 	% Merging these three input-output types into electricity production
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'Electricity Plants'	}
 {	'CHP Plants'	}
 ], 	"4 electricity"	);
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'Residential'	}
 {	'Commercial and public services'	}
 {	'Agriculture/forestry'	}
 {	'Fishing'	}], 	"3 other"	);
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'Industry'	}], 	"2 industry"	);
 ieaData.sector = renamecats(ieaData.sector, 	...
 [	"Transport"	], 	...
 [	"1 transport"	]);
 ieaDataSanity2b = ieaData;
 	% Summing over grouped categories
 	% Taking aboslute value to get rid of negative entries (transformation in
 raw data)
18
76

ieaData = varfun(@(x) sum(abs(x)), ieaData , 	"GroupingVariables"	,
["country"	 "sector"	], 	...
 	"InputVariables"	,3:8);
 ieaDataSanity3 = ieaData;
 	% The following for-loop is based on help from mathworks.com forum
 	% https://se.mathworks.com/matlabcentral/answers/1710975-how-to-
conditionally-and-by-groups-subtract-one-row-from-another-in-table
 	% Find the available countries
 countries = unique(ieaData.country);
 ieaData = removevars(ieaData, [	"GroupCount"	 ]);
 ieaData = renamevars(ieaData, 3:8,
 [	"coal"	 "oil"	 "natGas"	 "bio"	 "renew"	 "elAndHeat"	]);
 	% Loop through the available countries computing the net consumption for
 	% each
 endRow = size(ieaData,1); 	% current ending row number
 	for	 k = 1:numel(countries)
 	% sum the supplies and consumption
 idl = ieaData.sector == 	"Total energy supply"	 & ieaData.country ==
 countries(k);
 totalSupply = sum(ieaData{idl,3:end},1);
 idl = ieaData.sector == 	"Total final consumption"	 & ieaData.country ==
 countries(k);
 totalConsumption = sum(ieaData{idl,3:end},1);
 idl = ieaData.sector == 	"4 electricity"	 & ieaData.country ==
 countries(k);
 totalElectricity = sum(ieaData{idl,3:end},1);
 	% compute net
 misc = totalSupply -totalConsumption - totalElectricity;
 	% and add row
 endRow = endRow + 1;
 newRow = ieaData(find(idl),:); 	% base on any of the current matching
 rows
 newRow.sector = 	"misc"	;
 newRow{1,3:end} = misc;
 ieaData(endRow,:) = newRow;
 	end
 	% Does not make sense to perform misc. operation on electricity and total
 ieaData{ieaData.sector == 	"misc"	, [	"elAndHeat"	]} = 0;
 ieaData = sortrows(ieaData,	'country'	,'ascend'	);
 ieaDataSanity4 = ieaData;
 	% For each country, add misc to the sectors according to fuel shares
 endRow = size(ieaData,1); 	% current ending row number
 	for	 k = 1:numel(countries)
19
77

% take out vectors for each sector
 idl = ieaData.sector == 	"1 transport"	 & ieaData.country ==
 countries(k);
 transport = ieaData{idl,3:end};
 idl = ieaData.sector == 	"2 industry"	 & ieaData.country ==
 countries(k);
 industry = ieaData{idl,3:end};
 idl = ieaData.sector == 	"3 other"	 & ieaData.country == countries(k);
 other = ieaData{idl,3:end};
 idl = ieaData.sector == 	"Total final consumption"	 & ieaData.country ==
 countries(k);
 totalConsumption = ieaData{idl,3:end};
 idl = ieaData.sector == 	"misc"	 & ieaData.country == countries(k);
 misc = ieaData{idl,3:end};
 	% calculate shares (5x1) each
 trSh = transport ./ totalConsumption;
 inSh = industry ./ totalConsumption;
 otSh = other ./ totalConsumption;
 	% stack to get (3x5) to use dot product
 sh = [trSh;inSh;otSh];
 miscMatrix = repmat(misc, 3, 1);
 	% matrix of misc split across sectors and fuels
 miscSectors = sh .* miscMatrix;
 	% and add row for transport
 endRow = endRow + 1;
 newRow = ieaData(find(idl),:); 	% base on any of the current matching
 rows
 newRow.sector = 	"misc tr"	;
 newRow{1,3:end} = miscSectors(1,:);
 ieaData(endRow,:) = newRow;
 	% industry
 endRow = endRow + 1;
 newRow = ieaData(find(idl),:); 	% base on any of the current matching
 rows
 newRow.sector = 	"misc in"	;
 newRow{1,3:end} = miscSectors(2,:);
 ieaData(endRow,:) = newRow;
 	% and other
 endRow = endRow + 1;
 newRow = ieaData(find(idl),:); 	% base on any of the current matching
 rows
 newRow.sector = 	"misc ot"	;
 newRow{1,3:end} = miscSectors(3,:);
 ieaData(endRow,:) = newRow;
 	end
20
78

ieaDataSanity5 = ieaData;
 	% Merging misc. with corresponding sectors and summing again
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'1 transport'	 'misc tr'	}], 	"1 transport"	);
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'2 industry'	 'misc in'	}], 	"2 industry"	);
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'3 other'	 'misc ot'	}], 	"3 other"	);
 	% omitnan is important since previous step gave 0/0 for some fuels in some
 regions
 ieaData = varfun(@(x) sum(x, 	"omitnan"	), ieaData , 	"GroupingVariables"	,
["country"	 "sector"	], 	...
 	"InputVariables"	,3:8);
 ieaData = removevars(ieaData, [	"GroupCount"	 ]);
 ieaData = renamevars(ieaData, 3:8,
 [	"coal"	 "oil"	 "natGas"	 "bio"	 "renew"	 "elAndHeat"	]);
 ieaDataSanity6 = ieaData;
 	% For each country, add energy own use and losses in electricity to
 sectors
 endRow = size(ieaData,1); 	% current ending row number
 	for	 k = 1:numel(countries)
 	% take out vectors for each sector
 idl = ieaData.sector == 	"4 electricity"	 & ieaData.country ==
 countries(k);
 el = ieaData{idl,	"elAndHeat"	};
 idl = ieaData.sector == 	"1 transport"	 & ieaData.country ==
 countries(k);
 transport = ieaData{idl,	"elAndHeat"	};
 idl = ieaData.sector == 	"2 industry"	 & ieaData.country ==
 countries(k);
 industry = ieaData{idl,	"elAndHeat"	};
 idl = ieaData.sector == 	"3 other"	 & ieaData.country == countries(k);
 other = ieaData{idl,	"elAndHeat"	};
 idl = ieaData.sector == 	"Total final consumption"	 & ieaData.country ==
 countries(k);
 totalConsumption = ieaData{idl,	"elAndHeat"	};
 	% eiou is Energy industry own use + Losses = Elec - cons
 eiou = el - totalConsumption;
 	% calculate shares (1x1) each
 trSh = transport / totalConsumption;
 inSh = industry / totalConsumption;
 otSh = other / totalConsumption;
 	% stack to get (3x1) to use dot product
 sh = [trSh;inSh;otSh];
 eiouVector = repmat(eiou, 3,1);
21
79

% matrix of electricity energy own use split across sectors
 eiouSectors = sh .* eiouVector;
 eiouRow = zeros(3,6);
 eiouRow(:,6) = eiouSectors;
 	% and add row for transport
 endRow = endRow + 1;
 newRow = ieaData(find(idl),:); 	% base on any of the current matching
 rows
 newRow.sector = 	"eiou tr"	;
 newRow{1,3:end} = eiouRow(1,:);
 ieaData(endRow,:) = newRow;
 	% industry
 endRow = endRow + 1;
 newRow = ieaData(find(idl),:); 	% base on any of the current matching
 rows
 newRow.sector = 	"eiou in"	;
 newRow{1,3:end} = eiouRow(2,:);
 ieaData(endRow,:) = newRow;
 	% and other
 endRow = endRow + 1;
 newRow = ieaData(find(idl),:); 	% base on any of the current matching
 rows
 newRow.sector = 	"eiou ot"	;
 newRow{1,3:end} = eiouRow(3,:);
 ieaData(endRow,:) = newRow;
 	end
 ieaDataSanity7 = ieaData;
 	% Merging misc. with corresponding sectors and summing again
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'1 transport'	 'eiou tr'	}], 	"1 transport"	);
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'2 industry'	 'eiou in'	}], 	"2 industry"	);
 ieaData.sector = mergecats(ieaData.sector, 	...
 [{	'3 other'	 'eiou ot'	}], 	"3 other"	);
 	% omitnan is important since previous step gave 0/0 for some fuels in some
 regions
 ieaData = varfun(@(x) sum(x, 	"omitnan"	), ieaData , 	"GroupingVariables"	,
["country"	 "sector"	], 	...
 	"InputVariables"	,3:8);
 ieaData = removevars(ieaData, [	"GroupCount"	 ]);
 ieaData = renamevars(ieaData, 3:8,
 [	"coal"	 "oil"	 "natGas"	 "bio"	 "renew"	 "elAndHeat"	]);
 ieaDataSanity8 = ieaData;
 	% Cleaning and deleting
22
80

ieaData(ieaData.sector == 	"Total energy supply"	,:) = [];
 ieaData(ieaData.sector == 	"Total final consumption"	,:) = [];
 ieaData(ieaData.sector == 	"Energy industry own use"	,:) = [];
 ieaData(ieaData.sector == 	"misc"	,:) = [];
 ieaData.sector = removecats(ieaData.sector, [	"Total energy supply"	 ...
 	"Total final consumption"	 ...
 	"Energy industry own use"	 ...
 	"misc"	]);
 	% Unstacking to get sector-wise usage of each fuel (incl. electricity
 	% as sector)
 ieaDataUnstacked = unstack(ieaData,
 [	"coal"	 "oil"	 "natGas"	 "renew"	 "bio"	 "elAndHeat"	], 	"sector"	);
 	% Summing over fuel use in sectors to get total fuel use per country
 coalVars =
 contains(ieaDataUnstacked.Properties.VariableNames(:), 	"coal"	, 	"IgnoreCase"	,true);
 oilVars =
 contains(ieaDataUnstacked.Properties.VariableNames(:), 	"oil"	, 	"IgnoreCase"	,true);
 natGasVars =
 contains(ieaDataUnstacked.Properties.VariableNames(:), 	"natGas"	, 	"IgnoreCase"	,true);
 bioVars =
 contains(ieaDataUnstacked.Properties.VariableNames(:), 	"bio"	, 	"IgnoreCase"	,true);
 renewVars =
 contains(ieaDataUnstacked.Properties.VariableNames(:), 	"renew"	, 	"IgnoreCase"	,true);
 elAndHeatVars =
 contains(ieaDataUnstacked.Properties.VariableNames(:), 	"elAndHeat"	, 	"IgnoreCase"	,true);
 ieaDataUnstacked.coal = sum(ieaDataUnstacked{:,coalVars},2);
 ieaDataUnstacked.oil = sum(ieaDataUnstacked{:,oilVars},2);
 ieaDataUnstacked.natGas = sum(ieaDataUnstacked{:,natGasVars},2);
 ieaDataUnstacked.bio = sum(ieaDataUnstacked{:,bioVars},2);
 ieaDataUnstacked.renew = sum(ieaDataUnstacked{:,renewVars},2);
 ieaDataUnstacked.elAndHeat = sum(ieaDataUnstacked{:,elAndHeatVars},2);
 save(	"wrangling.mat"	, 	"ieaData"	, 	"-append"	);
 save(	"wrangling.mat"	, 	"ieaDataUnstacked"	, 	"-append"	);
 	%save("rawData.mat", "raw", "-append");
else
 load(	"wrangling.mat"	, 	"ieaData"	);
 load(	"wrangling.mat"	, 	"ieaDataUnstacked"	);
 	%load("rawData.mat", "raw");
end
% END IEA WRANGLING
PWT TRANSPORT CAPITAL DETAIL DATA
% PWT data is only for transport equipent, while OECD has for the whole
% sector. I think we are better off extrapolating OECD data than using
% negatively biased PWT data.
23
81

