Chapter 3. Previous Literature and Analytical Methodology
1. Previous Literature on the Analysis of City Gas Consumption in Korea
This section examines previous literature on the analysis of city gas consumption in Korea, and discusses the problems with the previous studies. In addition, it describes the differences between the purpose and the methodology of this study and those of previously conducted research.
1.1. Previous Research
Previous studies on the analysis of city gas consumption in Korea are listed in Table 3-1 (below). In terms of analytical target, these studies can be divided into those that analyzed the total demand for city gas and those that analyzed demand by usage type. Research conducted by Inmoo Kim et al. (2011), Jumsu Kim et al. (2011), Jinsoo Park et al. (2013), Seungjae Lee et al. (2013), and Sungro Lee (2017) fall into the former category, in which the total demand for city gas was analyzed by establishing a demand function and estimating the demand using an econometric analysis methodology.
Table 3-1. Summary of Previous Studies on the Analysis of City Gas Consumption in Korea Researchers Analytical Target Analytical Method Independent Variables Youngduk Kim
(1998)
City gas for household,
general, and industrial use
Multiple regression model Industrial production index, real city gas prices, and -ariables
Inmoo Kim, Changsik Kim, and Seonggeun Park (2011)
Total demand for city gas
Cointegrating regression model
Real GDP, relative price of city gas compared to electricity, and temperature variables
Jumsu Kim, Chunseung Yang, and
Junggu Park (2011)
Total demand for city gas
Cointegrating regression
model Real GDP and temperature variables
Gwangsu Park (2012)
City gas for household and general use
Multiple regression model
Real GDP, relative price of city gas compared to electricity, quarter dummy variables, and temperature variables Jinsoo Park, Yunbae
Kim, and Cheolwoo Jeong (2013)
Total demand for city gas
Autoregressive model, multiple regression model, and
weighted average model
Time difference variables of the dependent variable, holiday variables, and temperature variables
Seungjae Lee, Seungseob Eu, and Seunghoon Yoo (2013)
Total demand for
city gas Multiple regression model
Real GDP, real city gas prices, and time difference variables of the dependent variable
Myeongdeok Park and Sangyeol Lee (2015)
City gas for
industrial use Multiple regression model
Industrial production index, relative price of city gas compared to Bunker C, and temperature variables
Yujin Bae and Jaewoo Jeong (2017)
City gas for household use
Cointegrating regression
model Real GDP and temperature variables
Suktae Lee, Seulye Lim, and Seunghoon Yoo (2017)
City gas for industrial use
LMS (least median squares)
Real GDP, city gas price index, and time difference variables of the dependent variable
Sungro Lee (2017) Total demand for city gas
Cointegrating regression model
Relative price of city gas compared to electricity and Bunker C and temperature variables
Daeyong Kim and Sungro Lee (2018)
City gas for
household use Panel model
Regional GDP per capita, relative prices, temperature variables, number of household members, percentage of the elderly population, percentage of the employed, percentage of apartment complexes, and rate of natural increase of the population
Cheolwung Park and Cheolho Park (2018)
City gas for household,
general, and industrial use
ARDL, error correction model,
and cointegrating regression model
Real GDP, relative prices, and temperature variables
Sungro Lee and Jonghyun Ha (2019)
City gas for industrial use
Cointegrating regression
model Relative prices and temperature variables Source: Revised and reduced Table 1 in Cheolwung Park and Cheolho Park (2018).
However, since the characteristics of city gas consumption differ significantly by type of use, it may be problematic to establish a model and make estimations without making distinctions. For example, the amounts of city gas consumed by households and industries vary considerably depending on energy prices and temperatures. Household consumption does not respond significantly to fluctuations in city gas rates or oil prices; rather, it is most sensitive to changes in winter temperatures. On the other hand, industrial consumption is more affected by changes in relative energy prices, as companies look for ways to minimize their costs, than it is by temperature changes. Therefore, constructing a city gas demand function and estimating coefficients, such as price elasticity and income elasticity, while ignoring the characteristics of city gas consumption by usage type would be considered a problematic approach.
