Working Paper Series

Gert Bijnens, John Hutchinson, Jozef Konings, The interplay between

Arthur Saint Guilhem
green policy, electricity prices,
financial constraints and jobs:
firm-level evidence

No 2537 / April 2021

Disclaimer: This paper should not be reported as representing the views of the European Central Bank
(ECB). The views expressed are those of the authors and do not necessarily reflect those of the ECB.
 Abstract

Increased investment in clean electricity generation or the introduction of a carbon tax will
most likely lead to higher electricity prices. We examine the effect from changing electricity
prices on manufacturing employment. Analyzing firm-level data, we find that rising elec-
tricity prices lead to a negative impact on labor demand and investment in sectors most
reliant on electricity as an input factor. Since these sectors are unevenly spread across coun-
tries and regions, the labor impact will also be unevenly spread with the highest impact in
Southern Germany and Northern Italy. We also identify an additional channel that leads to
heterogeneous responses. When electricity prices rise, financially constrained firms reduce
employment more than less constrained firms. This implies a potentially mitigating role for
monetary policy.

Keywords: environmental regulation, labor demand, employment, manufacturing in-

dustry, monetary policy

JEL Classification: E52, H23, J23, Q48

ECB Working Paper Series No 2537 / April 2021 1

Non-technical summary

The EU has committed to tackling climate change and greenhouse gas emissions. To reach
the ambitious goals, significant investment in clean electricity generation in combination with a
meaningful carbon tax will be needed. There are, however, legitimate concerns that this will lead
to higher electricity prices and more broadly energy prices. Previous research suggests that a
$75 carbon tax is needed to keep global warming at 2 ◦ C. This could increase electricity prices in
Europe up to 20% depending on the emission intensity of generation. Whilst the overall impact
of such a price increase is complex, an individual manufacturing firm faces higher input costs
and lower competitiveness when electricity prices increase. This leads to reduced employment
and investment, predominantly within firms active in sectors most reliant on electricity as an
input factor.
In this paper we study over 200,000 manufacturing firms in Belgium, France, Germany,
Italy, the Netherlands and the UK over the period 2009 - 2017. We find that an electricity price
increase of 20% has a negative impact on employment of approx. 2% to 4% for the most impacted
industries. A $75 carbon tax could therefore lead to 150,000 affected jobs in the countries we
analyze. These jobs are either lost or need to be reallocated to other firms, industries and/or
regions.
We also find this impact to be highly heterogeneous. The impact is highest for industries
generally regarded as electricity intensive (e.g., chemicals, metals and paper). We find the highest
(most negative) country-wide impact for Belgium and the highest local impact in Southern
Germany and Northern Italy.
Should one worry about the 150,000 possible job losses in certain parts of the manufac-
turing industry in certain regions? In the past, Europe has successfully compensated millions
of lost manufacturing jobs by jobs in the services industry. There is nevertheless no room for
complacency. The impact could well reach beyond the 150,000 manufacturing jobs. One well
paid manufacturing job indirectly also creates 1 to 2.5 jobs in non-tradable services (hospitality,
food, retail, ...) in the immediate vicinity. Furthermore, the newly created services jobs are not
necessarily in the same region nor for the same skill-set. Given the rigid labor markets in several
European countries, increasing electricity prices could therefore have long-lasting negative labor
market effects for the affected regions and workers.
To ensure a positive public sentiment towards environment related price increases, there will

ECB Working Paper Series No 2537 / April 2021 2

be a need for targeted assistance to firms, workers, and disproportionately affected communities.
This is traditionally a role for fiscal policy makers. In this paper we also show there is a role for
monetary policy makers to cushion the negative employment effect from rising electricity prices.
We find indications that financially constrained manufacturing firms reduce employment more
when electricity prices rise. This implies a potentially mitigating role for monetary policy.

ECB Working Paper Series No 2537 / April 2021 3

1 Introduction

Many industrialized countries have committed themselves to tackling climate change and green-
house gas emissions. The European Union has set itself a long-term goal to become climate
neutral by 2050. To achieve these ambitious goals, the share of renewable energy sources (RES)
in total electricity generation has to rise substantially. Whilst investing in RES can have impor-
tant environmental and health benefits, it may increase electricity prices for firms and consumers
by altering the electricity generating mix. RES subsidies are often recovered by levies and sur-
charges paid by the electricity consumer.1 Furthermore, the introduction of a carbon tax, widely
regarded as an efficient means to curb greenhouse gas emissions,2 will most likely further in-
crease electricity prices. Several recent studies found evidence of a high degree of pass-through
of a carbon tax or emissions costs to wholesale electricity prices.3 There are hence legitimate
concerns that a climate neutrality objective could lead to increased electricity prices and more
broadly energy prices.
For the manufacturing sector, energy and more particularly, electricity costs are an important
component of total production costs. The EU defines energy intensive sectors as sectors where
energy costs amount to at least 3% of production value. For the most energy intensive industries
such as paper or metal production and processing the ratio can go up to 5–6% of production
cost. Consequently, electricity prices can have important effects on employment and investment
decisions as well. The impact of rising energy and more specific electricity prices on employment
remains, however, complex, heterogeneous and dependent on the perspective taken (firm-level,
sector-level, macro-level). When electricity prices rise, an individual manufacturing firm faces
higher input costs and lower competitiveness. This can translate into lower output and thus
lower employment. Furthermore, higher electricity prices make capital goods such as machinery
more expensive relative to labor. Since labor and capital are to a limited extent substitutes
(Henriksson et al. 2012), higher electricity prices can lower capital investment and increase
employment. Alternatively, firms could also decide to invest in energy saving technology that
1
Kreuz and Musgens (2017) calculate that Germany’s “Energiewende” or energy transition incurred an annual
gross costs in 2015 of ¿27.5 billion vs. a wholesale electricity value of ¿4.7 billion. The difference or RES cost is
paid by electricity consumers. They calculate that ∼22% of the final electricity price for private households and
∼35% for industrial users is related to the Energiewende.
2
See e.g., Stiglitz et al. (2017). Specifically for the EU Emission Trading System (ETS), Colmer et al. (2020)
found that ETS regulated firms did reduce carbon dioxide emissions by 8-12% compared to unregulated firms.
3
E.g., Fabra and Reguant (2014) for Spain, Hintermann (2016) for Germany and Lise et al. (2010) for 20
European countries.

ECB Working Paper Series No 2537 / April 2021 4

ultimately makes them more productive.4 This could lead to higher output and employment.
Finally, a firm might simply decide to relocate (parts of) its production to a region or country
with lower electricity prices.5 Which of these heterogeneous effects dominate is an empirical
question. The current literature suggests an employment elasticity as a function of energy and
electricity prices of around -0.2.6 This implies an electricity price increase of 10% has a negative
impact on employment within manufacturing firms of approx. 2%.
When one takes a more aggregate view, the negative impact becomes less clear. Hafstead
and Williams (2018) use a general equilibrium model and find that environmental policies lead
to shifts in employment but have little effect on total employment. Recently, Metcalf and Stock
(2020) found no support for the thesis that the EU emission trading system (ETS) had a negative
impact on aggregate employment. The lack of aggregate movements often hides substantial
heterogeneity under the hood. An example of underlying dynamics is given by Dussaux (2020).
He shows that increasing energy prices have a negative effect on average firm-level employment,
but a limited effect on aggregate sector-level employment. The reason is that jobs are shifted
from less energy efficient to more energy efficient firms. Marin and Vona (2017) find negative
effects from rising energy prices on French manufacturing establishments, but also find these
effects are to a certain extent mitigated by job reallocation between establishments of the same
firm. Subsequently Marin and Vona (2019) find that increasing energy prices lead to a higher
demand for technicians and at the same time to a lower demand for manual workers. Hille
and Möbius (2019) find positive net employment effects from increasing energy prices. Their
intuition is that while there might be job destruction in energy-intensive industries, these losses
are more than compensated in sectors producing, installing or consulting on energy saving or
pollution abatement technologies.
Whilst the economy-wide impact remains debated, it is safe to say rising electricity prices will
create some losers and these losers are likely to be active in electricity intensive manufacturing
4
Even if firms increase investment in energy saving technology, this does not necessarily mean they increase
overall investment. Weche (2018) found for Germany that environment related corporate investment crowds out
other business investment.
5
Saussay and Sato (2018) study the impact of energy prices on cross-border investment decisions and conclude
that a relative price increase does lead to increased investment from companies in the higher priced country
towards the lower priced country. They find an investment elasticity of -0.32.
6
For the US, Kahn and Mansur (2013) find that energy-intensive industries tend to locate in low electricity
price counties and estimate the elasticity between -1.65 to -0.17 depending on the sector with an average of -0.2.
Deschênes (2012) finds a negative relation between US state level electricity prices and employment and find
elasticity of -0.16 to -0.10. Cox et al. (2014) find, based on German data that higher input electricity prices lead
to lower employment due to output contractions. They estimate the elasticity between -0.69 and -0.06 depending
on the skill levels of the involved labor.

ECB Working Paper Series No 2537 / April 2021 5

industries and lower skilled workers. Even if the aggregate economic impact of rising electricity
prices is beneficial, there could be long-lasting negative and heterogeneous labor market effects
for some workers and/or regions, similar to the long-lasting negative effects from Chinese import
competition to the U.S. manufacturing industry, described by Autor et al. (2013, 2014).
To ensure a “just transition”7 towards a carbon neutral world, there is hence an important
role for fiscal policy makers. Recently, there is also an increased attention from monetary policy
makers towards climate change.8 Christine Lagarde, President of the ECB, stated that central
banks need to devote greater attention to understanding the impact of climate change.9
In our case, rising electricity prices are part of risks associated with the transition towards
a carbon neutral economy. Rising electricity and more broadly energy prices could e.g., cause
negative supply shocks, lead to stranded assets, and be a burden to employment and overall
economic activity in some parts of the economy.
In this study we confirm the negative and heterogeneous impact of rising electricity prices on
firm-level employment and investment in the manufacturing industry.10 The impact is highest
for industries generally regarded as electricity intensive (e.g., chemicals, metals and paper) where
we find employment elasticities between -0.1 and -0.2 and investment elasticities up to -1. We
find the highest (most negative) country-wide elasticity for Belgium. Since different regions have
different levels of specialization in the most affected industries, the impact from rising electricity
prices is also geographically diverse within a country. We find the highest impact on employment
in Southern Germany and Northern Italy.
Furthermore, we introduce a new source of firm-level heterogeneity. Firms that are more
financially constrained, are more responsive (i.e. experience a higher elasticity) towards changes
in electricity prices. These financially constrained firms will lower employment more when
electricity prices rise compared to firms that experience less constraints. It is already established
that firms expected to be more financially constrained react more to monetary policy shocks
as frictions in financial markets amplify the effects of monetary policy on borrowers with lower
access to external financial resources (e.g., Bernanke and Gertler 1995). We now also find
7
Just transition refers to social assistance programs for workers who lost their jobs as a result of environmental
policies, see e.g., Smith (2017).
8
See e.g. Batten et al. (2020) for the relevance of climate change to monetary policy.
9
Speech given at the launch of the COP 26 Private Finance Agenda, 27 February 2020.
10
Whilst the impact on households is not the subject of this study, Green and Knittel (2020) recently showed
that carbon taxes and more broadly climate policy also have an heterogeneous impact on households and could
lead to both progressive and regressive redistribution between households and regions.