% raw.pwtTransData = readtable("raw_data\pwt100-capital-detail.xlsx",
 Sheet="Data");
% pwtTransData = raw.pwtTransData(raw.pwtTransData.year == 2019,["countrycode"
 "Nc_Mach" "Nc_Struc" "Nc_TraEq" "Nc_Other"]);
% pwtTransData = categorizer(pwtTransData, 1);
% pwtTransData.capStock = pwtTransData.Nc_Mach + pwtTransData.Nc_Struc +
 pwtTransData.Nc_TraEq + pwtTransData.Nc_Other;
% pwtTransData.transCapShPWT = pwtTransData.Nc_TraEq ./ pwtTransData.capStock;
% pwtTransData = pwtTransData(:, ["countrycode" "transCapShPWT"]);
% END PWT TRANSPORT SHARE WRANGLING
OECD CAPITAL DETAILS DATA ()
if	 wrangle.OECD == 1
 raw.OECD = readtable(	"raw_data\oecd capital data.csv"	);
 oecdDataAll = raw.OECD(:,
["LOCATION"	 "Country"	 "Variable"	 "Industry"	 "Time"	 "Value"	]);
 oecdDataAll = categorizer(oecdDataAll, 4);
 oecdDataAll = oecdDataAll(oecdDataAll.Variable == 	'Net capital stock,
 volumes'	,:);
 	% For sectors
 keep = [{	'TOTAL'
 	'Agriculture, hunting, forestry and fishing [A]'
 	'Mining and quarrying [B]'
 	'Manufacturing [C]'
 	'Electricity, gas, steam and air conditioning supply [D]'
 	'Water supply; sewerage, waste management and remediation activities
 [E]'
 	'Construction [F]'
 	'Wholesale and retail trade, repair of motor vehicles and motorcycles
 [G]'
 	'Transportation and storage [H]'
 	'Accommodation and food service activities [I]'
 	'Information and communication [J]'
 	'Financial and insurance activities [K]'
 	'Real estate activities [L]'
 	'Professional, scientific and technical activities [M]'
 	'Administrative and support service activities [N]'
 	'Public administration and defence; compulsory social security [O]'
 	'Education [P]'
 	'Human health and social work activities [Q]'
 	'Arts, entertainment and recreation [R]'
 	'Other service activities [S]'	}];
 oecdData = oecdDataAll(ismember(oecdDataAll.Industry, keep),:);
 oecdData = renamevars(oecdData, 	...
 [	"LOCATION"	 "Country"	 "Industry"	 "Time"	], 	...
24
82

[	"countrycode"	 "country"	 "sector"	 "year"	]);
 transportSector = [{	'Transportation and storage [H]'	}];
 industrySector = [{	'Manufacturing [C]'	}
 {	'Construction [F]'	}
 {	'Mining and quarrying [B]'	}
 {	'Electricity, gas, steam and air conditioning supply [D]'	}];
 otherSector = [{	'Agriculture, hunting, forestry and fishing [A]'	}
 {	'Water supply; sewerage, waste management and remediation activities
 [E]'	}
 {	'Wholesale and retail trade, repair of motor vehicles and motorcycles
 [G]'	}
 {	'Accommodation and food service activities [I]'	}
 {	'Information and communication [J]'	}
 {	'Financial and insurance activities [K]'	}
 {	'Real estate activities [L]'	}
 {	'Professional, scientific and technical activities [M]'	}
 {	'Administrative and support service activities [N]'	}
 {	'Public administration and defence; compulsory social security [O]'	}
 {	'Education [P]'	}
 {	'Human health and social work activities [Q]'	}
 {	'Arts, entertainment and recreation [R]'	}
 {	'Other service activities [S]'	}];
 oecdData.sector = mergecats(oecdData.sector, transportSector, 	"Transport
 (and storage)"	);
 oecdData.sector = mergecats(oecdData.sector, industrySector, 	"Industry
 (manufacturing and construction)"	);
 oecdData.sector = mergecats(oecdData.sector, otherSector, 	"Other"	);
 	% Unstacking to get a variable per sector (a wider table, one country per
 row)
 oecdData = unstack(oecdData,	"Value"	,"sector"	,...
 	"VariableNamingRule"	,"preserve"	);
 oecdData = renamevars(oecdData, 	...
 [	"Transport (and storage)"	 "Industry (manufacturing and
 construction)"	 "Other"	 "TOTAL"	], 	...
 [	"transCap"	 "indCap"	 "othCap"	 "totalCap"	]);
 oecdData.transCapSh = oecdData.transCap ./ oecdData.totalCap;
 oecdData.indCapSh = oecdData.indCap ./ oecdData.totalCap;
 oecdData.othCapSh = 1 - oecdData.transCapSh - oecdData.indCapSh;
 oecdData.sanity = oecdData.transCapSh + oecdData.indCapSh +
 oecdData.othCapSh;
 	% Sanity check failed for several countries
 	% Some are just missing "other" categories --> now using reciprocal
 	% Delete Ireland and Israel. They are missing manufacturing.
 oecdData((oecdData.country == 	"Ireland"	 | oecdData.country == 	"Israel"	),:)
 = [];
 oecdData(isnan(oecdData.transCap),:) = [];
 	% Manually keeping most recent data [XXX Automate]
 oecdData = oecdData([2 4 6 7 9 11 13 15 16 18 20 22 23 25 27 28 30 32 33
 34 35 37 38 40 42 44 47 50 55],:);
25
83

oecdData = oecdData(:,[	"countrycode"	 "transCapSh"	 "indCapSh"	 "othCapSh"	]);
 save(	"wrangling.mat"	, 	"oecdData"	, 	"-append"	);
 	%save("rawData.mat", "raw.OECD", "-append");
else
 load(	"wrangling.mat"	, 	"oecdData"	);
 	%load("rawData.mat", "raw.OECD");
end
US BLS DETAILED CAPITAL DATA FOR FUEL
CAPITAL USE
% Getting total capital in US coal/oil/gas production and refinement.
% if wrangle.USBLS == 1
raw_USBLS = readtable(	"raw_data\USBLS cap_details.xlsx"	, Sheet=	"Data"	,
 VariableNamesRange=	"A1"	, DataRange=	"A2"	);
usblsVars = [	"sector"	 "type"	 "category"	 "unit"	 "value"	];
usblsDataAll = renamevars(raw_USBLS,
 [	"NAICSTitle"	 "MeasureTitle"	 "AssetCategory"	 "DurationTitle"	 "x2019"	],
 usblsVars);
usblsDataAll = usblsDataAll(:,usblsVars);
usblsDataAll = categorizer(usblsDataAll,4);
% Note: Convert to 2017 dollars (PWT uses 2017)
usblsDataAll = usblsDataAll(usblsDataAll.type == 	"Productive capital stock
 (direct aggregate-billions of 2012 dollars)"	, :);
usblsDataAll = usblsDataAll(usblsDataAll.category == 	"All assets"	, :);
usblsDataAll = usblsDataAll(usblsDataAll.unit == 	"Levels"	, :);
usblsData = usblsDataAll( 	...
 (usblsDataAll.sector == 	"Oil and gas extraction"	 | 	...
 usblsDataAll.sector == 	"Mining, except oil and gas"	 | 	...
 usblsDataAll.sector == 	"Petroleum and coal products"	 | 	...
 usblsDataAll.sector == 	"Pipeline transportation"	), :);
% The direct capital stock
% NOTE: This used PWT in millions, which is later changed to just ones.
usblsFossilCapSum = sum(usblsData.value); 	% In billions 2012 USD
usblsFossilCap = 1.0676 * usblsFossilCapSum; 	% In billions 2017 USD using
 https://www.in2013dollars.com/us/inflation/2012
usblsFossilCap = 1000 * usblsFossilCap; 	% PWT uses millions 2017 USD.
PWTcapUS = 69059088;
usblsFCsh = usblsFossilCap / PWTcapUS;
% About 5,8% of total US capital stock share of PWT. Seems high?
% Capital stock as share of total
usblsTotCap = sum([2620.40700000000;2913.24600000000; 	...
 3198.02200000000;840.280000000000;6422.82200000000; 	...
 5284.94200000000;1949.82800000000;3703.83500000000; 	...
 2097.87400000000;7141.95100000000;1033.88200000000; 	...
 753.843000000000;519.944000000000;908.600000000000; 	...
 1886.45500000000;300.480000000000]);
26
84

usblsTotCap = usblsTotCap * 1067,6;
usblsGovCap = 18739.076 * 1067,6;
usblsCapSh = usblsFossilCap / (usblsTotCap+usblsGovCap);
% This method gave 9% of total capital share. Maybe this data only
% has private sector?
% Added government, now it is 6,2% of total capital.
% Average of 5,8% and 6,2% is 6%. Six percent of total capital
% should go to fossil fuel extraction. This gives \bar(K), which
% must be subtracted from capital in indstry to avoid double
% counting
% save("calibData.mat", "usblsFossilCap", "-append");
% else
% load("calibData.mat", "usblsFossilCap");
% end
CHINA CAPITAL DETAILS
Manually extracted 2019 capital data from China Statistical Yearbook 2020 The following sectors: 'Mining and Wash-
ing of Coal' 'Extraction of Petroleum and Natural Gas' 'Professional and Support Activities for Mining' 'Processing of
Petroleum, Coal and Other Fuels'
raw_ChinaCapDetails = readtable(	"raw_data\china capital details.csv"	);
chinaCapData = categorizer(raw_ChinaCapDetails, 1);
chinaCapData.capStock = chinaCapData.totalAssets - chinaCapData.currentAssets;
chinaCapIndTotal = chinaCapData{1,	"capStock"	}; 	% 100 million 2019 Yuan
chinaCapFossil = sum(chinaCapData{2:5, 	"capStock"	}); 	% 100 million 2019 Yuan
chinaCapFossil = chinaCapFossil * 100; 	% To million 2019 Yuan
chinaCapFossil = chinaCapFossil / PWT{PWT.countrycode == 	"CHN"	, 	"xr"	} ; 	%
 Converted from million 2019 Yuan to million 2019 USD
chinaCapFossil = 0.96 * chinaCapFossil; 	% To 2017 USD
chinaCapFossilSh = chinaCapFossil / PWT{PWT.countrycode == 	"CHN"	, 	"cn"	};
% constists of 1% of total capital.
% Total assets method
chinaTotalAssetsFossil = sum(chinaCapData.totalAssets(2:5));
chinaTotalAssetsFossil = (chinaTotalAssetsFossil * 100 * 0.96) /
 PWT{PWT.countrycode == 	"CHN"	, 	"xr"	};
chinaCapShTotAsSh = chinaTotalAssetsFossil / PWT{PWT.countrycode
 == 	"CHN"	, 	"cn"	};
% this method gives 1.55% --> I use this one.
% Implies Russia 3%
% RUSSIA:
% https://www.investopedia.com/articles/markets/082615/5-biggest-russian-
natural-gas-companies.asp
% Got total assets from there
27
85

