Exchange Rate Effect on Carbon Credit Price via Energy Markets

by

Jongmin Yu, Ph.D Candidate

University of Illinois at Urbana-Champaign, Mumford Hall, MC-710, 1301West Gregory Drive,

Urbana, IL 61801, United States

Phone: 1-217-417-2216/ Email: yu35@illinois.edu

Mindy L. Mallory, Assistant Professor

University of Illinois at Urbana-Champaign, Mumford Hall, MC-710, 1301West Gregory Drive,

Urbana, IL 61801, United States

Phone: 1-217-372-4163/ Email: mallorym@illinois.edu

Abstract

This paper discusses the effect of currency exchange rates on the carbon market. We illustrate this

effect using the European Union Emission Trading Scheme (EU-ETS), which primarily uses two

substitutable energy inputs for the generation of electricity: coal and natural gas. The European coal

market is directly driven by global coal markets that are denominated in USD, whereas, natural gas is

mainly imported from Russia and is denominated in Euro. The impulse response functions of a Structural

Vector Autoregression (SVAR) model demonstrates that a shock in the EU/USD exchange rate is

delivered through the channel of energy substitution between coal and natural gas, and therefore,

produces influences on the carbon credit market.

Introduction

This paper shows the effect of a currency exchange rate on the carbon prices when substitutable

energies are denominated in different currencies. We illustrate this effect in the European Union Emission

Trading Scheme (EU-ETS) because this carbon market has the longest history and the largest trading

volume among carbon credit markets. The European Union primarily uses two substitutable fossil energy

inputs to generate electricity: coal and natural gas.1 The power generation industry has been the main

player in the EU-ETS because it consumes around 60% of the total credit allocation. Electricity

generators can substitute back and forth from coal to natural gas relatively easily, depending on the

relative prices of coal and natural gas and the demand for electricity. Burning coal generates almost twice

as much carbon dioxide emissions compared as natural gas in producing the same amount of energy. So,

making CO2 emissions more costly can influence the decision between fuels at the margin. Since the EU

ETS was introduced, power generators always select the fuels they are going to use by considering the

cost of using each fuel plus buying emission credits each day. This has been called fuel-switching, and it

affects the carbon emission allowance market. It is therefore reasonable to assume that electricity

producers decide their position in either the OTC or the emissions exchange markets by reacting to

changes in the relative prices of two competing fossil fuels. Theoretically, this selection decision can be

made on a daily basis because daily changes in fuel and carbon prices change the relative electricity

generation cost benefit by comparing marginal generation costs (Koenig, 2010), and power generators

also have traded in the carbon market on a daily basis (Bunn and Fezzi, 2007). This effect of relative

energy prices to the demand of the EU carbon market has been confirmed in numerous previous studies.

(Delarue, Voorspools and D’haeseleer, 2007; McGuinness and Ellerman, 2008; Alberola, Chevallier and

Chèze, 2008)

The European coal market is directly driven by global coal markets that are denominated in USD;

whereas, natural gas is denominated in Euro (Polański and Winkler, 2008; Timera Energy, 2011; Monthly

Bulletin of BlueNext ). Also, we know that the substitution between coal and natural gas affects the price

of carbon credits because coal generates more CO2 emissions than natural gas. Therefore, asymmetric

exposure to exchange rate risk of the two energy inputs changes the Euro denominated price spread

between coal and natural gas. Since exchange rate risk influences the relative price of coal to natural gas,

the substitution between the clean and dirty inputs affects the price of carbon. For example, if the Euro

depreciates against the USD, ceteris paribus, this increases the price of coal relative to the price of natural

gas in Euro. This causes energy substitution away from coal and toward natural gas thereby lowering

CO2 emissions; hence, carbon credit prices would fall.

To date, most studies of the carbon market in the European Union (EU) discuss how supply and

demand are affected by various factors such as temperature, emission cap policy, other energy markets,

and market regulations (Chevallier, 2009). Chevallier (2009; 2010 a,b,c) are the exceptions, and these

consider macro factors like the bond market, the stock market, and business cycle indices to understand

the EU carbon market using time-series models. However, Chevallier did not consider the effect of

international energy markets through currency exchange rates. On the other hand, Frank and Garcia (2010)

identified currency exchange rates as indicators of commodity prices. Their result shows that agricultural

1 According to the EU energy report (2009), 29.4% of electricity was generated from coal, and 22.6% was generated

from natural gas in 2007. (27.8% from nuclear, 15.6% from renewable, and 3.3% from oil)

goods are becoming more dependent on exchange rates. Thus, we propose a research to identify the effect

of exchange rate on the carbon credit price through the fuel switching process.