ECB Working Paper Series No 2537 / April 2021 6

evidence that firms react more to carbon shocks in the form of rising electricity prices. This
implies the credit channel of monetary policy could have a smoothing effect and absorb shocks
from rising electricity and more broadly energy prices.
To come to these findings we use firm-level financial and employment data for Belgium,
France, Germany, Italy, the Netherlands and the UK for the period 2008 - 2017. This selection
of 6 countries has a comparable level of industrialization of its manufacturing sector and it
includes both large and small countries. We estimate firm-level electricity prices based on
the industry’s firm size distribution. We use these prices as an explanatory variable in our
econometric model of equilibrium labor demand across countries and estimate the electricity
price elasticity of employment when the level of output can adjust as well.
The paper proceeds as follows. The next section explains the empirical estimation method
and the data used. Section 3 presents the main econometric findings and performs a range of
robustness checks. Section 4 discusses the policy implications and section 5 concludes.

2 Empirical data and method

2.1 Methodology

We are interested in the overall impact of changing electricity prices on employment and more
specifically the electricity price elasticity of firm-level labor demand. This elasticity can be
broken down into 2 components: the substitution and the scale effect (Hamermesh 1993). The
substitution effect captures the fact that (under a given level of output) electricity consuming
capital will be substituted for labor when electricity prices increase. The scale effect represents
the reduction in employment driven by lower output when increased electricity prices lead to
increased sales prices and/or lower profitability. We first use a reduced-form, static labor demand
model that allows for multiple input factors (Hamermesh 1993). In section 3.4 we also explore a
dynamic labor demand model. We focus on the impact on firm-level employment unconditional
on output. This implies the overall effect includes substitution effects as well as scale effects.
The model becomes:

empit = αi + µ1 priceit + µ2 wageit + ξXit + it (1)

Where empit , priceit and wageit stand for the natural logarithm of employment (emp), the

ECB Working Paper Series No 2537 / April 2021 7

electricity price (price) and wage for firm i in year t. The advantage of using this log-linear
model is that we can interpret the coefficient µ1 as a price elasticity. Firm-level wage is defined as
gross wage bill divided by the number of employees. Firm-level electricity prices are discussed in
the next section. αi is a firm-level panel fixed effect that accounts for firm-specific characteristics
that impact employment and are constant over time. Xit represents a vector of control variables.
In the most stringent specification this includes the natural logarithm of the firm’s capital stock,
country Ö year and sector Ö year fixed effects.
We also study the impact on firm-level investment. This is determined via a similar model
where (the logarithm of) employment is replaced by (the logarithm of) investment. priceit and
wageit are replaced by their lagged values priceit−1 and wageit−1 to account for the fact that
investment in year t is decided on in year t − 1.

2.2 Data

2.2.1 Background on electricity prices

The electricity cost for the end-user can generally be broken down into three parts that greatly
differ between different consumer profiles and countries. First, the end-user has to pay for the
electricity generation. Electricity is currently traded at electricity exchanges11 where the com-
modity can be bought on a spot or future basis. The share of the actual commodity cost ranges
from less than 50% to greater than 90% of the final price depending on the user profile and
location. Second, the end-user price includes a network cost. Network costs are the charges for
transmitting the electricity via the grid of transmission system operators (TSO) and distribu-
tion system operators (DSO). The breakdown of the transmission market is country specific.
Generally, a DSO manages a medium- to high-voltage grid and a TSO manages a high- and
very high-voltage grid. An end-user will pay a different network charge depending on the volt-
age used. Large consumers are directly connected to the high- or very high-voltage grid and
hence pay reduced or no DSO charges. Network charges are regulated and for industrial users
generally are the smallest part of the end price. Finally, there is a complex system of country
specific tariffs, charges, levies and exemptions thereof. These charges are characterized by a
large variance and in some cases can reach twice the cost of the underlying commodity.
When we refer to electricity price in this paper, this is the final cost towards an industrial
11
Examples are the European Power Exchange (EPEX) and the European Energy Exchange (EEX).

ECB Working Paper Series No 2537 / April 2021 8

 Figure 1: Electricity prices for industrial users

Italy

Germany

UK

Average

Belgium

Netherlands

France

Electricity price in Euro/kWh

0 0.05 0.10 0.15 0.20

2017 2012 2008

Note: Electricity prices (¿/kWh, excl. VAT and other recoverable taxes
and levies) for industrial users with a yearly consumption between 2,000
and 20,000 MWh, ranked based on 2017 price (descending).
Source: Eurostat.

consumer, i.e. the sum of the commodity, network costs and all non-recoverable taxes and levies,
excluding VAT.
Figure 1 shows the electricity prices for 6 European countries as well as the average price
for the year 2008, 2012 and 2017 for a commonly used industrial consumption band. We see
that prices can vary substantially.12 Electricity prices in Italy are almost 3 times higher than in
France.
Figure 2 shows the relative evolution of the same electricity price between 2008 and 2017.
While for France, Germany and the UK, the price has risen 30% to 40%, for Belgium and The
Netherlands, the price was stable or even came down.

2.2.2 Data sources

We obtain information on firm-level employment and financials from Orbis maintained by Bureau
Van Dijk. We include firms in the manufacturing sector of Belgium, France, Germany, Italy, The
Netherlands and the UK for the period 2008 - 2017. Micro firms with less than 10 employees are
excluded as these are generally not well reported in Orbis. Table 1 shows the summary statistics
12
Price differences and changes are driven by differences and changes in the underlying commodity, network
charges and/or taxes/levies and can vary for each consumption band and between different countries. The
importance of taxes/levies and changes thereof has risen over the past decade.

ECB Working Paper Series No 2537 / April 2021 9

 Figure 2: Relative evolution of electricity prices for industrial users

160

150

140
Average
130 UK
France
120 Germany
Italy
110 Belgium
Netherlands
100

90

80 Electricity price (2008 = 100)

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Source: Eurostat

Note: Relative evolution of electricity prices (excl. VAT and other re-
coverable taxes and levies) for industrial users with a yearly consumption
between 2,000 and 20,000 MWh. While on average we see an increase, the
evolution differs substantially between countries.
Source: Eurostat.

of firm-level employment per country. The UK firms that are included in our dataset are skewed
towards larger firms compared to the other countries.
Electricity prices are gathered from the electricity price statistics for non-household con-
sumers from Eurostat (2020a). It reports on a bi-annual basis weighted average prices for six
electricity consumption bands up to 150,000 MWh and for a consumption band above 150,000
MWh. Prices are very well reported for the bands up to 150,000 MWh. Price data for con-
sumption above 150,000 MWh is not available for all countries. Prices reflect the true cost to
the end-user and hence exclude VAT and other recoverable taxes and levies. Yearly prices are
calculated as the average over the 2 semesters.
The largest industrial electricity users, however, have a yearly consumption well over 150,000
MWh. For prices for these heavy users we rely on Deloitte (2018) who reports baseload and
peakload prices13 for Belgium,14 the Netherlands, France and Germany for the period 2013-2018
for 10 consumption bands between 100,000 MWh and 1,000,000 MWh. We use these prices and
how they compare to the Eurostat prices as the basis to extrapolate prices for the consumption
13
We calculate the average as 35% peakload and 65% baseload. This corresponds with baseload hours on
weekdays between 8h00 and 20h00.
14
We calculate the price for Belgium as 70% Flanders and 30% Wallonia.

ECB Working Paper Series No 2537 / April 2021 10

 Table 1: Firm-level employment statistics

Average employment
Country # firms Total employment
Mean Median p10 p90
Belgium 7,268 67 22 11 119 486,083
Germany 61,591 82 27 12 129 5,077,617
France 32,402 81 23 11 137 2,630,678
Italy 86,496 40 17 11 67 3,501,006
The Netherlands 12,031 40 19 11 69 477,088
UK 12,688 164 67 17 270 2,083,127
TOTAL 212,506 67 22 11 110 14,255,598

Note: The number of firms refers to the number of unique firms appearing 1 or more years in the
dataset. The employment statistics refer to the average number of employees over the years the
firm appears in the dataset.

bands not reported by Eurostat. In practice the Deloitte prices reported for Belgium, France
and the Netherlands for consumption of 100,000 MWh are comparable to the Eurostat prices for
the highest consumption bands and then linearly decrease reaching a price that is 20% to 30%
lower for the highest consumption band above 950,000 MWh. We assume prices for lower bands
relate to higher bands in a similar way in Italy and the UK as they do on average in Belgium,
Germany, France and the Netherlands. This brings us to electricity prices for 15 consumption
bands per country. The exact definition of the consumption bands can be found in Appendix
A.
We use the electricity consumption profile per sector for Belgium derived in Bijnens et al.
(2018). Based on firm-level electricity consumption profiles, they derive a consumption matrix
(given in Appendix B). This matrix shows for each manufacturing sector the share of electricity
consumption taken by each of the 15 consumption bands. We assume this consumption profile
to be similar across countries used in our study. This is a realistic assumption since technology
levels in manufacturing (with respect to electricity usage) in the most industrialized countries
in Europe are comparable.
This sectoral electricity consumption profile is used to estimate to which electricity consump-
tion band a firm belongs. We assume that, within a sector, a higher amount of tangible fixed
assets reflects a higher electricity consumption. We therefore rank, across countries, all firms in
a sector based on the average value for tangible fixed assets over the period. This ranking is
matched with the sectoral electricity consumption profile to determine the electricity price band
for each firm. This firm-level price band is subsequently linked with the relevant electricity price

ECB Working Paper Series No 2537 / April 2021 11

valid for the firm’s country in a certain year.

3 Results

3.1 Impact on employment

Table 2 estimates the impact of electricity prices on employment based on the model from
Equation 1. We sequentually include more control variables. Columns (1) includes country Ö
year and country Ö year fixed effects. Column (2) adds the wage and column (3) also controls
for the capital stock. All columns include firm panel fixed effects.

Table 2: Impact on employment

OLS panel regression results estimating Equation 1

(1) (2) (3)

employment employment employment
electricity price -0.0631∗∗∗ -0.0395∗∗∗ -0.0162∗
(0.00632) (0.00730) (0.00705)

wage -0.0641∗∗∗ -0.105∗∗∗

(0.000701) (0.000702)

tangible fixed assets 0.0612∗∗∗

(0.000286)
Firm FE yes yes yes
Country Ö year FE yes yes yes
Sector Ö year FE yes yes yes
N 1062182 836134 834172
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

Note: The elasticity of employment in function of the price of electricity is given by the coefficient for electricity price

and ranges between -0.02 and -0.06.

