에너지경제연구
에 너 지 경 제 연 구
제12권 제1호 12
권 제 1 호
한 국 자 원 경 제 학 회·
에 너 지 경 제
Price Transmission Channels of Energy and Exchange Rate on Food Sector: A Disaggregated Approach based on Stage of Process /
Dae Heum Kwon·Won W. Koo EU ETS의 장단기 가격결정요인 분석 /
백정호·김현석
북한 광물자원 현시대칭비교우위 분석 / 김세영·구영완
한국과 일본의 산업부문 에너지 소비에 대한 LMDI 요인분해 분석 / 박정욱·김수이
탄소배출저감을 위해 수입 화석연료에 부과하는 적정 조세에 관한 연구 / 이덕만
알뜰주유소 전환으로 인한 자영주유소의 휘발유가격 인하효과 분석 / 정준환·이지연·김형건
Korean Energy Economic Review
KOREAN RESOURCE ECONOMICS ASSOCIATION
Korean Energy Economic Review
2013. 3
March 2013 Volume 12, Number 1
Price Transmission Channels of Energy and Exchange Rate on Food Sector: A Disaggregated Approach based on Stage of Process /
Dae Heum Kwon and Won W. Koo
The relationship between allocations of carbon allowances and carbon prices: the case of EU ETS /
Jungho Baek and Hyun Seok Kim
An Analysis on the Revealed Symmetric Comparative Advantage of North Korea’s Mineral Resources /
Sae Young Kim and Young Wan Goo
LMDI Decomposition Analysis for Industry Energy Consumption of Korea and Japan /
Jungwook Park and Suyi Kim
Optimal Taxation for Reducing Carbon Emitted from the Consumption of Fossil Fuel Imported /
Dug Man Lee
The Conversion of an Unbranded Gas Station into a Thrifty Gas Station: Its Impact on Gas Prices /
Joonhwan Jung and Jiyon Lee and Hyunggun Kim
2 0 1 3
· 3
에너지경제연구
제12권 제1호
Korean Energy Economic Review
에너지경제연구
제12권 제1호Price Transmission Channels of Energy and Exchange Rate on Food Sector: A Disaggregated Approach based on Stage of Process ··· 1
Dae-Heum Kwon, Won W. Koo
EU ETS의 장단기 가격결정요인 분석 ··· 25 백정호, 김현석
북한 광물자원 현시대칭비교우위 분석 ··· 45 김세영, 구영완
한국과 일본의 산업부문 에너지 소비에 대한 LMDI
요인분해 분석 ··· 67 박정욱, 김수이
탄소배출저감을 위해 수입 화석연료에 부과하는
적정 조세에 관한 연구 ··· 105 이덕만
알뜰주유소 전환으로 인한 자영주유소의 휘발유가격
인하효과 분석 ··· 125
정준환, 이지연, 김형건
Price Transmission Channels of Energy and Exchange Rate on Food Sector:
A Disaggregated Approach based on Stage of Process
Dae-Heum Kwon*․Won W. Koo**1)
Abstract
The recent concurrent surges of food and energy prices renew our interest on the vulnerability of food system to sudden changes in the energy sector. Unlike previous studies focusing on the impacts of a single energy price on food sector, this study explores such dependency utilizing various food and energy prices classified by stage of processing.
Based on the method proposed by Toda and Yamamoto(1995) and Dolado and Lütkepohl(1996) of Granger causality tests, we identify how the movements of the exchange rate and the various energy prices affect on the food prices from farmers to consumers.
Key Words : Energy price, Exchange rate, Food price, Price transmission Channels, Processing stages of food. Granger causality test JEL Codes：Q40, Q10
* Assistant Professor, Department of Economics, Chosun University(first and corresponding author). dhkwon@chosun.ac.kr
** Professor, Department of Agribusiness and Applied Economics, North Dakota State University. Won.Koo@ndsu.edu
Ⅰ. Introduction
The United States(U.S.) economy heavily depends on energy consumption.
The overall food system, from farmers to consumers, is not an exception1). Despite such dependency of the food system on energy consumption, the energy price usually had been relatively low enough not to raise public concerns, except the oil price surge in the early 1970s and 1980s. For example, oil prices were roughly $20 per barrel during most of the 1990s and energy costs were smaller share of production costs and consumer budget than 1970s-80s(Bernanke, 2004).
However, the heavy dependency of the food system on the energy sector is demonstrated through the consequence vulnerability of the food system to sudden changes in the energy sector in several occasions. Especially the recent food inflation phenomenon raised public concerns with the almost simultaneous surge of energy price as the main factor for such food price hike(e.g., Abbott, Hurt, and Tyner 2008) and several studies have been conducted in this respects based on assumed causal structures. However, causal structures need to be
1) According to the Earth Policy Institute, the U.S. food system uses over 10 quadrillion Btu(10,551 quadrillion Joules) of energy each year, which is as much as France’s annual energy consumption and comprises about 10% of the total U.S. energy consumption (Murray, 2005). In addition to the agricultural production, the overall food processing and distribution systems also heavily rely on the energy sector. For example, Heller and Keoleian(2000) suggest that almost 41% is used in the processing and distribution system (14% goes to food transport, 16% to processing, 7% to packing, and 4% to food retailing) and 32% is used to home refrigeration and preparation, while approximately 21% of the total energy consumption in the food system is used in agricultural production.
inductively inferred for empirical investigations for such arguments and the detailed information of linkage channel among energy and food prices need to be empirically identified for figuring out possible countermeasures2).
Various studies are conducted to examine the effect of energy price fluctuation on the economy. In the literature on the nexus between oil price and macroeconomy, numerous studies provide empirical evidences that rising oil prices slow economic growth and stimulate inflation, and also identify various channels through which energy price fluctuations affect the overall economy (e.g., Brown and Yücel 2002, Jones, Leiby, and Paik 2004 and references in there). For example, Hamilton(1983) demonstrated that oil price shock had proceeded all but one recession in the terms of Granger causality3).
In the study regarding the impact of energy price on the agricultural sector, Hanson, Robinson, and Schluter(1993), among others, use Computable General Equilibrium(CGE) model to analyze the direct and indirect cost linkages among three energy sectors(crude oil and gas, petroleum refining, and electric and gas utilities) and various agricultural production and processing sectors. Under the scenario that the high oil price results in a depreciation of the dollar, which in turn stimulates agricultural exports4), they show that the rising energy and agricultural prices result in the increased consumer food prices. In the investigation on price linkage of oil and commodity prices, Baffes(2007) find that the price indexes for fertilizer and food commodities exhibit the highest pass-through of oil price changes among the various non-energy commodity 2) The disaggregate approach taken this study can be important if we are to prepare
countermeasure plans for each processing stage for the possibile price hike.