Among the studies that separated the city gas demand function by usage type, some analyzed city gas demand for all three types of usage—household, general, and industrial. These studies are Youngduk Kim (1998) and Cheolwung Park and Cheolho Park (2018). Other studies focused on one or two specific usage types, and these include Gwangsu Park (2012), Myeongdeok Park and Sangyeol Lee (2015), Yujin Bae and Jaewoo Jeong (2017), Suktae Lee et al. (2017), Daeyong Kim and Sungro Lee (2018), and Sungro Lee and Jonghyun Ha (2019).
It is impossible to avoid the quality issue regarding the data on city gas consumption when it comes to the analyses of the city gas demand function for household and general (or commercial) uses. There is a serious measurement error problem in the statistics on household and general city gas consumption, arising from the meter reading dates and self-metering. The problem with meter reading dates lies in the fact that the number of households and stores subject to meter reading is quite large, but the number of meter readers is limited. Since a meter reader needs to check the city gas consumption of a large number of households and stores, the meter reading dates for checking city gas consumption range from the first of the month to the 31st (or 28th, 29th, or 30th). This means that, while the meter readings checked earlier in the month are more likely to be from the previous month, the readings checked later in the month are for the current month. The problem with self-metering arises from the fact that the city gas meters are installed inside homes or businesses. Because it is difficult for a meter reader to check the city gas consumption from a meter inside these buildings, household members and store employees keep records of the meter readings. A meter reader visits these homes or businesses once a year to check the readings and calculate the total amounts consumed. In such cases, household members and business
owners tend to report less consumption in the winter, when city gas consumption is high, and inflate the figure in the summer, when gas consumption is low in order to smooth out their spending over time.
Due to these problems, data on household and commercial city gas consumption are less reliable, and the results analyzed with these data have low accuracy.
Industrial city gas consumption, however, is different. First, there not nearly as many industries as there are households or businesses, and consumption per industry is significantly higher than for households or businesses, allowing meter readers to directly check the amount consumed on the last day of the month.30 Therefore, analyses of industrial city gas consumption can avoid the measurement errors in statistical data.
Previous literature on the analysis of city gas consumption can also be divided according to analytical method into studies that account for the non-stationarity of the statistics and studies that ignore it. Jumsu Kim et al. (2011), Yujin Bae and Jaewoo Jeong (2017), Sungro Lee (2017), Cheolwung Park and Cheolho Park (2018), and Sungro Lee and Jonghyun Ha (2019) tested for the non-stationarity of the analytical data using the unit root test.31 The test revealed that a unit root was present in the analytical data used in their studies, and therefore these studies established city gas demand functions using regression models that recognize cointegration.
The rest of the previous studies examined ignore the problem of non-stationarity or unit roots. Since the data used to analyze the demand function of city gas are time series data, it is essential to check the non-stationarity of the data prior to conducting an analysis, and establish a model that takes this into consideration. If a unit root is not present in time series data (meaning that the time series data are stationary) through a unit root test, it is fine to construct a city gas demand function using the original data. However, if the unit root test reveals that a unit root is present, differentiated data should be used for the analysis, and the long-term relationship between variables should be expressed using cointegration. Therefore, studies that did not check for the non-stationarity of time series data have problems in terms of the application of a methodology.
There are no significant differences between the independent variables used in previous studies. Most used production or income variables, price variables, and temperature variables. However, Sungro Lee (2017) and Sungro Lee and Jonghyun Ha (2019) used regional data and omitted production or income variables, while Jinsoo Park (2013) used daily data and omitted income and price variables. Daeyong Kim and Sungro Lee (2018) added cross-sectional factors as independent variables using panel data, and Yujin Bae and Jaewoo Jeong (2017) excluded price variables by analyzing city gas demand for household use.
As such, the composition of independent variables varies between studies for particular reasons, such as the unavailability of certain variables. However, there seems to be little disagreement that income variables, price variables, and temperature variables are the main factors that explain the changes in city gas demand.