The coefficient of interest is µ1 from Equation 1 detailing the unconditional elasticity of labor
demand as a function of the electricity price. We find values in the range of -0.02 and -0.06
This implies an electricity price increase of 1% will, on average, reduce firm-level labor demand
0.02% to 0.06%. Cox et al. (2014) find unconditional demand elasticities between -0.06 and
-0.69 depending on the skill levels of the involved labor.
The estimates for the employment – wage elasticity of -0.06 to -0.10 are on the lower side of
the typical range of -0.15 to -0.75 reported by Hamermesh (1993) as we do not make a distinction

ECB Working Paper Series No 2537 / April 2021 12

between low and high skilled labor. In a more recent study Lichter et al. (2015) report, based
on a meta-regression analysis, point estimates for medium-term elasticities of -0.114 to -0.243.
As can be expected, the wage elasticity is higher (more negative) than the electricity elasticity
as wage costs represent a higher share of input costs than electricity for most firms.
The above elasticity is the average for all manufacturing firms in our dataset. Clearly, not all
firms will be impacted by changing electricity prices. Firms and sectors that do not significantly
rely on electricity as an input will only be impacted to a limited extent by changing electricity
prices. Furthermore, since the firm landscape will differ between countries, we also expect a
different impact between countries.
We therefore interact the electricity price (price in Equation 1) with the country the firm
is located and the sector the firm is active in. The detailed regressions results are included in
Table 7 in Appendix C. We use 2 different specifications: a reduced form model (similar to
column (1) from Table 2) and an extended model that controls for wages and tangible fixed
assets (similar to column (3) from Table 2).
Figure 3 shows a graphical representation of the country specific elasticities. We clearly find
a negative elasticity for Belgium, Germany and France with the highest impact in Belgium. For
Italy the elasticity is very close to zero and for the UK the elasticity becomes positive, albeit
not significantly different from zero. For the Netherlands, we find a negative elasticity only
for the reduced form model. The elasticity estimated by the extended model shows a large
confidence interval. A possible explanation of these differences between countries is a different
sectoral composition. Figure 4 now gives the sector specific elasticities. Sectors are defined
based on NACE 2-digit codes. We see a heterogeneous impact of electricity prices on firm-level
employment based on the sector the firm is active in. For the sectors which are generally regarded
as energy or electricity intense (e.g., chemicals, metals, paper) we find a negative elasticity.15 For
some industries, though, we find a positive elasticity. This suggests that the substitution effect,
where machinery is replaced by human labor, outweighs the scale effect for these industries.16
The heterogeneous sectoral impact leads to a heterogeneous geographical impact within
a country since the manufacturing industry is not evenly spread across countries, nor within
countries.17 Figure 5 and Figure 6 highlight this geographical diversity. Figure 5 shows the
15
An additional, though partial, explanation of sector-level differences for the elasticity is the level of competition
in a sector. This is further explored in a later section of this paper.
16
Other studies (e.g., Dechezleprêtre et al. 2017, Dussaux 2020) estimating sector level energy and electricity
elasticities also find positive coefficients for some industries.
17
Note that policy differences between countries that affect employment are absorbed by the country Ö year

ECB Working Paper Series No 2537 / April 2021 13

 Figure 3: Country specific elasticities of employment

Belgium

Germany

France

Italy

Netherlands

UK

-0.3 -0.2 -0.1 0.0 0.1 0.2

Reduced form model Extended model

Note: Graphical representation for the coefficient of pricei t from equation

(1) interacted with a country dummy. Detailed regression results can be
found in Table 7 in Appendix C. The horizontal lines around the point
estimate represent the 95% confidence interval.

ECB Working Paper Series No 2537 / April 2021 14

 Figure 4: Sector specific elasticities of employment

Food
Beverages
Textiles
Wearing apparel
Leather
Wood
Paper
Printing
Coke & petrol
Chemicals
Pharma
Plastics
Minerals
Basic metals
Metals
Computer
Elec. equip.
Machinery
Motor vehicles
Other transport
Furniture
Other
Repair & instal.
-0.3 -0.2 -0.1 0.0 0.1 0.2

Reduced form model Extended model

Note: Graphical representation for the coefficient of priceit from Equa-

tion 1 interacted with a NACE 2-digit sector dummy. Detailed regression
results can be found in Table 7 in Appendix C. The horizontal lines
around the point estimate represent the 95% confidence interval. The
coefficient for the tobacco industry (NACE 12) is omitted for graphical
clarity reasons.

ECB Working Paper Series No 2537 / April 2021 15

Figure 5: Relative employment in sectors where electricity prices have a negative impact on
employment

>40% >9%
35% - 40% 7% - 9%
30% - 35% 5% - 7%
25% - 30% 4% - 5%
20% - 25% 3% - 4%
<20% <3%

(a) Share of electricity sensitive employment (b) Share of electricity sensitive employment
as share of manufacturing employment as share of total employment

Note: Geographical areas defined based on NUTS1 code. Employment figures for 2016.

relative importance of electricity sensitive employment vs. overall manufacturing employment

(Figure 5a) and vs. total employment (Figure 5b).18 Electricity sensitive employment stands
for the employment in the manufacturing sectors for which we found a negative and significant
elasticity in Figure 4, i.e. sectors that experience a negative impact from rising electricity
prices.19 Figure 6 shows the absolute employment in these electricity sensitive sectors.
From Figure 5a we learn that these electricity sensitive sectors represent a high share of
overall manufacturing employment in Germany, the Netherlands, Belgium and Northern Italy.
Compared to total employment, the share of electricity sensitive sectors stands out in Germany
fixed effects. Country differences that affect the sectoral elasticity are not picked up as this requires the estimation
of country – sector specific elasticities.
18
We use the 2016 figures from Eurostat’s regional Structural Business Statistics (sbs r nuts06 r2) that reports
regional employment per NACE 2-digit sector. Missing data is replaced by data from our firm-level database.
Aggregate employment is taken from Eurostat Labor Force Survey (lfst r lfe2en2).
19
The sectors with negative and significant elasticity for both regression specification are tobacco (NACE
12), paper (NACE 17), chemicals (NACE 20), pharmaceuticals (NACE 21), metals (NACE 25), computers and
electronics (NACE 26), machinery and equipment (NACE 28) and equipment repair (NACE 33).

ECB Working Paper Series No 2537 / April 2021 16

Figure 6: Absolute employment in sectors where electricity prices have a negative impact on
employment

>400k
300k - 400k
200k - 300k
100k - 200k
50k - 100k
<50k

Note: Geographical areas defined based on

NUTS1 code. Employment figures for 2016.

ECB Working Paper Series No 2537 / April 2021 17

and Northern Italy (Figure 5b). When we look at absolute employment numbers in these
sectors (Figure 6), 5 regions stand out: Nordrhein-Westfalen, Baden-Württemberg and Bavaria
in Germany and North-West and North-East in Italy. These regions could face substantial job
losses and a high need for job reallocation driven by rising electricity prices.

3.2 Link with financial constraints

In the previous paragraph we showed that higher electricity prices will have a (substantial) neg-
ative effect on employment in certain industries. This also leads to a heterogeneous geographical
impact. In this section we examine to what extent an additional source of firm-level heterogene-
ity, i.e. financial constraints, lead to a heterogeneous response to changing electricity prices.
Over the past decades a large literature has been developed on how financially constrained firms
have larger reactions to monetary policy shocks (e.g., Bernanke and Gertler 1995, Hutchinson
and Xavier 2006) and how it affects corporate investment (e.g., Fezzari et al. 1987).
There are several possible explanations for the relevance of financial constraints. When
electricity prices rise, firms experiencing liquidity issues might have to increase output prices
quicker compared to their less financially constrained competitors.20 This leads to a lower output
and hence lower employment vs. the less constrained competitors. Financially constrained firms
might have fewer possibilities to invest in electricity saving technology. In addition, De Haas and
Popov (2019) showed that CO2 emissions per capita are lower in economies that are relatively
more equity-funded. Stock markets reallocate investment towards less polluting sectors and
push carbon-intensive sectors to develop and implement greener technologies.
Figure 721 below examines if there is a disproportionate impact from rising electricity prices
on firms that experience (possible) financial constraints. Since financial constraints are not
directly measurable, the literature has resorted to proxies or indicators such as age, profitability,
leverage and liquidity or combinations thereof that are easily observed (see e.g., Durante et
al. 2020 for more detail). A shortcoming of firm financial data as a proxy clearly is the fact
that they endogenously adjust to shocks. Age can therefore be regarded as the only exogenous
proxy (Cloyne et al. 2018). Furthermore, unobservable characteristics such as management
skills might well play a role.22 For the purpose of our study we limit ourselves to several proxies
20
A similar mechanism is explained in Amiti et al. (2019) who show that smaller firms have a higher cost pass
through mechanism than larger firms.
21
Detailed regression results can be found in Table 8 (Appendix C.)
22
Chercheye et al. (2020) recently proposed a new measure that recovers financial constraints beyond observable

ECB Working Paper Series No 2537 / April 2021 18

for financial constraints:23

ˆ Age: young firms (less than 10 years) vs. older firms

ˆ Gearing ratio : (total liabilities - equity) / equity

ˆ Interest cover: EBIT / interest paid

ˆ Return on equity (ROE) : (profit before tax + interest paid) / shareholder equity

We control for unobserved, time-invariant firm-level characteristics via firm-level fixed effects.
Figure 7 shows how the elasticity changes based on different measures for financial constraints.
We see that firms experiencing a higher (lower) level of financial constraints do experience a
higher (lower) impact from changing electricity prices. From Figure 7a we learn that young
firms face a higher elasticity than older firms.24 Figure 7b shows that firms with a low level of
gearing, i.e. a low level of debt vs. own equity face a lower elasticity. The more the debt level
is increased, the higher the negative impact on employment stemming from increased electricity
prices. Doubling the debt level increases (makes more negative) the elasticity with ∼1.5 per-
centage points. Figure 7c indicates a similar mechanism. The elasticity becomes smaller (less
negative) when the interest coverage (cash flow vs. interest payments) increases. Doubling the
interest cover decreases (makes less negative) the elasticity with ∼1 percentage point. Figure 7d
finally shows that when profitability, measured via the return on equity, increases, the elasticity
also decreases (becomes less negative).

3.3 Impact on investment

When it comes to investment, there are again several competing channels at play. The pollu-
tion haven hypothesis (McGuire 1982) predicts that firms will rather invest in areas with lower
environmental compliance costs. The Porter hypothesis (Porter and van der Linde 1995), how-
ever, argues that environmental costs trigger innovation in cost-cutting technology. While there
is certainly empirical proof that environmental regulation triggers innovation,25 the literature
suggests that the pollution haven effect outweighs the effects from the Porter hypothesis (see
firm characteristics.
23
Gearing, interest cover and ROE are log transformed.
24
Young firms are defined as firms that exist for 10 years or less.
25
E.g., De Jonghe et al. (2020) recently showed that, when the price of emission allowances is sufficiently high,
the EU ETS did lead to increased emission efficiency of highly polluting firms.