3) Hamilton(1983)'s such findings made a definitive contribution to expanding researchers’
attention to the entire period beyond several occasions of oil price shocks.
4) Hanson, Robinson, and Schluter(1993) use such scenario as a plausible assumption without empirical investigation. Thus we still need the empirical study for the detailed transmission channels among energy and food prices with the exchange rate.
indexes, based on annual data from 1960 to 2005.
In this study, we aim to identify the transmission mechanism of energy prices on overall food prices. For that purpose, three kinds of energy prices and four different kinds of food prices are adopted based on the stage of process(SOP) system classified by the Bureau of Labor Statistics(BLS). The main motivation to utilize the SOP system is that diverse kinds of energy can affect the overall food system at various stages in a number of different ways5).
Our approach is distinct from previous literature in several ways. and may contributes the literature on price linkages between energy and food sectors First, while previous studies focused on the single measure of energy price such as the crude oil price, this study uses diverse energy prices classified by the SOP system of producer price index(PPI) incorporating crude, intermediate, and finished stages. By using more disaggregated information than previous literature, we can obtain more detailed information on the energy-food price transmission mechanism.
Second, we explore the effects of energy price on the overall food sectors, covering processing and distributional systems not only the agricultural production, focused in the previous studies. The SOP system is further extended to integrate consumer price index(CPI) for food at home to examine the recent food inflation phenomenon at retail level. The use of broad food prices from crude, intermediate, finished, and retail stages allow us to investigate whether the recent food inflation is derived by the cost-push(e.g., Engel 1978) or
5) For example, the 34% of energy use in agricultural production is directly consumed as diesel and gasoline by farm vehicles for planting, tilling, and harvesting. On the other hand, the 35% is indirectly used in the form of fertilizer and pesticides, which are manufactured from natural gas and petroleum(Murray 2005). Beyond agricultural production, food processing and packing sectors use 23% of various kinds of energy consumed in the food system as the proliferation of processed food with small packages requires various energy consumptions.
demand-pull mechanism(e.g., Granger, Robinsons and Engle 1986).
Third, our analysis incorporates the exchange rate in analyzing the relationship between energy and food prices. Under the global economy, it is plausible that the high oil price can result in depreciation of the U.S. dollar, which in turn stimulate agricultural exports and hence boost food prices(e.g.
Hanson, Robinson, and Schluter 1993). It is also observed that the depreciation of the U.S. dollar is one of key factors contributing to the recent food inflation (e.g. Abbott, Hurt, and Tyner 2008). By incorporating the exchange rate, we can further investigate whether the exchange rate affects the energy price due to denomination effect of the U.S. dollar(e.g., Zhang, Fan, Tsai, and Wei 2008) or vice versa because of the impact of the energy price on current account (e.g., Chen and Chen 2007).
Finally, we utilize more robust method of the Granger causality test than the usual multivariate VAR approach used in the previous studies by adopting the method proposed by Toda and Yamamoto(1995) and Dolado and Lütkepohl (1996)(TYDL). The recent time-series studies(e.g., Yamada and Toda 1998, Giles and Mirza 1999, and Clarke and Mirza 2006) demonstrate robustness of the TYDL approach over a wide range of stationary, near-integrated, and cointegrated systems, compared to some drawbacks of the vector error correction model(VECM) or fully modified VAR methods, when the research objective is not detecting the presence/absence of unit roots or possible long-run(cointegrating) relationships.
In the following Empirical Procedure section, the Toda and Yamamoto(1995) and Dolado and Lütkepohl(1996)(TYDL) method is discussed with the description of data used in this study. In Empirical Results section, empirically inferred transmission channels from energy to food with exchange rate is provided.
The results are discussed in terms of the price transmission mechanism within the
food system and linkages of energy prices and exchange rate. The final section provides summary of overall findings with some concluding remarks for the future study.
Ⅱ. Empirical Procedure
1. Econometrics Considerations
The Granger causality(Granger 1969) is the most common concept for causality analysis in literature. For example, Hamilton(1983) established relationships between energy price and macroeconomic variables based on the Granger non-causality(GNC) test. There have been three approaches to implement GNC test, depending on time-series properties of variables: a VAR model in the level data(VARL), a VAR model in the first-differenced data (VARD), and a vector error correction model(VECM).
However, the non-stationary properties such as unit roots and cointegration can result in statistical complications for testing GNC. Under some conditions, the VARL can involve a singular covariance matrix that may result in a non-standard asymptotic null distribution(e.g., Toda and Phillips, 1993) and the Least Square regression involving variables with unit roots may give rise to a spurious regression(e.g., Granger and Newbold 1974). When the series are cointegrated, the VARD may be misspecified as potential causality from the long-run relationship and thus some forecastability or Granger causality from one variable to the other is ignored(Engel and Granger 1987).
In this respect, the VECM is frequently used when cointegration is suspected.
However, (i) GNC test in VECM involves the nonlinearity, where represents cointegrating vector and captures the speed of adjustment to such long-run relationship, and (ii) the asymptotic distribution of test statistics can be non-standard and may involve nuisance parameters unless the data meet the certain rank condition of submatrices in the cointegration space, which is not always satisfied under the null hypotheses(Toda and Phillips 1993)6).
To address this issue, Toda and Yamamoto(1995) and Dolado and Lutkepohl (1996)(TYDL) demonstrate that (i) given the nonstandard asymptotic properties of the test statistics are due to the singularity of the asymptotic distribution of the LS estimator, the main issue is to find alternative, which result in a nonsingular asymptotic distribution of the relevant estimator to overcome the complicated nonstandard limiting properties and (ii) the singularity in a nonstationary system can be removed by fitting a augmented VARL model. Its order exceeds the true order by the highest degree of integration in the system as follows:
(5)
where is the true lag length, is the maximal order of integration among variables in the system, represents to stack the row of a matrix in a 6) Toda and Phillips(1993) suggest a sequential test procedure involving non-stationary, cointegration, and rank condition of certain submatrices in the cointegration space.
However, such pretesting strategy has unknown overall properties with generally low statistical power, can leave the possibility to chose the inappropriate model for GNC test and lead to the misleading conclusion for GNC. Such possibilities are demonstrated by several simulation studies(e.g., Yamada and Toda 1998, Giles and Mirza 1999, Clarke and Mirza 2006).
column vector, is the appropriate selection vector corresponding to a specific GNC hypothesis and is vector of exchange rate and disaggregated energy and food prices based on the SOP system7).
Although there exist efficiency and power loss by augmenting extra lags, recent simulation studies(e.g., Yamada and Toda 1998, Giles and Mirza 1999, Clarke and Mirza 2006) demonstrate that (i) the TYDL method is better control the type I error probability than other methods based on the VARL, VARD, and VECM, (ii) the power loss in the TYDL approach is relatively minor for moderate and large sample sizes, and (iii) the TYDL approach results in a consistent performance over a wide range of systems, including stationary, near-integrated, and cointegrated systems, even for the mixed integrated systems.