Data

In this paper we use spot EU/USD exchange rates, European Emission Allowances (EUAs)

carbon credit prices, natural gas and coal prices. Our price data are nominal and weekly starting from

January 2007. The EU/USD currency exchange rates are spot rates from European Central Bank (ECB);

the carbon credit price (€/ton of CO2) is the nearby futures price of EUAs obtained from the

Intercontinental Exchange; the natural gas price (€/MMBtu) is obtained from the Zeebrugge Hub spot

price, which is the major short-term natural gas trading market in continental Europe, and the coal price

(€/ton) is from the ARA spot price, which is the major imported coal in northwest Europe.

Figure 1. Nominal prices of coal, natural gas and carbon credit prices in Euro

Figure 1 shows recent prices of EUA, natural gas, and coal in Euro. It is intuitive that these two

energy markets help us understand the carbon market because the relative prices of coal and natural gas

have been identified as a driver of EUA prices. Commonly, besides the price spread between coal and

natural gas, the overall price level of the two energy prices also affects EUA prices. Since higher energy

prices represent higher energy demand, higher energy consumption also indicates higher carbon dioxide

emissions and EUA prices, and vice versa. Figure 1 shows this standard co-movement of prices in EUA,

natural gas and coal; until mid-2008, there was an overall increase of those commodity market prices.

However, after mid-2008 and beginning in 2009, the prices dropped together steeply because of the

recession triggered by the default of Goldman Socks in U.S.

0

5

10

15

20

25

30

35

0

20

40

60

80

100

120

140

10

/10

/20

06

4/2

8/2

00

7

11

/14

/20

07

6/1

/20

08

12

/18

/20

08

7/6

/20

09

1/2

2/2

01

0

8/1

0/2

01

0

2/2

6/2

01

1

9/1

4/2

01

1

4/1

/20

12

Gas Coal EUA

Pri

ce o

f C

oal

an

d G

as

(un

it: E

uro

)

Pri

ce o

f EU

A (

un

it: E

uro

)

However, we may be able to identify different patterns of the carbon and energy markets’ co-

movement since 2009. The prices of energy are gradually recovering as time goes, but the price of EUA is

not catching up the other energy prices. Even between the energy markets, the price of natural gas

increases more slowly than the price of coal, which is an also different phenomenon compared to the

common idea of a highly correlated movement between energy markets before financial crisis. Since

there has been a consensus that energy market trends are the most important price driver because of

carbon emission demand, it would be worthwhile to look for other reasons why only the carbon market

has been decreasing compared to other energy market.

Figure2. Exchange rates

In figure 2, before mid-2008, the Euro appreciates against USD, which corresponds with the

bullish trend in energy sectors in figure 1; there was high demand for energy with the economy boom and

the Euro currency appreciates along with strong economic activities. After mid-2008, the Goldman

Sock’s bankruptcy and the European government’s budget deficit impacted the global financial markets,

which kept the Euro/USD exchange rate relatively high and unstable. Correspondingly, energy markets in

figure 1 also show a bearish trend compared to the period before mid-2008. The descriptive statistics for

all time-series data used in this paper are summarized in Table 1.

Table 1. Descriptive statistics

EU/USD Coal Natural gas EUA

Mean 0.72 96.46 48.23 18.06

Median 0.72 93.75 46.60 16.03

Maximum 0.84 192.50 95.00 31.42

Minimum 0.63 51.60 17.40 9.43

Std. Dev 0.04 30.35 16.15 4.98

Skewness -0.02 0.61 0.62 0.79

0.6

0.65

0.7

0.75

0.8

0.85

10

/10

/20

06

4/2

8/2

00

7

11

/14

/20

07

6/1

/20

08

12

/18

/20

08

7/6

/20

09

1/2

2/2

01

0

8/1

0/2

01

0

2/2

6/2

01

1

9/1

4/2

01

1

4/1

/20

12

Lehman bro. files for bankruptcy (15.Sept.2008)

Request for financial support - Greece (23.Apr.2010) - Ireland (21.Nov.2010)

Portugal (6.Apr.2011)

Kurtosis -0.43 -0.11 0.01 -0.46

Observation 249 249 249 249

Table 2 contains results of the Augmented Dickey Fuller test for stationarity, and we conclude

that each data series is I(1). Because all series are integrated of the same order, 1, it is possible to have

cointegration among those non-stationary variables. Here we determine whether a vector error correction

model or a vector autoregression model is appropriate. In table 3, we perform the Johansen trace test on

the system of 5 variables, but we did not find any cointegrating relationships among the 5 variables.

