Working Paper Series18/2019

Louison Cahen-FourotEmanuele Campiglio

Elena DawkinsAntoine Godin

Eric Kemp-Benedict

Capital stranding cascades:The impact of decarbonisation on

productive asset utilisation

Capital stranding cascades: The impact of decarbonisation on

productive asset utilisation∗

Louison Cahen-Fourot1, Emanuele Campiglio†1,2, Elena Dawkins3, Antoine Godin4,5

and Eric Kemp-Benedict3

1Institute for Ecological Economics, Vienna University of Economics and Business2Grantham Research Institute, London School of Economics and Political Science

3Stockholm Environment Institute4Agence Francaise de Developpement

5Institut fur Makrookonomie und Konjunkturforschung

This version: February 2019

Abstract

This article develops a novel methodological framework to investigate the exposure of eco-

nomic systems to the risk of physical capital stranding. Combining Input-Output (IO) and

network theory, we define measures to identify both the sectors likely to trigger relevant capital

stranding cascades and those most exposed to capital stranding risk. We show how, in a sample

of ten European countries, mining is among the sectors with the highest external asset strand-

ing multipliers. The sectors most affected by capital stranding triggered by decarbonisation

include electricity and gas; coke and refined petroleum products; basic metals; and transporta-

tion. From these sectors, stranding would frequently cascade down to chemicals; metal products;

motor vehicles water and waste services; wholesale and retail trade; and public administration.

Finally, we provide an estimate for the lower-bound amount of assets at risk of transition-related

stranding, which is in the range of 0.6-8.2% of the overall productive capital stock for our sample

of countries, mainly concentrated in the electricity and gas sector, manufacturing, and mining.

These results confirm the systemic relevance of transition-related risks on European societies.

Keywords: stranded assets, low-carbon transition, physical capital stocks, fossil fuels, input-output

analysis, networks

JEL codes: C67; E22; L71; O10; Q32

∗The research leading to these results has received funding from the Swedish Foundation for Strategic Environmen-

tal Research (Mistra) and the Oesterreichische Nationalbank (OeNB). The authors are grateful to Hanspeter Wieland

for useful comments. The usual disclaimer applies.†Corresponding author: Emanuele.campiglio@wu.ac.at.

1

1 Introduction

A low-carbon transition is now almost unanimously considered as a necessary and likely scenario

(IEA, 2017; IPCC, 2018; UNFCCC, 2016). However, less certainty exists on the socioeconomic

and financial implications of such a large-scale structural change. The decarbonisation process, if

not properly managed, might result in reduced economic prosperity, unemployment and financial

instability. In particular, a significant body of work has been developing to study the extent to

which the low-carbon transition might negatively affect economic and financial wealth by creating

“stranded assets” (Caldecott, 2018). Most of the research conducted so far on the topic has focused

on the stranding of fossil reserves (McGlade and Ekins, 2015; Mercure et al., 2018) or on the knock-

on financial effects of a drop in the market valuation of fossil fuel companies (Battiston et al., 2017;

Carbon Tracker, 2013). The exposure of the financial system to transition risks has also raised the

concerns of international financial regulators, who wish to avoid any risk that might create systemic

disturbance to the global system, still recovering from the global financial crisis (Carney, 2015;

Campiglio et al., 2018).

However, fossil reserves are only part of the picture. In a rapid low-carbon transition, a large

amount of built infrastructure, industrial plants and machinery would have to be abandoned or

entirely reconverted. This applies not only to the stock of extracting infrastructure, pipelines and

other forms of capital linked to fossil fuels, but also to a large amount of industrial plants whose

output requires fossil fuels as material inputs or for process heat (metals, coke, chemicals, steel,

etc). The reduction in production and consequent asset stranding (in the form of idle productive

capacity) would in turn cascade to physical stocks supporting the rest of the economic activity (e.g.

warehousing, transport infrastructure, etc.) following chains of intermediate exchange. While some

studies have explored the stranding of physical productive assets (Pfeiffer et al., 2018; IRENA, 2017),

the topic has not been thoroughly investigated, and never from a systemic perspective.

This article contributes to filling this gap in the literature by providing a novel analysis of the

process through which a low-carbon transition would affect the utilisation of productive capital

stocks at the sectoral level. More specifically, we employ data available for ten European economies

to achieve three main objectives.

First, we use Input-Output (IO) concepts to derive national matrices of “asset stranding mul-

tipliers”, including the entire range of productive sectors. These multipliers capture the monetary

value of physical capital stocks that would become idle (i.e. stranded) in a sector due to a unitary

drop in primary inputs1 utilised by another (or the same) sector, considering both direct and indirect

effects. To offer an example, these matrices are able to provide the monetary value of the capital

stock becoming unutilised in the transportation sector due to a drop in manufacturing, both directly

and through its intermediate effects on, say, wholesale trade. We identify the sectors most likely to

have large stranding effects and the sectors most exposed to the risk of capital asset stranding for

the countries in our sample. While the analysis developed here maintains a systemic perspective,

we highlight how mining - where fossil fuel extraction is included - is among the sectors with the

highest potential to trigger capital asset stranding in the rest of the economy.

Second, we focus our analysis on the mining sector in the attempt to identify the most relevant

channels of asset stranding propagation caused by a low-carbon transition. In order to do so, we treat

the matrix of asset stranding multipliers as an adjacency matrix for a directed weighted network and

show that, while countries exhibit different cascade processes depending on the peculiarities of their

1We define “primary inputs” as the main factors used in production (labour, capital, land, imports, and others).

IO tables report their factor costs (e.g. compensation of employees, consumption of fixed capital or net operating

surplus). See Miller and Blair (2009), Eurostat (2008a) and OECD (2001).

2

industrial structure, certain regular patterns emerge. The sectors most at direct risk of stranding

include electricity and gas; coke and refined petroleum products; basic metals; and transportation

and storage. The stranding in electricity and gas often cascades in a significant manner down to

public administration and water-related services. The coke and refined petroleum products sector

affects the capital stock of the chemical and land transport sectors the most, with further links to the

rest of transportation sectors (water and air transport, warehousing and postal services). Finally, the

stranding in the basic metals sector has significant impacts on fabricated metal products and motor

vehicles, which in turn cascade down to the trade and repair of motor vehicles. In contrast, the

service sectors appear to be the least at risk in terms of transition-related physical asset stranding.

Finally, we provide a ballpark estimate for the overall amount of productive capital stock at

risk of becoming stranded due to a complete transition away from fossil fuels. Results differ widely

across countries, ranging from less than 1% of capital stock in Belgium and Sweden to more than

8% in Slovakia. In absolute terms, the largest countries in terms of capital stock are also the most

affected ones. In the majority of countries in our sample at least half of the stranding takes place

in the electricity and gas sector. The main exceptions are the United Kingdom, where a large part

of the stranding happens in the mining industry, and Belgium, where manufacturing is the sector

proportionally more at risk. In general, these results confirm that capital stranding driven by a

transition away from fossil fuels could have significant effects on European economic systems.

To the best of our knowledge, our paper is the first to offer a systemic empirical analysis of

the productive capital stocks at risk of stranding due to a low-carbon transition. Previous work on

“committed cumulative emissions” has already shown that the greenhouse gas emissions embodied

in currently existing (and planned) capital infrastructure are incompatible with a 2°C-consistent

decarbonisation process (Davis et al., 2010; Davis and Socolow, 2014; Smith et al., 2019), thus

suggesting the likelihood of premature decommissioning of productive stocks in the event of a low-

carbon transition. Building on these results, Pfeiffer et al. (2018) calculate the amount of capital

stock that would become unutilised in the power generation sector, finding it to be equal to 20% of

total sectoral capacity even in the unlikely event of all currently scheduled projects being cancelled.

IRENA (2017) also provides estimates that suggest a significant capital stranding in several sectors,

including upstream energy, power generation, buildings and industry. However, none of the cited

studies considers the entire range of productive sectors, where capital stranding might take place

through cascade effects, nor disaggregates among industrial sub-sectors.