Consequently, the recent time-series literature(e.g., Yamada and Toda 1998, Giles and Mirza 1999, Clarke and Mirza 2006) recommends to use the TYDL approach, when the research objective is not detecting the presence(or absence) of unit roots or possible long-run(cointegrating) relationships.
In this study, we aims to infer detailed transmission channels among energy and food prices based on the Granger causality test of the possible cointegrated VAR models with I(0) /I(1) variables of food and energy prices with exchange rate. Thus the mentioned advantages of TYDL model over other Granger causality test models is utilized in this study, given that this study does not pursue to investigate the possible long-run cointegrating relationships,
7) TYDL further prove that (iii) the hypothesis can be tested based on asymptotic distribution by using modified Wald statistics while ignoring the coefficient matrix of the augmented lag in the estimated equation, which is a zero matrix by assumption, and (iv) it is valid to use the commonly used lag length selection procedure such as the general-to-specific method, based on sequential Likelihood Ratio(LR) test.
2.
Data DescriptionTo trace the impacts of various energy prices on the food prices at different stages of process, we collect several monthly price indexes based on the SOP system from January 1998 to July 2008: the PPI indexes of crude energy goods(denoted by CE), intermediate energy goods(IE), finished energy goods (FE), crude foodstuffs and feedstuffs(CF), intermediate foods and feeds(IF), and finished consumer foods(FF), and Consumer Price Index of food at home(HF).
According to BLS, the coverage of each index is as follows: crude petroleum, natural gas, coal, etc. for CE; diesel fuel, industrial natural gas, commercial electric power, etc. for IE; gasoline, residential natural gas, residential electric power, etc. for FE; wheat, corn, soybeans, fluid milk, etc.
for CF; flour, prepared animal feeds, fluid milk products, etc. for IF; pork, dairy products, processed fruits and vegetables, etc. for FF. The CPI index for food at home represents the food price at retail level, which encompasses the similar product coverage of PPI index for the finished food. All data are seasonally adjusted and log transformed. The real effective exchange rate variable(ER) from International Monetary Fund(IMF) is also incorporated to investigate the claimed nexus of exchange rate with energy and/or food prices as discussed.
Ⅲ. Empirical Results
1.
Preliminary AnalysisThe main objective of this study is to understand how various energy prices affect the food prices at different stages of process, not vice versa. In addition, to avoid the multicollinearity problem among three kinds of energy prices and to address the degree-of-freedom issue in the multivariate Granger causality test framework, we develop three cases of granger causality test model among ER, CF, IF, FF, FF, HF, and one of CE, IE, and FE for each case. The case I incorporates relationship between the crude energy price and the crude, intermediate, finished food price indexes and exchange rate variable. And the case II(and III) encompasses the relationship between the intermediate(finished) energy price index with the common variables of the food prices at various stages of process and the exchange rate.
Following Toda and Yamamoto(1995), the general-to-specific method, based on sequential Likelihood Ratio(LR) test, is applied to determine appropriate lag length. For the case I, the hypothesis test of reduction of lag length from 3 to 2 results in a LR test statistic of 53.56 with a p-value of 0.03, while those from 2 to 1 are 85.10 and 0.00, respectively. Diagnostic statistics of the Lagrange Multiplier(LM) test for the absence of auto-correlation in residual show that the p-value of LM test for order 1(and 2) are 0.32(and 0.05) for the
two lag length VAR specification. For the case II(and III), the LR test statistic is 47.00 with p-value 0.10(45.87 with 0.13) for lag length reduction from 3 to 2, while those from 2 to 1 are 87.35 and 0.00(83.43 and 0.00), respectively.
The p-values of LM tests against order 1 and 2 are 0.20 and 0.13(0.49 and 0.14) for two lag length specification in the case II(and III). These results suggest that lag length of two is appropriate for the subsequent analyses without concern for the autocorrelation problem for all the three cases.
2.
Price Transmission Mechanism within the Food SystemBased on the above preliminary results, the Granger non-causality(GNC) tests are conducted based on the TYDL method, using the two lag length specification and assuming maximum integration order() of one. The modified Wald statistics and p-values are reported in Tables 1, 2, and 3 for cases I, II, and III, respectively. The identified causal flows in Granger sense in Tables 1, 2, and 3 are summarized in the corresponding Figures 1, 2, and 3.
In addition, the overall causal structure we might draw from the overall results of three cases is recapitulated in Figure 4. In Figures 1-4, each(and dotted) arrow represents the identified causal flow at least 5%(and 10%) significant level in Granger sense, given that the p-value of less than 5%
indicates rejection of the null hypothesis of Granger non-causality at the 95%
confidence level.
The identified price transmission mechanism within the food system is quite robust with regard to the variations among the three cases. The crude(and finished) food price Granger causes the intermediate food(home food) price at
the 1% significance level. Despite the absence of causal link between the intermediate food and the finished food prices, the causal flow from crude to finished food prices and causal relationship from intermediate to home food prices connect the overall cost-push mechanism at least 6.1% significance level8).
Such results can be explained by the facts that the intermediate stage is defined as residuals after defining the crude and finished stage(BLS 2008). As Gaddie and Zoller(1988) pinpoint, part of output at a given stage of process can be used by stages of process beyond the next sequential stage of process (skip mechanism in SOP system), since the complicated industrial relationships preclude the clear division of goods into three stages. Considering such aspects, the non-robust results of the demand-pull mechanism from the intermediate to crude food prices can be also explained, given that causal relationship is statistically significant only at the 9.9% significant level in the case I.
8) Although the Granger causality is the most common concept for causality analysis in literature and used in this study, there exist conceptual difference between the philosophical notion based on manipulation and the statistical concept based on predictability and thus we need to be cautions against over-interpreting the empirical results based on the Granger causality concept. For example, the causal direction form crude(and finished) to intermediate(home) food prices inferred here only suggest the cost-push mechanism in terms of Granger causality concept.
Table 1. Modified Wald Test Result for Case I Dependent
Variable CE CF IF FF HF ER
CE -
-
1.354 0.508
1.096 0.578
2.150 0.341
2.822 0.244
5.464*
0.065*
CF 8.661**
0.013**
- -
4.629*
0.099*
0.762 0.683
0.454 0.797
1.576 0.455
IF 1.998
0.368
1.532***
0.003***
- -
1.134 0.567
1.551 0.460
0.563 0.755
FF 2.222
0.329
5.610*
0.061*
3.660 0.160
- -
0.834 0.659
1.825 0.402
HF 3.445
0.179
2.834 0.242
11.566***
0.003***
28.921***
0.000***
- -
1.894 0.388
ER 3.720
0.156
0.348 0.840
0.771 0.680
0.096 0.953
0.452 0.798
- - Note: 1) CE, IE, FE, CF, IF, FF, HF, and ER denote the PPI index of crude energy and
intermediate energy, finished energy, crude foodstuffs and feedstuffs, intermediate foods and feeds, and finished consumer foods, and CPI indexes of food at home, and Exchange Rate, respectively.