PART 2: COMBINING PRE-PROCESSED DA-
TA INTO USABLE SPREADSHEETS FOR
CALIBRATION %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%
MERGING FOR FINAL GOOD SECTORS
CALIBRATION
Combining data: PWT, UN, WIOD, ILOSTAT, OECD
if	 wrangle.sectors == 1
 	% WIOD/UN MERGE
 finalGoodSectors = outerjoin(wiodData, unData, Keys=	"countrycode"	,
 MergeKeys=true);
 	% Manually merging variables (XXX Could automate or shorten?)
 	% Year variables
 isUndefined = isundefined(finalGoodSectors.year_wiodData);
 finalGoodSectors.year_wiodData(isUndefined) =
 finalGoodSectors.year_unData(isUndefined);
 finalGoodSectors(:,	"year_unData"	) = [];
 	% Sector variables
 isUndefined = isundefined(finalGoodSectors.sector_wiodData);
 finalGoodSectors.sector_wiodData(isUndefined) =
 finalGoodSectors.sector_unData(isUndefined);
 finalGoodSectors(:,	"sector_unData"	) = [];
 	% Value variables
 isUndefined = isnan(finalGoodSectors.sum_value_wiodData); 	% Numbers use
 NAN, categories <undefined>
 finalGoodSectors.sum_value_wiodData(isUndefined) =
 finalGoodSectors.sum_value_unData(isUndefined);
 finalGoodSectors(:,	"sum_value_unData"	) = [];
 	% Renaming variables to tidy up
 finalGoodSectors = renamevars(finalGoodSectors,
["country_unData"	 "year_wiodData"	 "sector_wiodData"	 "sum_value_wiodData"	], 	...
 [	"country"	 "year"	 "sector"	 "output"	]);
 	% Exporting list of countries with missing sector data
 countriesWithoutSectors =
 [finalGoodSectors.country(isnan(finalGoodSectors.output))
 finalGoodSectors.countrycode(isnan(finalGoodSectors.output))];
 	% For now, I 'm just dropping the small countries for which there is no
 	% PWT data.
28
86

finalGoodSectors(isundefined(finalGoodSectors.countrycode),:) = [];
 	% Unstacking sectorData to get a variable per sector (a wider table)
 finalGoodSectors = unstack(finalGoodSectors,	"output"	,"sector"	,...

"AggregationFunction"	,@(x)x(~isempty(x)),	"VariableNamingRule"	,"preserve"	);
 	% Getting sectors as shares of output
 finalGoodSectors.totalShare = finalGoodSectors{:,	"Total Economy"	} ./
 finalGoodSectors{:,	"Total Economy"	};
 finalGoodSectors.transportShare = finalGoodSectors{:,	"Transport (and
 storage)"	} ./ finalGoodSectors{:,	"Total Economy"	};
 finalGoodSectors.industryShare = finalGoodSectors{:,	"Industry
 (manufacturing and construction)"	} ./ finalGoodSectors{:,	"Total Economy"	};
 finalGoodSectors.otherShare = finalGoodSectors.totalShare -
 finalGoodSectors.transportShare - finalGoodSectors.industryShare; 	% As
 reciprocial because poor data
 finalGoodSectors.sanityShare = finalGoodSectors.industryShare +
 finalGoodSectors.transportShare + finalGoodSectors.otherShare;
 	% % Some countries do not sum to unity if reciprocal method
 isn't
 	% % used
 	% poorDataCountries = sectorData(sectorData.sanityShare <
 0.9,:);
 	% goodDataCountries = sectorData(sectorData.sanityShare >
 0.99,:);
 	% poorDataSum = summary(poorDataCountries);
 	% goodDataSum = summary(goodDataCountries);
 	% Doing single year to use 2019 only (and 2014 for those few WIOD
 	% countries). Full data: 2019 has 114 countries, 2018 125, 2017 130.
 	% NOTABLE missing shares from 2019:
 	% Canada, Egypt, Japan, Korea, Philippines, UK
 	% Missing shares if 2018:
 	% Canada, Egypt, Phillipines
 	% If 2017:
 	% Egypt, Phillipines (non-recoverable
 	% I choose 2019 as base year, but use older data for those with no
 	% 2019 data. NOTE: Year 9999 are missing countries from UN, but kept for
 	% regional extrapolation.
 finalGoodSectors = finalGoodSectors( 	...
 (finalGoodSectors.year == 	"2014"	 | 	...
 finalGoodSectors.year == 	"2019"	 | 	...
 finalGoodSectors.year == 	"9999"	 | 	...
 finalGoodSectors.country == 	"Japan"	 | 	...
 finalGoodSectors.country == 	"Canada"	 | 	...
 finalGoodSectors.country == 	"United Kingdom"	 | 	...
 finalGoodSectors.country == 	"Republic of Korea"	 | 	...
 finalGoodSectors.country == 	"Yemen"	),:);
 finalGoodSectors((finalGoodSectors.country == 	"Japan"	 &
 (finalGoodSectors.year == 	"2017"	 | isnan(finalGoodSectors.otherShare))), :) =
 [];
29
87

finalGoodSectors((finalGoodSectors.country == 	"Republic of Korea"	 &
 (finalGoodSectors.year == 	"2017"	 | isnan(finalGoodSectors.otherShare))), :) =
 [];
 finalGoodSectors((finalGoodSectors.country == 	"United Kingdom"	 &
 (finalGoodSectors.year == 	"2017"	 | isnan(finalGoodSectors.otherShare))), :) =
 [];
 finalGoodSectors((finalGoodSectors.country == 	"Canada"	 &
 isnan(finalGoodSectors.otherShare)),:) = [];
 finalGoodSectors((finalGoodSectors.country == 	"Yemen"	 &
 isnan(finalGoodSectors.otherShare)),:) = [];
 	%sectorData = sectorData((sectorData.year == "2014" | sectorData.year ==
 "2018"),:);
 	%sectorData = sectorData(sectorData.sector == 'Total Economy',:);
 	% Keeping only needed vars (dropping year, even if not all ==2019.
 	% See previous point).
 finalGoodSectors = finalGoodSectors(:,
["countrycode"	 "country"	 "rice_code"	 "rice_region"	 "transportShare"	 "industryShare"	 "otherShare"	]);
 	% GDP, CAPITAL, EMPLOYMENT MERGE FROM PWT
 finalGoodSectors =
 outerjoin(finalGoodSectors,PWT,	"Keys"	,"countrycode"	,"MergeKeys"	,true);
 finalGoodSectors = finalGoodSectors(:,
["country_PWT"	 "countrycode"	 "rice_code_PWT"	 "rice_region_PWT"	 	...

"thesisCode"	 "cgdpo"	 "cn"	 "emp"	 "transportShare"	 "industryShare"	 "otherShare"	 "labsh"	 "labshExt"	]);
 finalGoodSectors = renamevars(finalGoodSectors,
 [	"country_PWT"	 "rice_code_PWT"	 "rice_region_PWT"	],
 [	"country"	 "rice_code"	 "rice_region"	]);
 	% MERGING WITH ILOSTAT ALSO
 finalGoodSectors = outerjoin(finalGoodSectors, ilostatData);
 	%sectorLabor.Sanity = sectorLabor.otherLabSh + sectorLabor.industryLabSh +
 sectorLabor.transportLabSh;
 	% sanity check passed
 finalGoodSectorsSanity = finalGoodSectors;
 	% Deleting small countries without PWT data

finalGoodSectors(isundefined(finalGoodSectors.country_finalGoodSectors), :) =
 [];
 	%sectorData = removevars(sectorData, "country_ilostatData");
 	% MERGING WITH OECD CAPITAL SHARE DATA
 finalGoodSectors = outerjoin(finalGoodSectors,oecdData);
 finalGoodSectors = renamevars(finalGoodSectors, 	...
 [	"country_finalGoodSectors"	 "countrycode_finalGoodSectors"	], 	...
 [	"country"	 "countrycode"	]);
30
88

% MERGING WITH PWT CAPITAL IN TRANSPORT DATA
 	% sectorData = outerjoin(sectorData,pwtTransData);
 	% sectorData = renamevars(sectorData, "countrycode_sectorData",
 "countrycode");
 	% sectorData.tranSanity = sectorData.transCapSh -
 sectorData.transCapShPWT;
 	% Sanity check failed. OECD >> PWT. Off by up to 10 percetage points.
 	% PWT data is only for 'transport equipment', OECD has all capital in
 	% the sector which could include more than just the equipment.
 	% Conclusion: Dropping PWT data.
 	% % MERGING WITH IEA VOLUME DATA
 	% sectorDataWithIEA = outerjoin(sectorData, ieaDataUnstacked);
 	%
 	% MERGING WITH REGION CODES
 finalGoodSectors = outerjoin(finalGoodSectors, countriesPWTRICE,
 Keys=	"countrycode"	, MergeKeys=true);
 	% Deleting small countries without PWT data
 	%sectorData(isundefined(sectorData.country_sectorData), :) = [];
 	%sectorData = removevars(sectorData, ["country"]);
 finalGoodSectors = renamevars(finalGoodSectors, 	...

