Open access peer-reviewed chapter - ONLINE FIRST

The Dynamic of Residential Energy Demand Function: Evidence from Natural Gas

Written By

Mohamed Jaouad Malzi

Submitted: October 22nd, 2021 Reviewed: January 3rd, 2022 Published: April 11th, 2022

DOI: 10.5772/intechopen.102451

Natural Gas - New Perspectives and Future Developments Edited by Maryam Takht Ravanchi

From the Edited Volume

Natural Gas - New Perspectives and Future Developments [Working Title]

Dr. Maryam Takht Ravanchi

Chapter metrics overview

10 Chapter Downloads

View Full Metrics


This analysis uses annual data on residential gas use for 29 Organization for Economic Cooperation and Development nations from 2005 to 2016 to look at per capita energy demand. The effect of price and income on natural gas demand elasticities has been studied in the past, but most research have ignored demographic aspects. The goal of this study is to incorporate these characteristics into natural gas demand modeling. A dynamic panel system dubbed the Generalized Method of Moments (GMM) estimator was used to address the endogeneity issue. The following are the study’s main findings: First, the residential sector consumes more natural gas per capita as the population grows. Second, the consumption of per capita residential natural gas in Organization for Economic Cooperation and Development countries is decreasing as the population ages. Finally, as the population density rises, so does per capita gas consumption.


  • dynamic function
  • energy demand
  • generalized method of moments

1. Introduction

The OECD1 countries have encountered many challenges throughout the years, including rapidly aging societies and diminishing fertility rates. The share of the senior population (those over 65 years old) climbed from less than 9% in 1960 to 17% in 2015, and it is anticipated to continue to rise, reaching 28% in 2050.2

Furthermore, since 1970, most OECD countries have faced the difficulty of a primarily urban population. In OECD countries, particularly Australia, Korea, Chile, France, and Japan, urbanization is higher, and the trend is anticipated to continue.3

Because demographic shifts pose serious distributional issues and are projected to have significant economic effects, OECD nations must take these factors into account.

Indeed, natural gas’ adaptability, low price, and lower greenhouse gas emissions from combustion than coal and oil have pushed natural gas demand to rise rapidly, gaining share in all sectors, particularly residential. Residential natural gas usage has risen consistently in OECD countries over time. The amount has increased from around 8 million terajoules (TJ) in 1980 to more than 11 million TJ in 2016. As shown in Figure 1, the total has moved from about 8 million terajoules (TJ) in 1980 to more than 11 million TJ in 2016. Moreover, in order to understand the individual trend of per capita residential natural gas consumption considered as the dependent variable, during the period of study (2005–2016), per capita natural gas demand was plotted in a time-series graph for each country in Figure 2 and note that each country has its own specific trend.

Figure 1.

Evolution of aggregate residential consumption for natural gas in OECD countries (terajoules, 1980–2016). Reference: own elaboration based on IEA data.

Figure 2.

Evolution of per capita residential consumption for natural gas by country (MWh, 2005–2016). Reference: own elaboration based on IEA data.

These two figures are useful for econometric reasons since they show the evolution of aggregate and per capita individual demand for natural gas in the residential sector over time.

In fact, knowing the numerous factors of residential natural gas consumption and thus estimating the demand equation accurately is critical to creating natural gas legislation and corporate strategies for investors in the natural gas residential sector. To estimate household natural gas demand, the literature has concentrated on the effect of price and income. The majority of these research used static or dynamic models, or both, to simulate natural gas consumption behavior.

In previous energy research ([1, 2, 3, 4, 5]), demographic factors such as the elderly, population density, and urbanization have received little attention, despite the fact that they influence household natural gas consumption explicitly or implicitly. Furthermore, with the exception of Gautam et al. [5], the majority of these research are based on data from before 2010, and it is critical to update studies, especially in this economic field where inputs are rapidly changing.

Policymakers need to know not only how natural gas demand will respond to income and price changes in order to make holistic decisions.