2) The asterisks of ***, **, and * represent statistically significant at 1, 5, and 10
%, respectively. For each cell, first and second number is and corresponding p-value, respectively.
Figure 1. Price Transmission Mechanism for Case I
Note: see note in Table 1 and refer Table 1 for a specific significant level. Each(and dotted) arrow represents the identified causal flow at least 5%(and 10%) significant level in Granger sense.
Table 2. Modified Wald Test Result for Case II Dependent
Variable IE CF IF FF HF ER
IE -
-
0.362 0.834
0.720 0.698
0.533 0.766
0.174 0.916
1.583 0.453 CF 10.730***
0.005***
- -
4.311 0.116
0.295 0.863
0.511 0.774
1.257 0.533 IF 5.535*
0.063*
10.976***
0.004***
- -
0.744 0.689
2.013 0.365
0.769 0.681 FF 6.286**
0.043**
6.173**
0.046**
3.846 0.146
- -
1.073 0.585
0.246 0.536 HF 5.442*
0.066*
1.960 0.375
10.619***
0.005***
27.967***
0.000***
- -
1.718 0.424 ER 8.807**
0.012**
0.333 0.846
1.030 0.597
0.136 0.934
0.213 0.899
- - Note: 1) CE, IE, FE, CF, IF, FF, HF, and ER denote the PPI index of crude energy and
intermediate energy, finished energy, crude foodstuffs and feedstuffs, intermediate foods and feeds, and finished consumer foods, and CPI indexes of food at home, and Exchange Rate, respectively.
2) The asterisks of ***, **, and * represent statistically significant at 1, 5, and 10
%, respectively. For each cell, first and second number is and corresponding p-value, respectively.
Figure 2. Price Transmission Mechanism for Case II
Note: see note in Figure 1 and refer Table 2 for a specific significant level.
Table 3. Modified Wald Test Result for Case III Dependent
Variable FE CF IF FF HF ER
FE -
-
0.493 0.782
0.703 0.704
0.386 0.824
0.703 0.704
1.520 0.468
CF 6.685**
0.035**
- -
4.396 0.111
0.302 0.860
0.591 0.744
1.281 0.527
IF 4.878*
0.087*
10.608***
0.005***
- -
0.805 0.669
2.364 0.307
1.019 0.601
FF 3.972
0.137
5.724*
0.057*
3.927 0.140
- -
0.910 0.635
1.662 0.436
HF 7.473**
0.024**
2.321 0.313
11.787***
0.003***
28.485***
0.000***
- -
2.470 0.291
ER 3.264
0.196
0.212 0.899
0.871 0.647
0.087 0.958
0.178 0.915
- - Note: 1) CE, IE, FE, CF, IF, FF, HF, and ER denote the PPI index of crude energy and
intermediate energy, finished energy, crude foodstuffs and feedstuffs, intermediate foods and feeds, and finished consumer foods, and CPI indexes of food at home, and Exchange Rate, respectively.
2) The asterisks of ***, **, and * represent statistically significant at 1, 5, and 10
%, respectively. For each cell, first and second number is and corresponding p-value, respectively.
Figure 3. Price Transmission Mechanism for Case III
Note: see note in Figure 1 and refer Table 3 for a specific significant level
Figure 4. Overall Inferred Price Transmission Mechanism
Note: see note in Figure 1 and refer Table 1-3 for a specific significant level.
The identified cost-push transmission mechanism summarized in Figure 4 is consistent to several studies(e.g., Boyd and Brorsen 1985, Goodwin and Holt 1999, and Goodwin and Harper 2000), which show the price transmission mechanism from farm to wholesale to retail market for a specific commodity such as pork or beef. On the other hand, our results, based on broad food price indexes beyond a specific commodity, identify more detailed price transmission channels and reveal the cost-push mechanism along the sequential stages of process with some skip mechanisms. These finding can contribute to understand the recent food inflation phenomenon, by providing empirical evidences of the notion that the increase of farm commodity prices is large enough to affect retail food prices, despite the small portion of agricultural commodity values in retail food prices(e.g., Abbott, Hurt, and Tyner 2008).
3. Linkages of Energy prices and Exchange rate
The effects of the energy price and exchange rate on the food price at the retail level are another common aspect to the various analyses for the recent food inflation, as discussed. In this respect, our results provide empirical evidences such that (i) the crude energy price Granger causes the crude food price at about 1% significance level(Table 1), (ii) the intermediate energy price causes the crude food price at the 1% significance level, the finished food price at the 5% significance level, and the intermediate and home food prices at the 10% significance level(Table 2), (iii) the finished energy price leads the crude and home food prices at the 5% significance level and the intermediate food price at the 10% significance level(Table 3).
The results are consistent with the previous findings(e.g., Reed, Hanson, Elitzak, and Schluter 1997, Baffes 2007) for the heavy dependence of food sector on energy consumption. The crude, intermediate, and finished energy prices significantly affect the crude food price at least 3.5%(1.3%, 0.5%, and 3.5%, respectively) significance level, whose effects are transmitted to all the food prices at the various stages of process through the cost-push mechanisms within the food system(Figure 4).
In addition, our results further identify the heterogeneous paths of energy and food price linkages at various stages of process. The intermediate(finished) energy price Granger causes the finished(home) food price at the 4.3%(2.4%) significance level, which are analogical to the forward sequential input-output relationship in the SOP system. The complex interdependencies and/or the difficulties in defining the intermediate stage also results in the impacts of the intermediate energy price on the intermediate and finished food prices and the
effects of the finished energy price on the intermediate food price(Figure 4).
With the multifarious causal relationships between energy and food prices, the results also show that the intermediate energy price Granger causes the exchange rate at the 1.2% significance level, while the crude energy price is caused by the exchange rate at the 6.5% significance level(Table 1 and 2). The identified linkages of exchange rate with energy and/or food prices provide additional empirical evidence to understand the recent food inflation phenomenon. For example, Abbott, Hurt, and Tyner(2008) argue that the depreciating dollar is related with the over half of the crude oil price increases, because most commodities such as crude oil are denominated in the U.S.
dollars, but are purchases in the local currency. They further claim that the link between the U.S. dollar and commodity prices is more important than many other studies imply, since the high oil prices can bring expanding current account deficits, which in turn bring depreciating currency especially when the large trade deficits exist. Our findings are also consistent with previous literature on the nexus of energy price and the exchange rate(e.g., Amano and Norden 1998, Chaudhuri and Daniel 1998, and Chen and Chen 2007). For example, Chen and Chen(2007) show that the real oil prices contribute significant forecasting power for real exchange rate movements based on the panel cointegration analysis.