1.2. Differences between This Study and Previous Studies
The differences between this study and previous ones can be summarized as follows. First, this study focuses on the structural changes in city gas consumption. Previous research focused on estimating key
30 This was based on the consultation provided by KyungDong City Gas and Samchully, which are two of the city gas providers with a high share of the industrial city gas supply.
31 Inmoo Kim et al. (2011) did not perform a unit root test but used a cointegrating regression model with the assumption of a unit root in the time series data.
coefficients, such as price elasticity, by analyzing a city gas demand function or assessing the predictive power of a demand function. It was difficult to find a previous study that researched the structural changes of city gas consumption. This study does not aim to simply create a city gas demand function and estimate specific coefficients, but rather qualitatively investigate and econometrically verify the changes in the current structure of city gas consumption with a focus on energy-intensive industries.
In addition to purpose of research, this study differs from previous studies in that it more accurately estimates the city gas demand function by accounting for the structural changes in the city gas demand function, which was overlooked in previous studies. As discussed in Chapter 2, the consumption structure of industrial city gas has changed recently due to the rapid changes in relative energy prices and the market environment. If such changes in the consumption structure are not reflected, the models would not be able to sufficiently explain the data, and it would be impossible to avoid bias in the estimates of key factors, such as price elasticity.
Second, this study differs from previous studies in that a different statistical methodology is used, since the focus of this study is structural changes in city gas consumption. Previous studies did not account for structural changes when conducting the unit root test on time series data.
Table 3-2. Unit Root Testing Methods Used in Previous Studies
Researchers Unit Root Testing Method
Jumsu Kim, Chunseung Yang, and
Junggu Park (2011) ADF test
Yujin Bae and Jaewoo Jeong (2017) ADF test
Sungro Lee (2017) ADF test, KPSS test
Cheolwung Park and Cheolho Park (2018) ADF test
Sungro Lee and Jonghyun Ha (2019) ADF test, KPSS test Source: referred to studies mentioned in the table
Note: The ADF test refers to the Augmented Dickey-Fuller test, and the KPSS test refers to the Kwiatkowski-Phillips- Schmidt- Shin test.
This study, on the other hand, uses a unit root testing method that accounts for structural changes.32 Figure 3-1 (below) shows the trend of city gas consumption in energy-intensive industries, which are the subject of this study. The trend of city gas consumption in the petrochemical industry has been changing considerably, and the trend of city gas consumption in the steel and fabricated metal industries has changed since 2010 or 2011. If a general unit root testing method is applied to each time series data without taking these changes into account, the test results would be unreliable.
In addition, this study established a city gas demand function and used the Bai-Perron test, which is most widely used to check for structural changes, to detect structural changes and estimate structural breakpoints. Moreover, a new city gas demand function was estimated by taking these structural breakpoints into account.
32 This study uses the unit root testing methods proposed by Perron (1989, 1990, and 1994) and Perron and Vogelsang (1992a, 1992b, 1993a, and 1993b).
Figure 3-1. City Gas Consumption Trend in Energy-Intensive Industries
Source: Monthly Energy Statistics
석유화학 Petrochemical
철강 Steel
조립금속 Fabricated metal
Third, this study analyzed the city gas demand function of each energy-intensive industry among consumers of industrial city gas. As mentioned above, it is not advisable to bundle and analyze the demand for city gas of each industry all together in one total demand, because the characteristics of the demand function vary by the usage type of city gas. As examined in Chapter 2, each industry has different characteristics in terms of its consumption of industrial city gas. In addition, the statistical characteristics of time series data on city gas consumption differ by industry. This can be confirmed by Figure 3-1 (above). As each industry reacts differently to the relative energy prices and temperature, it is problematic to group all of these industries together and analyze them using a single industrial city gas demand function. Therefore, unlike previous studies, this study uses different city gas demand functions for each of the three industries with high industrial city gas consumption, and analyzes them separately.