We complement this literature by providing an alternative methodology to study transition-

related stranding risks based on concepts and methods borrowed from the IO and network analysis

literature. Following Blochl et al. (2011), Acemoglu et al. (2012), Joya and Rougier (2019), and

others, we treat input-output linkages as the edges of a directed weighted network representing the

complex web of economic interconnections among productive sectors. Unlike most of the research in

this field, however, we do not study the effect of demand-driven sectoral shocks on aggregate output

volatility, nor are we interested in determining the “centrality” of sectors. Given the focus of our

analysis, the origin of the shock is by definition located into the mining sector and, we claim, is more

suitably interpretable as a supply, rather than demand, disruption. For this reason, we adopt the

supply-driven IO model (Ghosh, 1958), complementing it with sectoral data on capital stocks.

The Ghosh model can be used to identify the relevance of productive sectors in supporting

downstream economic activity through the calculation of sectoral “forward linkages” (see for instance

Aldasoro and Angeloni, 2015; Antras et al., 2012; Cahen-Fourot et al., 2019) or, in the case of

this paper, their relevance in keeping downstream capital stocks in operation. In contrast, its

plausibility as a theoretical approach to investigate the inter-industry impact of supply changes -

3

especially large ones - has been questioned due to some of its limiting assumptions, such as perfect

elasticity of demand in reacting to changes in supply and perfect substitutability among input factors

(Oosterhaven, 1988; Galbusera and Giannopoulos, 2018). Some of these criticisms can be solved

when treating the framework as a price model (Dietzenbacher, 1997), and are further mitigated

by the features of our specific research questions, which, for instance, do not include the issue of

allocating an excess output supply. Nonetheless, the figures provided in section 4.2 do assume perfect

input substitutability and abstract from dynamic considerations, and should thus be interpreted as

a ballpark estimate of the lower-bound of the capital stock currently at risk of stranding due to the

decarbonisation process.

We also aim to contribute to two additional streams of work. First, our results offer new insights

to the burgeoning literature on the macro-financial implications of a low-carbon transition (Battiston

et al., 2017; Stolbova et al., 2018; Vermeulen et al., 2018). These studies usually investigate how the

value loss of financial assets issued by firms in fossil or fossil-related sectors would affect the portfolio

of financial investors, with potential knock-on effects within the financial network. A more sophisti-

cated assessment of sectoral physical capital stranding would contribute to improving the accuracy

of the real-financial cascade effects. Second, the stress put on the relevance of potential capital stock

underutilisation during a low-carbon transition could contribute to the climate economic modelling

literature, which traditionally focuses on flow variables when evaluating the economic implications

of mitigation scenarios (“GDP losses” or, more recently, “investment needs”), rather than stock vari-

ables (Tavoni et al., 2015; McCollum et al., 2018; Tol, 2009)2. Productive capital assets are often

present in the analysis as the main factor of production (“K”) but, despite a few exceptions (see for

instance Rozenberg et al., 2018; Baldwin et al., 2018), the possibility of their idleness is not usually

contemplated. This methodological approach is justified by the century-long perspective taken by

most of these modelling exercises, a horizon within which shorter-term phenomena (e.g. low capacity

utilisation or a drop in firms’ market valuation) can be netted out by longer-term trends driven by

capital accumulation and technological innovation. However, addressing the complexity of safely

managing the decarbonisation process cannot abstract from these short-term phenomena affecting

both flow and stock variables.

The remainder of the article is as follows. Section 2 introduces the method to compute the

matrices of sectoral asset stranding multipliers. Section 3 presents the results of the analysis for ten

European countries, discussing the sectors most likely to create large stranding and the ones most

exposed to stranding risk. Section 4 focuses on a low-carbon transition, identifies relevant channels

of asset stranding and provides an estimate of the total amount of capital stocks at risk of stranding.

Finally, Section 5 concludes.

2 The asset stranding multiplier matrix

The starting point of the proposed methodology is the inter-industry matrix Z, a square matrix

recording intermediate consumption of industries and thereby showing all the transactions of goods

and services among industrial sectors measured in monetary units (Miller and Blair, 2009).

In an IO table (IOT), the Z matrix is complemented by a set of column vectors representing

final consumption i.e. demand (f) and by a set of row vectors representing value added items (v)

such as compensation of employees, consumption of fixed capital and gross operating surplus. All

sectors appear twice in Z: as producers of goods and services (rows); and as users of intermediate

2This is less the case for studies focusing on climate change impacts, where the effects on stock variables have been

often incorporated. See Piontek et al. (2018) for a recent example.

4

Intermediate uses Final uses (f)Total use

(TU)Inter-Industry matrix (Z) Sector A Sector B Cons. Inv. Exp.

Production

Sector A

Products of A

used as inputs

by A

Products of A

used as inputs

by B

Final use of products by A

Total use of

products of

A

Sector B

Products of B

used as inputs

by A

Products of B

used as inputs

by B

Final use of products by B

Total use of

products of

B

Total Total intermediate inputs Total final uses Total uses

Value

added (v)

Comp. of

employees

Total value addedCons. of

fixed capital

Operating

surplus

Output Total domestic output

Imports Total imports

Total supply (TS) Total supply

Table 1: A stylised Input-Output (IO) table

inputs (columns). The core principle in IOTs are monetary industry balances, where total supply

xT = iTZ + v must be equal to total use x = Zi + f per industry, where i is a column vector of 1’s

of the same dimension of Z3. In other words, the sum of all flows over a row (total industry output

broken down by type of use, i.e. intermediate use and final consumption) must equal the sum over

the corresponding column (total industry input broken down by “source”, i.e. other industries and

value added items). Two main types of IOTs are usually included in a single-region IO dataset: the

“domestic” IOT reports only goods and services produced domestically; the“total” IOT reports also

imported goods and services, which are again used either as intermediate inputs or as final demand.

Table 1 shows a stylized version of a “total” IO table.

The most common use of IOTs in economic analysis is aimed at estimating the direct and indirect

effects of changes in final demand using the Leontief model, where the matrix is often referred to as

the Leontief “inverse” (Leontief, 1951). However, while both demand and supply patterns will be

crucial in defining the process of decarbonisation, we believe that the demand-driven Leontief model

would not be the most appropriate methodological approach to study the low-carbon transition and

its implications for asset stranding. The shifts in consumption patterns due to climate-related drivers

(e.g. changes in relative taxation, rapid technological innovation, more widespread environmental

awareness) will most likely materialise in shifts of final demand from high- to low-carbon versions

of the same category of goods, rather than a shift between distinct categories. In other words,

consumers will mostly shift their preferences towards energy-efficient durable goods and electricity

produced via low-carbon energy sources, rather than shifting preferences from one category of goods

to another (e.g. from manufactured goods to food products). This is also due to the relatively low

proportion of final demand for fossil fuels, which are instead mainly used by other productive sectors

as intermediate goods. Hence, the main effect of low-carbon transition drivers will concern the way

3Pre-multiplying a matrix by iT returns its column sum; post-multiplying a matrix by i returns its row sum.

5

in which goods and services are produced. In this context, the demand-oriented Leontief model is

not suitable for studying rapid processes of substitution away from fossil-based input factors. We

should instead focus more on the supply side and the production process.

A better methodological approach for our purposes is provided by the Ghosh (1958) supply-

driven IO system (Augustinovics, 1970; Beyers, 1976). Instead of calculating a matrix A of technical

input coefficients as in the Leontief model, the Ghosh model defines a matrix B = x−1Z of output

allocation coefficients, whose elements represent the allocation of the output of a sector to all other

sectors. In other words, each element bij quantifies the share of industry i’s output that is used by

industry j. The Ghosh matrix G is then defined as:

G = (I−B)−1. (1)

For convenience, we transpose G to be able to read the effects of changes in sectoral primary

inputs over the columns (similarly to the Leontief system) of GT , where T denotes the matrix

transposition. Each element gi,j of GT describes the change in output x in sector i that would

result from a unitary change of primary inputs flowing into sector j. In other words, an increase

(decrease) of one monetary unit of primary inputs contributing to production in sector i will increase

(decrease) the output of sector j by an amount equal to gi,j , where gi,j includes both direct and

indirect effects. “Primary inputs” refers to any item appearing on the rows below the inter-industry

matrix. Traditionally, this has been meant to represent compensation of employees (and thus labour

input) but, more generally, it can be used to represent any form of societal effort put in producing

the output of a specific sector, as represented by factor payments.

We then combine the Ghosh matrix with sectoral data of physical capital stocks k. We define

κi = ki/xdi as the capital intensity of sector i, where xd is the domestic output of the sector. By

multiplying the diagonalised form of the vector of capital intensities by the Ghosh matrix, we find

the matrix S of asset stranding multipliers:

S = κGT . (2)

Each element sij of matrix S represents the change in the utilisation of capital in sector i triggered

by a unitary change of primary inputs used by sector j. For our purposes, the elements of S define

the amount of capital stock of a sector i that could become stranded because of a unitary decrease

in the primary inputs used in the production of goods and services of another sector j (e.g. fossil

fuels).