Ⅴ. Summary
This study explores the price transmission channels of energy prices and exchange rate on food prices. Unlike previous studies focusing on the impacts of a single energy price such as crude oil price, our analysis is based on the disaggregated information based on the various stage of process(SOP). By utilizing the TYDL method of Granger causality tests, we identify how the movements of the exchange rate and the various energy prices affect on the food prices at different stages of process from farmers to consumers.
The overall findings can be summarized as follows. First, the crude(and finished) food price Granger causes the intermediate food(retail food) price. The causal flow from crude(and intermediate) to finished(home) food prices connects the overall cost-push mechanism, despite the absence of causal link between the intermediate and the finished food prices. Overall results provide detailed information on price transmission channels and reveal the cost-push mechanism along the sequential stages of process with some skip mechanisms.
Second, the crude, intermediate, and finished energy prices significantly affect the crude food price, whose effects are transmitted to all the food prices at the various stages of process through the cost-push mechanisms within the food system. In addition, we identify the heterogeneous paths of energy and food price linkages at various stages of process. The intermediate(and finished) energy price Granger causes the finished(retail) food price, which are analogical to the forward sequential input-output relationship in the SOP system.
Finally, with the multifarious causal relationships between energy and food
prices, the identified linkages of exchange rate with energy and/or food prices provide additional empirical evidence to understand the recent food inflation phenomenon. The intermediate energy price Granger causes the exchange rate, while the crude energy price is caused by the exchange rate. For example, Abbott, Hurt, and Tyner(2008) argue that the depreciating dollar is related with the over half of the crude oil price increases, because the crude oil is denominated in the U.S. dollars. They also claim that the high energy prices can bring expanding current account deficits, which in turn bring depreciating currency especially when the large trade deficits exist. In this respect, our findings provide empirical evidences of causal mechanisms among the causes of the recent surge of food prices identified by previous studies, contributing to understand how the recent food inflation phenomenon happens.
Given that the surge of energy price is considered as the main factor for the recent food inflation, this study aims to empirically identified the causal direction and structures among various energy and food prices and exchange rate. We hope the results of this study provides a starting points for the future directions of further studies9) such as to empirically measure the magnitude of causal linkages among energy and food prices and to more fully investigate causal relationships by incorporating export food prices and other related variables.
접수일(2012년 10월 30일), 게재확정일(2012년 12월 4일)
9) As mentioned in footnote 8, the Granger causality is the statistical concept based on predictability and different from the philosophical notion based on manipulation. Thus alternative empirical approach for inductive causal inferences based on manipulation and considering the effects of unobserved variables need to be also pursued in the future studies(see Kwon, D.H. and Bessler, D.A. 2011)
◎ References ◎
Abbott, P.C., C. Hurt, and W.E. Tyner. 2008. “What’s Driving Food Prices?” Issue Report, Farm Foundation. July.
Amano, R.A. and S. Norden. 1998. “Oil Prices and the Rise and Fall of the U.S. Real Exchange Rate.” Journal of International Money and Finance. 17(2):299-316.
Baffes, J. 2007. “Oil spills on Other Commodities.” Resource Policy. 32:126-134.
Bernanke, B.S. 2004. “Oil and the Economy.” Speech at the Distinguished Lecture series, Darton College, Albany, Georgia, October.
Boyd, M.S. and B.W. Brorsen. 1985. “Price Asymmetry in the U.S. Pork Marketing Channel.” North Central Journal of Agricultural Economics 10(1 January):103-110.
Brown, S.P.A., M.K. Yücel, 2002. “Energy Prices and Aggregate Economic Activity: An Interpretative Survey.” The Quarterly Review of Economics and Finance. 42:193-208.
Bureau of Labor Statistics. 2008. “Producer Price Indexes.” in Handbook of Methods ch.14.
Available at: http://www.bls.gov/opub/hom/homch14_itc.htm
---. 2011. Consumer and Producer price indexes. Accessed August 30, 2011 at:
http://www.bls.gov/bls/inflation.htm. 2011.
Chaudhuri, K. and B.C., Daniel, 1998. “Long-run Equilibrium Real Exchange Rates and Oil Prices.” Economic Letters. 58(2):231-238.
Chen, S.S. and H.C., Chen. 2007. “Oil Prices and Real Exchange Rates.” Energy Economics.
29:390-404.
Clarke, J.A. and S. Mirza. 2006. “A Comparison of Some Common Methods for Detecting Granger Noncausality.” Journal of Statistical Computation and Simulation. 76(3 March):
207-231.
Colclough, M.J. and M.D. Lange. 1982. “Empirical Evidence of Causality from Consumer to Wholesale Prices.” Journal of Econometrics. 19:379-384
Dolado, J.J. and H. Lütkepohl. 1996. “Making Wald Tests Work for Cointegrated VAR Systems.” Econometric Reviews 15:369-386
Engel, R.F. 1978. “Testing Price Equations for Stability across Spectral Frequency Bands.”
Econometrica, 46:869-881.
Engel, R.F. and C.W.J. Granger. 1987. “Cointegration and Error Correction: Representation, Estimation, and Testing.” Econometrica 55:251-276
Gaddie, R. and M. Zoller. 1988. “New Stage of Process Price System Developed for the Producer Price Index.” Monthly Labor Review. April:3-16
Giles, J.A. and S. Mirza. 1999. “Some Pretesting Issues on Testing for Granger Noncausality.” Economic Working Paper No. 9914, Department of Economics, University of Victoria.
Goodwin, B.K. and M.T. Holt. 1999. “Asymmetric Adjustment and Price Transmission in the U.S. Beef Sector.” American Journal of Agricultural Economics 81(August):630-637.
Goodwin, B.K. and D.C. Harper. 2000. “Price Transmission, Threshold Behavior, and Asymmetric Adjustment in the U.S. Pork Sector.” Journal of Agriculture and Applied Economics 32(December):543-553.
Granger, C.W.J. 1969 "Investigating Causal Relations by Econometric Methods and Cross-Spectral Methods." Econometrica 34: 424-438.
Granger, C.W.J. and P. Newbold. 1974. “Spurious Regressions in Econometrics.” Journal of Econometrics 2:111-120.
Granger, C.W., Robinsons, R.P., and Engle, R.F. 1986. “Wholesale and Retail Prices:
Bivariate Time-series Modeling with Forecastable Error Variance.” in Modeling Reliability.
pp. 1-17. eds. D.A. Belsley and K. Edwin. MIT Press, Cambridge, Massachusetts, Gordon 1975
Hamilton, J.D. 1983. “Oil and the Macroeconomy Since World War II.” Journal of Political Economy. 91:28-248.