[	"country_finalGoodSectors"	 "rice_code_finalGoodSectors"	 "rice_region_finalGoodSectors"	 "thesisCode_finalGoodSectors"	], 	...
 [	"country"	 "rice_code"	 "rice_region"	 "thesisCode"	]);
 finalGoodSectors = sortrows(finalGoodSectors, 	"countrycode"	);
 finalGoodSectors = sortrows(finalGoodSectors, 	"countrycode"	, 	"ascend"	);
 sectorDataPreInterpol = finalGoodSectors;
 save(	"calibData.mat"	, 	"sectorDataPreInterpol"	, 	"-append"	);
else
 load(	"calibData.mat"	, 	"sectorDataPreInterpol"	);
end
MERGING AND PREPARING INTERMEDIATE
AND REFINED ENERGY SECTORS
Data: IEA
if	 wrangle.energy == 1
 valueset = 1:4;
 catnames = {	'transport'	,'industry'	,'other'	, 	'electricity'	};
 energyData = outerjoin(countriesPWTRICE, ieaData);
 energyDataSanity = energyData;
31
89

energyData(isundefined(energyData.countrycode),:) = [];
 energyData(isnan(energyData.coal),:) = [];
 energyData.sector = string(energyData.sector);
 energyData = sortrows(energyData, [	"countrycode"	 "sector"	], 	"ascend"	);
 	%sizes2 = categorical(A2,valueset,catnames,'Ordinal',true)
 energyDataUnstacked = outerjoin(countriesPWTRICE, ieaDataUnstacked);
 energyDataUnstacked(isundefined(energyDataUnstacked.countrycode),:) = [];
 	% Countries with no PWT data.
 energyDataUnstacked =
 sortrows(energyDataUnstacked, 	"countrycode"	, 	"ascend"	);
 energyDataUnstacked(isnan(energyDataUnstacked.coal),:) = []; 	% Countries
 with no IEA data
 save(	"calibData.mat"	, 	"energyData"	, 	"-append"	);
 save(	"calibData.mat"	, 	"energyDataUnstacked"	, 	"-append"	);
else
 load(	"calibData.mat"	, 	"energyData"	, 	"energyDataUnstacked"	);
end
MERGING INTO ONE BIG SHEET
wrangledData = outerjoin(finalGoodSectors, energyDataUnstacked,
 Keys=	"countrycode"	);
Functions %%
%%%%%%%%%%%%%%%
% Changes table variables to category-type. Easier to work with.
function	 tab = categorizer(tabInput, numOfVars)
for	 i = 1:numOfVars
 tabInput =
 convertvars(tabInput,tabInput.Properties.VariableNames{i},	'categorical'	);
end
tab = tabInput;
end
Published with MATLAB® R2022a
32
90

Appendix F
Calibration script
91

Table of Contents
 ........................................................................................................................................................ 1
1. SETUP	 .......................................................................................................................................... 2
GROWTH RATES	 .............................................................................................................................. 6
MISSING DATA INTERPOLATION IN FINAL GOOD SECTOR DATA	 ................................................... 7
REGIONAL AGGREGATION (of levels)	 ............................................................................................. 11
Calibration read-ins (bottom-->top)	 ...................................................................................................... 13
Calibration coefficient equations	 .......................................................................................................... 15
Initialization of other coefficients	 ......................................................................................................... 17
Production calculation	 ........................................................................................................................ 18
Need input vectors to carry out calculation:	 ........................................................................................... 19
Refined primary energy production:	 ...................................................................................................... 19
Electricity sector	 ............................................................................................................................... 20
Intermediate sector (sector specific energy composite):	 ............................................................................ 20
Final good production sector:	 .............................................................................................................. 20
Final good aggregation sector:	 ............................................................................................................. 21
Diffusion model	 ................................................................................................................................ 21
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% ACE native production system
%
% last edited 05/24/22
% ANDREAS SYNC CHECK 16:22 CT 5:01
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% README %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%
% - Required add-ons: Statistical and Machine learning package (for regress
 function) %
% - Pick correct region before running code, at first headline in SETUP.
 %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%
% MATRIX ORDERS %%%
%
% a) Sectors are ALWAYS ordered (1) Transport (2) Industry (3) Other
%
% b) Fuels are ordered
% EITHER
% Coal, oil, natgas, bio, renew [into elec]
% OR
% Coal, oil, natgas, bio, elect [into int (d)]
% so it is the fifth input that changes.
%
% c) *Regions* are always columns, while *sectors* or *fuels* are rows.
% Table of contents
% 1. Setup (load data, set code logicals, pick region and substitutabilities)
1
92

% 2. Growth rate calculations
% 3. Missing data interpolations
% 4. Regional aggregation
% 5. Calibration input data read-in
% 6. Calibration calculations
% 7. Production calculations
% LOADED DATASETS
% pwtDynamic: Used for growth rates. PWT 2010-2019.
% finalGoods Data: Final good sectors. PWT, UN, WIOD, OECD, ILOSTAT.
% energyData: IEA energy data (matrix form, 4 rows per country)
% energyDataUnstacked: IEA energy data (one row per country, wider)
% wPriceLevel: Vector of regional price levels for thesis
% CALIBRATION INPUT TABLE NAMES (Interpolated and regionalized)
% growthRates = Regional growth rates [XXX and World coming]
% finalGoods = Regional final good sector
% energyMatrix & energy = energy data in matrix and one-row form, resp.
% UNITS (after conversions)
% Final good sector: Real GDP (in PPP 2017USD) -- NOTE: Original PWT is in
 millions. This is just in ones.
% Energy: Terajoules (can consider Megajoules instead)
% Emissions: kg CO2
% Price of energy: USD/TJ (PPP)
% Price of emissions: USD/kg
% Notes:
% - all production parameters are in the prod structure, e.g.: prod.a_fin,
 prod.alpha_fin
% - all calibration inputs are in the calib structure.
% - initial state values are initial.
% Code is called by RegionalACE, requiring calibrate_calculate being defined:
% = 1 : calibration
% = 2 : calculation
% = 3 : both (if using file as stand-alone, uncomment next line)
%clear all; calibrate_calculate = 3; recycle_mat_files = 0;
% NOTE CT:
%
% Overwritten alpha in the main code by hand for the moment subracting
 nu*alpha_bar
%
if	 calibrate_calculate == 1 | 3
1. SETUP
 	%clear all
 	%clc
2
93

if	 recycle_mat_files == 0
 load(	'calibData.mat'	); 	% Data import from wrangling.m
 prod.structure = 	'ACE native'	;
 production.gr = 1;
 production.interpolate = 1;
 production.regionalize = 1;
 production.calibration_input=1; 	% Initialize parameters and variables
 with data input
 production.coefficients=1; 	% to calibrate the production system
 	%production.calculation=1; % to use the file for simulating production
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% Remove comment for selected region %
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 regionSelected = 	"thesisCode"	; 	% is WORLD, 1 is LOW, 2 is HIGH
 	%regionSelected = "rice_code"; % NOTE: Will be biased until alpha bar
 and prices are RICE-regionalized
 	%regionSelected = "country_code"; % NOTE: Will not work until
 correspondance between IEA data and rest is sorted.
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% Substitutabiity parameters in different sectors: %
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% World parameters
 prod.agg_s = -1; 	% Subst. at aggregation level
 prod.agg_sbar = 1; 	% Returns to scale
 prod.int_s = [-4 -.0001 .1667]'; 	% Subst. energy intermediate sectors
 prod.int_s_0 = .444; 	% Subst. in electricity production
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% Avg. Fossil industry capital share of total capital %
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% Used to get alpha bars, which are extrapolated
 fossilCapShareUS = 0.055; 	% Source USBLS (2019 data), China, Russia
 fossilCapShareCHN = 0.0155; 	% Source Chn. stat. yearbook 2020.
 fossilCapShareRUS = 0.03; 	% Source various.
 	%%%%%%%%%%%%%
 	% Alpha bar %
 	%%%%%%%%%%%%%
 	% US alpha bar: 0.7723 (based on 5.5% capital)
 	% Chinese: 0.283 (1.55%)
 	% Russian: 0.469 (3%)
 alphaBarH = 0.712 	% 80/20 avg of US and Russia
 alphaBarL = 0.283 	% Chinese data.
 alphaBarH = 0.1
 alphaBarL = 0.05
 alphaBarW = 0.8*alphaBarH + 0.2*alphaBarL; 	% Weighed by share of total
 fossil.
3
94

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% CO2 coefficients per fuel type %
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% CO2 emissions should be in kg per TJ
 	% [Coal, oil, gas, bio, renew]'
 co2Factors = [90762.95;
 69247.51;
 50148.00;
 3462.38;
 0];
 	% Observed emissions correction (IEA = 33.2 Gt)
 co2Factors = co2Factors .* 1.0463;
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% Average world refined energy prices %
 	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 	% Original units %%%
 	% Coal: 83.16 USD per million tonne
 	% Oil prodcuts: 71.64 USD per barrel
 	% Nat. gas: 3.13 USD per million Btu
 	% Bio (ethanol): 1.25 USD per gallon
 	% Bio (biodiesel): 800 USD per ton
 	% Bio (wood pell): 166.25 per ton
 	% Renew: 0,055 USD/kWh
 	% 1 short ton = 0.90718474 tonne
 	% Conversion factors (https://www.justintools.com/unit-conversion/
energy.php)
 	% COAL
 	% 1 Tonne of coal equivalent = 0.029288 TJ
 	% OIL PRODUCTS
 	% Barrel to tonnes (oil products basket) = 0.124. BP Conversion
 factors
 	% Tonne to gigajoule (oil products basket) = 43.076 BP Conversion
 factors
 	% Gigajoule to terajoule = 0.001
 	% ===> 1 barrel = 0.005341424 TJ
 	% NATURAL GAS
 	% 1 000 000 Btu = 0.00105505585 TJ
 	% BIO
 	% Ethanol: 80000 Btu per gallon whivh gives 8.440448E-5 TJ
 	% Biodiesel: 1 t biodiesel = 0,86 toe which is 0.03600648 TJ
 	% Pellets: 1 t pellets = 5 MWh = 0.018 TJ
 	% RENEW:
 	% 1 kWh = 3.6E-6 TJ
 	% Converting idiosyncratic units to USD/TJ
 price.coal = 83.16/0.029288;
 price.oil = 71.64/0.005341424;
 price.natGas = 3.13/0.00105505585;
 	%price.bioE = 1.3/8.440448E-5;
4
95

%price.bioD = 850 / 0.03600648;
 price.bioW = 166.25 / 0.018;
 price.renew = 0.055 / 3.6E-6;
 	% Vector of refined energy prices
 enePricesWorld = [
 price.coal;
 price.oil;
 price.natGas;
 price.bioW;
 price.renew
 ];
 	% Electricity = price of renewables
 elPriceWorld = price.renew;
 	% Setting prices for the two thesis regions
 enePricesHigh = enePricesWorld;
 elPriceHigh = elPriceWorld;
 	% bioenergy price half in low-income countries
 enePricesLow = enePricesHigh .* [1 1 1 .5 1]';
 elPriceLow = elPriceHigh;
 	% Merging, converting from 2019 to 2017 and adjusting for PPP
 	% Energy
 enePrices = [enePricesWorld enePricesLow enePricesHigh]; 	% regional
 matrix
 enePrices = enePrices .* 0.9588; 	% Inflation adjustment
 enePrices = enePrices ./ wPriceLevel'; 	% PPP correction
 	% Electricity
 elPrices = [elPriceWorld elPriceLow elPriceHigh]; 	% Nominal 2019 USD.
 Row, one price for each region.
 elPrices = elPrices .* 0.9588 ; 	% Inflation adjusted. Row, one price
 for each region.
 elPrices = elPrices ./ wPriceLevel'; 	% PPP adjusted 2017 USD.
 	%%%%%%%%%%%%%%%%%%%%
 	% Emissions prices %
 	%%%%%%%%%%%%%%%%%%%%
 	% Finding world average emissions price
 emCoverage = 0.145; 	% World Bank estimate coverage rate
 emPriceCovered = 2.74887 	% Weighted avg price of covered emissions, in
 2017USD per TONNE co2
 emPriceCovered = emPriceCovered / 1000; 	% to per kg co2 (emissions
 factors are in kg per TJ)
 emPriceWorld = emCoverage * emPriceCovered; 	% Avg world emissions
 price.
 	% Setting emissions prices for two-region model
 emPriceL = 0; 	% No price initiatives in these countries
 emPriceH = emPriceWorld / 0.807; 	% 80.7 percent of tot. emissions are
 in H.
5
96