The column sum of matrix S gives a measure of the total amount of stranded physical assets

resulting from a unitary reduction of primary inputs in a sector j. We define this as the total asset

stranding multiplier of a sector:

sTOTj = iTS (3)

where n is the dimension of matrix S. We expect the values of sTOTi to be strongly affected by

sectoral capital intensities, and hence by the amount of internal asset stranding. We thus also

define a measure for the external asset stranding multiplier, to calculate the effect of a unitary

reduction of primary inputs of a sector on the capacity utilisation of capital stocks in all other

sectors:

sEXTj = sTOT

j − sdiagj , (4)

where sdiagj represents the j-th element of the diagonal of S.

Finally, we can interpret the sum of the rows of S as the exposure of a sector to stranding risk

(i.e. the loss in capital utilisation resulting from a unitary loss in primary inputs used in all the rest

6

of the sectors):

sEXPi = Si. (5)

3 Physical capital stranding networks

We apply the methodology described in section 2 to a selection of European countries4. The main

source of data for both IO tables and capital stock data is the Eurostat statistical database. More

specifically, we use symmetric input-output tables at basic prices5 and the cross-classification of

fixed assets by industry and by asset6. In the case of Sweden, we complemented Eurostat data with

retrieving C20-21, H52-53 and M71-72 sectors fixed assets data from Statistics Sweden. Sectors

are classified using the NACE classification system (Eurostat, 2008b). Table 2 lists NACE level 1

categories, while Table A1 in the Appendix offers a more detailed disaggregation. We consider both

physical productive infrastructure (N112N - Other buildings and structure) and machinery (N11MN

- Machinery and equipment and weapons systems)7. This will be referred to as productive capital

hereafter. The geographical representation of our sample is mainly determined by availability of

data, especially for capital stock data. After harmonizing sectoral aggregation between IO tables

and capital stock data, we are able to offer results for ten European countries using 2010 data:

Austria, Belgium, Czech Republic, Germany, Greece, France, Italy, Sweden, Slovakia and the United

Kingdom. In 2010, these countries represented approximately 74% of the European Union Gross

Domestic Product. The disaggregation of data among NACE sectors differs across countries, with

some (e.g. Germany) reporting only NACE level 1 values and others (e.g. Austria) offering a more

disaggregated perspective. For each of them, we use the most detailed disaggregation level available.

Table 3 reports the results for: i) total stranding multipliers; ii) external stranding multipliers;

iii) exposure to stranding risk. We focus on the top 5 sectors for each country. Given the capital

stock composition of the economy and leaving all else equal, the first two sets of multipliers show

the sectors that are likely to create the largest amount of stranded assets in the economic system

following a unitary drop in their primary inputs. The third set of results report instead the sectors

that are likely to be most affected by capital asset stranding from a unitary drop distributed equally

across all sectors8.

Regarding total multipliers, sectors in the E category (Water supply; sewerage; waste manage-

ment and remediation activities) are by far the most prevalent, appearing among the top 5 sectors

for nine out of ten countries, and as the very top sector for six of them. Where further disaggregation

is possible, we observe this to be particularly true for sector E36 (Natural water; water treatment

and supply services). Studying the S matrix, one can notice that sectors often significantly affected

by the stranding originating in E sectors include L (Real estate activities) and O (Public admin-

istration and defence; compulsory social security). In certain countries (e.g. Sweden and Austria)

this is likely to be mainly driven by the high proportion of hydroelectric power in the energy mix9.

4The R code files used to obtain the results shown in this article, as well as the full set of results, are available as

an online appendix at: github.com/capital-stranding-cascades/online material.5naio 10 cp1700, total economy, product by product in million e.6nama 10 nfa st, stocks at current replacement costs in million e.7We assume that most dwellings will continue being inhabited even in the event of an abrupt low-carbon transi-

tion, and thus exclude the dwelling category (N111N). We also exclude cultivated biological resources (N115N) and

intellectual property products (N117N).8A large value in this column may represent small contributions from a large number of sectors. In that case, the

risk is mitigated to the extent that changes in primary inputs to those sectors are uncorrelated.9The share of hydropower in the electricity mix was 56.5% for Austria and 44.6% for Sweden in 2010 according to

the World Development Indicators database of the World Bank.

7

Sector code Sector description

A Agriculture, forestry and fishing

B Mining and quarrying

C Manufacturing

D Electricity, gas, steam and air conditioning

E Water supply; sewerage; waste management and remediation activities

F Constructions and construction works

G Wholesale and retail trade; repair of motor vehicles and motorcycles

H Transportation and storage

I Accommodation and food service activities

J Information and communication

K Financial and insurance activities

L Real estate activities

M Professional, scientific and technical activities

N Administrative and support service activities

O Public administration and defence; compulsory social security

P Education

Q Human health and social work activities

R Arts, entertainment and recreation

S Other services activities

Table 2: NACE level 1 sectors

In addition to E sectors, activities included in categories A (Agriculture, forestry and fishing), D

(Electricity, gas, steam and air conditioning), H (Transportation and storage) and O (Public ad-

ministration and defence; compulsory social security) all appear among the top 5 sectors in terms

of total asset stranding multipliers for six of the examined countries. For those countries in which

further disaggregation of category H is possible, we observe a particular relevance of sectors H50

(Water transport) and H52 (Warehousing and support activities for transportation). These results

appear to be strongly driven by the high capital intensity of the sectors, thus highlighting significant

potential internal asset stranding. Significant exception to these patterns are Greece and France,

where sectors N77 (Rental and leasing activities) and B (Mining and quarrying) represent the top

sectors, respectively.

External asset stranding multipliers, which abstract from the internal stranding of a sector and

thus offer a more accurate representation of the stranding effect of a sector on the rest of the

economy, exhibit a significantly different pattern. The relevance of capital-intensive activities is

drastically reduced, although water and waste management activities (E) still exhibit significantly

high stranding effects in Austria and Sweden, and transportation and storage sectors (H) appear

at the top of the ranking of both Czechia and Slovakia. Sectors B (Mining and quarrying), C

(Manufacturing), M (Professional, scientific and technical activities) and N (Administrative and

support service activities) become the most recurrent sectors, appearing among the top 5 sectors for

six, five, seven and nine countries, respectively. For those countries in which further disaggregation

is possible, we notice a particular prevalence of sectors C33 (Repair and installation services of

machinery and equipment), M74 75 (Other professional, scientific and technical activities) and N80-

82 (Security and investigation activities; buildings and landscape; office support). All these sectors

appear high in the ranking of external asset stranding multipliers because they provide significant

amount of inputs to capital-intensive sectors. For instance, both sectors C33 and N80-82 provide

substantial intermediate goods to transportation (H), real estate (L) and public administration (O).

Table 4 offers a closer look at the mining sector, reporting the top 5 sectoral values for the external

stranding multipliers originating in B (excluding B itself, to abstract from internal stranding). Sector

8

Austria

Belgiu

mCzechia

Germ

any

Greece

France

Italy

Sweden

Slovakia

UK

Tota

lst

ran

din

gm

ult

iplier

s

E36

(11.3

2)

H(3

.13)

E36

(13.6

6)

E(6

.65)

N77

(5.8

9)

B(4

.20)

D(5

.52)

E36

(12.6

5)

E36

(29.8

3)

E36

(8.6

1)

A03

(9.9

2)

E(2

.52)

O(7

.88)

R(3

.7)

A02

(4.6

2)

O(3

.79)

A(5

.07)

D(5

.04)

H52

(13.6

6)

H52

(4.7

9)

A01

(8.4

8)

D(2

.51)

H52

(6.7

4)

A(3

.58)

E36

(4.0

8)

E(2

.26)

O(4

.07)

A01

(4.8

2)

H50

(9.8

1)

E37-3

9(3

.3)

E37-3

9(8

.44)

N(1

.65)

P(6

.01)

O(3

.21)

O(3

.59)

D(2

.07)

B(3

.99)

O(4

.81)

B(8

.96)

A02

(3.2

1)

H52

(8.0

6)

C16-1

8(1

.65)

H50

(5.8

6)

N(2

.84)