Hanson, K., S. Robinson, and G. Schluter. 1993. “Sectoral Effects of a World Oil Price Shock: Economywide Linkages to the Agricultural Sector.” Journal of Agricultural and Resource Economics. 18(1):96-116.
Heller, M.C. and G.A., Keoleian. 2000. “Life Cycle-based Sustainability Indicators for Assessments of the U.S. Food System.” Ann Arbor, M.I. CSS00-04. Center for Sustainable Systems, University of Michigan.
Hendrickson, J. 1996. “Energy Use in the U.S. Food System: A Summary of Existing Research and Analysis.” Madison, W.I. Center for Integrated Agricultural Systems, University of Wisconsin.
International Monetary Fund. International Financial Statistics, Washington, D.C. Accessed August 30, 2011 at: http://www.imfstatistics.org/imf, 2011.
Jones, D.W., P.N., Leiby, and I.K. Paik. 2004. “Oil Price Shocks and the Macroeconomy:
What Has Been Learned Since 1996.” Energy Journal. 25(2): 1-32.
Kauppi, H. 2004. “On the Robustness of Hypothesis Testing Based on Fully Modified Vector Autoregression When Some Roots Are Almost One.” Econometric Theory 20:341-359.
Kwon, D.H. and Bessler, D.A. 2011. "Graphical Methods, Inductive Causal Inference, and Econometrics: A Literature Review." Computational Economics 38:85-106.
Lee, T.H. and S. Scott. 1999. “Investigating Inflation Transmission by Stages of Processing.”
in Cointegration, Causality, and Forecasting: A Festschrift in Honor of Clive W.J.
Granger, Ch.12. pp.283-200 eds. R. Engel and H. White. Oxford University Press, England.
Lütkepohl, H. 1993. Introduction to Multiple Time Series Analysis, 2nd edition, Berlin, Springer-Verlag.
Murray, D. 2005. “Oil and Food: A Rising Security Challenge.” Earth Policy Institute.
Phillips, P.C.B. 1995. “Fully Modified Least Squares and Vector Autoregression.”
Econometrica 63:1023-1078.
Reed, A.J., K. Hanson, H. Elitzak, and G. Schluter, 1997. “Changing Consumer Food Prices:
A User’s Guide to ERS Analysis.” Technical Bulletin No. 1862. Economic Research Service, U.S. Department of Agriculture.
Silver, J.L. and T.D. Wallace. 1980. “The Lag Relationship between Wholesale and Consumer Prices.” Journal of Econometrics. 12:375-387.
Toda H.Y. and P.C.B. Phillips. 1993. “Vector Autoregressions and Causality.” Econometrica 61(6):1367-1393
Toda, H.Y. and T. Yamamoto. 1995. “Statistical Inference in Vector Autoregressions with Possibly Integrated Processes.” Journal of Econometrics 66:225-250
Yamada, H. and H.Y. Toda. 1998. “Inference in Possibly Integrated Vector Autoregressive Models: Some Finite Sample Evidence.” Journal of Econometrics 86:55-95
Weinhagen, J. 2005. “Price Transmission within the PPI for Intermediate Goods.” Monthly Labor Review. May: 41-49.
Zhang, Y.J., Fan, Y., Tsai, H.T., and Wei, Y.M. 2008. “Spillover Effect of US Dollar Exchange Rate on Oil Prices.” Journal of Policy Modeling. 30(6):973-991.
요 약 식품부문에 대한 에너지와 환율의 가격전달경로:
가공단계에 따른 세부적 인과구조분석
권대흠*․ 구원회**
최근에 발생한 에너지와 식료품의 동시적 가격급등현상으로 인해 에너지 부 문의 급격한 변화가 식품가격에 어떻게 영향을 미치는가에 대한 관심이 높아졌 다. 이 연구에서는 그랜져 인과관계 개념을 바탕으로 에너지 가격과 환율이 식 품가격에 영향을 주는 인과구조를 실증적으로 밝히고자 하였다.
이전의 연구와 달리 분석대상변수의 시계열적 특성에 영향을 덜 받는 Toda and Yamamoto(1995) and Dolado and Lütkepohl(1996)의 그랜져 인과관계 검증 모형을 도입하였다. 또한 에너지와 식품 각각의 가공단계별(Stage of Processing) 가격자료를 활용함으로써 에너지와 식품 각각의 단일 가격자료만 을 분석한 선행연구 보다 더욱 상세한 인과구조분석을 수행하였다.
주요 단어 : 에너지 가격, 환율, 식품가격, 가격전달경로, 가공단계별 분석, 그랜져 인과관계
경제학문헌목록 주제분류：Q40, Q10
* 조선대학교 경상대학 경제학과 조교수(주․교신저자). dhkwon@chosun.ac.kr
** 노스다코다 주립대학 교수. Won.Koo@ndsu.edu
EU ETS의 장단기 가격결정요인 분석 *
백정호**․김현석***1)
요 약
본 연구는 EU ETS 시장에서 전력가격, 석유가격, 기후 및 제도 더미변수 등 의 결정요인이 배출권 거래가격에 미치는 장단기적 영향을 살펴보았다. 이를 위해 Johansen 공적분 검정을 변수 간의 장기관계 분석에 사용하였으며, 오차 수정모형(VECM)을 통해 각 변수들 간의 단기적 조정과정도 분석하였다. 특히, EU ETS의 배출권의 과다할당이 거래가격에 직접적인 영향을 미쳤는지 여부를 계량모형을 통해 명시적으로 분석하였다. 실증분석 결과에 따르면 배출권 거래 가격은 전력 및 석유가격과 장기적으로 양의 관계를 갖는 것으로 나타났다. 또 한, 외부충격으로부터 배출권 거래가격이 안정화되는 데는 길게는 약 1년 가까 이 소요되는 것으로 나타났다. 이는 시장의 불확실성이 큰 배출권 시장의 특성 상 외부충격으로 인한 사회적 피해가 생각보다 크다는 것을 의미하는 것이다.
주요 단어 : 과다할당, 배출권, EU ETS 경제학문헌목록 주제분류：Q54, C32
* 본 논문은 에너지경제연구원 연구보고서 배출권 할당이 거래가격에 미치는 영향 분석 – EU-ETS를 중심으로(2011) 의 일부 내용을 발췌하여 수정․보완한 논문임.