The goal of this research is to estimate the dynamics of per capita residential natural gas demand in 29 OECD nations from 2005 to 2016. The rest of the paper is organized as follows. A brief survey of the literature is presented in Section 2. Section 3 gives a description of the data as well as some descriptive statistics. In Section 4, the estimation findings are shown. A brief conclusion is included in the concluding section.


2. Literature review

The impact of population characteristics, particularly age, density, and proportion of urban population, on residential natural gas demand is rare; this could be due to a lack of data. To estimate natural gas demand, most studies focused on price and income elasticities. Several studies, however, have shed light on the impact of population factors on overall household energy usage over the last two decades.

According to several studies, there is a link between household age and space heating energy use. That is, because elderly people are more sensitive to temperature, they consume more energy for space heating than younger people because they spend more time at home. Meanwhile, Chen et al. [6] found that age has a greater impact than wealth in a study of Hangzhou, China, and that there is a negative relationship between occupant age and residential energy usage, particularly for heating and cooling. It was discovered that older housewives are more supportive of economic conduct than younger housewives.

Kronenberg [7] observes in his study on energy consumption and greenhouse gas emissions in Germany that demographic changes, defined by a rise in the number of elderly persons, have a favorable impact on energy demand, particularly for heating.

Liao and Chang [8] used the discrete technique to estimate the space heating and water heating energy demands of senior residents in the United States using data from the 1993 Residential Energy Consumption Survey. They believe that the elderly consumes more natural gas and electricity to heat their homes. However, there is a considerable negative association between water heating energy use and age.

Ota et al. [9] found that the aging of the society has no substantial impact on residential electricity and city gas demand in 47 Japanese prefectures every 5 years between 1990 and 2010. Furthermore, population decline and the rise of nuclear households raise electricity usage while lowering city gas consumption.

Residents who live in densely populated locations (such as Dublin) use less energy for space heating than those who live in less densely populated areas. According to Elnakat et al. [10], a socioeconomic and demographic study on the residential sector in San Antonio, Bexar County, and Texas, areas with higher population density spend less energy per capita than those with lower population density. Furthermore, Arbabi and Mayfield [11] found that for increasing population densities, falling per capita gas consumption patterns are observed in a study aimed at investigating consumption behavior within the transport and home sectors in England and Wales.

Furthermore, He et al. [12] conclude, based on data from 2001 to 2011, that the greater the urban population, the greater the total natural gas consumption.

Rather than calculating natural gas demand price and income elasticities, as most research has done, the focus of this work is on demographic characteristics, particularly the elderly, population density, and urbanization rate.


3. Empirical framework

3.1 Data

Between 2005 and 2016, annual data for OECD nations was used. The dataset is based on two main sources: the International Energy Agency (IEA) dataset for residential natural gas consumption and residential natural gas and electricity prices, and the World Bank dataset for per capita income, overall population, population density, urban population percentage, aged population, heating and cooling degree days.

Due to a lack of data, five of the 34 current members of the Organization for Economic Cooperation and Development were excluded from the sample. Due to lacking price data or no reported natural gas demand, Estonia, Iceland, Israel, Norway, and Slovenia were omitted from the list. Figure 3 depicts the study’s precise countries. Every year, each country in the sample is observed, ensuring that the data set is balanced. The following are the descriptive data for the remaining 29 members, as shown in Table 1.

Figure 3.

Countries in the study. Reference: own elaboration with tableau public software.

VariableLabelMeanStandard deviationMinimumMaximum
Per capita residential natural gas consumption, MWhGT2.2760221.8355150.0360867.55551
Elderly population (total)ELD6,316,6119,100,80267,0794.86e+07
Urban population (% of total population)URB77.19139.72660753.46897.897
Population density (Inhabitants/Km2)DEN139.3273128.58612.654778525.7048
Population (total)POP4.23e+076.10e+07465,1583.23e+08
Natural gas end-user price (US$ per MWh)GP75.9657331.6834915169.6404
Electricity end-user price (US$ per MWh)EP201.982367.5457263.73405.56
Per capita income (current US $)INC38234.4121450.67384.258119225.4
Annual Heating degree Daysa (baseline: 18°C)HDD3078.3151573.367128.90468929.906
Annual Cooling degree Daysb (baseline: 18°C)CDD252.071608.864501875.273

Table 1.