D(3

.46)

R(2

.01)

H(3

.38)

J61

(3.3

3)

D(7

.54)

H49

(3.1

2)

Exte

rnal

stra

nd

ing

mu

ltip

lier

s

E36

(3.6

4)

M69-7

1(0

.91)

H50

(3.9

3)

N(1

.12)

M74

75

(1.5

7)

B(1

.02)

B(1

.79)

S95

(1.6

8)

H50

(6.3

)C

23

(1.4

5)

E37-3

9(2

.44)

N(0

.81)

H53

(3.0

8)

B(0

.89)

N77

(1.5

)N

(0.8

3)

C19

(1.2

9)

E36

(1.5

8)

B(4

.03)

N77

(1.4

3)

N78

(2.0

5)

M73-7

5(0

.71)

N80-8

2(3

.02)

M(0

.73)

C18

(1.5

)M

73-7

5(0

.77)

M69-7

1(1

.28)

C33

(1.4

5)

C33

(3.5

1)

N79

(1.4

2)

N80-8

2(1

.89)

J62

63

(0.5

9)

M74

75

(2.8

5)

J(0

.72)

N80-8

2(1

.48)

E(0

.74)

N(1

.19)

N80-8

2(1

.25)

M74

75

(3.4

)C

33

(1.3

7)

B(1

.86)

K(0

.56)

B(2

.12)

D(0

.68)

C33

(1.4

7)

K(0

.67)

D(1

.14)

E37-3

9(1

.23)

M71

(2.5

1)

C16

(1.3

4)

Exp

osu

reto

stra

nd

ing

risk

L(9

.32)

H(2

.18)

O(1

0.5

2)

C(1

.56)

O(7

.63)

O(2

.62)

H(2

.97)

O(5

.71)

D(1

1.6

9)

F(6

.36)

O(3

.19)

G(1

.28)

H52

(7.5

8)

O(1

.41)

H50

(4.8

)L

(0.9

7)

D(2

.3)

L(3

.93)

O(1

1.1

9)

H49

(2.6

4)

F(2

.82)

F(0

.96)

H49

(2.9

2)

H(0

.83)

H49

(3.3

2)

D(0

.94)

O(2

.28)

D(3

.09)

H52

(11.1

6)

O(1

.97)

A01

(2.8

1)

C10-1

2(0

.67)

L(2

.6)

E(0

.71)

I(1

.76)

H(0

.77)

G(1

.61)

J61

(2.4

2)

E36

(2.6

1)

H52

(1.7

2)

H52

(2.7

4)

D(0

.55)

D(2

.43)

D(0

.69)

D(1

.51)

A(0

.73)

A(1

.54)

E36

(1.5

8)

C29

(2.3

3)

G47

(1.6

8)

Tab

le3:

Top

5se

ctor

sfo

r:to

tal

stra

nd

ing

mu

ltip

lier

s;ex

tern

al

stra

nd

ing

mu

ltip

lier

s;an

dex

posu

reto

stra

nd

ing

risk

Austria

Belgiu

mCzechia

Germ

any

Greece

France

Italy

Sweden

Slovakia

UK

D(0

.71)

H(0

.10)

D(0

.79)

D(0

.32)

D(0

.29)

D(0

.45)

D(0

.59)

D(0

.18)

D(2

.42)

D(0

.49)

C19

(0.1

1)

C19

(0.0

9)

H52

(0.1

5)

C(0

.3)

C19

(0.1

4)

H(0

.09)

C19

(0.2

7)

C24

(0.0

9)

C19

(0.4

2)

F(0

.11)

A01

(0.1

1)

C20

(0.0

6)

O(0

.14)

O(0

.06)

H50

(0.0

9)

O(0

.09)

H(0

.25)

C19

(0.0

7)

C24

(0.1

9)

H49

(0.0

6)

L(0

.09)

C24-2

5(0

.04)

C24

(0.1

3)

H(0

.04)

H49

(0.0

6)

A(0

.05)

A(0

.1)

A01

(0.0

5)

H52

(0.1

5)

C24

(0.0

5)

H49

(0.0

9)

C22-2

3(0

.03)

C19

(0.1

2)

Q(0

.03)

O(0

.06)

C19

(0.0

4)

O(0

.08)

O(0

.05)

O(0

.13)

C19

(0.0

4)

Tab

le4:

Sec

tora

las

set

stra

nd

ing

mu

ltip

lier

sof

Min

ing

an

dqu

arr

yin

gse

ctor

B(T

op

5se

ctors

excl

ud

ing

B)

9

D (Electricity, gas, steam and air conditioning) appears as the sector more at risk of physical asset

stranding in all the countries in our sample except for one. Several manufacturing sectors appear

among the top 5, most notably C19 (Coke and refined petroleum products) and C24 (Basic metals).

Other recurrent sectors are in the A, H and O categories.

Finally, looking at the values of total sectoral exposure to asset stranding, we can identify three

main sectors at risk, repeatedly appearing among the sectors with the highest row sums in S: O

(Public administration and defence; compulsory social security); H (Transportation and storage);

and D (Electricity, gas, steam and air conditioning). Exceptions include the real estate sector (L) in

Austria, manufacturing (C) in Germany, and construction (F) in the United Kingdom. These sectors,

in addition to having high capital intensities, are affected by multiple relevant inward stranding links.

To observe this, we treat S matrix as an adjacency matrix for a directed network (Godsil and Royle,

2013), interpreting productive sectors as the vertices of the network and the si,j elements of S as

the weight of the edges going from vertex j to vertex i. It is possible to represent visually the

network as a chord diagram. Figure 1 shows the outcome of this procedure for Germany, whose low

sectoral granularity contributes in improving the readability of results10. The size of each sectoral

segments is a function of the aggregation of both inward and outward stranding links. While most

sectors exhibit both inward and outward links, there are several (B, F, I, J, M, S) which only have

outward links (i.e. they affect capital stranding in other sectors but are not themselves affected by

any other sector), and one sector (Q) which only has inward links (i.e. it doesn’t create stranding

in any other sector). The figure offers a more detailed and disaggregated picture of the data showed

in Table 3. Sector N has the highest stranding multipliers, with the ones directed towards C, E and

H being particularly strong. The stranding effects triggered by the mining sector (B) focus mainly

on manufacturing (C) and electricity and gas (D). As for sectoral exposure, the sectors most at

risk appear to be manufacturing (with the strongest stranding coming from A and B) and public

administration (with a particularly relevant stranding effect stemming from R).

4 The stranding effects of a low-carbon transition

4.1 Cascades of physical asset stranding

After having shown the stranding potential and the stranding risk exposure of the entire range of

productive sectors, we now shift our attention to sector B (Mining and quarrying). Our aim is to

investigate the relevance of the potential stranding of physical assets due to a transition away from

fossil fuels, and understand how the stranding process originating in the fossil fuel sector might

propagate throughout the economic system. In order to do so, we start by identifying the most

relevant stranding links originating from a unitary loss of primary inputs supporting the production

of B (i.e. the largest values appearing on the B column of matrix S). We retain only the top q

percentile of the values, and position the affected sectors on the first layer of our cascade network.

We repeat the procedure for the sectors in the first layer, identifying the sectors within the top q

percentile of stranding links originating in the layer. The weight of the resulting network edges are

re-weighted to take into account that the loss in primary inputs in these sectors will be lower than

one and a function of the strength of the upper edges. In other words, the stranding links tend

to be stronger the closer they are to the shock origin, and get gradually weaker as they cascade

downwards. We then repeat the same procedure for each layer, excluding the sectors that had

10To further ease readability, we first eliminate less relevant network connections computing the “minimal fully-

connected network” of S as in Campiglio et al. (2017).

10

A

B

C

D

E

F

G

H

IJ

K

L

M

N

O

P

Q

RS

00.

40

0.4

0

0.4

0.8

1.2

00.40.8

0

0.4

0.8

0

0

0.4

0

0.4

0.8

00

0.4

0

0.4

0

0.40

0.40

0.4 0.8

0

0.4

0.8

1.2

00

00

Figure 1: Chord diagram of S minimal fully-connected network for Germany (2010)

already appeared in upper layers, until no new sectors appear. The results of this procedure for a

selection of countries11 are shown in the pyramid-like networks of Figure 2, for q = 0.05. The width

of the edges as represented in Figure 2 are proportional to their weight. The numerical weight of

the top 10 edges is shown for reference.