** University of Alaska at Fairbanks 경제학과 부교수(주저자). jbaek3@alaska.edu
*** 경북대학교 농업경제학과 조교수(교신저자). hyun.kim@knu.ac.kr
Ⅰ. 서 론
우리나라는 국가 온실가스 감축목표를 달성하기 위한 수단의 하나로 배출 권거래제를 도입하였다. 그러나 일부에서는 배출권거래제가 경제에 부정적인 영향을 미칠 수 있다는 주장을 하고 있다. 또한 EU 배출권거래제(Emission Trading Scheme, ETS) 1기의 배출권 과다할당에 따른 가격 폭락으로 EU 배 출권시장의 효율성 및 온실가스 감축 실효성에 대한 의문이 제기됨에 따라 우리나라에서도 배출권 거래시장의 효율성에 대한 우려의 목소리가 나타나고 있다. 이러한 시점에서 EU ETS의 분석은 효율적 정책수립이라는 측면에서 중요할 것이다. 따라서 본 연구는 효율적이고 안정된 배출권거래제의 도입을 위해 EU ETS의 가격구조를 분석하고, 이를 바탕으로 보다 정확한 배출권 가 격의 분석을 그 목적으로 한다.
현재까지 EU ETS 및 기타 배출권 거래시장의 실증적 분석을 시도한 연구 는 대부분 계량 분석을 통해 임시방편적으로 설정된 요인들이 배출권 가격에 미친 영향을 분석하는데 초점을 두었다. Mansanet-Bataller 외(2007)는 기후 및 에너지관련 변수 등을 사용하여 2005년의 배출권 거래가격에 대한 결정요 인을 최소자승법(OLS)을 통해 분석하였다. 이들의 연구결과, 석유 및 천연가 스 가격 등 에너지가격 변수들이 거래가격에 큰 영향을 미친 것으로 나타난 반면, 기후관련 변수는 매우 춥거나 매우 더운 경우에만 거래가격에 영향을 준 것으로 나타났다.
Rickels 외(2007)는 Mansanet-Bataller 외(2007)의 시계열자료를 1년 확장하 여 EU ETS 1기의 초기 2년(2005∼2006)간 배출권 거래가격에 대한 결정요인 을 분석하여 Mansanet-Bataller 외(2007)와 동일한 결과를 제시하였다. 한편, Alberola 외(2008)는 다른 연구들과는 달리 2006년 4월을 전후로 발생한 배출
권 거래가격 폭락에 실증분석의 초점을 맞추고,1) EU ETS의 1기를 가격 폭 락 이전(2005년 7월∼2006년 4월)과 폭락 이후(2006년 6월∼2007년 4월)로 구 분하여 거래가격의 결정요인을 분석하였다. 이들은 가격폭락 이전의 경우 배 출권 거래가격은 주로 전력가격에 의해 결정된 반면, 가격폭락 이후에는 전력 가격 이외에 석유 및 천연가스 가격과 기후에도 큰 영향을 받았다는 결과를 제시하였다.
Bunn과 Fezzi(2008)는 앞에서 소개된 세 편의 선행연구들이 변수 간의 동 태적 관계를 고려하지 않았다는 점에 착안하여 독일 및 영국의 천연가스가격, 전력가격 및 배출권 거래가격 간의 동태적 관계를 오차수정모형을 이용해 실 증 분석하였다. 이들의 연구결과에 의하면 영국이나 독일의 경우, 석탄을 이용 한 전력발전이 많은 관계로 탄소가격의 변화가 이들 나라의 전력가격 변화에 큰 영향을 주는 것으로 나타난 반면, 천연가스가격 변화에는 별다른 영향을 주지 않는 것으로 나타났다. 또한 Keppler와 Mansanet-Bataller(2010)는 EU ETS 1기와 2기 배출권의 선물가격과 에너지 및 기후관련 변수의 장단기 관 계를 분석하였다. 이들의 연구결과에 따르면 EU ETS 1기에는 에너지 스프레 드(클린 스파크 스프레드와 클린 다크 스프레드)와 예상치 못한 기온의 변화 가 배출권의 선물가격에 영향을 주고, 배출권의 선물가격이 현물가격에 영향 을 주는 것으로 나타났다. EU ETS 2기에서는 에너지 스프레드와 배출권의 선물가격이 쌍방향으로 영향을 미치는 것으로 나타났으며, 배출권의 선물가격 은 가스 가격에 영향을 주는 것으로 나타났다.
한편, Hintermann(2010)은 기존의 실증연구들이 계량 분석에서의 결정변수 들이 임시방편적으로 결정되어 배출권 가격 결정요인에 대한 경제이론적 기 반이 매우 취약하다고 비판하며, 배출권 가격 결정함수에 대한 경제이론을 도 출한 후 다각적인 방법으로 실증분석을 시도하였다. 실증분석 결과 천연가스 1) EU ETS는 당해 연도 참여자들의 의무준수 결과를 다음 연도 4월에 발표하고 있다. 1 기의 경우 2005년 참여자들의 의무준수 결과가 2006년 4월 발표됨에 따라 수일 만에 약 15%의 가격 하락이 발생하였다. 이는 2005년 의무준수 결과 할당된 배출권이 실제 배 출량보다 많아 배출권의 과잉할당 사실이 알려지면서 가격이 급락하게 된 결과이다.
및 석탄가격, 여름철의 고온, 강수량 등이 EU ETS 1기의 배출권 가격에 큰 영향을 미친 것으로 나타났다.
이들 선행연구는 주로 단일방정식을 기본으로 정태적 분석을 통해 배출권 가격의 결정요인을 분석하였다. 그러나 배출권 가격과 그 결정변수들은 계량 모델 내에서 서로 내생적으로 결정되는 경향이 강할 뿐만 아니라, 결정요인들 이 거래가격에 미치는 영향이 장기와 단기적으로 상이할 가능성도 매우 크다 고 볼 수 있다. 따라서 본 연구에서는 보다 엄격한 계량분석을 통한 배출권 가격의 결정요인을 분석하기 위하여 다변량 계량분석을 사용하였다. Johansen (1995)에 의해 제시된 다변량 공적분 검정법을 이용하여 배출권 가격의 결정 요인에 대한 장기적 효과를 분석하였다. 이를 바탕으로 오차수정모형(VECM) 을 이용하여 각 변수들 간의 단기적인 관계도 분석하였다. 특히, 선행연구들이 주로 EU ETS 1기만을 대상으로 실증분석을 한 것과는 달리 본 연구는 EU ETS 시행 전(全) 기간(2005년 5월∼2011년 8월)에 대한 시계열자료를 가지고 배출권 가격의 장단기 결정요인을 분석하였다. 또한, 본 연구는 선행연구들에 서 고려하지 않은 EU ETS 1기 배출권 과다할당 및 잉여배출권 이월금지와 같은 정책적 요인이 거래가격 폭락에 직접적인 영향을 미쳤는지를 계량모델 을 통해 명시적으로 분석하였다.