Table 2. Augmented Dickey Fuller unit root test

Model Lags 5% statistics Result

EUA CNT 0 -3.41 -2.14 Non-stationary

Coal CNT 2 -2.86 -2.01 Non-stationary

Natural gas NCNT 1 -1.94 0.02 Non-stationary

EU/USD CNT 0 -2.86 -2.19 Non-stationary

d. EUA NCNT 0 -1.94 -9.50 Stationary

d. Coal NCNT 2 -1.94 -5.27 Stationary

d. Natural gas NCNT 0 -1.94 -8.93 Stationary

d. EU/USD NCNT 0 -1.94 -14.95 Stationary

CT: constant and trend, CNT: constant and no trend, NCNT: no constant and no trend.

We start estimating the VAR model with 3 lags. Once the full model is estimated with a given

number of lags, we applied the sequential parameter elimination is applied based on Top-Down procedure

with AIC criteria. With the regressors that have survived from the sequential elimination process, the

model is estimated again with only selected regressors. Then, we finally get the information criteria

indexes as we could see in table 3 for given number of lags. We repeat this process for various numbers

of lags and compared those to get the optimal model.

Table 3. Johansen trace test

Rank Trace statistic 5% critical Value

0* 34.28 53.94

1 19.63 35.07

2 8.04 20.16

3 2.49 9.14

Table 4 shows the somewhat perplexing results: the more lags we have, the lower the AIC and the

higher the SIC. Both AIC and BIC have pro and cons. Having too small a number of lags based on SIC is

likely to result in autocorrelation of the estimated VAR model, and having large number of lags based on

the AIC steeply increases the number of estimated coefficients of the model. Furthermore, this paper

seeks to examine the effect of short-term macro shock within a couple of months, so we limit our trial to

up to 5 lags. In short, we choose to use the Hannan-Quinn criterion which indicates 4 lags in the

differenced VAR model, which is optimal.

Table 4.

# of lags of endogenous various AIC SIC HQ

5 -9.071854e+00* -8.496865e+00 -8.840254e+00

4 -9.011207e+00 -8.638558e+00 -8.861125e+00*

3 -8.981066e+00 -8.652377e+00 -8.848703e+00

2 -8.940307e+00 -8.711654e+00 -8.848229e+00

1 -8.871413e+00 -8.714214e+00* -8.808109e+00

Model Specification

We will use a structural VAR model since we want to look at the interdependency of exchange

rates and other energy commodities across time, and the structural VAR model has been used to identify

the market linkage in many previous studies. We estimate a first differenced VAR model to test the

hypotheses developed earlier. In this paper, we need to estimate a structural VAR after estimating reduced

VAR because the energy substitution occurs at a daily frequency; the instantaneous effect found in

weekly data (that encompasses 5 trading days) will represent the daily energy substitution of the industry.

We can identify the instantaneous causality by estimating the matrix B below by estimating the structural

VAR model.

More explicitly,

[

⁄

] [

]

[

]

[

]

[

] ( ) ( )

Here, the number of lags, J, to include in this model will be determined based on AIC, BIC or HQC. We

repeat this process until we get the optimal number of lags, which minimizes the information criteria.

Then, we apply sequential parameter elimination based on Top-Down procedure with AIC criteria. With

the regressors that have survived from the sequential elimination process, the model is estimated again

with only selected regressors. As diagnostic tests, we will perform Jarque-Bera tests to ensure normality,

ARCH-LM for heteroscedasticity, and the Portmanteau test for autocorrelation.

As proposed by Sims (1980), VAR model estimation has routinely orthogonalized shocks by

using a lower triangular matrix B. Since the Choleski decomposition of the matrix B imposes an ordering

restriction of the variables, we identify the instantaneous interdependencies among the carbon market,

energy markets, and exchange rates in the VAR model.