The sectors in the first layer of the network overlap with the ones reported in Table 4. The

strongest stranding link is the one flowing to the D sector (Electricity, gas, steam and air condition-

ing) for all countries with the only exception of Belgium. Manufacturing activities, especially coke

and refined petroleum products (C19) and basic metals (C24), as well as transportation and storage

(H), also frequently appear among the sector most strongly affected by the immediate stranding

caused by B. From the electricity and gas sector (D), the stranding cascade often continues affecting

public administration (O) and water services (E36). Given the strength of the original stranding

from B to D, these links are often the most relevant after the ones affecting sectors in the first layer,

and are justified by both the high capital intensity of the sectors and their large consumption of

D products (e.g. water pumping and purification are energy intensive). From the coke and refined

petroleum products sector (C19), the most common cascades proceed through the chemical sector

(C20) and the land transport and pipelines sector (H49). The stranding in the basic metals sector

(C24) often propagates through to the fabricated metal products (C25) and the motor vehicles sector

(C29), with the latter frequently further affecting trade and repair of motor vehicles (G45). Finally,

when the disaggregation among H subsectors is available, we observe a relevant stranding clustering

11The choice of the four countries represented in Figure 2 was mainly determined by the granularity of national

data and the richness of the resulting networks. The cascades for the rest of the countries in our sample can be found

in the appendix.

11

0.7090.11 0.108

0.094

0.023 0.021 0.066 0.024

0.009

0.017

B

D C19 A01 L

H49 O H52 C10-12 I Q87_88 G47

G46 F E37-39 M69_70 H51 Q86 J61

C24 K64 P

C25 C31_32

C28 C26

C29 G45

N77

AT

(a) Austria (q=0.05)

0.79 0.148 0.145

0.038 0.01

0.009

0.019 0.017 0.014

0.005

B

D H52 O

P H49 E36 A01

L E37-39 C10-12 I

C24

C28 C25 C29

C23 G45

F

CZ

(b) Czechia (q=0.05)

0.183 0.09 0.066

0.009 0.0080.006 0.04 0.039

0.016

0.02

B

D C24 C19

O L E36 C25 C29 C20-21 H49

P J61 C28 G45 A01 H52-53 C17

Q86 C10-12 I H50 C18

SE

(c) Sweden (q=0.05)

0.485

0.108 0.056

0.048

0.022 0.015 0.005

0.005

0.004 0.004

B

D F H49 C24

E36 G47 L O H52 G46 E37-39 C25 C29

K65 P H51 I C20 Q86 G45 N77

K64 Q87_88 C22 A01

R93 C10-12

J59_60 R90-92

S94

UK

(d) United Kingdom (q=0.05)

Figure 2: Cascades of productive capital stranding for selected countries

among them, and especially between land transport and pipelines (H49) and warehousing (H52).

In addition to the sectors mentioned above, several other sectors frequently appear in the cas-

cade networks. For instance, public administration (O) often appears on the second layer of the

network affected by multiple stranding links (originating from electricity and gas, transport sectors,

and others). The sectors in the E category also often appear, sometimes with E36 (water services)

affecting E37-39 (sewerage and waste services). The latter in turn sometimes appears to have rele-

vant stranding links towards basic metals (C24). Of the primary sectors, only agriculture (A01) is

12

present in the cascade networks (presumably due to high fossil-fuel inputs in modern agricultural

systems), while forestry (A02) and fishing (A03) never appear. Service sectors such as information

and communication (J), finance and insurance (K), professional services (M), administrative services

(N), Human health and social work services (Q) and arts and recreation services (R) also tend not

to be present in the networks, or only to appear at lower layers, suggesting that the decarbonisation

process might not be particularly detrimental for them in terms of capital asset stranding. Accomo-

dation and food services (I), on the other hand, is often present and usually affected by stranding

cascading from agriculture (A01).

Studying the structure of the networks in conjunction with the weights of its edges, we can identify

particularly relevant cascades. Both in Austria and Czechia it is possible to observe a deep stranding

cascades passing through sewerage and waste (E37-39) and basic metals (C24), and then affecting a

significant number of manufacturing sectors (fabricated metal products, machinery and equipment,

motor vehicles, and others) as well as trade of motor vehicles (G45). Sweden exhibits several fairly

long stranding cascades, with the most prominent one passing through coke and refined petroleum

products (C19), land transport and pipelines (H49), warehousing and postal services (H52-53) and,

finally, water transport (H50). The cascade network of the United Kingdom is peculiar, in that it

has two particularly deep cascades originating in wholesale trade (G46), with the first one involving

several manufacturing sectors and agriculture, and the second one mainly involving service sectors

(accommodation and food, residential care and social work, sports, arts and others).

4.2 The capital stock at risk of stranding in a low-carbon transition

The aim of this section is to provide an estimate of the overall productive capital stock at risk of

becoming stranded from a switch away from fossil fuels. In order to do so, we must distinguish fossil

fuels from metals and minerals in the use of B sector products by each sector in each country12.

However, the Eurostat database does not provide the required granularity. To overcome this issue,

we use the data provided by Exiobase (Wood et al., 2015), a multi-regional IO database covering 200

types of products13, to calculate the proportion of fossil fuels within the overall use of B products

by the sectors of a national economic system. This proportion tends to be close to 100% in the

electricity (D) and manufacturing (C) sectors of many countries, while it is usually closer to zero

for sectors such as construction (F) and wholesale and retail trade (G). Calling fi the fossil fuel

proportion of sector B product use by sector i, we define the amount of capital at risk of being

stranded due to the transition away from fossil fuels as

Strandi = fisi,BZB,ji

GB,B, (6)

where si,B is the element of matrix S representing the stranding effect on sector i of a unitary drop

in primary inputs used in sector B, ZB,ji is the sum of the elements of row B in the inter-industry

matrix Z representing the total use of products of sector B in the economic system, and GB,B is

the B element of the diagonal in matrix G representing the loss in production in B due to a unitary

drop of primary inputs in the same sector. Division by GB,B transforms the projected loss in output

of sector B into the equivalent loss of primary inputs required to produce it, making it compatible

with the units used by S.

12Sector B includes several distinct activities: Mining of coal and lignite (B05); Extraction of crude petroleum

and natural gas (B06); Mining of metal ores (B07); Other mining and quarrying (B08); and Mining support service

activities (B09). While only some of them concern fossil fuels (B06, B07 and, partially, B09), available Eurostat data

for the sector is usually presented in an aggregated manner at the NACE level 1 category.13The database is available at www.exiobase.eu.

13

Total capital Mining (B) Manufacturing (C) Electricity/gas (D)

Austria 5,689 (0.8%) 431 (16.0%) 1,706 (2.4%) 3,315 (12.5%)

Belgium 3,181 (0.6%) 1 (0.1%) 2,692 (3.0%) 285 (1.2%)

Czechia 17,536 (3.7%) 4,075 (60.9%) 2,772 (3.3%) 6,718 (25.7%)

Germany 40,752 (1.0%) 3,629 (29.6%) 12,702 (2.8%) 21,627 (12.2%)

Greece 8,774 (2.7%) 1,313 (48.7%) 1,800 (8.1%) 2,683 (17.1%)

France 35,514 (1.4%) 3,644 (21.4%) 3,877 (2.1%) 21,913 (23.3%)

Italy 58,589 (2.1%) 2,252 (10.7%) 19,776 (4.9%) 30,565 (14.0%)

Sweden 3,970 (0.8%) 55 (1.4%) 1,762 (2.2%) 1,856 (3.1%)

Slovakia 18,749 (8.2%) 473 (15.1%) 3,220 (7.7%) 13,458 (35.1%)

UK 84,678 (3.6%) 45,900 (69.3%) 7,385 (2.9%) 28,384 (35.7%)

Table 5: Productive capital stock at risk of stranding (million e at current prices in 2010 and share

of total/sectoral capital stocks)

This procedure must be amended in the case of the B sector itself, where the stranding of the

capital stock triggered by a transition away from fossil fuels will originate from the phasing-out of

production (e.g. oil platforms remaining inactive) rather than a lack of input factors as in the case

of all other sectors (e.g. electricity plants interrupting production because of missing fuels). Using

Exiobase, we calculate the ratio of fossil fuels production on the overall production of sector B and

assume that the same proportion of capital stock of sector B is to be considered at risk of stranding.