Ⅱ. 분석방법론 및 데이터
2.1. 분석방법론
배출권 가격의 결정요인에 대한 분석을 위하여 본 연구는 다변량 계량분석 을 사용하였다. Johansen 공적분 검정(Johansen, 1995)은 모형 내의 모든 변 수를 내생변수로 간주하는 벡터자기회귀모형(vector autoregression model,
VAR)을 기반으로 공적분 관계를 검정하고 공적분 계수를 추론하는 방법이다.
Johansen이 제시한 공적분 검정은 다음과 같은 동태적 VAR 모형에서 출발 한다.
(1)여기서
는 내생변수 벡터,
는 추정모수 벡터,
는 평균값 벡터, 는 추세 또는 더미변수 등의 결정적 변수를 나타내는 벡터이다. 식 (1)을 오차수 정모형으로 재모수화하면 다음과 같다.∆
∆
∆
(2)여기서
,
(
는 단위행렬을 표 시함),
,∆
는 각 변수의 차분을 나타낸다. 식 (2) 에서
항은 변수 간의 장기적 공적분 관계를 나타내는 반면,
는 변수간의 단기 동태적 관계를 나타낸다.
의 계수행렬은 부하행렬을 나타내는
와 공적분 벡터를 내포하는 행렬인
로 분해될 수 있다( ′
). 여기서
는 균형으로의 조정속도를 의미하고,
는 장기계수행렬을 의미한다. 식 (2)에서 변수들 간의 공적분 행렬은
항이 축소된 위수(reduced rank)
을 가질 때 성립하게 된다. 이때 0과 다른 행렬
의 고유치(eigenvalue)의 개수가
이 되며, 최대
개의 공적분관계가 나타나게 된다.2) 또한
의 랭크라고 불 리는 공적분 벡터의 수는 우도비(Likelihood ratio) 검정이 기반이 되며, 통상 적으로는 추적검정(trace test)을 통해 결정된다.여기서 꼭 언급해야 할 사항으로는, 시계열 자료는 대개의 경우 임의보행
2) 이 경우
≤
임.(random walk)을 따르는 비정상적(nonstationary)인 특성을 보인다는 것이다.
이런 비안정성적인 변수들의 경우, 유한한 분산을 가지지 않는 관계로 통상적 인 최소자승법을 통해서는 일관성 있는 회귀계수를 추정할 수 없다는 문제점 이 있다. 즉, 최소자승법을 이런 변수들에게 적용하여 모형을 추정할 경우, 소 위 말하는 가성회귀(spurious regression)의 문제가 발생하게 된다. 그러나 Engle과 Granger(1987)에 따르면 계량모형 내의 변수들이 모두 비정상적인 특징을 보이더라도 이들 변수들의 선형결합은 안정성을 가질 수가 있는데 이 를 ‘비정상적인 변수들 간에 공적분 관계가 성립한다’라고 말하며, 이 경우 가 성회귀 문제도 발생하지 않게 된다. 따라서 공적분 검정은 시계열 자료의 단 위근 검정을 실시하여 시계열들의 (비)정상 검정을 밝혀내는 것으로부터 시작 된다.
만일
벡터의 모든 변수가 단위근을 가지고, 서로 간 공적분 관계를 형성 하는 것이 검정된다면, 식 (2)는 다음과 같은 단기적 동태모델로 전환될 수 있다.∆
∆
∆
′
(3)
여기서
′
는 균형으로부터의 이탈을 측정하는 오차수정항을 의미한다.3) 따라서 식 (3)은 내생변수 간의 단기 및 장기적 효과를 동시에 분석할 수 있는 방법을 제공하게 된다. 예를 들어, 장기균형이 성립이 되고 있다면,
′
의 계수는 0을 나타내게 된다. 반면 현재 불균형상태에 있다면′
의 계수는 0이 아니게 되며, 그 계수의 값은 불균형상태에서 균형 상태로 수 렴하는데 소요되는 기간의 측정을 나타내게 된다. 따라서
의 추정치는 조정 속도에 대한 정보를 제공하게 된다.3) 오차수정항은 공적분 벡터의 잔차로부터 구해진다.
2.2. 데이터
본 연구에서는 배출권 거래가격의 결정요인으로 전력가격, 석유가격, 기후 및 제도적 요인을 사용하였다. EU 내 전원별 발전에 사용되는 화석연료의 대 부분이 석탄과 가스이므로, 석탄과 가스가격은 전력가격과 상관관계가 높다 (김현석, 2011, pp.50). 따라서 가스와 석탄가격의 경우 전력가격을 통해 배출 권 가격에 영향을 준다고 가정하였다. 그 이외에 석유가격과 극한 기후, 과잉 할당과 잉여배출권의 이월금지 등의 제도적 요인에 따라 배출권 가격이 변화 한다는 가정 하에 모형을 추정하였다. 배출권 거래가격은 Point Carbon에서 수집한 장외거래(OTC) 현물가격을 사용하였으며, 석유가격은 Intercontinental Exchange에서 제공하는 브렌트유(Brent) 가격을 사용하였다. 전력가격은 European Power Exchange에서 제공하는 독일․오스트리아 전력가격을 사용 하였다. 배출권 거래가격 및 전력가격이 유로화로 표시된 관계로 유럽중앙은 행에서 제공하는 달러 대 유로 환율을 이용해서 달러화로 표시된 브렌트유 가격을 유로화로 전환하여 사용하였다. 기후관련 변수는 Hintermann(2010)의 선행연구를 참고하여 추운 달(11월부터 3월까지)과 더운 달(6월에서 9월까지) 로 구분한 더미변수를 사용하였으며, 마지막으로 EU ETS 1기의 배출권 과다 할당 및 잉여배출권의 이월금지가 거래가격에 미친 영향을 명시적으로 고려 하기 위해 두 개의 정책더미를 사용하였다. 첫 번째 더미변수는 Alberola 외 (2008)의 연구를 참조하여 배출권 가격이 본격적으로 폭락하기 시작한 ’06년 4월 26일을 기준점으로 하여 그 이전은 0, 그 이후는 1로 설정하였다. 두 번 째 더미변수는 EU ETS 1기 전체 기간 동안의 배출권 과다할당 및 잉여배출 권의 이월금지로 인해, 동 기간의 거래가격이 2기 동안의 거래가격보다는 전 반적으로 하락했을 것이라는 점에 착안하여 제 1기 동안은 1, 그 이후는 0으 로 설정하였다. 즉 첫 번째 더미변수()는 ’05년의 할당 및 감축에 관 한 결과가 ’06년 4월 공개된 이후 배출권 거래가격의 변화를 반영하기 위한 더미변수이며, 두 번째 더미변수()는 1기와 2기의 전체적인 배출권