[

] [

]

The terms in matrix represent the unrestricted parameters values and 0 terms enforce the

contemporaneous effect of endogenous variable to be zero. In terms of ordering, the 1st row means the

exchange rates are not affected by any other endogenous terms at least contemporaneously. This setup

follows Akram (2009) since usually macro variables reflect the overall economy of a certain country so

that those would work more independently from energy markets and carbon market. Thus, the energy

markets from the 2nd

to 3rd

row of matrix are subject to the effects of exchange rate shocks. The 4th row,

the carbon credit market, is the last term after the energies since the scale of the energy markets are bigger

than the carbon credit market; the price transmission from energy to carbon market would be more

acceptable than the other way around. Hence, the carbon market is the last of the ordering. Between the

coal and natural gas markets in the EU, the coal market has been regarded more susceptible to

international shocks since the main currency in coal trading is U.S. dollar. However, the natural gas

market is based on the domestic currency, i.e. Euro, so that the market has been regarded to be isolated

from currency markets. We will see later in this paper how this different exposure of energy markets to

international currency markets affects to the carbon credit price determination.

Empirical Results

(1) Impulse response function

In the following, we run impulse response functions based on the previously described differenced

SVAR model. 95% confidence intervals were estimated with respect to the impulse response to different

shocks. Even though the size of the data is not small (n=249), we used the bootstrapping of 1500

iterations suggested by Hall (1992) to obtain the more reliable confidence interval. The impulse response

which describes effects of macro policy variables on energy and carbon markets appear in figure 3.

Figure 4 shows the accumulative effects on impulse response. Both figures demonstrate the effect of

energy substitution. Figure 3 shows the marginal impulse response of coal, natural gas, and carbon market

to a shock of one positive standard deviation of the exchange rate (depreciated Euro against USD), and

figure 4 shows the accumulative response, respectively.

Figure 3. Impulse to Euro/USD exchange rate; response to coal (1st row), natural gas (2

nd row), and

carbon credit (3rd

row).

Figure 4. Accumulative impulse to Euro/USD exchange rate; response to coal (1st row), natural gas (2

nd

row), and carbon credit (3rd

row).

In the coal market, the first row of figure 3 and figure 4, a depreciated Euro against the USD leads to an

initial increase in the coal price. Currently, the European coal market has been dominated by global

markets which use the USD. Since the Euro/USD exchange rate impacts the purchasing power of coal, the

depreciated Euro drives increases the domestic coal price in Euro increases, even though the price based

on the USD is at a standstill. Figure 3 shows the marginal positive effect and figure 4 confirms that the

positive accumulative effects also persist.

In the natural gas market, the second rows of figure 3 and figure 4 show unclear results. There is no

immediate response from the Euro/USD exchange rate shock, and the accumulative effect is also not

critically different from the mean of 0. The reason is that the European natural gas market is not exposed

to the Euro/USD exchange risk. The biggest supplier of natural gas to the EU has been Russia, and this

constitutes 41% of total imports (EU-energy report, 2010). Based on the agreement between the EU and

Russia, natural gas supply contract has opted for the Euro currency instead of the USD (Occasional paper

series, European Central Bank, 93-2008).

In the carbon market, the third row of figure 3 and figure 4, a depreciated Euro against the USD decreases

carbon credit prices. Figure 3 shows the marginal negative effect and figure 4 also confirms that the

negative accumulative effects also persist. This occurs due to the energy substitution effect resulting from

relative energy price changes. A depreciated Euro increases the price of coal relative to natural gas, which

induces less consumption of coal. Since natural gas is a less carbon-intensive fossil fuel than is the case

for coal, the depreciated Euro again the USD would lower both the demand for carbon dioxide emission

and the demand for emission credit.

Now let us examine the interactions among coal, natural gas and the carbon market. Again, the

substitution effect from the relative price difference between coal and natural gas has been regarded as the

main price driver of carbon pricing. There would be more energy demand for the cheaper fossil fuel

source. For example, if coal price in Euro increases due to the depreciated Euro, coal would be consumed

less than would be the case for natural gas. Less consumption of coal leads to less CO2 emissions so that

the demand on CO2 emission credit would decrease, which would lower the price of carbon credits. On

the contrary, if natural gas prices increase due to use of the appreciated Euro, coal would be consumed

more than natural gas. More consumption on coal increases CO2 emissions so that the demand for CO2

emission credits would rise, which would increase the price of carbon credits. Figure 5 shows the

marginal impulse responses and figure 6 shows the accumulative effects in a comparative manner.