ECB Working Paper Series No 2537 / April 2021 19

 Figure 7: Elasticity of employment and financial constraints

-0.05
-0.02

-0.03
-0.06

-0.04
-0.07
-0.05

-0.08 -0.06

-0.07
-0.09
-0.08

-0.10 Gearing (debt/equity ratio, logarithmic scale)

Brackets mark the 95% confidence interval -0.09 Grey area marks the 95% confidence interval

Young firms Old firms 50% 100% 200% 400% 800%

(a) Elasticity for young vs. old firms (b) Elasticity and gearing

-0.05

-0.03
-0.06
-0.04

-0.07
-0.05

-0.06 -0.08

-0.07
-0.09
-0.08

-0.10
-0.09

Interest coverage (EBITDA/interest paid, logarithmic scale) Return on equity (logarithmic scale)
-0.10 Grey area marks the 95% confidence interval -0.11 Grey area marks the 95% confidence interval

0.25 0.5 1 2 4 8 16 32 1% 2.5% 5% 10% 20% 40% 80%

(c) Elasticity and interest coverage (d) Elasticity and return on equity

Note: Graphs based on the reduced form model, i.e. columns (1), (3), (5) and (7) from Table 8 (Appendix C).

ECB Working Paper Series No 2537 / April 2021 20

Dechezleprêtre and Sato 2017 for an overview). In addition, electricity prices can increase output
prices and reduce output, which can lead to lower investment.

Table 3: Impact on investment

OLS panel regression results estimating Equation 1

(1) (2) (3)

investmentt investmentt investmentt
electricity pricet-1 -0.503∗∗∗ -0.257∗ -0.330∗∗
(0.0878) (0.106) (0.105)

waget-1 -0.603∗∗∗ -0.302∗∗∗

(0.0119) (0.0123)

tangible fixed assetst-1 -0.467∗∗∗

(0.00530)
Firm FE yes yes yes
Country Ö year FE yes yes yes
Sector Ö year FE yes yes yes
N 944857 664441 664441
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

Note: The elasticity of investment in function of the price of electricity is given by the coefficient for electricity price

and ranges between -0.2 and -0.4.

Table 3 gives the impact on investment26 from changing electricity prices. Also for investment
we find an elasticity that is higher (more negative) compared to the effect on employment. For
investment we find a negative elastic between -0.2 and -0.5. This implies that a 10% rise in
electricity prices reduces next year’s investment with 2% to 5%. Our result is line with figures
found in other studies. E.g., Saussay and Sato (2018) find that a 10% increase in the industrial
energy price differential between two countries is associated with a 3.2% increase in the number
of acquisitions of firms or assets located in the countries where the energy price is lower.
Figure 8 also shows the sector specific elasticities. Elasticities range between approx. -1
and +1, which is a broader range than we found for the sector level elasticities of employment
(Figure 4). The confidence intervals are broader as well.27 Nevertheless, the point estimates are
predominantly negative. This indicates a negative impact on investment from rising electricity
prices for most sectors.
26
Investment in year t is calculated from the firm’s financial accounts and is defined as the difference between
tangible fixed assets in year t and year t − 1 plus the amount depreciated in year t.
27
Detailed regression results can be found in Table 9 in Appendix C. Table 9 also includes country specific
elasticities.

ECB Working Paper Series No 2537 / April 2021 21

 Figure 8: Sector specific elasticities of investment

Food
Beverages
Textiles
Wearing apparel
Leather
Wood
Paper
Printing
Coke & petrol
Chemicals
Pharma
Plastics
Minerals
Basic metals
Metals
Computer
Elec. equip.
Machinery
Motor vehicles
Other transport
Furniture
Other
Repair & instal.
-2.0 -1.0 0 1.0 2.0

Reduced form model Extended model

Note: Graphical representation for the coefficient of priceit from Equation

1 interacted with a NACE 2-digit sector dummy. Detailed regression re-
sults can be found in Table 7 in Appendix C. The horizontal lines around
the point estimated represent the 95% confidence interval. The coeffi-
cient for the tobacco industry (NACE 12) is omitted for graphical clarity
reasons.).

ECB Working Paper Series No 2537 / April 2021 22

 Table 4: Impact on employment – dynamic effects
OLS panel and Arellano-Bond regression results estimating Equation 1

(1) (2) (3) (4) (5) (6)

employmentt employmentt employmentt employmentt employmentt employmentt
employmentt-1 0.558∗∗∗ 0.581∗∗∗ 0.551∗∗∗ 0.686∗∗∗ 0.631∗∗∗ 0.602∗∗∗
(0.00153) (0.00165) (0.00162) (0.00942) (0.00846) (0.00825)

electricity pricet -0.0295∗∗∗ -0.0286∗∗ -0.0223∗ -0.0245∗ -0.0238∗ -0.0209+

(0.00811) (0.00908) (0.00883) (0.0107) (0.0113) (0.0109)

waget -0.117∗∗∗ -0.139∗∗∗ -0.206∗∗∗ -0.220∗∗∗

(0.00119) (0.00117) (0.00161) (0.00158)

tangible fixed assetst 0.0424∗∗∗ 0.0372∗∗∗

(0.000398) (0.000521)
Firm FE yes yes yes yes yes yes
Country Ö year FE yes yes yes yes yes yes
Sector Ö year FE yes yes yes yes yes yes
N 379159 294521 293728 290132 229856 229143
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

Note: Regression (1), (2) and (3) are OLS panel regressions; (4), (5) and (6) are Arellano-Bond regres-
sions including 1 lag and treating electricity prices as endogenous. Included sectors are tobacco (12), paper (17),
chemicals (20), pharmaceuticals (21), metals (25), computers and electronics (26), machinery and equipment
(28) and equipment repair (33).

We also examine the impact of financial constraints (Table 10 in Appendix C). While we can
confirm that financial constraints have a direct negative impact on the level of investment, we do
not find that financial constraints amplify the negative employment effect from rising electricity
prices. The coefficients for the interaction terms are not significant. If financial constraints
would have an impact on the investment elasticity, the direct channel (the fact that financially
constrained firms have less access to capital) clearly outweighs this potential impact.

3.4 Dynamic impact on employment

Table 4 shows the results when we include dynamic effects. We focus on the sectors where
we previously found a negative elasticity.28 In our empirical approach so far, we have esti-
mated static labor demand. Static labor demand does not consider the existence of adjustment
costs. Dynamic labor demand, on the contrary, explicitly accounts for the costs associated with
changing the level of employment.
Adjustment costs may arise from either institutional (e.g., firing costs), economic (e.g., hiring
28
These sectors are tobacco (NACE 12), paper (NACE 17), chemicals (NACE 20), pharmaceuticals (NACE 21),
metals (NACE 25), computers and electronics (NACE 26), machinery and equipment (NACE 28) and equipment
repair (NACE 33).

ECB Working Paper Series No 2537 / April 2021 23

costs, training) or technological adjustment obstacles (e.g., capital stock is rather fixed). These
costs may lead to the situation that firms do not change their demand for labor significantly
after an exogenous shock (in our case a change of the electricity price) because the adjustment
costs outweigh the benefits of a change of the level of employment. Furthermore, when a
firm experiences an input price shock, it does not know yet to what extent this shock will be
permanent. These factors make firms merely adjust their workforce slowly and employment will
be rather persistent.
Empirically, dynamic effects are accounted for by including the lagged value of the dependent
variable (employment) as an explanatory variable into the regression (Nickell 1986).
Table 4 retakes the specification from Table 2 and includes the lagged level of employment
(employmentt−1 ) into the estimation. Column (1), (2) and (3) show the results for a standard
panel fixed effects regression. Our estimates include the widest possible set of fixed effects. This
allows to control for unobserved characteristics such as the fact that better managed companies
are larger and hence employ more workers, but are also able to negotiate better electricity prices.
Nevertheless, due to possible remaining issues arising from omitted variables, this fixed effects
estimator might be biased downwards. We hence also include estimates based on Arellano and
Bond’s (1991) method in columns (4), (5) and (6). This is a generalized method of moments
(GMM) method that also controls for reverse causality by instrumenting the lagged levels, as
well as the lagged differences of the endogenous regressors. A firm experiencing a positive
demand shock increases output (and employment) and hence electricity use. The increased
electricity usage might lead to a lower electricity price. We therefore treat the electricity price
as endogenous.
The results from the dynamic model confirm the negative elasticity. We find an average
short run elasticity if -0.02 to -0.03 and a long run elasticity of approx. -0.07.29 This means that
a sustained increase of 10% of the electricity price will, after several years, decrease employment
with 0.6%.
29
The long run elasticity is calculated by dividing the coefficient for electricity pricet by 1 minus the value of
the coefficient for employmentt−1 .

ECB Working Paper Series No 2537 / April 2021 24

 Table 5: Impact on employment – dynamic effects
OLS panel and Arellano-Bond regression results estimating Equation 1

(1) (2) (3) (4)

employment employment ∆ employment ∆ employment
electricity price -0.0763∗∗∗ -0.00520
(0.0136) (0.0137)

∆ electricity price -0.0312∗∗∗

(0.00257)

∆ electricity price (up) -0.0405∗∗∗

(0.00456)

∆ electricity price (down) -0.0201∗∗∗

(0.00519)
N 387418 469195 798427 798427
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001
Note: Regression (1), (2) retake regression (1) from Table 2, but on a split sample where electricity
prices go up (1) and electricity prices come down (2). Column (3) give the results for the first difference
specification from Equation 2 and column (4) gives the results for the same Equation where Deltapriceit is
interacted with a dummy that take the value 1 if the price goes up and 0 otherwise and a dummy that takes the
value 1 if the price comes down and 0 otherwise.

3.5 Robustness checks and additional analysis

3.5.1 Symmetric price effect

So far we have assumed that the price effect on employment is symmetric. This implies that
the size of the negative effect on employment from an electricity price increase is the same as
the size of the positive effect on employment from a similar electricity price decrease. As we
predominantly expect future price increases driven by, among others increasing carbon taxes,
the results found in the previous paragraphs would not be useful if they were predominantly
driven by price decreases.
The empirical analysis of this section therefore differentiates between a price drop and a price
increase. We initially retake the base specification from column (1) in Table 2 and split the data
sample between observations that saw an electricity price increase in year t vs. year t − 1,
this is given in column (1) from Table 5, and observations that experienced an electricity price
decrease, this is given in column (2). From these results we learn that the negative elasticity
found so far is predominantly driven by price increases.
To confirm this finding we specify a reduced form model written in first differences:

ECB Working Paper Series No 2537 / April 2021 25

 ∆empit = µ1 ∆priceit + it (2)

Where ∆empit and ∆priceit stand for empit − empit−1 and priceit − priceit−1 . Taking first
differences controls for all factors that affect firm-level employment that are fixed over time and
is comparable to a panel fixed effect. This specification directly relates a year-on-year change in
the price of electricity with a year-on-year change in employment. The coefficient of µ1 is given
in column 3 of Table 5 and confirms the negative elasticity found in the previous paragraphs.
This specification allows to split the explanatory variable ∆priceit into a variable when the price
goes up (and 0 if the price goes down) and a the same variable when the price goes down (and
0 otherwise). These results are given in column 4 of Table 5 and confirm as well the elasticity
is certainly not predominantly driven by a price decrease.