Definition of variables and descriptive statistics. N*T (number of observations × time series) =348.

Heating degree day (HDD) is a quantitative index reflecting demand for energy to heat buildings or businesses.

Cooling degree day (CDD) is a quantitative index reflecting demand for energy to cool buildings or businesses.

The elderly, the urban population, and population density are three crucial variables in the model, taking into account the paper’s demographic methodology. The mid-year population is divided by the land area in square kilometers to get the density.

Figure 4 shows the evolution of the senior population (over 65 years) and the urbanization rate (the fraction of the population living in cities) in OECD countries. The percentage of the population over 65 years old and the population living in urban regions are both growing over time, as seen in Figure 4. These two variables are critical since they account for around 17 and 79% of the total in 2016.

Figure 4.

Elderly and population rate (percentage, 2000–2016). Reference: own elaboration based on World Bank data (with tableau public software).

Then, the population density4 of different OECD countries in 2016 was exposed as the year of reference (the most recent one). Figure 5 shows that population density varies dramatically across countries and Korea, Japan, Belgium and Netherlands are the densest countries.

Figure 5.

Population density (inhabitants/km2, 2016). Reference: own elaboration based on World Bank data (with tableau public software).

The entire population aged 65 or older in the total population at the national level is included in the empirical model, indicating that each country is aging. The urbanization rate is also included to look into the role of the growing rate of people moving from rural to urban areas. Furthermore, the study considers population density when examining the effects of densely populated countries on residential natural gas use. Other control variables are also included in the model. To determine the income level, the study takes into account the total population (POPit), end-user natural gas price (GPit), and gross domestic product per capita (INCit). In addition, to account for weather effects, the price of electricity (EPit) as the closest replacement, as well as the Heating and Cooling Degree Days (HDDit) and CDDit) were included.

The correlation matrix was used to show the statistical correlation between the dependent variable and the regressors. Table 2 shows that per capita natural gas demand is positively correlated with the fraction of the elderly (ELDit) and population density (DENit) (GTit). The matrix also reveals a negative relationship between urbanization rate (URBit) and natural gas consumption. Furthermore, natural gas usage is inversely connected with its own prices (GPit) and cooling degree days (CDDit), but favorably with per capita income (INCit), population (POPit), electricity price (EPit), and heating degree days (HDDit). The correlation matrix, on the other hand, is a basic statistical link between two variables; as a result, a more precise specification is required to investigate the impact of demographic variables on natural gas demand.


Table 2.

Correlation matrix.

3.2 Econometric technique

Household production theory, which considers the consumer as a firm, assumes that households employ inputs (natural gas in this case) to manufacture nonmarket commodities or utility-yielding items. Thus, the demand for welfare services such as space heating, water heating, cooking, and so on is not directly produced from natural gas. Households use natural gas to meet these demands.

The production function of the welfare services S can be written as:


Natural gas is denoted by the letter G. The quantity of natural gas acquired determines the output, which is welfare services (S). In fact, welfare services S, as well as total consumption X, are regarded to be an element of the household’s utility function. The demographic factors Z and the weather of the household’s country, designated W, have an impact on this utility function. As a result,


The above utility function is maximized by the household under a budget constraint:


Where Yis the household income and Pis the price of natural gas. Solving this optimization issue involves demand function for natural gas:


Based on the Eq. (4) and employing a log linear specification, The static model is as follows:


Where GTit is per capita residential natural gas consumption, ELDit is the elderly population, URBit is the urban population, DENit is the population density in the country I in year t, Xit is the sum of the control variables that are likely to influence per capita natural gas consumption, and eit is the error component to account for unobserved factors. The parameters were simply interpreted as demand elasticities after the dependent variable and regressors were transformed into logarithms.

When estimating a static energy demand model using panel data, the endogeneity problem is often addressed by using the fixed or random effects with the Within estimator or GLS [13], respectively, to avoid the heterogeneity bias with a constant term that the OLS may suffer from.

