Estimating the residential demand function for natural gas in Seoul with correction for sample selection bias
Introduction
Over the last two decades, the consumption of natural gas in Korea has increased at a high annual growth rate of 18.7%. This is due to the expansion of the infrastructure for gas distribution and an increase in individual income. Crucial factors that underlie the recent rise in the consumption of natural gas have been an upsurge in the urban consumption of gas3 and an increase of 55% in the proportion of the urban consumption of gas that occurs in the residential sector [1].
Owing to additional factors, in the capital city of Seoul, which accounts for 23.3% of the residential consumption of energy in Korea, a significant number of households have switched from oil and coal to gas and electricity. Especially in urban areas, where, in 1987, natural gas began to be supplied and the rate of consumption increased dramatically thereafter, the rate of consumption has increased at an average rate of 6.9% p.a. As of 2001, natural gas was the main source of energy for households in Seoul, accounting for 55.4% of the residential consumption of energy [2].
However, Korea has scarce reserves of natural gas, which is therefore imported entirely from natural gas-producing countries. The supply chain for gas in urban areas in Korea is structured as follows. Natural gas is imported in liquid form (LNG) from foreign countries and is then supplied to domestic gas companies (wholesale companies and a monopolized public utilities company that is owned by the Korean government). It is also supplied to regular, urban, gas companies (for-profit retail companies and private firms) that supply to households, businesses, and industries.
Thus, the consumer’s price for urban gas is comprised of the cost of importation and the wholesale and retail firms’ expenses for supplying gas. Among the various components of cost, the cost of importation fluctuates due to fluctuations in the international price of oil and the exchange rate. Being public fees, the expenses of wholesale and retail firms are regulated by local governments and legislative assemblies. In the case of Seoul, the local government suppressed price increases by lowering and freezing fees; however, after 2006, due to inflation and other factors, the price of the gas was increased for the first time [2].
On the other hand, the international price of LNG has recently been at a very high level due to the high price of oil. The high price of oil aside, the actual prices of goods are higher than ever before due to the impact of technological and seasonal factors upon the short-term supply and demand for natural gas. In particular, the price levels, which in the past several years showed a downward trend, are now showing an upward trend, as per a recently concluded long-term contract. This reverse trend is due to an increase in demand, increased uncertainty in the ability to increase supply, and the expectation that the price of oil will remain high, besides various other interdependent factors [4].
Therefore, the international LNG market is expected to be unfavorable to importing countries such as Korea that procure most of their natural gas in the form of LNG. In this context, to balance supply with demand, the demand for natural gas (NGD) needs to be accurately predicted, especially for the residential sector. This can be achieved by analyzing the pattern of residential NGD. In addition, the estimation of the demand function for natural gas will provide useful information for pricing-related policies relating to importation and the costs of supply.
Many studies have been conducted on the estimation of the residential demand for natural gas elsewhere in the world [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15]. Most of them have been based upon the analysis of panel data but some studies [5], [6], [16] have used cross-sectional data. In Korea, to the best of the authors’ knowledge, neither panel data sets nor investigations of the demand for natural gas are available, because of the short history of the use of LNG. Therefore, the goal of this paper is to estimate the Korean residential NGD function using cross-sectional data. Microeconomic approaches to the modeling of the residential demand for natural gas enable us to acquire a better understanding of the nature of consumer responses to a change in the price of natural gas.
To gather data, a survey was conducted of households in Seoul. As with many other cross-sectional data sets and associated surveys, the survey of households manifested a high rate of non-response [17]4: of the 380 households sampled, 216 households (56.8%) failed to answer the question in the survey on natural gas consumption. In the present study, failure to incorporate all the missing observations was an issue.
To determine whether non-response in the present study results in bias, two questions need to be asked: Do response rates differ across identifiable groups of households? Are there systematic differences within a particular group between those who answered the questions on NGD and those who did not? Bias will occur to the extent that these between- and within-group differences exist and furthermore, are related to NGD [18].
A given cross-sectional study may suffer from either a between-group sample non-response bias and/or a within-group sample selection bias.5 Non-response biases can generally be compensated for by weighting and imputation procedures, which assume that those participants in a given subgroup of the population who do not respond have the same expected values for NGD as those in the same subgroup who do respond. The correction procedures for those biases rely on the assumption of random non-response within the categories. In the case of sample selection bias, however, non-response is nonrandom: the individuals who do not respond to either the survey or the question hold different expected values for NGD than comparable individuals who do respond. To deal with sample selection bias, a sample selection model proposed by Heckman [19] has been commonly employed to rectify the problems emanating from the bias. However, empirical applications of the model in the estimation of the residential NGD function are lacking.