The stranding impact on the remainder of the capital stock in B, used to extract metals and other

materials, is calculated using the same methodological approach as the rest of the sectors14.

Table 5 reports the results of this procedure. The overall amount of capital stock at risk of

stranding due to a transition away from fossil fuels is shown in the first column (values in million e;

share of total capital in brackets). The largest absolute exposure is in the United Kingdom, where

around e84 billion worth of capital stock (3.6% of the value of the total capital stock) may become

idle if fossil fuels were not used in production anymore. The other large economies in the sample

(Italy, Germany, France) follow, with approximately e59, e40 and e36 billion of capital at risk.

The least exposed countries in proportional (and absolute) levels appear to be Belgium, Sweden and

Austria (0.6%, 0.8% and 0.8% of total capital at risk, respectively), while Slovakia is by far the most

exposed economy, with around 8.2% of its productive capital stock at risk. The remainder of the

columns report the stranding values for the three sectors that tend to exhibit the highest sectoral

stranding across countries: B (Mining and quarrying), C (Manufacturing) and D (Electricity, gas,

steam and air conditioning).

The stranding of physical capital in sector B strongly varies across countries. In absolute terms,

the United Kingdom is by far the most exposed, with around e46 billion (69.3% of total capital

in sector B) at risk of stranding in the form of built infrastructure for the extraction of oil and

gas, especially in the North Sea. No other country in our sample gets close to these values. In

contrast, mining sectors in countries such as Austria and Sweden, are only minimally exposed,

as they are relatively small and mainly oriented towards the production of metals and minerals.

In proportional terms, the United Kingdom, Czechia and Greece appear to be the most exposed

countries. Computations based on World Development Indicators natural resources rents and current

prices GDP data for 2010 show that oil amounts for 84% of the total natural resources rents (that

14Lacking a specific value in the G matrix capturing the output effect of a unitary loss in primary inputs for fossil

fuel production on the production of metals and minerals, we assume that the internal output effect in sector B is

equal to its default value less than unitary drop in production. In other words, GB,B = GB,B − 1.

14

include coal, forest, mineral, natural gas and oil) of the UK while coal amounts for respectively 78%

and 45.7% of Czechia and Greece total natural resources rents. This indicates that the mining sector

of those countries is dominated by fossil fuels15, which explains the high stranding values of the B

sector capital stock.

The manufacturing sectors in our ten countries are all quite significantly exposed, in a range

that has France as the lower bound (2.1% of total C capital at risk of stranding) and Slovakia and

Greece as the upper bound (7.7% and 8.1%, respectively). Such a high exposure may be explained

by the share of the coke and refined petroleum products sector (C19) in the whole manufacturing

productive capital stock: 11% for Greece and 8.1% for Slovakia, with a sample average of 4.3% 16.

Conversely, the French C19 sector only amounts to 1.9% of its whole manufacturing sector capital

stock. For Slovakia, the motor vehicles, trailers and semi-trailers and other transport equipment

sector (C29 C30) also amounts to 18.1% of the capital stock, with a sample average of 11.2%.

Moreover, it is important to note that the figures in table 5 take into account the indirect effects

of a complete decarbonization, that is the effects through the non mining sectors supplying the

manufacturing sectors. Therefore, these percentages capture the reliance of the manufacturing

sectors both on raw fossil fuels and on fossil fuels dependent sectors, i.e. transportation.

Turning to electricity and gas, the capital at risk of stranding in this sector is the largest sectoral

value in absolute terms for all the countries in our sample with the exception of the United Kingdom.

Italy is exposed by around e31 billion worth of capital, followed by the UK (e28 billion), France

and Germany (both around e21 billion). In proportional terms, the most exposed countries are

the UK and Slovakia, followed by Czechia and France. Part of the explanation lie in the electricity

production capacities. Generation capacity based on fossil fuels amounts to 77.5%, 40.6%, 56.8%

and 20.5% of the total electricity production capacities in, respectively, the UK, Slovakia, Czechia

and France. The electricity generation origins vary considerably between these countries: the UK

and Czechia produce respectively 76.4% and 60.1% of their electricity from oil, gas and coal sources

but this figure is only 25.1% for Slovakia and 10% for France. This level of exposure of the Slovak

and French D sectors to decarbonization is consistent with their electricity generation, which is,

respectively, 53.1% and 76% from nuclear power and 21.6% and 14% from renewables, according to

World Development Indicators data.

Due to data availability and limitations of IO data, the estimates provided should be taken as

rough ballpark figures. First, it is likely that the decarbonisation process would entail a gradual

phase-out of fossil fuels utilisation as production inputs rather than a sudden and complete disap-

pearance (Kemp-Benedict, 2018). In this sense, the figures provided should be interpreted as the

upper bound of capital stocks that would potentially become idle due to a low-carbon transition. In

a dynamic perspective, a forward-looking technological shift to low-carbon inputs, combined with

the gradual depreciation of current fossil-intensive capital stocks, would lead to lower stranding

impacts. Second, the analysis attributes shares of sectoral capital stranding in proportion of the

relative use of fossil fuel among all the production input factors of a sector, thus implicitly assum-

ing complete substitutability among input factors and ignoring complementarity effects. In reality,

however, the relative proportion of capital asset stranding due to, say, diesel is likely to be much

larger than its monetary share among input factors, for the simple reason that without diesel a

significant proportion of machinery cannot operate at all. In this sense, hence, the figures provided

represent the lower bound of capital asset stranding due to the loss of essential production factors.

15This needs to be nuanced for Greece: non-fossil fuels mineral resources amounts for 48.4% of the country total

natural resources rents.16We exclude Germany from this computation because disaggregated data is not available for C sectors.

15

Third, using static IO data, it is not possible to fully capture the macroeconomic feedback triggered

by a low-carbon transition, which would also affect employment and aggregate demand, with more

complex implications for the overall capital stranding.

5 Conclusions

The transition to a decarbonised economic system will involve a large-scale reallocation of material

and economic resources. While an expanding literature has been investigating the economic and

financial repercussions of leaving fossil reserves in the ground and the consequent loss in market

capitalisation of fossil companies, less attention has been devoted to the understanding of the poten-

tial stranding of productive physical capital stocks, and how this would cascade within the network

of economic interdependencies. This article contributes to this effort by providing an original mea-

sure to quantify the monetary value of productive capital stock of a sector that is at risk of being

unutilised because of a reduction of primary inputs flowing into another sector.

Analysing the data available for ten European countries, we have shown the sectors with the

highest “asset stranding multipliers” to be linked to water and waste management (E); real estate

(L); and public administration (O). This result is mainly driven by the high capital intensity of the

sectors and the consequent significance of sectoral internal capital stranding. When abstracting from

internal effects with the aim of identifying the sectors that are likely to have the strongest stranding

impacts on the rest of the economic system, we have found the mining and quarrying sector (B) to

be particularly relevant, together with sectors linked to manufacturing (C); professional, scientific

and technical activities (M); and administrative and support service activities (N). These sectors are

key in providing essential inputs to other sectors with high capital intensity. Finally, we have shown

how the sectors most exposed to the risk of capital asset stranding include public administration

(O); transportation and storage (H); and electricity, gas, steam and air conditioning (D).

When focusing more specifically on the mining and quarrying sector in the attempt of studying

the stranding effects of a low-carbon transition, we have shown how moving away from fossil fuel

would have a particularly strong stranding effect on sectors linked to electricity, gas, steam and air

conditioning (D); coke and refined petroleum products (C19); and basic metals (C24). We have

identified regular patterns in cascade structures, such as the links from electricity and gas (D) to

water services (E36) and public administration (O); from waste services (E37-39) to basic metals

(C24) and then to fabricated metal products (C25) and motor vehicles sector (C29); from the coke

and refined petroleum products sector (C19) to the chemical sector (C20) and the land transport

and pipelines sector (H49). Other sectors usually present in the stranding cascade networks include:

agriculture (A01), wholesale and retail trade (G), transportation and storage (H), accommodation

and food services (I) and public administration (O).