Nerlove [14], on the other hand, claims that economic behavior models are dynamic in nature, and that current behavior is dependent on the state of the system defining it. Furthermore, according to Gutiérrez [15], disregarding the influence of path-dependency can lead to erroneous estimations of the entire variables. The lagged dependent variable was inserted on the right-hand side of the equation to compensate for the intrinsic dynamic feature of the demand function, assuming that natural gas demand in the residential sector is affected by prior levels. As a result, the dynamic version of the natural gas demand model is:


According to Achen [16], the lagged dependent variable will capture not only the impact of the omitted variables, but also the impact of the variables that have previously been included, with the possibility of modifying or decreasing their impact, sometimes to the point of being inconsequential.

In reality, adding a lagged dependent variable to a static model will result in skewed results because the latter variable may be associated with the error component eit. Thus, the within transformation and GLS will be biased since (Yi,t−1-1) is linked with (eit-i.) and its consistency is dependent on Tbeing big, where yis the log natural gas consumption per capita and. -1 are the average lagged log per capita natural gas consumption inside nation i. [13]. To deal with this problem, one can first change the model to remove the country-fixed effects:


Then using Yi,t−2 as an instrument for Yi,t-1 [17]. Because it does not use all of the available moment conditions, this instrumental variable estimation approach produces consistent but not necessarily efficient estimates of the model’s parameters [18]. Arellano and Bond [19] presented a generalized method of moment (GMM), which entails using the orthogonality criteria that exist between lagged values of Yit and the disturbances eit in Eq to add additional instruments (7). As a result of this discussion, as well as the fact that the dataset had N = 29 and T = 12, the dynamic demand function was estimated and stated in Eq. (6) using a dynamic system GMM using differenced and lagged variables as instruments for the differenced and level equations, respectively. The GMM system uses the entire set of instruments and puts cross-equation limitations on the coefficients entering the two models (corresponding to the full set of orthogonality conditions for both models). The validity of the orthogonality assumptions in the estimate procedure determines the consistency of the system GMM estimator. Two specification tests are used, as recommended by Arellano and Bond [19, 20]; and Blundell and Bond, [21]. The Arellano-Bond tests (AR1) and (AR2) were used to assess the first and second serial correlation among error terms and the Sargan/Hansen test was used to check the validity of the instruments. These experiments assist us in determining the most appropriate model for national natural gas demand.


4. Empirical results

This section presents the estimated results and their implications on how the demand for natural gas responds to demographic and non-demographic factors in OECD countries. The estimation results for the static model utilizing the fixed effect to adjust for unobserved heterogeneity are shown in Table 3. This indicates that the majority of the coefficients are statistically significant and are nearly identical to previous research findings. In terms of the overall picture, per capita residential demand is statistically significant and positively correlated with urbanization rate, electricity prices, heating and cooling degree days, while elderly, population density, and natural gas prices are negatively correlated with per capita natural gas consumption.

Dependent variable: log residential natural gas consumption per capitaCoefficients
lELD−0.63*** (0.16)
IURB1.96** (0.68)
IDEN−0.31 ** (0.15)
IPOP0.54 (0.45)
IGP−0.12** (0.06)
IINC0.07 (0.09)
IEP0.2** (0.08)
IHDD0 .38*** (0.1)
CDD0.01** (0.1)
Intercept−10.76* (6.5)
Sample size348
R2 within0.13
R2 between0.06

Table 3.

Estimation results: static model—FE.

Significant at the10% level.

Significant at the5% level.

Significant atthe1%level.

Note: Figures in () are the standard error.

However, because static models aren’t ideal for observing economic trends over time, a dynamic panel model was employed to estimate energy demand, which may be more accurate than a static model. In addition, using a lagged dependent variable as a regressor to investigate residential natural gas demand violates the rigorous exogeneity constraint in static models; thus, the lagged dependent variable is included in the explanatory variables in the dynamic model in this work. The dynamic estimating model for residential natural gas demand is shown in Table 4.