% Collecting and adjusting for PPP
 emPrices = [emPriceWorld emPriceL emPriceH;] 	% In USD/tCO2 - row
 vector
 emPricesKG = emPrices .* wPriceLevel'; 	% PPP adjusted (kg CO2)
 	% Need same units as ene prices, for additivity.
 	% This is a matrix: rows are fuels, columns are regions
 	%emPricesTJ = co2Factors * emPrices; % In nom USD/TJ, for PPP
 sensitivity check
 emPricesTJ = co2Factors * emPricesKG; 	% in PPP USD/TJ
 	% end setup
GROWTH RATES
(1) Takes pwtDynamic as input (2) Groups data into regions (3) Calculates growth rates for GDP and capital per region
(4) Adds world growth rate on top of table (5) "growthRates" is output table (region is World)
 	if	 production.gr == 1
 	% Aggregates GDP and capital over regions for each year and takes
 log transformations.
 	% Runs a regression of logvars on constant and year to get avg.
 yearly growth rates.
 pwtDynamicRegional = varfun(@(x) sum(x, 	"omitnan"	),
 pwtDynamic, 	'InputVariables'	 , 	...
 [	"cgdpo"	 "cn"	] , 	'GroupingVariables'	 ,
 [regionSelected 	"year"	]);
 logs = varfun(@log, pwtDynamicRegional,	"InputVariables"	,
["Fun_cgdpo"	 "Fun_cn"	]);
 pwtDynamicRegional = [pwtDynamicRegional logs];
 reg_gdp = varfun(@(x) regress(x, [ones(10,1)
 unique(pwtDynamicRegional.year)]), 	...
 pwtDynamicRegional, 	'InputVariables'	 , 	'log_Fun_cgdpo'
 , 	'GroupingVariables'	 , regionSelected);
 	% pwtDynamicRegional, InputVariables="log_Fun_cgdpo",
 GroupingVariables=regionSelected);
 reg_cap =varfun(@(x) regress(x, [ones(10,1)
 unique(pwtDynamicRegional.year)]), 	...
 pwtDynamicRegional, 	'InputVariables'	 , 	'log_Fun_cn'
 , 	'GroupingVariables'	 , regionSelected);
 	% pwtDynamicRegional, InputVariables="log_Fun_cn",
 GroupingVariables=regionSelected);
 gr_gdp = reg_gdp(2:2:end, :);
 gr_cap = reg_cap(2:2:end, :);
 growthRates = join(gr_gdp,gr_cap);
 	% World
 	% pwtDynamicWorld = varfun(@(x) sum(x, "omitnan"), pwtDynamic,
 InputVariables=["cgdpo" "cn"], GroupingVariables="year");
 pwtDynamicWorld = varfun(@(x) sum(x, 	"omitnan"	),
 pwtDynamic, 	'InputVariables'	 , [	"cgdpo"	 "cn"	] , 	'GroupingVariables'
 , 	'year'	);
6
97

logs = varfun(@log, pwtDynamicWorld,	"InputVariables"	,
["Fun_cgdpo"	 "Fun_cn"	]);
 pwtDynamicWorld = [pwtDynamicWorld logs];
 reg_gdpW = regress(pwtDynamicWorld.log_Fun_cgdpo, [ones(10,1)
 unique(pwtDynamicWorld.year)]);
 reg_capW = regress(pwtDynamicWorld.log_Fun_cn, [ones(10,1)
 unique(pwtDynamicWorld.year)])
 growthRatesWorld = [reg_gdpW(2) reg_capW(2)];
 newRow = growthRates(1,:);
 newRow{1,regionSelected} = categorical(0);
 newRow{1,3:4} = growthRatesWorld;
 growthRates = [newRow; growthRates];
 	end	 % growth rate calculations
MISSING DATA INTERPOLATION IN FINAL
GOOD SECTOR DATA
I use the regional averages to interpolate missing country-level data. This code interpolates over RICE regions, but it
is possible to use other regions (after some cumbersome re-coding)
 	if	 production.interpolate == 1
 	% Saving an untouched version
 finalGoodsData = sectorDataPreInterpol;
 	% Finding GDP per capital stock and GDP per worker to use for
 	% extrapolation
 finalGoodsData.capPerGdp = finalGoodsData{:,	"cn"	} ./
 finalGoodsData{:,	"cgdpo"	};
 finalGoodsData.empPerGdp = finalGoodsData{:,	"emp"	} ./
 finalGoodsData{:,	"cgdpo"	};
 	% Finding mean of sectors in each region
 sectorDataRegionsMean = varfun(@(x) mean(x,	'omitnan'	),
 finalGoodsData, 	...
 	'GroupingVariables'	 , [	"rice_region"	 "rice_code"	], 	...
 	'InputVariables'	 , [ 	...
 	"cgdpo"	 ...
 	"cn"	 ...
 	"emp"	 ...
 	"capPerGdp"	 ...
 	"empPerGdp"	 ...
 	"transportShare"	 ...
 	"industryShare"	 ...
 	"otherShare"	 ...
 	"transportLabSh"	 ...
 	"industryLabSh"	 ...
 	"otherLabSh"	 ...
 	"transCapSh"	 ...
7
98

"indCapSh"	 ...
 	"othCapSh"	 ...
 	"labsh"	]);
 	% "othCapSh"]); and above:
 	% GroupingVariables=["rice_region" "rice_code"], ...
 	% InputVariables=[ ...
 	% Applying regional mean shares to countries without sectoral data
 	% Missing GDP shares
 	for	 i = 1:12
 	if	 ~isempty(finalGoodsData((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.transportShare)),:))
 	% Filling in missing GDP shares from regional mean
 finalGoodsData.transportShare((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.transportShare))) = 	...
 sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_transportShare"	};
 finalGoodsData.industryShare((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.industryShare))) = 	...
 sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_industryShare"	};
 finalGoodsData.otherShare((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.otherShare))) = 	...
 sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_otherShare"	};
 	else
 disp(i)
 	end
 	end
 	% Missing capital data
 	for	 i = 1:12
 	if	 ~isempty(finalGoodsData((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.cn)),:))
 	% Filling in missing capital data from cap/gdp mean
 	% extrapolation of regional mean
 finalGoodsData.capPerGdp((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.capPerGdp))) = 	...
 ectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_capPerGdp"	};
 	% Getting capital and labor from gdp*(x/gdp) [last
 fraction is regional mean]
 finalGoodsData.cn((finalGoodsData.rice_code == string(i) &
 isnan(finalGoodsData.cn))) = 	...
 finalGoodsData.cgdpo((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.cn))) .* 	...
 finalGoodsData.capPerGdp((finalGoodsData.rice_code ==
 string(i) & 	...
 isnan(finalGoodsData.cn)));
 	else
 disp(i)
 	end
 	end
8
99

% Missing emp data
 	for	 i = 1:12
 	if	 ~isempty(finalGoodsData((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.emp)),:))
 	% Filling in missing labor data from emp/gdp mean
 	% extrapolation of regional mean
 finalGoodsData.empPerGdp((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.empPerGdp))) = 	...
 sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_empPerGdp"	};
 	% Getting capital and labor from gdp*(x/gdp) [last
 fraction is regional mean]
 finalGoodsData.emp((finalGoodsData.rice_code == string(i)
 & isnan(finalGoodsData.emp))) = 	...
 finalGoodsData.cgdpo((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.emp))) .*	...
 finalGoodsData.empPerGdp((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.emp)));
 	else
 disp(i)
 	end
 	end
 	% Missing labor share in sectors data
 	for	 i = 1:12
 	if	 ~isempty(finalGoodsData((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.transportLabSh)),:))
 	% Filling in missing labor shares from regional mean
 finalGoodsData.transportLabSh((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.transportLabSh))) = 	...
 sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_transportLabSh"	};
 finalGoodsData.industryLabSh((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.industryLabSh))) = 	...
 sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_industryLabSh"	};
 finalGoodsData.otherLabSh((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.otherLabSh))) =	...
 sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_otherLabSh"	};
 	else
 disp(i)
 	end
 	end
 	% Missing labor share of GDP data
 	for	 i = 1:12
 	if	 ~isempty(finalGoodsData((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.labsh)),:))
 	% Filling in missing labor shares from regional mean
 finalGoodsData.labsh((finalGoodsData.rice_code ==
 string(i) & isnan(finalGoodsData.labsh))) =	...
9
100

sectorDataRegionsMean{sectorDataRegionsMean.rice_code
 == string(i),	"Fun_labsh"	};
 	else
 disp(i)
 	end
 	end
 	% Getting capital share of GDP as extensive variable preparing for
 regional aggregation
 finalGoodsData.capPropExtPWT = finalGoodsData.cgdpo .* (1 -
 finalGoodsData.labsh);
 	% Missing capital share data. For now, all NaNs get the OECD
 average, irrespective of region
 finalGoodsData.transCapSh(isnan(finalGoodsData.transCapSh)) =
 mean(finalGoodsData.transCapSh, 	"omitnan"	);
 finalGoodsData.indCapSh(isnan(finalGoodsData.indCapSh)) =
 mean(finalGoodsData.indCapSh, 	"omitnan"	);
 finalGoodsData.othCapSh(isnan(finalGoodsData.othCapSh)) =
 mean(finalGoodsData.othCapSh, 	"omitnan"	);
 finalGoodsData = sortrows(finalGoodsData,	'rice_code'	,'ascend'	);
 	% Getting nominal value added instead of shares of output
 finalGoodsData.transVA = finalGoodsData{:,	"transportShare"	} .*
 finalGoodsData{:,	"cgdpo"	};
 finalGoodsData.indVA = finalGoodsData{:,	"industryShare"	} .*
 finalGoodsData{:,	"cgdpo"	};
 finalGoodsData.othVA = finalGoodsData{:,	"otherShare"	} .*
 finalGoodsData{:,	"cgdpo"	};
 	%sectorData.sanity = sectorData.transVA + sectorData.indVA +
 sectorData.othVA - sectorData.cgdpo;
 	% sanity check passed
 	%sectorData.Sanity = sectorData.otherLabSh +
 sectorData.industryLabSh + sectorData.transportLabSh;
 	% sanity check passed
 	% Getting number employed in sector instead of fraction
 finalGoodsData.transLab = finalGoodsData{:,	"transportLabSh"	} .*
 finalGoodsData{:,	"emp"	} / 100;
 finalGoodsData.indLab = finalGoodsData{:,	"industryLabSh"	} .*
 finalGoodsData{:,	"emp"	} / 100;
 finalGoodsData.othLab = finalGoodsData{:,	"otherLabSh"	} .*
 finalGoodsData{:,	"emp"	} / 100;
 	%sectorData.sanityLab = sectorData.emp -sectorData.transLab -
 sectorData.indLab - sectorData.othLab;
 	% sanity check passed
 	% Getting capital value instead of shares
 finalGoodsData.transCap = finalGoodsData{:,	"transCapSh"	} .*
 finalGoodsData{:,	"cn"	};
 finalGoodsData.indCap = finalGoodsData{:,	"indCapSh"	} .*
 finalGoodsData{:,	"cn"	};
10
101