Figure 5. Impulse to coal (1st column), natural gas (2

nd column) and carbon credit (3

rd column); response

to coal (1st row), natural gas (2

nd row) and carbon credit (3

rd row)

Figure 6. Accumulative Impulse to coal (1st column), natural gas (2

nd column) and carbon credit (3

rd

column); response to coal (1st row), natural gas (2

nd row) and carbon credit (3

rd row)

The first columns of figure 5 and figure 6 show the effect of coal price shock and subsequent

responses of the natural gas market and the carbon market. A shock of one positive standard deviation of

the coal price leads to an initial increase in the natural gas price. Energy prices are highly correlated, so

the positive response of natural gas to the shock of coal is not surprising. However, the response of the

carbon credit price is not critically different from 0. This is the case because energy users can mitigate the

effect of a coal price shock on the carbon market by shifting their energy demand, so the total demand

effect on emissions is not certain. For example, the rise in coal price is not directly transmitted to the

carbon price because energy users may replace expensive coal with cheaper natural gas with less CO2

emission demands. In an opposite manner, a decrease in coal price is not directly transmitted to the

carbon price because energy users may choose to use cheaper coal instead of expensive natural gas, which

induces more CO2 emissions.

However, the second column regarding the effect of a natural gas price shock contrasts the effect

of the price shock on coal. A shock of one positive standard deviation in the coal price leads to an

increase in the natural gas price, which also can be explained as energy prices correlation. In a manner

that is contrary to the effect of the coal price shock on carbon price, a price shock in natural gas generates

a response in carbon credit prices that is significantly different than zero. A natural gas price shock makes

energy users replace natural gas with coal, which inspires more emissions needs and increases the carbon

price. In this case, the energy substitution effect makes the price of carbon credit correlate in a positive

manner with the price of natural gas. For example, an increase in natural gas prices increases carbon

prices because energy users may replace expensive gas with cheaper coal that have more CO2 emission

demands. Likewise, a decrease in the price of natural gas lowers the carbon price because energy users

may use cheaper gas instead of expensive coal, which has less CO2 emission needs.

The third column shows the effect of the carbon market on the other energy markets. A shock of one

positive standard deviation in the carbon price does not result in a significant price change in coal, but

leads to a significant price increase for natural gas. In principle, high emissions costs proxies for high

emissions demand, which is expected to increase both energy prices. However, it is more likely to

produce a higher demand effect on natural gas than coal because energy users can save on compliance

costs by using less carbon-intensive energy.

(2) Forecast error variance decomposition of shocks

The following figures decompose the contribution of different structural shocks to price fluctuations

over the course of 50 weeks. The forecast error variance decomposition (FEVD) displays the percentile

contribution of four different sources of shocks: exchange rates, coal prices, natural gas prices and carbon

prices. Each row shows the coal, natural gas and carbon market, respectively.

The FEVD generally confirms the causality result obtained from the IRF analysis. The EU/USD

exchange rate has the highest explanatory power regarding the price of coal, the second highest

explanatory power regarding the price of carbon credits, and almost negligible explanatory power

regarding the price of natural gas. This is because the price of coal is converted from USDs to Euro when

purchased from the international coal market. Thus the nominal coal price in Euro is directly affected by

the Euro/USD exchange rate. In the case of the carbon market, the Euro/USD exchange rate indirectly

affects carbon prices via coal market because the price spread between coal and natural gas affects the

carbon price. On the other hand, European natural gas market is unglued from the international energy

market so the exchange rate has little impact on natural gas prices.

With regard to interactions between the carbon market and the different types of energy, the FEVD

results remain consistent with the IRF, which demonstrates the energy substitution effect and the positive

correlation between substitutable energies. These results allow us to ascertain that coal is more relatively

exogenous than natural gas and carbon credits, and about 20% of the forecast error in natural gas is

explained by coal. On the other hand, carbon credit is the most endogenous variable in the sense that it is

explained by all of the other variables. Natural gas explains about 25% of the forecast error in carbon

credit as the most important price driver. These results are consistent with the assumption of the ordering

restriction in Choleski decomposition.