Finally, we have provided a ballpark estimate of the overall and sectoral productive capital at risk

of stranding due to a transition away from fossil fuels. The figure is in the range of 0.6-8.2% of the

overall productive capital stock for our sample of countries. In absolute terms, the United Kingdom

has the largest stock of capital at risk (around e85 billion), a result mainly driven by the stranding

of its fossil extraction infrastructure. The UK is followed by Italy, Germany and France, where more

than half of the capital stranding takes place in the electricity and gas sector. In proportional terms,

Slovakia is the most affected country in our sample, with 8.2% of its productive capital stock at

risk of stranding. In contrast, the least affected countries are Belgium, Sweden and Austria, with

0.6-0.8% of overall productive capital at risk. Looking at specific sectors, the stranding in mining

and quarrying varies widely across countries, depending on the size of the sector and the relative

16

share of fossil fuels over metals and minerals. While the UK mining would suffer a loss of almost e46

billion of its capital stock, only e1 million is at risk of stranding in Belgium. The manufacturing

sector is significantly affected in all countries in the sample (never less than 2% of its capital stock

is at risk). Greece is the country with most manufacturing capital at risk in proportional terms

(8.1%), while Italy would lose the largest amount in absolute terms (almost e20 billion). In the

electricity and gas sector, the relevance of capital stranding strongly depends on the energy mix

used to produce electricity, going from 1-3% of sectoral capital in Belgium and Sweden to as high

as 35% in Slovakia and United Kingdom.

The results of our analysis suggest that the drop in capital utilisation triggered by an abrupt

and unplanned transition might be substantial and systemic. While our methodological framework

is not able to provide insights regarding the transition dynamics, the large proportion of productive

assets directly or indirectly dependent on fossil inputs, together with the planned capital expansion

in coming years, indicate that cascading physical capital stranding is a likely scenario to consider.

This offers valuable insights for two main areas of work. First, current research studying the impli-

cations of climate mitigation trajectories through numerical simulations, usually predicated on the

assumption of full capital utilisation, might be underestimating the economic effects of a low-carbon

transition. Second, the burgeoning literature on the macro-financial repercussions of the decarbon-

isation process might improve its analytical power by considering the effects of financial stranded

assets in a wider range of productive sectors than just fossil extraction and power generation. En-

hancing these strands of research with the inclusion of capital utilisation considerations will support

policy-makers in the management of a rapid and smooth transition to a low-carbon economic system.

References

Acemoglu, D., V. M. Carvalho, A. Ozdaglar, and A. Tahbaz-Salehi (2012, September). The network

origins of aggregate fluctuations. Econometrica 80 (5), 1977–2016.

Aldasoro, I. n. and I. Angeloni (2015, April). Input–output-based measures of systemic importance.

Quantitative Finance 15 (4), 589–606.

Antras, P., D. Chor, T. Fally, and R. Hillberry (2012, May). Measuring the Upstreamness of

Production and Trade Flows. American Economic Review 102 (3), 412–416.

Augustinovics, M. (1970). Methods of International and Intertemporal Comparison of Structure.

In A. P. Carter and A. Brody (Eds.), Contributions To Input-Output Analysis, pp. 249–269.

Amsterdam, the Netherlands and London, UK: North-Holland Publishing Company.

Baldwin, E., Y. Cai, and K. Kuralbayeva (2018). To Build or Not to Build? Capital Stocks and

Climate Policy. CESifo Working Paper 6884, CESifo, Munich.

Battiston, S., A. Mandel, I. Monasterolo, F. Schutze, and G. Visentin (2017, March). A Climate

Stress-Test of the Financial System. Nature Climate Change 7, 283–288.

Beyers, W. B. (1976, April). Empirical Identification of Key Sectors: Some Further Evidence.

Environment and Planning A: Economy and Space 8 (2), 231–236.

Blochl, F., F. J. Theis, F. Vega-Redondo, and E. O. Fisher (2011, April). Vertex centralities in input-

output networks reveal the structure of modern economies. Physical Review E 83 (4), 046127.

17

Cahen-Fourot, L., E. Campiglio, E. Dawkins, A. Godin, and E. Kemp-Benedict (2019). Looking for

the inverted pyramid: An application using input-output networks. Working Paper Forthcoming,

WU Institute for Ecological Economics, Vienna.

Caldecott, B. (2018). Stranded Assets and the Environment: Risk, Resilience and Opportunity.

Routledge Explorations in Environmental Studies. Oxon: Routledge.

Campiglio, E., Y. Dafermos, P. Monnin, J. Ryan-Collins, G. Schotten, and M. Tanaka (2018,

June). Climate Change Challenges for Central Banks and Financial Regulators. Nature Climate

Change 8 (6), 462–468.

Campiglio, E., A. Godin, and E. Kemp-Benedict (2017). Networks of stranded assets: A case for

a balance sheet approach. AFD Research Papers 2017-54, Agence Francaise de Developpement,

Paris.

Carbon Tracker (2013). Unburnable Carbon 2013: Wasted Capital and Stranded Assets. Technical

report, Carbon Tracker Initiative and the Grantham Research Institute, London.

Carney, M. (2015, September). Breaking the Tragedy of the Horizon – Climate Change and Financial

Stability. Speech given at Lloyd’s of London by the Governor of the Bank of England, Bank of

England, London.

Davis, S. J., K. Caldeira, and H. D. Matthews (2010, September). Future CO2 Emissions and

Climate Change from Existing Energy Infrastructure. Science 329 (5997), 1330–1333.

Davis, S. J. and R. H. Socolow (2014, August). Commitment accounting of CO2emissions. Envi-

ronmental Research Letters 9 (8), 084018.

Dietzenbacher, E. (1997). In vindication of the Ghosh model: A reinterpretation as a price model.

Journal of regional science 37 (4), 629–651.

Eurostat (2008a). Eurostat Manual of Supply, Use and Input-Output Tables. Methodologies and

Working Papers, Eurostat, Luxembourg.

Eurostat (2008b). NACE Rev.2 Statistical classification of economic activities in the European

Community. Technical report, Eurostat, Luxembourg.

Galbusera, L. and G. Giannopoulos (2018, September). On input-output economic models in disaster

impact assessment. International Journal of Disaster Risk Reduction 30, 186–198.

Ghosh, A. (1958). Input-Output Approach in an Allocation System. Economica 25 (97), 58–64.

Godsil, C. and G. F. Royle (2013). Algebraic Graph Theory, Volume 207. Springer Science &

Business Media.

IEA (2017). Energy Technology Perspectives 2017. Technical report, International Energy Agency,

Paris.

IPCC (2018). Global Warming of 1.5◦C. Technical report, Intergovernmental Panel on Climate

Change.

IRENA (2017). Stranded assets and renewables: How the energy transition affects the value of

energy reserves, buildings and capital stock. Technical report, International Renewable Energy

Agency, Abu Dhabi.

18

Joya, O. and E. Rougier (2019, April). Do (all) sectoral shocks lead to aggregate volatility? Empirics

from a production network perspective. European Economic Review 113, 77–107.

Kemp-Benedict, E. (2018, November). Investing in a Green Transition. Ecological Economics 153,

218–236.

Leontief, W. W. (1951). Part IV: Application of Input-Output Technique to the American Economic

System in 1939. In The Structure of the American Economy, 1919-1939: An Empirical Application

of Equilibrium Analysis (2nd ed.)., pp. 139–218. New York: Oxford University Press.

McCollum, D. L., W. Zhou, C. Bertram, H.-S. de Boer, V. Bosetti, S. Busch, J. Despres, L. Drouet,

J. Emmerling, M. Fay, O. Fricko, S. Fujimori, M. Gidden, M. Harmsen, D. Huppmann, G. Iyer,

V. Krey, E. Kriegler, C. Nicolas, S. Pachauri, S. Parkinson, M. Poblete-Cazenave, P. Rafaj,

N. Rao, J. Rozenberg, A. Schmitz, W. Schoepp, D. van Vuuren, and K. Riahi (2018, July). Energy

Investment Needs for Fulfilling the Paris Agreement and Achieving the Sustainable Development

Goals. Nature Energy 3 (7), 589–599.

McGlade, C. and P. Ekins (2015). The geographical distribution of fossil fuels unused when limiting

global warming to 2 ◦C. Nature 517 (7533), 187–190.

Mercure, J.-F., H. Pollitt, J. E. Vinuales, N. R. Edwards, P. B. Holden, U. Chewpreecha, P. Salas,

I. Sognnaes, A. Lam, and F. Knobloch (2018, July). Macroeconomic Impact of Stranded Fossil

Fuel Assets. Nature Climate Change 8 (7), 588–593.

Miller, R. E. and P. D. Blair (2009). Input-Output Analysis: Foundations and Extensions. Cambridge

university press.