finalGoodsData.othCap = finalGoodsData{:,	"othCapSh"	} .*
 finalGoodsData{:,	"cn"	};
 	%sectorData.sanityLab = sectorData.cn - sectorData.transCap -
 sectorData.indCap - sectorData.othCap;
 	% sanity check passed
 	end	 %interpolation
REGIONAL AGGREGATION (of levels)
This section takes country level macro and energy data and aggregates it into regions. Region is World. For thesis,
region 1 is Poor and 2 is Rich.
 	if	 production.regionalize == 1
 	% For pre-aggregation deletion of share variables
 notShareVars =
 ~contains(finalGoodsData.Properties.VariableNames(:), 	"Sh"	);
 	%%%%%%%%%
 	% WORLD %
 	%%%%%%%%%
 regionCode = categorical(	"0"	);
 GroupCount = 1;
 	% Final good sectors
 finalGoodsWorld = finalGoodsData(:,notShareVars);
 finalGoodsWorld = varfun(@(x) sum(x, 	"omitnan"	),
 finalGoodsWorld, 	...
 	"InputVariables"	, @isnumeric);
 finalGoodsWorld = addvars(finalGoodsWorld, regionCode,
 GroupCount, 	'Before'	, 1);
 	% Energy unstacked (Whole region in one row)
 energyWorld = varfun(@(x) sum(x, 	"omitnan"	),
 energyDataUnstacked, 	...
 	"InputVariables"	, @isnumeric);
 energyWorld = addvars(energyWorld, regionCode,
 GroupCount, 	'Before'	, 1);
 	% Energy stacked (Four rows per region)
 energyWorldStacked = varfun(@(x) sum(x, 	'omitnan'	),
 energyData, 	...
 	"InputVariables"	,@isnumeric, 	"GroupingVariables"	,[	"sector"	]);
 energyWorldStacked.regionCode(:) = regionCode;
 energyWorldStacked.GroupCount(:) = GroupCount;
 energyWorldStacked = movevars(energyWorldStacked, 	...
 [	"regionCode"	 "GroupCount"	], 	'Before'	, 1);
 	%%%%%%%%%%%
 	% REGIONS %
 	%%%%%%%%%%%
11
102

% Final good sectors
 finalGoodsRegional = finalGoodsData(:,notShareVars);
 finalGoodsRegional = varfun(@(x) sum(x, 	"omitnan"	),
 finalGoodsRegional, 	...
 	"InputVariables"	, @isnumeric, 	'GroupingVariables'	 ,
 [regionSelected]);
 	%"InputVariables", @isnumeric,
 GroupingVariables=[regionSelected]);
 finalGoodsRegional.Properties.VariableNames(1) = 	"regionCode"	;
 	% Energy unstacked (Whole region in one row)
 energyRegional = varfun(@(x) sum(x, 	"omitnan"	),
 energyDataUnstacked, 	...
 	"InputVariables"	, @isnumeric, 	"GroupingVariables"	,
[regionSelected]);
 energyRegional.Properties.VariableNames(1) = 	"regionCode"	;
 	% Energy stacked (Four rows per region)
 energyRegionalStacked = varfun(@(x) sum(x, 	'omitnan'	),
 energyData, 	...
 	"InputVariables"	,@isnumeric, 	"GroupingVariables"	,
[regionSelected 	"sector"	]);
 energyRegionalStacked.Properties.VariableNames(1) = 	"regionCode"	;
 finalGoods = [finalGoodsWorld;finalGoodsRegional];
 finalGoods.capShPWT = finalGoods.Fun_capPropExtPWT ./
 finalGoods.Fun_cgdpo;
 energy = [energyWorld;energyRegional];
 energyMatrix = [energyWorldStacked;energyRegionalStacked];
 	% Adding total fuel use
 energy.totalFossil = energy.Fun_coal + energy.Fun_oil +
 energy.Fun_natGas;
 energy.totalPrimary = energy.Fun_coal + energy.Fun_oil +
 energy.Fun_natGas + 	...
 energy.Fun_bio + energy.Fun_renew;
 	if	 regionSelected == 	"thesisCode"
 	% Manually appending bunkers to thesis emissions, since this
 data was made available right before the deadline.
 energy.Fun_oil_x1Transport = energy.Fun_oil_x1Transport +
 [1.8E+07; 0.3E+07; 1.5E+07];
 energy.Fun_oil = energy.Fun_oil + [1.8E+07; 0.3E+07; 1.5E+07];
 	end
 	end	 % region making
12
103

Calibration read-ins (bottom-->top)
This section takes cleaned and regionalized data and reads it into vectors used for parameter calibration and model
initialization
 	if	 production.calibration_input == 1
 	% Loops over all regions
 numOfRegions = size(finalGoods, 1);
 	for	 i = 1:numOfRegions
 	% CO2Factors
 prod.co2Factors(:,i) = co2Factors;
 	% REFINED ENERGY SECTOR: prices and energy volumes
 	% Rows: Energy sector. Column: Region.
 	if	 regionSelected == 	"thesisCode"
 	% (p^e) (5x1) vector [coal; oil; natGas; bio; renew]
 calib.price_ene_e(:,i) = enePrices(:,i) +
 emPricesTJ(:,i); 	% PPP USD/TJ
 	elseif	 regionSelected == 	"rice_code"
 calib.price_ene_e(:,i) = enePricesWorld +
 emPricesTJ(:,1); 	% TJ
 	end
 	% (e_i) (5x1) vector [coal; oil; natGas; bio; renew]
 calib.quant_ene_e(:,i) = energy{i,
["Fun_coal"	; 	"Fun_oil"	; 	"Fun_natGas"	; 	"Fun_bio"	; 	"Fun_renew"	]}; 	% TJ
 	% REFINED ENERGY SECTOR: emissions
 	% Emissions are in kg, to match real emission data.
 	% Rows: Energy sector. Column: Region.
 	if	 regionSelected == 	"thesisCode"
 	% (p^E) Emission prices. (5x1) vector, one per fuel.
 calib.price_ene_E(i) = emPricesKG(i); 	% PPP USD/kgCO2
 	% (E) Emissions per fuel. (5x1) vector, one per fuel
 calib.quant_ene_E(:,i) = co2Factors .*
 calib.quant_ene_e(:,i); 	% kg CO2
 	elseif	 regionSelected == 	"rice_code"
 	% (p^E) Emission prices. (5x1) vector, one per fuel.
 calib.price_ene_E(i) = emPricesKG(1); 	% PPP USD/kgCO2
 (World avg for all regions)
 	% (E) Emissions per fuel. (5x1) vector, one per fuel
 calib.quant_ene_E(:,i) = co2Factors .*
 calib.quant_ene_e(:,i); 	% kg CO2
 	end
 	% ELECTRICITY SECTOR
 	% Rows: Fuel inputs to electricity. Column: Region.
 	if	 regionSelected == 	"thesisCode"
 calib.price_int_0(i) = elPrices(i) +
 emPricesTJ(i); 	%(p^e_0) electricity price scalar
 	elseif	 regionSelected == 	"rice_code"
13
104

calib.price_int_0(i) = elPrices(1) + emPricesTJ(1);
 	end
 	%calib.quant_ene_0 = energyMatrix{28,
 calib.quant_ene_0(:,i) = 	...
 energy{i,
 [	"Fun_coal_x4Electricity"	; 	"Fun_oil_x4Electricity"	; 	"Fun_natGas_x4Electricity"	; 	"Fun_bio_x4Electricity"	; 	"Fun_renew_x4Electricity"	]};
 calib.quant_ene_0_inputSum(i) = sum(calib.quant_ene_0(:,i)); 	%
 (e_0)(sum of inputs before losses)
 calib.quant_int_0Test(i) = energy{i, [	"Fun_elAndHeat"	]}; 	%
 (e_0) the total produced electricity (read-in of output from data)
 	% INTERMEDIATE SECTOR
 	% Rows: Inputs and outputs. Column: Region.
 calib.price_int_d(i) = 1; 	% (p_l^d) Intermediate good prices
 (unity by normalization)
 calib.price_int_e(:,i) = [calib.price_ene_e(1:4,i);
 calib.price_int_0(i)]; 	% (N/A) Vector of prices of inputs to intermediate
 production [c o g b el]'
 	%calib.quant_int_e = energyMatrix{25:27, ["Fun_coal" "Fun_oil"
 "Fun_natGas" "Fun_bio" "Fun_elAndHeat"]}; % (N/A) (3x5)
 calib.quant_int_e_tr(:,i) = energy{i,
 [	"Fun_coal_x1Transport"	; 	"Fun_oil_x1Transport"	; 	"Fun_natGas_x1Transport"	; 	"Fun_bio_x1Transport"	; 	"Fun_elAndHeat_x1Transport"	]}; 	%(5x1)
 calib.quant_int_e_in(:,i) = energy{i,
 [	"Fun_coal_x2Industry"	; 	"Fun_oil_x2Industry"	; 	"Fun_natGas_x2Industry"	; 	"Fun_bio_x2Industry"	; 	"Fun_elAndHeat_x2Industry"	]};
 calib.quant_int_e_ot(:,i) = energy{i,
 [	"Fun_coal_x3Other"	; 	"Fun_oil_x3Other"	; 	"Fun_natGas_x3Other"	; 	"Fun_bio_x3Other"	; 	"Fun_elAndHeat_x3Other"	]};
 	% Stacking sectors to get matrix for easier calc. in next
 step. Remember, happens for each country so matrix is ok.
 calib.quant_int_e = [calib.quant_int_e_tr(:,i)';
 calib.quant_int_e_in(:,i)'; calib.quant_int_e_ot(:,i)'];
 calib.quant_int_d_val(:,i) = calib.quant_int_e *
 calib.price_int_e(:,i); 	% Matrix multiplication (3x5)(5x1) = (3x1) vector of
 value of intermediate by sector
 	%calib.quant_int_d_test = sum(calib.quant_int_e,2);
 calib.quant_int_0(i) = calib.quant_int_e_tr(5,i) +
 calib.quant_int_e_in(5,i) + calib.quant_int_e_ot(5,i);
 	% FINAL GOODS AND AGGREGATION
 	% Rows: Inputs and outputs. Column: Region.
 calib.price_fin_c(i) = 1; 	% (p_l) Sector prices (column
 vector)
 calib.price_K(i) = 1; 	% (p^K) (economy-wide)
 	%calib.price_N(i) = 1; % (omega) (economy-wide
 calib.quant_fin_c(:,i) = [finalGoods{i, 	"Fun_transVA"	};
 finalGoods{i, 	"Fun_indVA"	};
 finalGoods{i, 	"Fun_othVA"	}]; 	% (c) (3x1) Sector
 quantities.
 calib.quant_fin_K(:,i) = finalGoods{i, 	"Fun_cn"	}; 	% (K)
 Scalar. Total capital in region, used to calibrate alpha.
 calib.quant_fin_K_sectors(:,i) = finalGoods{i,
 [	"Fun_transCap"	; 	"Fun_indCap"	; 	"Fun_othCap"	]};
14
105