This paper, therefore, has two major purposes. The main purpose of this paper is to estimate the residential NGD function. The second purpose is to explore, during the estimation of the residential NGD function, the sample selection model that produces consistent parameter estimates and unbiased mean estimates of NGD. Moreover, the paper compares the results with those from a model that assumes no sample selection bias. It also employs two procedures, described in Section 4, to test for sample selection bias.
The remainder of the paper is organized as follows. Section 2 reviews the literature on analyses of residential NGD. Section 3 explains the data that are employed and presents an overview of the proposed methodology. The empirical findings are reported in the penultimate section, Section 4. Some concluding remarks are made in Section 5.
Section snippets
Literature review of analyses of the demand for natural gas
The empirical estimation of the NGD function has received considerable attention in the literature. Table 1 summarizes the findings. Interestingly, we could not find case-studies of Asian and African regions. Most studies focus on North American and European regions that have a long history of the consumption of natural gas. The main goal of most of these studies, especially those for North American provinces and states, was to estimate the short-run and long-run elasticities of residential
Data
A survey of households was conducted in the month of May 2005. The survey was restricted to Seoul, whose population is a fourth of the national population. The survey was administered to residents, who were heads of households and/or housewives between 20 and 65 years of age. To obtain a representative sample of this population, the city of Seoul was divided into 25 residential districts. A random sample was drawn by a Seoul-based, professional polling firm (Dongseo Research, Inc.) according to
Test for non-response bias
In the first place, non-response bias can be examined by comparing the characteristics of those who responded in the interview (the response sample) with the characteristics of those who did not respond (the non-response sample). To test whether there is a significant difference in the characteristics between the two samples, the Z-test was employed for each characteristic. The Z-test statistic follows a standard normal distribution under the null hypothesis that there is no difference in the
Concluding remarks
The estimation of the residential NGD function offers a useful framework for analyzing consumer behavior with regard to residential NGD, including the calculation of elasticities. This paper has attempted to estimate the NGD function in the residential sector using data from a survey conducted of households in Seoul. In the survey, 57.8% of sampled households did not respond to questions on their consumption of natural gas. The resulting lack of response can cause sample selection bias, leading
References (23)
Residential demand for natural gas
J Urban Econ
(1980)- et al.
Dealing with bottled water expenditures data with zero observations: a semiparametric specification
Econ Lett
(2000) - Park GS. Mid-term and short-term analysis of energy balance 2007–2012. Seoul: Korea Energy Economics Institute (KEEI);...
- Seoul Metropolitan Government. Regional energy plan of Seoul city, Seoul;...
- Do HJ. Changes in LNG trade background. KEEI Issue Paper;...
- et al.
Estimation of rural demand for natural gas
Management Science
(1973) - et al.
Patterns in residential gas and electricity consumption: an econometric analysis
J Econ Soc Measure
(1994) - et al.
Pooling cross section and time series data in the estimation of a dynamic model: the demand for natural gas
Econometrica
(1996) - et al.
Demand for natural gas: residential and commercial markets in Ontario and British Columbia
Can J Econ
(1977) The demand for natural gas: a survey of price and income elasticities
Energy J
(1989)
Analyzing demand behavior: a study of energy elasticities
Cited by (49)
A novel grey Lotka–Volterra model driven by the mechanism of competition and cooperation for energy consumption forecasting
2023, EnergyCitation Excerpt :Wang et al. [12] proposed an interval forecasting system, and quantify the change in forecast results due to uncertainty. Yoo et al. [13] analysed the reason for the sharp increase in natural gas consumption in Seoul due to the increase in residential consumption, and predicted the consumption demand of natural gas in Seoul. While, artificial intelligences, which describe time series data via a complicated and flexible way and interpret the relationships between input and output data through simulating the structure of biological neural networks, have been used for energy consumption forecasting.
The daily price and income elasticity of natural gas demand in Europe
2022, Energy ReportsPrice and income elasticity of natural gas demand in Europe and the effects of lockdowns due to Covid-19
2022, Energy Strategy ReviewsModeling energy flow in natural gas networks using time series disaggregation and fuzzy systems tuned by particle swarm optimization
2020, Applied Soft Computing Journal