3.5.2 Sector-level elasticities and competition

Next to the energy intensity of a sector, sector-level competition could also play a role in explain-
ing the differences in elasticities across sectors given in Figure 4. In non-competitive sectors,
producers might be able to pass-through the electricity price increase to their customers. In
such a sector, a producer will predominantly defend its profit, rather than its output (and the
employment associated with a certain level of output). In a very competitive sector, a producer
does not have the ability to rise its output price and needs to absorb a price increase by lowering
its profit. In such a case, employment will not change as the output price and the associated
demand did not change. To account for sector-level competition we calculate the Herfindahl-
index30 for a sector in country in a certain year.31 The Herfindahl-index takes a value from 0
to 1, moving from a huge number of very small firms to a single monopolistic producer.
We use the same specification as for the regressions from Table 2, but interact the electricity
price with the Herfindahl-index. The regressions results are included in Table 6. We indeed
find that a higher Herfindahl-index does increase (makes more negative) the elasticity. The
median value for the index is approx. 0.02 and the 75th percentile value is approx. 0.6. From
the coefficients for electricity price × Herf indahl − index from Table 6 we learn that if the
30
The index is defined as the sum of the squares of the market shares of the firms within the industry. We use
the share of employment of a firm vs. total employment in the sector.
31
We hence regard the level of competition within a sector as country specific. If the level of competition is
rather defined on a European level, i.e. when the market is predominantly global in stead of local, the level of
competition is already accounted for by the country Ö year fixed effects included in the regressions of Table 2

ECB Working Paper Series No 2537 / April 2021 26

level of competition within an industry decreases (i.e. the Herfindahl-index increases) from the
median value to the 75th percentile value, the elasticity increases (becomes more negative) by
0.01 (result from column 1) to 0.02 (result from column 3). We hence do find an impact on
the elasticity from the level of competition within a sector, but this plays only a minor role in
explaining the different sector-level elasticities found in Figure 4.

Table 6: Impact on employment

Role of level of competition in explaining sector-level elasticities

(1) (2) (3)

employment employment employment
electricity price -0.0598∗∗∗ -0.0337∗∗∗ -0.00971
(0.00637) (0.00740) (0.00714)

electricity price Ö Herfindahl-index -0.240∗∗∗ -0.454∗∗∗ -0.508∗∗∗

(0.0508) (0.0847) (0.0820)

wage -0.0641∗∗∗ -0.105∗∗∗

(0.000701) (0.000702)

tangible fixed assets 0.0612∗∗∗

(0.000286)

Herfindahl-index -0.535∗∗∗ -1.075∗∗∗ -1.208∗∗∗

(0.124) (0.213) (0.206)
Firm FE yes yes yes
Country Ö year FE yes yes yes
Sector Ö year FE yes yes yes
N 1062182 836134 834172
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

Note: OLS panel regression estimating Equation (1) where the price of electricity is interacted with the Herfindahl-

index. The Herfindahl-index measures the level of competition within a sector. We calculated the index based on

the employment share of an individual firm vs. the total employment for each NACE 2-digit sector, country, year

combination. The elasticity of employment in function of the price of elasticity increases (becomes more negative)

if the Herfindahl-index increases (i.e. when the level of competition decreases).

ECB Working Paper Series No 2537 / April 2021 27

 Figure 9: Relative evolution of gas prices for industrial users

120

110

100
Average
Netherlands
90 France
UK

80 Italy
Germany
Belgium
70

60

50 Gas price (2008 = 100)

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Note: Relative evolution of gas prices (excl. VAT and other recoverable
taxes and levies) for industrial users with a yearly consumption between
100,000 GJ and 1,000,000 GJ. The evolution of gas prices shows similar
trends across countries.
Source: Eurostat

3.5.3 High energy intensity and the use of gas

In the previous paragraphs (see Figure 4 with sector specific elasticities of employment) we have
not consistently found lower (less negative) elasticities for sectors with high energy intensity.32
Especially for the textiles, minerals and basic metals industries we found close to zero or even
positive elasticities. This is counter-intuitive as these heavy users should be impacted the hardest
when electricity prices increase. Next to the fact that these industries are very competitive,
another possible explanation is that large electricity users can easily switch to gas to generate
electricity in situ or to generate steam used in the production process. While electricity prices
were on the rise over the period 2008-2017, gas prices have come down in the countries we study
(Figure 9). This offers on incentive to substitute electricity by gas.
We analyze the impact of gas prices on employment in the sectors with highest electricity
intensity, i.e. textiles (NACE 13), paper (NACE 17), coke and refined petroleum (NACE 19),
chemicals (NACE 20), rubber and plastics (NACE 22), minerals (NACE 23) and basic metals
(NACE 24). As we do not have the necessary data to construct firm-level gas prices, we base
32
We regard textiles (NACE 13), paper (NACE 17), coke and refined petroleum (NACE 19), chemicals (NACE
20), rubber and plastics (NACE 22), minerals (NACE 23) and basic metals (NACE 24) as high energy intensity
sectors. See e.g. OECD (2020) for sector level energy intensity for the French manufacturing sector.

ECB Working Paper Series No 2537 / April 2021 28

ourselves on Eurostat (2020b) and use yearly, country specific gas prices.33
We use a similar model as Equation 1 where the firm-level electricity price is replaced by the
country-level gas price. Note that since the gas price is country-year specific, we cannot include
country-year fixed effects.
Table 11 in Appendix C gives the results of the econometric estimates. These indicate that
the elasticity of employment as a function of gas prices is around -0.1. This is higher (more
negative) than the elasticity we found for the electricity price. For the analysis of the gas
price, we only include, however, the sectors that have the highest energy intensity. Firms in
other sectors are not very likely to make the investment for in situ electricity production. The
different elasticities for the different sectors are represented in Figure 10.

3.5.4 Firm-level electricity prices

We extrapolated the prices for the very high electricity consumption bands based on Deloitte
(2018), which does not report on prices for the UK and Italy, nor for the period before 2013.
Furthermore, we have allocated an electricity consumption band to each firm based on its amount
of tangible fixed assets vs. other firms in the same industry.
We now change how we construct firm-level electricity prices based on the user profile of
Appendix B. We assume that the electricity price for the consumption bands above band F
(150,000 MWh) are the same as the price for band F. Band F is the highest consumption band
for which Eurostat reports accurate data for all countries.
Table 12 in Appendix C shows the same regressions as Table 2, but with the electricity
price calculated as described above. We find an elasticity estimate that is slightly higher (more
negative) than the elasticity we found initially.

3.5.5 Serial correlation and clustering

Table 13 in Appendix C shows the same regressions as Table 2, but now standard errors are
clustered at the frim-level. Clustering deals with concerns of serial correlation of the error term
in panel time series (Bertrand et al. 2004). It is not illogical to assume that when Equation
1 e.g., underestimates employment for a certain firm in a certain year, it will make a similar
33
We take the average of the 5 consumption bands that are well reported: consumption below 1,000 GJ,
consumption between 1,000 GJ and 10,000 GJ, consumption between 10,000 GJ and 100,000 GJ, consumption
between 100,000 GJ and 1,000,000 GJ and consumption 1,000,000 GJ and 4,000,000 GJ.

ECB Working Paper Series No 2537 / April 2021 29

 Figure 10: Sector specific gas price elasticities of employment

Textiles

Paper

Coke & petrol

Chemicals

Plastics

Minerals

Basic metals

-0.3 -0.2 -0.1 0.0 0.1 0.2

Reduced form model Extended model

Note: Graphical representation for sector-level elasticity of employment

as a function of the price for gas. The horizontal lines around the point
estimate represent the 95% confidence interval.

ECB Working Paper Series No 2537 / April 2021 30

underestimation for all subsequent years. While the confidence levels of the estimates reduce
after clustering, we still find a negative and significant elasticity.

3.5.6 The impact of small firms

We also perform a weighted regression in Table 14. We weigh each firm by the square root
of employment. A traditional regression does not distinguish between small and large firms.
Potentially our initial results could be predominantly driven by small firms that have a relative
unimportant share in aggregate employment. We take the square root as the absolute number
would simply mean only the very large firms are taken into account as firm size shows large
differences. The results remain comparable.

4 Policy discussion

In the previous section we have shown that rising electricity prices lead to a loss of employment
in manufacturing industries, especially in that part of manufacturing that is most reliant on
electrical energy. IMF (2019) estimates that an ambitious climate change scenario (i.e. keeping
global warming at 2 ◦ C) requires a carbon tax of $75 a ton CO2 .
Under such a scenario, the IMF also estimate energy prices will rise considerably. Coal prices
could typically rise by more than 200 percent as coal has a high carbon content. They expect
electricity prices34 to rise more modestly. Furthermore, the increase will vary across countries
depending on the emission intensity of generation. With a $75 a ton CO2 carbon tax, the IMF
expects electricity prices to rise with 2% in France, 18% in Germany, 18% in Italy and 16% in
the UK.35 France will experience the lowest rise as a large part of its power generation is nuclear
based.
Figure 11 shows the negative employment impact from the scenario with a $75 a ton CO2
carbon tax and its country specific impact on electricity prices. We again focus on employment
in electricity sensitive sectors and use the sector-level elasticities from the reduced form model
as shown in Figure 4. We combine the IMF (2019) estimates on the electricity price increase, the
sector-level elasticities and the regional employment numbers in these sectors. We subsequently
34
IMF (2019) discusses the impact of a carbon tax on retail electricity prices. We assume the impact on
industrial electricity prices to be similar. Since the underlying commodity price represents a higher share of the
industrial price, this is likely to be a lower bound for the impact.
35
IMF (2019) does not give an estimate for Belgium and The Netherlands. For our analysis, we assume the
average of the 4 reported countries, i.e. 14%.

ECB Working Paper Series No 2537 / April 2021 31

Figure 11: Negative employment impact on the manufacturing industry from rising electricity
prices associated with a $75 a ton CO2 tax

>0.3% >15,000
0.2% - 0.3% 10,000 - 15,000
0.15% - 0.2% 5,000 - 10,000
0.1% - 0.15% 2,500 - 5,000
0.05% - 25% 1,000 - 2,500
<0.05% <1,000

(a) Affected jobs as percentage of total em- (b) Affected jobs in absolute numbers
ployment

Note: Geographical areas defined based on NUTS1 code. Employment figures for 2016. The total amount of
affected jobs amounts to approx. 150,000.