calib.quant_fin_N(:,i) = finalGoods{i, 	"Fun_emp"	}; 	% (N)
 Scalar.Total labor in region (not vector!)
 calib.quant_fin_N_sectors(:,i) = finalGoods{i,
 [	"Fun_transLab"	; 	"Fun_indLab"	; 	"Fun_othLab"	]};
 calib.price_agg_Y(i) = 1; 	% (p) Aggregate price level
 (scalar)
 calib.quant_agg_Y(:,i) = finalGoods{i, 	"Fun_cgdpo"	};
 calib.gr_Y(i) = growthRates{i, 	"Fun_log_Fun_cgdpo"	};
 calib.gr_K(i) = growthRates{i, 	"Fun_log_Fun_cn"	};
 	% end calibration read-ins
Calibration coefficient equations
Each set of sectors has a matrix containing varieties in columns and production coefficients on different inputs in a row
 	% Alpha bar is extrapolated from only 3 countries.
 	% High is 80/20 split of US and Russia.
 	% Low is China
 	% I use world average if RICE regions are used
 	if	 regionSelected == 	"thesisCode"
 	if	 i == 1
 prod.ene_alpha(i) = alphaBarW;
 	elseif	 i == 2
 prod.ene_alpha(i) = alphaBarL;
 	elseif	 i == 3
 prod.ene_alpha(i) = alphaBarH;
 	end
 	elseif	 regionSelected == 	"rice_code"
 prod.ene_alpha(i) = alphaBarW;
 	end
 	% (a_{l,i}^tilde) Intermediate energy good CES shares (NOTE:
 col 5 is electricity. Does not include renew) (B.14)
 	% % Rows: Fuel-specific parameters. Column: Region.
 	% NOTE: currently manually selecting substitutability
 parameter. Should be automated to have dynamic number of sectors.
 prod.int_atildestar_tr(:,i) = (calib.price_int_e(:,i) /
 calib.price_int_d(i)) .* (calib.quant_int_e_tr(:,i) ./ 	...
 calib.quant_int_d_val(1,i)).^(1-prod.int_s(1));
 prod.int_asum_tr(i) = sum(prod.int_atildestar_tr(:,i));
 prod.int_Atilde_tr(i) = prod.int_asum_tr(i).^(1./
prod.int_s(1));
 prod.int_atilde_tr(:,i) = prod.int_atildestar_tr(:,i) ./
 prod.int_asum_tr(i);
 prod.int_atildestar_in(:,i) = (calib.price_int_e(:,i) /
 calib.price_int_d(i)) .* (calib.quant_int_e_in(:,i) ./ 	...
 calib.quant_int_d_val(2,i)).^(1-prod.int_s(2));
15
106

prod.int_asum_in(i) = sum(prod.int_atildestar_in(:,i));
 prod.int_Atilde_in(i) = prod.int_asum_in(i).^(1./
prod.int_s(2));
 prod.int_atilde_in(:,i) = prod.int_atildestar_in(:,i) ./
 prod.int_asum_in(i);
 prod.int_atildestar_ot(:,i) = (calib.price_int_e(:,i) /
 calib.price_int_d(i)) .* (calib.quant_int_e_ot(:,i) ./ 	...
 calib.quant_int_d_val(3,i)).^(1-prod.int_s(3));
 prod.int_asum_ot(i) = sum(prod.int_atildestar_ot(:,i));
 prod.int_Atilde_ot(i) = prod.int_asum_ot(i).^(1./
prod.int_s(3));
 prod.int_atilde_ot(:,i) = prod.int_atildestar_ot(:,i) ./
 prod.int_asum_ot(i);
 	% (a_{0,i}^tilde) Electricity sector CES share on fuel inputs
 (incl. renew) (missing in theory, from B.19)
 	% % Rows: Fuel-specific parameters. Column: Region.
 	% Note:
 prod.int_atildestar_0(:,i) = (calib.price_ene_e(:,i) ./
 calib.price_int_0(i)) .* (calib.quant_ene_0(:,i) ./ 	...
 calib.quant_int_0(i)).^(1-prod.int_s_0); 	% (1x5) vector.
 Electricity sector. idx 5 is RENEW, not ELECTRICITY
 prod.int_asum_0(i) = sum(prod.int_atildestar_0(:,i));
 prod.int_Atilde_0(i) = prod.int_asum_0(i).^(1./prod.int_s_0);
 prod.int_atilde_0(:,i) = prod.int_atildestar_0(:,i) ./
 prod.int_asum_0(i);
 	% (alpha, nu) Final consumption good CD function share
 calibration (B.12)
 	% alpha is sector INdependent. CHECK: Should be (3x1)
 identical entries or scalar.
 	% NOTE: I multiply the FLOW Y by 10 to get decadal timestep
 prod.fin_alpha(i) = (calib.price_K(i) ./
 calib.price_fin_c(i)) .* (calib.quant_fin_K(i) ./ 	...
 (10*calib.quant_agg_Y(i))); 	% Scalar or (3x1) vector with
 identical entries.
 	% nu is sector DEpendent
 prod.fin_nu(:,i) = (calib.price_int_d(i) ./
 calib.price_fin_c(i)) .* (calib.quant_int_d_val(:,i) ./ 	...
 calib.quant_fin_c(:,i)); 	% Should be a (3x1) vector.
 Originally 0.1.
 	% getting the overall nu for wage calibration
 prod.fin_nuOverAll(i) = sum(calib.quant_int_d_val(:,i)) ./
 sum(calib.quant_fin_c(:,i));
 prod.fin_nuOverAll(i) = mean(prod.fin_nu(:,i));
 	%prod.fin_nuOverAll(i) = 3 * prod.fin_nuOverAll(i);
 prod.fin_alpha(i) = prod.fin_alpha(i) - prod.fin_nuOverAll(i);
 	%Debug
 	%prod.fin_alpha(i) = alphaTest(i);
16
107

calib.shareN(i) = 1 - prod.fin_alpha(i)-
 prod.fin_nuOverAll(i);
 calib.GDPtoN(i)=calib.quant_agg_Y(i) / calib.quant_fin_N(i);
 calib.price_N(i) = (1 - prod.fin_alpha(i)-
 prod.fin_nuOverAll(i)) .* (calib.quant_agg_Y(i) / 	...
 calib.quant_fin_N(i));
 	% (a^bar) The coefficient on E inside the min() function
 	% Rows: Fuel-specific parameters. Column: Region.
 prod.ene_abar(:,i) = ((1-prod.ene_alpha(i)).*
 (calib.price_ene_e(:,i) .* calib.quant_ene_e(:,i)) ./ 	...
 (calib.price_N(i) .* calib.quant_ene_E(:,i))) - 	...
 (calib.price_ene_E(i) ./ calib.price_N(i)); 	% (5x1) vector
 of Leontief coefficients, one for each fuel.
 	% (a) Final aggregation CES function share calibration (B.11)
 prod.agg_astar(:,i) = (calib.quant_fin_c(:,i) ./
 calib.quant_agg_Y(i)).^(1-prod.agg_s);
 prod.agg_asum(:,i) = sum(prod.agg_astar(:,i))
 prod.agg_A(i) = prod.agg_asum(i).^(1/prod.agg_s);
 	% This is the important output
 prod.agg_a(:,i) = prod.agg_astar(:,i) / prod.agg_asum(i); 	%
 Should sum to one
Initialization of other coefficients
(A^bar) TFP in refined energy production % Rows: Fuel-specific parameters. Column: Region.
 prod.ene_Abar(:,i) = ((calib.price_N(i)
 + (calib.price_ene_E(:,i) ./ prod.ene_abar(:,i))).^(1-
prod.ene_alpha(i))) ./ 	...
 (calib.price_ene_e(:,i) .*
 (prod.ene_alpha(i)^prod.ene_alpha(i))*((1-prod.ene_alpha(i))^(1-
 prod.ene_alpha(i))));
 	% With co2factors
 prod.ene_Abar(:,i) = ((calib.price_N(i)
 + (calib.price_ene_E(:,i) .* prod.co2Factors(:,i))).^(1-
prod.ene_alpha(i))) ./ 	...
 (calib.price_ene_e(:,i) .*
 (prod.ene_alpha(i)^prod.ene_alpha(i))*((1-prod.ene_alpha(i))^(1-
 prod.ene_alpha(i))));
 	% No alpha version
 	%prod.ene_Abar(:,i) = (calib.price_N(i) +
 (calib.price_ene_E(:,i) ./ prod.ene_abar(:,i))) ./ calib.price_ene_e(:,i);
17
108

% prod.fin_A(:,i) = calib.quant_fin_c(:,i) ./ ...
 	% (calib.quant_fin_c(:,i).^(1-
prod.fin_nuOverAll(:,i)) .* prod.fin_alpha(i).^prod.fin_alpha(i) .* ...
 	% ((1 - prod.fin_alpha(i) - prod.fin_nu(:,i))/
calib.price_N(i)).^(1-prod.fin_alpha(i) - prod.fin_nu(i)) .* ...
 	% calib.quant_int_d_val(:,i).^prod.fin_nu(:,i)); %
 TFP final good production (column vector)
 	% % A version 2
 	% prod.fin_A2(:,i) = prod.fin_alpha(i).^(-
prod.fin_alpha(i)) .* ...
 	% ((calib.price_N(i) ./ (1-
prod.fin_alpha(i) - prod.fin_nuOverAll(i))).^(1-prod.fin_alpha(i) -
 prod.fin_nuOverAll(i))) .* ...
 	% (calib.quant_fin_c(:,i) ./
 calib.quant_int_d_val(:,i)).^(prod.fin_nuOverAll(i));
 	%
 	%
 	% % Trying another version of A
 	% prod.fin_A(:,i) = prod.fin_alpha(i).^(-
prod.fin_alpha(i)) .* ...
 	% prod.fin_nuOverAll(:,i).^(-
prod.fin_nuOverAll(:,i)) .* ...
 	% (calib.price_N(i) ./ (1 - prod.fin_alpha(i) -
 prod.fin_nuOverAll(:,i))).^(1 - prod.fin_alpha(i) - prod.fin_nuOverAll(:,i));
 	% Trying another version of A (sector wise nu)
 prod.fin_A(:,i) = prod.fin_alpha(i).^(-
prod.fin_alpha(i)) .* 	...
 prod.fin_nu(:,i).^(-prod.fin_nu(:,i)) .* 	...
 (calib.price_N(i) ./ (1 - prod.fin_alpha(i) -
 prod.fin_nu(:,i))).^(1 - prod.fin_alpha(i) - prod.fin_nu(:,i));
 	% prod.fin_A2(:,i) = prod.fin_alpha(i).^(-
prod.fin_alpha(i)) .* ...
 	% ((calib.price_N(i) ./ (1-prod.fin_alpha(i) -
 prod.fin_nu(:,i))).^(1-prod.fin_alpha(i) - prod.fin_nu(:,i))) .* ...
 	% (calib.quant_fin_c(:,i) ./
 calib.quant_int_d_val(:,i)).^(prod.fin_nu(:,i));
 	end	 % for loop
 	end
 save 	Prod_ACEn.mat
 	else
 load 	Prod_ACEn.mat
 	end	 % direct read-in option
end	 % end of calibrate loop
Production calculation
if	 calibrate_calculate == 2 | 3
18
109