Dependent variable: log residential natural gas consumption per capitaTwo-step system
Lagged lGT0.1* (0.05)
lELD−0.61** (0.19)
IURB1.1 * (0 .65)
IDEN−0.16* (0.09)
IPOP0.82** (0.26)
IGP−0.21*** (0.06)
IINC0.15 (0.10)
IEP0.08 (0.10)
IHDD0.78 *** (0 .11)
CDD0.01 *** (0.01)
Sample size319
Hansen test3.96 (0.79)
Arellano-Bond test for AR(1)−1.62 (0.11)
Arellano-Bond test for AR(2)−0.71 (0.48)

Table 4.

Estimation results: dynamic GMM estimation.

Significant at the10% level.

Significant at the5% level.

Significant atthe1%level.

Note: Figures in () are corrected standard error.

The regression results for the dynamic model generated using the two-step system GMM estimator are shown in Table 4. First, the findings suggest that the lagged value of natural gas has a beneficial impact on demand. Furthermore, the calculated results suggest that the elderly coefficient (ELDit) in the residential natural gas demand dynamic equation is significantly negative. That is, as the population ages, natural gas usage declines. The senior elasticity of per capita natural gas demand is estimated to be −0.6. This appears counterintuitive, given that older people are more sensitive to temperature and consume more natural gas to suit their comfort demands, notably for space and water heating, and they spend more time at home [22]. However, possible explanations for the negative effect of ELD on residential natural gas consumption could be related to the fact that either old people in OECD countries have an economic behavior, or that OECD old people do not spend much time inside the house and prefer to do more activities outside. It could also be due to the fact that older folks prefer electric equipment to gas appliances. Another explanation is that, with the recent ubiquity of electrified houses, a major share of appliances used for daily life at home in OECD countries may be electric appliances. The fact that the estimation revealed a negative influence on residential natural gas demand backs up the previous claims.

Second, in the residential natural gas demand dynamic equation, the urbanization rate (URBit) coefficient is notably positive. The elasticity of per capita natural gas demand in relation to urbanization rate, in other words, is expected to be +1.1. The result appears to be consistent with previous research, as natural gas is widely used in cities and less so in rural areas, where coal and wood are commonly used. Coal, wood, and other conventional fuels are being replaced by cleaner energy sources, particularly natural gas and electricity, as the population of rural areas migrates to cities and towns.

Third, when it comes to the influence of population density (DENit), the study reveals that the DENit coefficient is strongly negative. A decrease in per capita natural gas usage occurs when population density rises. The population density elasticity of per capita natural gas demand is estimated to be around −0.16. Residential natural gas consumption per capita is lower in countries with dense populations. A probable explanation for the finding is that the majority of OECD countries apply energy saving processes and choose to use central heating systems, which provide warm space and water to the entire interior of the building.

Almost all of the control variable estimated coefficients are statistically significant, have the predicted sign, and have appropriate magnitudes. Gas prices (GPit) have a negative impact on per capita natural gas demand, whereas population (POPit), heating degree days (HDDit), and cooling degree days (CDDit) have positive impacts on residential natural gas consumption, indicating that natural gas demand is more sensitive to hot than cold weather.

When it comes to assessing the dynamic model, the AR1 and AR2 tests show that there is no significant autocorrelation in the model, which is a need for the instruments’ validity. Furthermore, the Hansen test demonstrates that the null hypothesis, namely, that the over-identifying constraints are valid, is not rejected.


5. Conclusion

Using a static and dynamic model, this study looked at per capita natural gas demand in the residential sector in OECD countries from 2005 to 2016. The goal of this study is to add to the empirical literature on residential natural gas demand research by analyzing the impact of demographic characteristics on natural gas consumption in the OECD environment, specifically urbanization rate, density, and elderly population.

In fact, no previous study in the OECD has employed a comprehensive model to estimate residential natural gas demand. Previous research has frequently focused on price and income. It is suggested that adding demographic variables will be helpful for policymakers.