ECB Working Paper Series No 2537 / April 2021 32

estimate the potential job losses in these electricity sensitive sectors associated with a $75 carbon
tax. We only study the employment impact stemming from rising electricity prices. Via this
method, we estimate the total number of affected jobs in the studied countries at approx.
∼150,000.36 These jobs are either lost or need to be reallocated to other more electricity efficient
firms, industries and/or regions.37
This number, however, is again unevenly spread over different regions (Figure 11) with
Germany (∼80,000 affected jobs) and Italy (∼35,000 affected jobs) experiencing the highest
impact. Figure 11a shows the relative number of affected jobs as a percentage of the total
workforce. We again see substantial regional variation. As expected, there is little impact on
France as electricity prices are not expected to rise substantially. The highest impact can be
found in Southern Germany and Northern Italy. Figure 11b shows the absolute number of
affected jobs. Again the burden is highest for Southern Germany and Northern Italy and the
German region of Nordrhein-Westfalen.
Should one worry about the 150,000 possible job losses in certain parts of the manufacturing
industry in certain regions? Over the past decades, millions of manufacturing jobs have disap-
peared whilst unemployment rates did not rise.38 Whilst the employment impact clearly seems
manageable, there is, however, no need for complacency:

ˆ The recent COVID-19 pandemic led to an unprecedented decline of economic activity and
CO2 emissions with 2020 emissions 4% to 7% lower than previously expected.39 Yet, to
keep global warming at the 1.5 ◦ C or 2 ◦ C temperature targets of the Paris Agreement,
UN (2019) estimates that an emission reduction of respectively 7.6% and 2.7% is needed
each year between 2020 and 2030. A $75 carbon tax, used for our estimations and only
roughly double the current price,40 might well be a lower bound for the carbon tax needed
36
One could argue that the firm-level elasticities we calculate do not hold anymore if all firms experience the
same price shock from e.g., a carbon tax. Here it is important to note that the electricity price impact from a
carbon tax will be highly heterogeneous between countries and firms differ substantially in electricity efficiency.
Furthermore, in the short term the question remains to what extent non-EU countries will also adopt a carbon
tax. In the past “carbon off-shoring” has been limited, but EU ETS carbon prices have also not been very binding
yet.
37
Note that even if the jobs created in green parts of the economy outweigh the jobs lost in energy intensive
manufacturing, this will still require a potentially painful reallocation process as these jobs are unlikely located
in the same area nor for the same skill-set.
38
According to Eurostat’s Labor Force Survey overall employment in the manufacturing industry in Belgium,
France, Germany, Italy, Netherlands and UK declined from 27.5M (1992) to 22M (2008).
39
Speech by Isabel Schnabel, Member of the Executive Board of the ECB, at a virtual roundtable on “Sustain-
able Crisis Responses in Europe” organized by the INSPIRE research network, 17 July 2020.
40
The EU ETS price stands around ¿30/ton CO2 (September 2020) despite the sharp contraction in energy
consumption during the COVID-19 crisis. Bayer and Aklin (2020) estimate that the ETS system reduced CO2

ECB Working Paper Series No 2537 / April 2021 33

 to trigger such a fundamental economic shift. A higher tax could well lead to substantially
higher electricity prices and hence job losses.41

ˆ Job losses in the manufacturing industry have historically been compensated by newly
created jobs in services industries, albeit not necessarily in the same region nor for the
same skill-sets. Goos and Manning (2007) have shown, however, that in the UK there was
growth in employment for the highest-skilled and lowest-skilled services with declining
employment in the middle of the distribution. This is the so called job polarization where
we end up with only “lovely” and “lousy” jobs (Goos and Manning 2007). Clearly not all
workers that shift from manufacturing to services will be better off.42

ˆ The local job losses could well reach beyond the manufacturing industry. Well paid man-
ufacturing jobs indirectly also create a higher demand for local non-tradable services in
the immediate vicinity. Moretti (2010) estimates that 1 manufacturing job in a given city,
creates 1 to 2.5 jobs in non-tradable sectors in the same city. As the number of workers and
the equilibrium wage increase in a city, the demand for local goods and services increases.
Goos et al. (2018) even estimate this effect to be as large as 5 for high-tech jobs.

These factors contribute to the fact that the impact from rising electricity and more broadly
energy prices could well be similar to the “China syndrome”. This refers to the insight from
Autor et al. (2013) who show that different U.S. regions were exposed differently to Chinese
import competition and “rising exposure increased unemployment, lowered labor force partici-
pation, and reduced wages in local labor markets.” Exposure to Chinese competition affected not
only local manufacturing employment but also numerous other factors. In Autor et al. (2014)
they add “earnings losses are larger for individuals with low initial wages, low initial tenure,
and low attachment to the labor force.” Pressure on China exposed industries and regions led
to the fact that a part of the labor force was worse off than before, even many years after the
shock occurred. Possibly, drastically rising electricity prices -driven by increased (or properly
emission with 3.8% between 2008 and 2016, i.e. a reduction far below the levels needed for the Paris agreement
targets. It has to be noted, though, that during 2012-2016 the ETS price was ¿5 - ¿10/ton CO2 .
41
E.g., in a recent study, EnergyVille, a cooperation between Belgian universities, concluded that the Belgian
wholesale electricity price would double in real terms by 2030 driven by an increased carbon price and investment
in clean energy generation.
42
E.g., Bijnens and Konings (2017) argue that in Belgium higher paying manufacturing jobs were predominantly
replaced by lower paying services jobs. In absolute numbers, the so called less knowledge intensive services showed
an increase in employment of 3-to-1 compared to the increase in employment of the knowledge intensive services
during 1997-2013.

ECB Working Paper Series No 2537 / April 2021 34

allocated) environmental costs- will create a similar result in the regions that are most affected.
There clearly will be a need for targeted assistance to firms, workers, and disproportionately
affected communities.

5 Concluding remarks

Increasing the share of renewables in the electricity generation mix is an important lever to
reduce global warming and air pollution. This will require continued and large investment in
clean electricity generation capacity. The costs associated with these investments are generally
borne by the end user and result in a higher electricity price. These investments could go hand
in hand with a carbon tax. A meaningful carbon tax could increase electricity prices with 20%
depending on the carbon emission intensity of the electricity generation mix.
In this paper we have shown that such an increased price of electricity will lead to a reduced
level of employment within manufacturing firms. We estimate the elasticity of employment with
respect to the price of electricity and find values up to -0.2, depending on the industry. This
implies a 10% increase of the electricity price can lead to a 2% reduction of firm-level employment
in the hardest hit industries. These jobs are either lost or need to be reallocated to other firms,
industries and/or regions. To come to this conclusion we analyze over 200,000 firms representing
∼14 million employees in 5 countries over the period 2008-2017.
Manufacturing that is most reliant on electricity as an input factor is not evenly spread across
countries nor regions. Therefore, the impact of rising electricity prices will also be unevenly
spread. Especially Southern Germany and Northern Italy could experience significant job losses
if electricity prices meaningfully rise. If the impact of past trade shocks serves as an example,
the impact of rising electricity prices could have long lasting distributional consequences (similar
to Autor et al. 2013). Although the overall benefits of environment related taxes are clear, some
workers (especially manual workers, Marin and Vona 2019) in some regions could end up worse
off even years after the price rise occurred. The labor market effects will be most negative
countries and regions where labor mobility is insufficient.
To mitigate these heterogeneous effects and ensure a positive public sentiment towards en-
vironment related price increases there is hence clearly a role for fiscal policy makers. In this
paper we also show there is a role for monetary policy makers to cushion the negative employ-
ment effect from rising electricity prices. We find that financially constrained firms react more

ECB Working Paper Series No 2537 / April 2021 35

to rising electricity prices. Doubling the debt level of a firm increases (makes more negative) the
elasticity ∼1.5 percentage points. The credit channel of monetary policy could therefore have a
smoothing effect to absorb shocks from rising electricity prices.

ECB Working Paper Series No 2537 / April 2021 36

References

Amiti, M., Itskhoki, O., & Konings, J. (2019). International shocks, variable markups, and
domestic prices. The Review of Economic Studies, 86(6), 2356-2402.

Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations. The Review of Economic Studies, 58(2),
277-297.

Autor, D. H., Dorn, D., & Hanson, G. H. (2013). The China Syndrome: Local Labor Market
Effects of Import Competition in the United States. American Economic Review, 103(6), 2121-
2168. doi: 10.1257/aer.103.6.2121

Autor, D. H., Dorn, D., Hanson, G. H., & Song, J. (2014). Trade adjustment: Worker-level
evidence. The Quarterly Journal of Economics, 129(4), 1799-1860.

Batten, S., Sowerbutts, R., & Tanaka, M. (2020). Climate change: Macroeconomic impact
and implications for monetary policy. In Ecological, Societal, and Technological Risks and the
Financial Sector (pp. 13-38). Palgrave Macmillan, Cham.

Bernanke, B. S., & Gertler, M. (1995). Inside the black box: the credit channel of monetary
policy transmission. Journal of Economic perspectives, 9(4), 27-48.

Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differences-in-
differences estimates? The Quarterly Journal of Economics, 119(1), 249-275.

Bijnens, G., & Konings, J. (2017). An Enterprise Map of Belgium. KU Leuven - VIVES Policy
Paper.

Bijnens, G., Konings, J., & Vanormelingen, S. (2018). The impact of electricity prices on jobs
and investment in the Belgian manufacturing industry. KU Leuven - VIVES Policy Paper.

Cherchye, L., De Rock, B., Ferrando, A., Mulier, K., & Verschelde, M. (2020). Identifying
Financial Constraints. ECB Working Paper 2420.

Cloyne, J., Ferreira, C., Froemel, M., & Surico, P. (2018). Monetary policy, corporate finance
and investment (No. w25366). National Bureau of Economic Research.

ECB Working Paper Series No 2537 / April 2021 37

Colmer, J., Martin, R., Muûls, M., & Wagner, U. J. (2020). Does pricing carbon mitigate climate
change? Firm-level evidence from the European Union emissions trading scheme. Centre for
Economic Performance Discussion Paper No. 1728.

Cox, M., Peichl, A., Pestel, N., & Siegloch, S. (2014). Labor demand effects of rising electricity
prices: Evidence for Germany. Energy Policy, 75, 266-277.

Dechezleprêtre, A., Lovo, S., Martin, R., & Sato, M. (2017). Does climate change pose a risk to
competitiveness? Global firm-level evidence. Working paper, London School of Economics and
Imperial College London.

Dechezleprêtre, A., & Sato, M. (2017). The impacts of environmental regulations on competi-
tiveness. Review of Environmental Economics and Policy, 11(2), 183-206.

De Jonghe, O., Mulier, K., & Schepens, G. (2020). Going Green by Putting a Price on Pollution:
Firmlevel Evidence from the EU. National Bank of Belgium Working Paper No. 390.

Deloitte (2018). Benchmarking study of electricity prices between Belgium and neighboring
countries.

Deschênes, O. (2012). Climate Policy and Labor Markets. In D. Fullerton & C. Wolfram (Eds.),
The Design and Implementation of U.S. Climate Policy. Chicago: University of Chicago Press.

Dussaux, D. (2020). The joint effects of energy prices and carbon taxes on environmental and
economic performance: Evidence from the French sector. OECD Environment Working Papers
No. 154.

Eurostat (2020a). Electricity price statistics.

Eurostat (2020b). Natural gas price statistics.

Fabra, N., & Reguant, M. (2014). Pass-through of emissions costs in electricity markets. Amer-
ican Economic Review, 104(9), 2872-99.

Fazzari, S., Hubbard, R. G., & Petersen, B. C. (1987). Financing constraints and corporate
investment (No. w2387). National Bureau of Economic Research.

De Haas, R., & Popov, A. A. (2019). Finance and carbon emissions. European Cental Bank
Working Paper 2318.

ECB Working Paper Series No 2537 / April 2021 38

Durante, E., Ferrando, A., & Vermeulen, P. (2020). Monetary policy, investment and firm
heterogeneity. European Cental Bank Working Paper 2390.

Green, T. W., & Knittel, C. R. (2020). Distributed Effects of Climate Policy: A Machine
Learning Approach. MIT Center for Energy and Environmental Policy Research Working Paper.