Figure 7. FEVD based on SVAR model with coal(1st row), natural gas(2

nd row) and carbon(3

rd row).

Conclusion

This paper discusses the link between carbon pricing and exchange rates. Our paper verifies that

depreciated (appreciated) local currencies can cause low (high) carbon prices through the energy

substitution mechanism. The exchange rate depreciation shock on the carbon market is due to non-

homogeneous effects on three commodities. Our empirical analysis shows that the depreciation in Euro

leads to a price increase for coal, a neutral price change for natural gas, and a price decrease for carbon

credits. This occurs because all three commodities have different degrees of exposure to exchange rate

risks. Coal has been traded in USDs, natural gas has been imported mainly from Russia and contracted in

Euro, and carbon price has been determined by the price spread between coal and natural gas. Then, the

substitutability between coal and natural gas serves as the key component that determines the carbon price

in an asymmetric manner.

If the European debt crisis persists and the contagion spreads to other European countries, we can

expect the Euro currency to become weaker and for exchange rate Dollars to again increase. This makes it

more probable that the spread between coal and natural gas in Europe market will increase, and this

makes industries prefer natural gas to coal along with their demand for carbon dioxide emission. If

exchange rates increase due to the potential risk of bankruptcies or moratoriums in many European firms,

then this may magnify the slump of carbon markets. In conclusion, the carbon market may show bearish

movements because of the effect of depreciated Euro, although energy markets are recovering. This

monetary policy aspect may explain the de-trending of carbon markets away from energy markets.

Furthermore, we can generalize our results such that the different types of energy dependency in

the international markets may characterize the effects of exchange rate on carbon market in different

ways. If the European coal market is also separated from the international market and not denominated in

USDs, this may invalidate the effect of exchange rate shocks on the carbon market.

References

Alberola, E., Chevallier, J., and Chèze, B., 2008. Price drivers and structural breaks in European carbon

prices 2005-2007 . Energy Policy 36(2), 787-797.

Akram, Q. Farooq, (2009). "Commodity prices, interest rates and the dollar," Energy Economics,

Elsevier, vol. 31(6), pages 838-851, November.

Bunn, D. and C. Fezzi (2007), Interaction of european carbon trading and energy prices. Fondazione Eni

Enrico Mattei Working Paper 123.

Chevallier J (2009) Carbon futures and macroeconomic risk factors: a view from the EU ETS. Energy

Econ 31:614–625

Chevallier J (2011) Macroeconomics, finance, commodities: interactions with carbon markets in a data-

rich model. Econ Model 28:557–567

Chevallier J (2011) Evaluating the carbon-macroeconomy relationship: Evidence from threshold vector

error-correction and Markov-switching VAR models. Econ Model (forthcoming).

oi:10.1016/j.econmod.2011.08.003

Chevallier J (2011) A model of carbon price interactions with macroeconomic and energy dynamics.

Energy Econ (forthcoming). doi:10.1016/j.eneco.2011.07.012

Delarue E, Voorspools KR, D’haeseleer WD (2008) Fuel switching in the electricity sector under the EU

ETS: review and prospective. J Sol Energy Eng 134:40–58

Frank, J., and Philip Garcia. 2010. “How Strong are the Linkages among Agricultural, Oil, and Exchange

Rate Markets?” Proceedings of the NCCC-134 Conference on Applied Commodity Price Analysis,

Forecasting, and Market Risk Management. St. Louis, MO.

Koenig, P., 2011. Modelling Correlation in Carbon and Energy Markets. Cambridge Working Papers in

Economics 1123, Faculty of Economics, University of Cambridge.

Market Observatory for Energy report (2009) Europe’s Energy Position; markets and supply. European

Commission

McGuinness M, Ellerman AD (2008) CO2 abatement in the UK power sector: evidence from the EU ETS

trial period.Working paper 2008-04, Centre for Energy and Environmental Policy Research, MIT,

USA

Polański and Winkler (2008) Russia, EU Enlargement and the Euro. Occasional paper series. No93-2008.

European Central Bank

Timera Energy, http://www.timera-energy.com/commodity-prices/exchange-rate-impacts-on-european-

power-and-gas-markets/