OECD (2001). Measuring productivity. Measurement of aggregate and industry-level productivity

growth. Technical report, Organisation for Economic Co-operation and Development, Paris.

Oosterhaven, J. (1988). On the Plausibility of the Supply-Driven Input-Output Model. Journal of

Regional Science 28 (2), 203–217.

Pfeiffer, A., C. Hepburn, A. Vogt-Schilb, and B. Caldecott (2018). Committed emissions from

existing and planned power plants and asset stranding required to meet the Paris Agreement.

Environmental Research Letters 13 (5), 054019.

Piontek, F., M. Kalkuhl, E. Kriegler, A. Schultes, M. Leimbach, O. Edenhofer, and N. Bauer (2018,

December). Economic growth effects of alternative climate change impact channels in economic

modeling. Environmental and Resource Economics.

Rozenberg, J., A. Vogt-Schilb, and S. Hallegatte (2018, November). Instrument choice and stranded

assets in the transition to clean capital. Journal of Environmental Economics and Management .

Smith, C. J., P. M. Forster, M. Allen, J. Fuglestvedt, R. J. Millar, J. Rogelj, and K. Zickfeld (2019,

January). Current fossil fuel infrastructure does not yet commit us to 1.5 ◦C warming. Nature

Communications 10 (1), 101.

Stolbova, V., I. Monasterolo, and S. Battiston (2018, July). A Financial Macro-Network Approach

to Climate Policy Evaluation. Ecological Economics 149, 239–253.

19

Tavoni, M., E. Kriegler, K. Riahi, D. P. Van Vuuren, T. Aboumahboub, A. Bowen, K. Calvin,

E. Campiglio, T. Kober, J. Jewell, et al. (2015). Post-2020 Climate Agreements in the Major

Economies Assessed in the Light of Global Models. Nature Climate Change 5 (2), 119–126.

Tol, R. S. J. (2009, June). The economic effects of climate change. Journal of Economic Perspec-

tives 23 (2), 29–51.

UNFCCC (2016, January). Report of the Conference of the Parties on Its Twenty-First Session, Held

in Paris from 30 November to 13 December 2015. Addendum. Part Two: Action Taken by the Con-

ference of the Parties at Its Twenty-First Session. Technical Report FCCC/CP/2015/10/Add.1,

United Nations Framework Convention on Climate Change, Paris.

Vermeulen, R., E. Schets, M. Lohuis, B. Kolbl, D.-J. Jansen, and W. Heeringa (2018). An energy

transition risk stress test for the financial system of the Netherlands. Occasional Studies 16-7, De

Nederlandsche Bank, Amsterdam.

Welsch, H. (2001, December). The determinants of production-related carbon emissions in West

Germany, 1985–1990: assessing the role of technology and trade. Structural Change and Economic

Dynamics 12 (4), 425–455.

Wood, R., K. Stadler, T. Bulavskaya, S. Lutter, S. Giljum, A. De Koning, J. Kuenen, H. Schutz,

J. Acosta-Fernandez, A. Usubiaga, M. Simas, O. Ivanova, J. Weinzettel, J. H. Schmidt, S. Merciai,

and A. Tukker (2015, January). Global Sustainability Accounting—Developing EXIOBASE for

Multi-Regional Footprint Analysis. Sustainability 7 (1), 138–163.

20

A Sector codes and descriptions

Table A1: Sector codes and descriptions

Sector code Sector description

A Agriculture, forestry and fishing

A01 Crop and animal production, hunting and related service activities

A02 Forestry and logging

A03 Fishing and aquaculture

B Mining and quarrying

C Manufacturing

C10-12 Food, beverages and tobacco products

C13-15 Textiles, wearing apparel, leather and related products

C16 Wood and products of wood and cork, except furniture

C17 Paper and paper products

C18 Printing and reproduction of recorded media

C19 Coke and refined petroleum products

C20 Chemicals and chemical products

C21 Basic pharmaceutical products and pharmaceutical preparations

C22 Rubber and plastic products

C23 Other non-metallic mineral products

C24 Basic metals

C25 Fabricated metal products, except machinery and equipment

C26 Computer, electronic and optical products

C27 Electrical equipment

C28 Machinery and equipment n.e.c.

C29 Motor vehicles, trailers and semi-trailers

C30 Other transport equipment

C31 32 Furniture and other manufactured goods

C33 Repair and installation services of machinery and equipment

D Electricity, gas, steam and air conditioning

E Water supply; sewerage; waste management and remediation activities

E36 Natural water; water treatment and supply services

E37-39 Sewerage services; sewage sludge; waste collection, treatment and disposal ser-

vices; . . .

F Constructions and construction works

G Wholesale and retail trade; repair of motor vehicles and motorcycles

G45 Wholesale and retail trade and repair services of motor vehicles and motorcycles

G46 Wholesale trade, except of motor vehicles and motorcycles

G47 Retail trade services, except of motor vehicles and motorcycles

H Transportation and storage

H49 Land transport and transport via pipelines

H50 Water transport

H51 Air transport

H52 Warehousing and support activities for transportation

Continued on next page

21

Table A1: Sector codes and descriptions (continued)

Sector code Sector description

H53 Postal and courier activities

I Accommodation and food service activities

J Information and communication

J58 Publishing activities

J59 60 Motion picture, video and television production, sound recording, broadcasting,

. . .

J61 Telecommunications

J62 63 Computer programming, consultancy; Information service activities

K Financial and insurance activities

K64 Financial services, except insurance and pension funding

K65 Insurance, reinsurance and pension funding services, except compulsory social

security

K66 Activities auxiliary to financial services and insurance services

L Real estate activities

M Professional, scientific and technical activities

M69 70 Legal and accounting services; Activities of head offices; management consul-

tancy activities

M71 Architectural and engineering activities; technical testing and analysis

M72 Scientific research and development

M73 Advertising and market research

M74 75 Other professional, scientific and technical activities; Veterinary activities

N Administrative and support service activities

N77 Rental and leasing activities

N78 Employment activities

N79 Travel agency, tour operator and other reservation service and related activities

N80-82 Security and investigation activities; buildings and landscape; office support

O Public administration and defence; compulsory social security

O84 Public administration and defence; compulsory social security

P Education

P85 Education

Q Human health and social work activities

Q86 Human health activities

Q87 88 Residential care activities; social work activities without accommodation

R Arts, entertainment and recreation

R90-92 Creative, arts and entertainment activities; libraries, museums, archives; gam-

bling and betting

R93 Sports activities and amusement and recreation activities

S Other services activities

S94 Activities of membership organisations

S95 Repair of computers and personal and household goods

S96 Other personal service activities

22

B Cascade networks for all countries

0.102 0.0870.057 0.038

0.011 0.009 0.005

0.005

0.0130.007

B

H C19 C20 C24-25

G N O F C22-23 C16-18 C21 C29-30 C28

C10-12 Q86 K D L M69-71

I A Q87_88

BE

(a) Belgium (q=0.1)

0.323 0.303 0.059 0.036

0.011 0.016 0.014 0.001 0.002

0.001

B

D C O H

Q A E N G

R P M

J S K

L

DE

(b) Germany (q=0.2)

0.29 0.144 0.086 0.064

0.0130.007 0.05

0.0070.007 0.006

B

D C19 H50 H49

O E36 C24 A01 S94 J61 G46

E37-39 C25 F C27 C10-12 I

G47

EL

(c) Greece (q=0.05)

0.452 0.093 0.091 0.051

0.01 0.005 0.002 0.005 0.001 0

B

D H O A

R G L C10-12 I C16-18

K

FR

(d) France (q=0.1)

Figure A1: Cascades of productive capital stranding for selected countries

23

0.594 0.272 0.249 0.096

0.028 0.012 0.029 0.009 0.009 0.005

B

D C19 H A

O G C10-12 I

IT

(a) Italy (q=0.1)

2.416 0.42 0.1910.145

0.1090.042 0.027 0.0440.038

0.015

B

D C19 C24 H52

O E36 P H49 C20 C28 C29 C25

J59_60 L S94 R93 A01 C22 C17 G45 C31_32

C18 J61 R90-92 Q86 C10-12 C26 I

SK

(b) Slovakia (q=0.05)

Figure A2: Cascades of productive capital stranding for selected countries

24

WU ViennaInstitute for Ecological Economics

Welthandelsplatz 2/D5A-1020 Vienna

+43 (0)1 313 36 4848ecolecon@wu.ac.at