Goos, M., & Manning, A. (2007). Lousy and lovely jobs: The rising polarization of work in
Britain. The review of economics and statistics, 89(1), 118-133.

Goos, M., Konings, J., & Vandeweyer, M. (2018). Local high-tech job multipliers in Europe.
Industrial and Corporate Change, 27(4), 639-655.

Hamermesh, D. S. (1993). Labor demand. Princeton University Press.

Hafstead, M. A., & Williams III, R. C. (2018). Unemployment and environmental regulation in
general equilibrium. Journal of Public Economics, 160, 50-65.

Henriksson, E., Söderholm, P., & Wårell, L. (2012). Industrial electricity demand and energy
efficiency policy: The role of price changes and private R&D in the Swedish pulp and paper
industry. Energy Policy, 47, 437-446.

Hille, E., & Möbius, P. (2019). Do energy prices affect employment? Decomposed international
evidence. Journal of Environmental Economics and Management, 96, 1-21.

Hintermann, B. (2016). Pass-through of CO2 emission costs to hourly electricity prices in

Germany. Journal of the Association of Environmental and Resource Economists, 3(4), 857-
891.

Hutchinson, J., & Xavier, A. (2006). Comparing the impact of credit constraints on the growth
of SMEs in a transition country with an established market economy. Small business economics,
27(2-3), 169-179.

IMF (2019). Fiscal Monitor: How to Mitigate Climate Change. International Monetary Fund.
Washington.

Kahn, M. E., & Mansur, E. T. (2013). Do local energy prices and regulation affect the geographic
concentration of employment? Journal of Public Economics, 101, 105-114.

ECB Working Paper Series No 2537 / April 2021 39

Kreuz, S., & Müsgens, Fs. (2017). The German Energiewende and its roll-out of renewable
energies: An economic perspective. Frontiers in Energy, 11(2), 126-134.

Lichter, A., Peichl, A., & Siegloch, S. (2015). The own-wage elasticity of labor demand: A
meta-regression analysis. European Economic Review, 80, 94-119.

Lise, W., Sijm, J., & Hobbs, B. F. (2010). The impact of the EU ETS on prices, profits and emis-
sions in the power sector: simulation results with the COMPETES EU20 model. Environmental
and Resource Economics, 47(1), 23-44.

Marin, G., & Vona, F. (2017). The impact of energy prices on employment and environmental
performance: Evidence from french manufacturing establishments. Technical Report, Observa-
toire Francais des Conjonctures Economiques (OFCE).

Marin, G., & Vona, F. (2019). Climate policies and skill-biased employment dynamics: Evidence
from EU countries. Journal of Environmental Economics and Management, 98, 102253.

McGuire, M. C. (1982). Regulation, factor rewards, and international trade. Journal of public
economics, 17(3), 335-354.

Metcalf, G. E., & Stock, J. H. (2020). Measuring the Macroeconomic Impact of Carbon Taxes.
In AEA Papers and Proceedings (Vol. 110, pp. 101-06).

Moretti, E. (2010). Local multipliers. American Economic Review, 100(2), 373-77.

Nickell, S. J. (1986): “Dynamic Models of Labor Demand,” in Handbook of Labor Economics,

Vol(1), ed. by O. Ashenfelter, and D. Card, pp. 473–523, Amsterdam, NL. Elsevier North-
Holland.

Porter, M. E., & Van der Linde, C. (1995). Toward a new conception of the environment-
competitiveness relationship. Journal of economic perspectives, 9(4), 97-118.

Saussay, A., & Sato, M. (2018). The impacts of energy prices on industrial foreign investment
location: Evidence from global firm level data. Centre for Climate Change Economics and Policy
Working Paper 344.

Smith, S. (2017). Just transition: A report for the OECD. Just Transition Centre.

Stiglitz, J. E., Stern, N., Duan, M., Edenhofer, O., Giraud, G., Heal, G. M., ... & Shukla, P. R.

ECB Working Paper Series No 2537 / April 2021 40

(2017). Report of the high-level commission on carbon prices.

UN (2019). United Nations Environment Programme. Emissions Gap Report 2019. UNEP,
Nairobi.

Weche, J. P. (2018). Does green corporate investment crowd out other business investment?.
Industrial and Corporate Change, 28(5), 1279–1295, https://doi.org/10.1093/icc/dty056

ECB Working Paper Series No 2537 / April 2021 41

A Definition of electricity consumption bands

Band Yearly consumption range (in Mega Watt Hour, MWh)

A Below 20
B 20 – 500
C 500 – 2000
D 2,000 – 20,000
E 20,000 – 70,000
F 70,000 – 150,000
G 150,000 – 250,000
H 250,000 – 350,000
I 350,000 – 450,000
J 450,000 – 550,000
K 550,000 – 650,000
L 650,000 – 750,000
M 750,000 – 850,000
N 850,000 – 950,000
O Above 950,000

ECB Working Paper Series No 2537 / April 2021 42

 B Sectoral Electricity consumption profile (% of electricity consumption in certain band)

A B C D E F G H I J K L M N O
10 Food 3% 6% 32% 35% 13% 9% 2%
11 Beverages 1% 3% 28% 51% 16%
12 Tobacco 7% 25% 68%
13 Textiles 3% 8% 46% 15% 12% 15%
14 Wearing Apparel 6% 49% 20% 25%
15 Leather 1% 6% 2% 90%
16 Wood 4% 5% 15% 13% 28% 35%
17 Paper 1% 2% 11% 19% 31% 10% 6% 10% 11%

ECB Working Paper Series No 2537 / April 2021

18 Printing 1% 15% 17% 46% 21%
19 Coke and petrol 1% 6% 30% 63%
20 Chemicals 5% 11% 11% 15% 8% 7% 1% 6% 4% 31%
21 Pharma 1% 10% 22% 2% 32% 33%
22 Rubber and plastic 2% 6% 55% 35% 2%
23 Minerals 1% 2% 13% 22% 28% 30% 3%
24 Basic metals 3% 2% 4% 10% 5% 3% 7% 4% 7% 4% 9% 41%
25 Metals 1% 22% 24% 47% 6%
26 Computer 5% 8% 51% 37%
27 Electrical equip. 5% 7% 46% 43%
28 Machinery 6% 8% 21% 34% 23% 7%
29 Motor vehicles 1% 2% 18% 30% 30% 19%
30 Other transport 3% 2% 37% 58%
31 Furniture 1% 25% 30% 43%
32 Other 3% 18% 11% 29% 39%
33 Repair and installation 1% 18% 18% 35% 27%

43
C Detailed regression results
Table 7: Impact on employment
Country and sector specific elasticities
(1) (2) (3) (4)
employment employment employment employment
wage -0.105∗∗∗ -0.105∗∗∗
(0.000702) (0.000702)

tangible fixed assets 0.0612∗∗∗ 0.0612∗∗∗

(0.000286) (0.000286)

electricity price Ö Belgium -0.183∗∗∗ -0.139∗∗∗

(0.0255) (0.0253)

electricity price Ö Germany -0.129∗∗∗ -0.0501∗

(0.0116) (0.0209)

electricity price Ö France -0.126∗∗∗ -0.0748∗∗∗

(0.0131) (0.0132)

electricity price Ö Italy -0.0206∗∗ 0.00610

(0.00783) (0.00798)

electricity price Ö Netherlands -0.0703∗∗ 0.00262

(0.0267) (0.0787)

electricity price Ö UK 0.0746+ 0.0468

(0.0387) (0.0387)

electricity price Ö Food -0.0453∗∗∗ -0.00133

(0.0114) (0.0122)

electricity price Ö Beverages -0.115∗∗∗ -0.0642+

(0.0332) (0.0344)

electricity price Ö Tobacco -0.291∗ -0.486∗∗

(0.139) (0.153)

electricity price Ö Textiles 0.0529∗∗ 0.106∗∗∗

(0.0189) (0.0199)

electricity price Ö Wearing apparel -0.0227 0.0540

(0.0307) (0.0341)

electricity price Ö Leather 0.0888+ 0.0889

(0.0479) (0.0547)

electricity price Ö Wood 0.0980∗∗∗ 0.158∗∗∗

(0.0202) (0.0234)

electricity price Ö Paper -0.129∗∗∗ -0.0959∗∗∗

(0.0206) (0.0216)

electricity price Ö Printing -0.0137 0.0499∗

(0.0196) (0.0247)

electricity price Ö Coke & petrol -0.121∗ -0.0473

(0.0614) (0.0641)

electricity price Ö Chemicals -0.157∗∗∗ -0.119∗∗∗

(0.0160) (0.0168)

electricity price Ö Pharma -0.167∗∗∗ -0.108∗∗∗

(0.0289) (0.0296)

electricity price Ö Plastics -0.0588∗∗∗ 0.0186

(0.0168) (0.0197)

electricity price Ö Minerals 0.0587∗∗∗ 0.0921∗∗∗

(0.0172) (0.0187)

electricity price Ö Basic metals -0.0339 -0.0198

(0.0219) (0.0229)

electricity price Ö Metals -0.0690∗∗∗ -0.0342∗∗

(0.0105) (0.0131)

electricity price Ö Computer -0.240∗∗∗ -0.168∗∗∗

(0.0194) (0.0233)

electricity price Ö Elec. equip. -0.0456∗ 0.0157

(0.0208) (0.0243)

electricity price Ö Machinery -0.139∗∗∗ -0.0950∗∗∗

(0.0108) (0.0122)

electricity price Ö Motor vehicles 0.0581∗∗ 0.110∗∗∗

(0.0218) (0.0237)

electricity price Ö Other transport -0.0471 -0.0579

(0.0317) (0.0402)

electricity price Ö Furniture 0.00864 0.137∗∗∗

(0.0227) (0.0290)

electricity price Ö Other -0.0873∗∗∗ -0.0361

(0.0183) (0.0222)

electricity price Ö Repair & instal. -0.183∗∗∗ -0.185∗∗∗

(0.0193) (0.0249)
Firm FE yes yes yes yes
Country Ö year FE yes yes yes yes
Sector Ö year FE yes yes yes yes
N 1062182 834172 1062182 834172
r2 0.0274 0.102 0.0278 0.103
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

ECB Working Paper Series No 2537 / April 2021 44

 Table 8: Impact on employment
Link with financial constraints

(1) (2) (3) (4) (5) (6) (7) (8)

employment employment employment employment employment employment employment employment
electricity price -0.0663∗∗∗ -0.0212∗∗ -0.0448∗∗∗ -0.00475 -0.0673∗∗∗ -0.0335∗∗∗ -0.0658∗∗∗ -0.0266∗∗∗
(0.00633) (0.00706) (0.00635) (0.00699) (0.00783) (0.00745) (0.00845) (0.00808)

wage -0.105∗∗∗ -0.143∗∗∗ -0.146∗∗∗ -0.135∗∗∗

(0.000701) (0.000787) (0.000828) (0.000877)

tangible fixed assets 0.0612∗∗∗ 0.0659∗∗∗ 0.0677∗∗∗ 0.0643∗∗∗

(0.000286) (0.000312) (0.000327) (0.000348)

young=1 × electricity price -0.0214∗∗∗ -0.0116∗∗

ECB Working Paper Series No 2537 / April 2021

(0.00364) (0.00382)

electricity price × gearing -0.0144∗∗∗ -0.00807∗∗∗

(0.00113) (0.00123)

electricity price × interest cover 0.00798∗∗∗ 0.00726∗∗∗

(0.000509) (0.000484)

electricity price × ROE 0.00373∗∗∗ 0.00778∗∗∗

(0.00108) (0.00104)

young=1 -0.0952∗∗∗ -0.0752∗∗∗

(0.00844) (0.00885)

gearing -0.0332∗∗∗ -0.0192∗∗∗

(0.00266) (0.00290)

interest cover 0.0229∗∗∗ 0.0218∗∗∗

(0.00120) (0.00114)