%if production.calculation == 1
 	for	 i = 1:numOfRegions
Need input vectors to carry out calculation:
 controltest.K_fin(:,i) = calib.quant_fin_K_sectors(:,i); 	%(3,1) per
 region
 controltest.N_fin(:,i) = calib.quant_fin_N_sectors(:,i); 	%(3,1) per
 region
 controltest.K_ene(:,i) = ones(5,1);
 controltest.N_ene(:,i) = ones(5,1);
 controltest.K_ene(:,i) = calib.quant_fin_K(i) .*
 [.0014;.0058;.0012;.0009;.0029];
 controltest.N_ene(:,i) = calib.quant_fin_N(i) .*
 [.0068;.0271;.0055;.0041;.0138];
 controltest.E(:,i) = calib.quant_ene_E(:,i); 	%(5,1) per region
 controltest.sums(:,i) = [calib.quant_ene_e(1:4,i);
 calib.quant_int_0(i)]; 	% Rows: Total energy prod. per fuel (5th is
 electricity).
 controltest.el_e_shares(:,i) = calib.quant_ene_0(:,i) ./
 calib.quant_ene_e(:,i);
 controltest.int_e_shares_tr(:,i) =calib.quant_int_e_tr(:,i) ./
 controltest.sums(:,i);
 controltest.int_e_shares_in(:,i) =calib.quant_int_e_in(:,i) ./
 controltest.sums(:,i);
 controltest.int_e_shares_ot(:,i) =calib.quant_int_e_ot(:,i) ./
 controltest.sums(:,i);
 	%controltest.int_e_shares = ones(3,5)/5; % Each sector's coefficient
 on fuel inputs. Col5 is electricity
 controltest.elTest(i) = controltest.int_e_shares_tr(5,i) +
 controltest.int_e_shares_in(5,i) + 	...
 controltest.int_e_shares_ot(5,i);
 	%controltest.elTest(i) = calib.quant_int_0(i);
 	% Each set of sectors has a matrix containing varieties in columns and
 production coefficients on different inputs in a row
Refined primary energy production:
 	% Debugging:
 	% prod.ene_Abar(:,i) = 1;
 	% column vector of dimenions 'fuels'.
 f_ene(:,i) = min(controltest.N_ene(:,i) , prod.ene_abar(:,i) .*
 controltest.E(:,i)); 	% min of elements in each row, returns (5x1) vector
 	%f_ene(5,i) = controltest.N_ene(5,i);
 e_ene(:,i) = prod.ene_Abar(:,i) .*
 controltest.K_ene(:,i).^prod.ene_alpha(i) .* f_ene(:,i).^(1 -
 prod.ene_alpha(i));
19
110

Electricity sector
 	% Debugging:
 	% prod.int_Atilde_0(i) = 1;
 e_ene_0(i) = prod.int_Atilde_0(i) .* sum(prod.int_atilde_0(:,i) .*
 (controltest.el_e_shares(:,i) .* e_ene(:,i)).^prod.int_s_0).^(1./
prod.int_s_0); 	% Scalar
Intermediate sector (sector specific energy
composite):
 	% Debugging:
 	% prod.int_Atilde_tr(i) = 1;
 	% prod.int_Atilde_in(i) = 1;
 	% prod.int_Atilde_ot(i) = 1;
 	% ---> Setting to one fixes Rich > World problem for d_int.
 d_int_tr(i) = prod.int_Atilde_tr(i).* (prod.int_atilde_tr(5,i) .*
 (controltest.int_e_shares_tr(5,i) .* e_ene_0(i)).^prod.int_s(1) + 	...
 sum(prod.int_atilde_tr(1:4,i) .*
 (controltest.int_e_shares_tr(1:4,i) .* e_ene(1:4,i)).^prod.int_s(1))).^(1./
prod.int_s(1));
 d_int_in(i) = prod.int_Atilde_in(i).* (prod.int_atilde_in(5,i) .*
 (controltest.int_e_shares_in(5,i) .* e_ene_0(i)).^prod.int_s(2) + 	...
 sum(prod.int_atilde_in(1:4,i) .*
 (controltest.int_e_shares_in(1:4,i) .* e_ene(1:4,i)).^prod.int_s(2))).^(1./
prod.int_s(2));
 d_int_ot(i) = prod.int_Atilde_ot(i) .* (prod.int_atilde_ot(5,i) .*
 (controltest.int_e_shares_ot(5,i) .* e_ene_0(i)).^prod.int_s(3) + 	...
 sum(prod.int_atilde_ot(1:4,i) .*
 (controltest.int_e_shares_ot(1:4,i) .* e_ene(1:4,i)).^prod.int_s(3))).^(1./
prod.int_s(3));
 d_int(:,i) = [d_int_tr(i); d_int_in(i); d_int_ot(i)];
 	% XXX double check that equation above is raising right parts to right
 power...
Final good production sector:
column vector 'goods' (e.g. transport, industry, other)
 	% Debugging:
 	% prod.fin_A(:,i) = 1;
 	%Sectorwise nu
20
111

c_fin(:,i) = (prod.fin_A(:,i) .*
 controltest.K_fin(:,i).^prod.fin_alpha(i) .* 	...
 controltest.N_fin(:,i).^(1-prod.fin_alpha(i)-prod.fin_nu(:,i)) .*
 d_int(:,i).^prod.fin_nu(:,i));
Final good aggregation sector:
scalar
 	% Debugging:
 	% prod.agg_A(:,i) = 1;
 c_agg(i) = prod.agg_A(i).*(sum(prod.agg_a(:,i) .*
 c_fin(:,i).^prod.agg_s))^(prod.agg_sbar/prod.agg_s);
 emi(:,i) = e_ene(:,i) ./ prod.ene_abar(:,i);
 compEn(:,i) = e_ene(:,i) ./ (calib.quant_ene_e(:,i));
 compEl(:,i) = e_ene_0(:,i) ./ (calib.quant_int_0(i));
 compInt(:,i) = d_int(:,i) ./ calib.quant_int_d_val(:,i);
 compSec = c_fin(:,i) ./ [finalGoods{:,	"Fun_transVA"	}
 finalGoods{:,	"Fun_indVA"	} finalGoods{:,	"Fun_othVA"	}]';
 	end	 % for loop
 c_agg' ./ finalGoods{:,	"Fun_cgdpo"	}
 disp([	'Aggregate consumption is '	 num2str(c_agg)])
 Y_gross_Andreas=c_agg(i)
end	 	% of calculation loop
countrylist = PWT(:, [	"country_PWT"	 "thesisCode"	]);
%writetable(finalGoodsRegional, 'finalGoods.xlsx');
% writetable(energyRegionalStacked, 'energyRegional.xlsx');
writetable(countrylist, 	'countries.xlsx'	);
Diffusion model
Diffusion init
prod.ene_Abar_dyn(:,i)= prod.ene_Abar(:,i); 	% initializing vector of TFP in
 energy
g_ren_H= 0.01; 	% high income renewable growth
g_renDecline = 0.02;
theta_diffu = 0; 	% diffusion parameter (in 1)
lambda_diffu = 0.00; 	% inherent growth in low income (in >0)
welfare(i) = 0;
L
% Growth in renewable energy TFP
Ahat = (prod.ene_Abar_dyn(5,3) - prod.ene_Abar_dyn(5,2))./
 prod.ene_Abar_dyn(5,2)
21
112

breakcheck = Ahat > 0.01
if	 diffusion == 1
 prod.ene_Abar_dyn(5,1) =
 prod.ene_Abar_dyn(5,1).*exp(g_ren_H(t).*timestep);
 	if	 breakcheck
 g_ren_L = theta_diffu .* Ahat + lambda_diffu;
 	else
 g_ren_L = 0;
 	end
 prod.ene_Abar_dyn(5,2) = prod.ene_Abar_dyn(5,2)*exp(g_ren_L*timestep);
 prod.ene_Abar_dyn(5,3) =
 prod.ene_Abar_dyn(5,3).*exp(g_ren_H(t).*timestep);
end
g_ren_LTime(t) = g_ren_L;
AhatOverTime(t) = Ahat;
E(i,t)=sum(Emi);
E_field(i,:,t)=Emi';
C_fin_field(i,:,t) = c_fin;
C_fin_tr(i,t)=c_fin(1);
C_fin_in(i,t)=c_fin(2);
C_fin_ot(i,t)=c_fin(3);
d_fin_tr(i,t)=d_int(1);
d_fin_in(i,t)=d_int(2);
d_fin_ot(i,t)=d_int(3);
% Total energy use in regions
ene_field(i,:,t) = ene';
e_ene_field(i,:,t) = e_ene;
e_ene_tot(i,t)=sum(e_ene_field(i,:,t),2);
ene_tot(i,t)=sum(ene_field(i,:,t),2);
e_ene_c(i,t) = e_ene(1); 	% coal
e_ene_o(i,t) = e_ene(2); 	% oil
e_ene_g(i,t) = e_ene(3); 	% gas
e_ene_b(i,t) = e_ene(4); 	% bio
e_ene_r(i,t) = e_ene(5); 	% renew
e_ene_e(i,t) = e_el(1); 	% el
% How to find energy use within a sector?
ene_tr(:,i,t) = cont(:,1) .* ene;
ene_in(:,i,t) = cont(:,2) .* ene;
ene_ot(:,i,t) = cont(:,3) .* ene;
ene_el(:,i,t) = [cont(1:4,4) .* ene(1:4); e_ene(5)];
% debug
%ene_sum = ene_tr + ene_in +ene_ot + ene_el;
utility(i,t) = log(Y_net(i,t)*x(i)); 	% utils in that period
if	 t == 1
 welfare(i,t) = utility(i,t);
else
 welfare(i,t) = welfare(i,t-1)+beta(i)^t .* utility(i,t); 	% the final t
 gives the total discounted
end
Published with MATLAB® R2022a
22
113

Husholdningers betalingsvillighet for å unngå strømbrudd

Vista Analyse	 | Rapport 2022/23	 	3

Vista Analyse AS
Meltzers gate 4
0257 Oslo

post@vista	-analyse.no
vista	-analyse.no