A considerable effect of urbanization, density, and elderly population on residential natural gas usage was discovered using a dynamic framework. Due to policy efficiency, rapid urbanization leads to the use of more natural gas per capita, whereas population density leads to the use of less natural gas per capita, especially in buildings. Furthermore, older adults use less natural gas per capita and are more likely to use electric appliances.

Although previous studies have shown that older persons use more energy for heating, these findings appear to be counterintuitive in terms of economic behavior, preference, or the ubiquity of such appliances in OECD buildings.


  1. 1. Alberini A, Gans W, Velez-Lopez D. Residential consumption of gas and electricity in the US: The role of prices and income. Energy Economics. 2011;33(5):870-881
  2. 2. Al-Sahlawi MA. The demand for natural gas: A survey of price and income elasticities. The Energy Journal. 1989;10(1)
  3. 3. Burke PJ, Yang H. The price and income elasticities of natural gas demand: International evidence. Energy Economics. 2016;59:466-474
  4. 4. Dagher L. Natural gas demand at the utility level: An application of dynamic elasticities. Energy Economics. 2012;34(4):961-969
  5. 5. Gautam TK, Paudel KP. The demand for natural gas in the northeastern United States. Energy. 2018;158:890-898
  6. 6. Chen J, Wang X, Steemers K. A statistical analysis of a residential energy consumption survey study in Hangzhou, China. Energy and Buildings. 2013;66:193-202
  7. 7. Kronenberg T. The impact of demographic change on energy use and greenhouse gas emissions in Germany. Ecological Economics. 2009;68(10):2637-2645
  8. 8. Liao HC, Chang TF. Space-heating and water-heating energy demands of the aged in the US. Energy Economics. 2002;24(3):267-284
  9. 9. Ota T, Kakinaka M, Kotani K. Demographic effects on residential electricity and city gas consumption in the aging society of Japan. Energy Policy. 2018;115:503-513
  10. 10. Elnakat A, Gomez JD, Booth N. A zip code study of socioeconomic, demographic, and household gendered influence on the residential energy sector. Energy Reports. 2016;2:21-27
  11. 11. Arbabi H, Mayfield M. Urban and rural—Population and energy consumption dynamics in local authorities within England and Wales. Buildings. 2016;6(3):34
  12. 12. He GJ, Xiao RG, Liang S. Prediction and Influencing Factors Analysis of Natural Gas Consumption in China Based on SPSS. Mexico: AMCCE; 2015
  13. 13. Baltagi BH, Baltagi BH, editors. A Companion to Theoretical Econometrics. Oxford: Blackwell; 2001
  14. 14. Nerlove M (2000, June). An essay on the history of panel data econometrics. In Panel Data Conference in Geneva
  15. 15. Gutiérrez LH. The effect of endogenous regulation on telecommunications expansion and efficiency in Latin America. Journal of Regulatory Economics. 2003;23(3):257-286
  16. 16. Achen CH. Why lagged dependent variables can suppress the explanatory power of other independent variables. In Annual Meeting of the Political Methodology Section of the American Political Science Association, UCLA. 2000;20(22):07-2000
  17. 17. Anderson TW, Hsiao C. Formulation and estimation of dynamic models using panel data. Journal of Econometrics. 1982;18(1):47-82
  18. 18. Ahn SC, Schmidt P. Efficient estimation of models for dynamic panel data. Journal of Econometrics. 1995;68(1):5-27
  19. 19. Arellano M, Bond S. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies. 1991;58(2):277-297
  20. 20. Arellano M, Bover O. Another look at the instrumental variable estimation of error-components models. Journal of Econometrics. 1995;68(1):29-51
  21. 21. Blundell R, Bond S. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics. 1998;87(1):115-143
  22. 22. Yamasaki E, Tominaga N. Evolution of an aging society and effect on residential energy demand. Energy Policy. 1997;25(11):903-912


  • Organization for Economic Co-operation and Development.
  • Health at a Glance 2017.
  • Trends in Urbanization and Urban Policies in OECD Countries 2010.
  • Inhabitants/km2.

Written By

Mohamed Jaouad Malzi

Submitted: October 22nd, 2021 Reviewed: January 3rd, 2022 Published: April 11th, 2022