ROE 0.0119∗∗∗ 0.0261∗∗∗

(0.00252) (0.00241)
Firm FE yes yes yes yes yes yes yes yes
Country Ö year FE yes yes yes yes yes yes yes yes
Sector Ö year FE yes yes yes yes yes yes yes yes

45
N 1062182 834172 993549 790792 758607 756266 646651 641734
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001
 Table 9: Impact on investment
Country and sector specific elasticities
(1) (2) (3) (4)
investmentt investmentt investmentt investmentt
L.waget-1 -0.301∗∗∗ -0.301∗∗∗
(0.0123) (0.0123)

L.tangible fixed assetst-1 -0.467∗∗∗ -0.467∗∗∗

(0.00530) (0.00530)

L.electricity pricet-1 Ö Belgium -1.842∗∗∗ -1.826∗∗∗

(0.385) (0.374)

L.electricity pricet-1 Ö Germany -0.972∗∗ -1.039∗∗

(0.347) (0.362)

L.electricity pricet-1 Ö France -0.858∗∗∗ -0.810∗∗∗

(0.126) (0.200)

L.electricity pricet-1 Ö Italy -0.251∗ -0.141

(0.106) (0.117)

L.electricity pricet-1 Ö Netherlands 0.184 -0.690

(1.607) (1.680)

L.electricity pricet-1 Ö UK 0.0512 -0.0470

(0.552) (0.540)

L.electricity pricet-1 Ö Food -1.017∗∗∗ -0.537∗∗

(0.156) (0.182)

L.electricity pricet-1 Ö Beverages -1.000∗ -0.943+

(0.470) (0.522)

L.electricity pricet-1 Ö Tobacco -1.089 -3.296

(2.294) (2.292)

L.electricity pricet-1 Ö Textiles -0.403 -0.223

(0.248) (0.286)

L.electricity pricet-1 Ö Wearing apparel -0.656 -0.979+

(0.436) (0.513)

L.electricity pricet-1 Ö Leather -0.583 -0.0423

(0.645) (0.798)

L.electricity pricet-1 Ö Wood -0.234 0.183

(0.298) (0.345)

L.electricity pricet-1 Ö Paper -0.636∗ -0.717∗

(0.306) (0.320)

L.electricity pricet-1 Ö Printing -0.514+ -0.255

(0.294) (0.362)

L.electricity pricet-1 Ö Coke & petrol -0.154 0.256

(0.900) (0.896)

L.electricity pricet-1 Ö Chemicals -0.563∗ -0.457+

(0.232) (0.238)

L.electricity pricet-1 Ö Pharma -0.891∗ -0.940∗

(0.419) (0.414)

L.electricity pricet-1 Ö Plastics -0.804∗∗ -0.381

(0.258) (0.291)

L.electricity pricet-1 Ö Minerals 0.935∗∗∗ 0.952∗∗∗

(0.250) (0.274)

L.electricity pricet-1 Ö Basic metals -0.353 -0.182

(0.316) (0.323)

L.electricity pricet-1 Ö Metals -0.416∗∗ -0.564∗∗

(0.160) (0.194)

L.electricity pricet-1 Ö Computer -0.594+ -0.661+

(0.326) (0.354)

L.electricity pricet-1 Ö Elec. equip. -1.384∗∗∗ -1.162∗∗

(0.332) (0.365)

L.electricity pricet-1 Ö Machinery -0.426∗ -0.331+

(0.167) (0.182)

L.electricity pricet-1 Ö Motor vehicles -0.688∗ -0.376

(0.337) (0.350)

L.electricity pricet-1 Ö Other transport 0.895 1.016

(0.595) (0.623)

L.electricity pricet-1 Ö Furniture 0.360 1.087∗

(0.355) (0.429)

L.electricity pricet-1 Ö Other -0.589+ -0.0570

(0.301) (0.352)

L.electricity pricet-1 Ö Repair & instal. -0.640∗ -0.726+

(0.277) (0.380)
Firm FE yes yes yes yes
Country Ö year FE yes yes yes yes
Sector Ö year FE yes yes yes yes
N 944857 664441 944857 664441
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

ECB Working Paper Series No 2537 / April 2021 46

 Table 10: Impact on investment
Link with financial constraints

(1) (2) (3) (4) (5) (6) (7) (8)

investmentt investmentt investmentt investmentt investmentt investmentt investmentt investmentt
L.electricity pricet-1 -0.481∗∗∗ -0.322∗∗ -0.645∗∗∗ -0.362∗∗∗ -0.420∗∗∗ -0.232∗ -0.573∗∗∗ -0.321∗∗
(0.0882) (0.105) (0.0893) (0.106) (0.0912) (0.111) (0.0990) (0.119)

L.waget-1 -0.302∗∗∗ -0.343∗∗∗ -0.333∗∗∗ -0.385∗∗∗

(0.0123) (0.0127) (0.0134) (0.0147)

L.tangible fixed assetst-1 -0.467∗∗∗ -0.493∗∗∗ -0.495∗∗∗ -0.497∗∗∗

(0.00530) (0.00551) (0.00576) (0.00610)

-0.128∗∗ -0.0872

ECB Working Paper Series No 2537 / April 2021

young=1 × L.electricity pricet-1
(0.0469) (0.0579)

L.electricity pricet-1 × L.gearingt-1 0.107∗∗∗ 0.0105

(0.0162) (0.0188)

L.electricity pricet-1 × L.interest covert-1 -0.0204∗∗∗ 0.00104

(0.00601) (0.00709)

L.electricity pricet-1 × L.ROEt-1 -0.0293∗ -0.0121

(0.0127) (0.0149)

young=1 -0.300∗∗ -0.191

(0.113) (0.135)

L.gearingt-1 0.0486 -0.140∗∗

(0.0390) (0.0445)

L.interest covert-1 0.0891∗∗∗ 0.122∗∗∗

(0.0145) (0.0167)

L.ROEt-1 0.0282 0.0591+

(0.0305) (0.0348)
Firm FE yes yes yes yes yes yes yes
Country Ö year FE yes yes yes yes yes yes yes
Sector Ö year FE yes yes yes yes yes yes yes

47
N 944857 664441 898959 639263 885773 609331 737827 515824
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001
 Table 11: Impact on employment from changes in gas price
OLS panel regression results estimating Equation 1 with electricity price replace by gas price

(1) (2) (3)

employment employment employment
gas price -0.105∗∗∗ -0.126∗∗∗ -0.118∗∗∗
(0.0112) (0.0127) (0.0123)

wage -0.0533∗∗∗ -0.0982∗∗∗

(0.00143) (0.00144)

tangible fixed assets 0.0662∗∗∗

(0.000609)
Firm FE yes yes yes
Country Ö year FE no no no
Sector Ö year FE yes yes yes
N 229558 190146 189374
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

Note: only firms from textiles (NACE 13), paper (NACE 17), coke and refined petroleum (NACE 19), chemicals
(NACE 20), rubber and plastics (NACE 22), minerals (NACE 23) and basic metals (NACE 24) included.

ECB Working Paper Series No 2537 / April 2021 48

 Table 12: Impact on employment – robustness checks
Prices for electricity consumption bands F and above kept constant

(1) (2) (3)

employment employment employment
electricity price -0.0853∗∗∗ -0.0644∗∗∗ -0.0467∗∗∗
(0.00711) (0.00857) (0.00828)

wage -0.0642∗∗∗ -0.105∗∗∗

(0.000702) (0.000702)

tangible fixed assets 0.0612∗∗∗

(0.000286)
Firm FE yes yes yes
Country Ö year FE yes yes yes
Sector Ö year FE yes yes yes
N 1062184 836135 834172
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

Table 13: Impact on employment – robustness checks

Standard errors clustered the firm level

(1) (2) (3)

employment employment employment
electricity price -0.0631∗∗∗ -0.0395∗∗∗ -0.0162
(0.00937) (0.0109) (0.0104)

wage -0.0641∗∗∗ -0.105∗∗∗

(0.00282) (0.00301)

tangible fixed assets 0.0612∗∗∗

(0.00106)
Panel yes yes yes
Country Ö year FE yes yes yes
Sector Ö year FE yes yes yes
N 1062182 836134 834172
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

ECB Working Paper Series No 2537 / April 2021 49

 Table 14: Impact on employment – robustness checks
Weighted regression

(1) (2) (3)

employment employment employment
electricity price -0.0460∗∗ -0.0386∗ -0.00918
(0.0173) (0.0184) (0.0169)

wage -0.0912∗∗∗ -0.142∗∗∗

(0.00688) (0.00739)

tangible fixed assets 0.0821∗∗∗

(0.00297)
Firm FE yes yes yes
Country Ö year FE yes yes yes
Sector Ö year FE yes yes yes
N 1062182 836134 834172
Standard errors in parentheses
+ ∗ ∗∗ ∗∗∗
p < 0.10, p < 0.05, p < 0.01, p < 0.001

ECB Working Paper Series No 2537 / April 2021 50

Acknowledgements
This paper should not be reported as representing the views of the European Central Bank (ECB) nor the National Bank of Belgium
(NBB). The views expressed are those of the authors and do not necessarily reflect those of the ECB nor the NBB.

Gert Bijnens
National Bank of Belgium, Brussels, Belgium; KU Leuven, Leuven, Belgium; email: [email protected]

John Hutchinson
European Central Bank, Frankfurt am Main, Germany; email: [email protected]

Jozef Konings
KU Leuven, Leuven, Belgium; University of Liverpool Management School, Liverpool, United Kingdom;
Nazarbayev University Graduate School of Business, Astana, Kazakhstan; email: [email protected]

Arthur Saint Guilhem

European Central Bank, Frankfurt am Main, Germany; email: [email protected]

© European Central Bank, 2021

Postal address 60640 Frankfurt am Main, Germany
Telephone +49 69 1344 0
Website www.ecb.europa.eu
All rights reserved. Any reproduction, publication and reprint in the form of a different publication, whether printed or produced
electronically, in whole or in part, is permitted only with the explicit written authorisation of the ECB or the authors.
This paper can be downloaded without charge from www.ecb.europa.eu, from the Social Science Research Network electronic library or
from RePEc: Research Papers in Economics. Information on all of the papers published in the ECB Working Paper Series can be found
on the ECB’s website.
PDF ISBN 978-92-899-4537-0 ISSN 1725-2806 doi:10.2866/302949 QB-AR-21-028-EN-N