Residential Natural Gas Demand in Illinois - AgEcon Search

5 downloads 0 Views 179KB Size Report
This study estimates the demand for residential natural gas in the state of Illinois using ... Long-run elasticity estimates show that only real residential natural gas.
JRAP 41(2): 138-147. © 2011 MCRSA. All rights reserved.

Residential Natural Gas Demand in Illinois: Evidence from the ARDL Bounds Testing Approach James E. Payne#, David Loomis*, and Renardo Wilson* University of South Florida Polytechnic#, Illinois State University* – USA Abstract. This study estimates the demand for residential natural gas in the state of Illinois using an autoregressive distributed lag (ARDL) bounds testing approach based on annual data from 1970 to 2007. The ARDL bounds testing approach reveals a long-run equilibrium relationship between natural gas consumption per capita and real residential natural gas prices, real personal disposable income per capita, real residential electricity prices, real fuel oil prices, and heating degree days. Long-run elasticity estimates show that only real residential natural gas prices, real residential electricity prices, and heating degree days are statistically significant. The results from the corresponding error correction model indicate that only real residential natural gas prices and heating degree days are statistically significant. While the long-run elasticity estimates are larger than the short-run elasticity estimates, both the short-run and long-run elasticity estimates are less than one in absolute terms. Furthermore, the speed of adjustment towards long-run equilibrium is approximately 1.42 years.

1. Introduction Residential natural gas consumption in the U.S. has exhibited a steady downward trend after peaking in 1996. According to the Energy Information Administration, this downward trend in residential natural gas consumption can be attributed to several factors: (1) improvement in the energy efficiency of new home construction and heating equipment; (2) an increase in the number of natural gas customers who use alternative energy sources for primary heating; (3) population migration to warmer climates; and (4) an increase in natural gas prices due to the deregulation of the natural gas industry with the Natural Gas Policy Act of 1978 (Krichene, 2002; Kliesen, 2006; EIA, 2010). To gain a better understanding of residential natural gas consumption behavior for the entire U.S., it is worthwhile to gain a regional perspective. As pointed out by Herbert and Barber (1988), an examination of regional residential natural gas demand and corresponding elasticity estimates can

be used to examine issues surrounding price discrimination and the effect of natural gas price increases on the real incomes of households. Furthermore, elasticity estimates are also useful to gas utilities in rate proceedings before state public utility commissions. Of the four U.S. census regions, the Midwest region yields the greatest amount of residential natural gas consumption. With the largest population (12.9 million) in the Midwest, this study undertakes modeling residential natural gas consumption for the state of Illinois. Illinois’ energy demands are substantial, with over 81% of all homes within Illinois heated using natural gas. Moreover, residential natural gas consumption accounts for over 47% of all natural gas consumption in Illinois. As shown in Figure 1, Illinois residential natural gas consumption per capita has been trending downward, a pattern similar to the residential natural gas consumption per capita for the U.S. as a whole.

Residential Natural Gas Demand in Illinois

139

46 44 42 40 38 36 34 32 30

1970

1980

1990

2000

2007

Figure 1. Illinois residential natural gas consumption (million cubic feet) per capita. This study estimates the demand for residential natural gas in Illinois using an autoregressive distributed lag (ARDL) model. A survey of the residential natural gas demand literature reveals the absence of any published articles specific to the state of Illinois. Moreover, the literature also shows the absence of more recent studies on residential natural gas demand at the state level in general, which is particularly surprising given the growing interest in natural gas as a bridge to alternative renewable energy sources. To address the gap in the natural gas literature, the study will do the following: (1) simultaneously model the short-run and long-run residential natural gas demand dynamics; (2) provide short-run and long-run income, own price, and cross price elasticity estimates; (3) determine the impact of heating degree days on residential natural gas demand; and (4) evaluate the residential natural gas demand model’s performance in tracking actual residential natural gas consumption behavior. This analysis should provide a platform for similar studies of residential natural gas demand at the state level. Section 2 provides a brief overview of the literature on residential natural gas demand in the U.S. Section 3 discusses the data, methodology, and results. Concluding remarks are given in Section 4.

2. Brief overview of the literature While studies on residential natural gas demand have been examined across a number of countries, the literature review will highlight research pertaining to residential natural gas demand at the subnational level in the U.S. 1 Balestra and Nerlove (1966) recognize that in the short-run the change in the relative price of natural gas does not impact consumer choices once the stock of gas appliances are in place. However, the relative price of alternative fuels, including natural gas, is relevant in the planning stage of the type of fuel to be used in heating. Balestra and Nerlove (1966) pool annual data from 1950 to 1962 for 36 states to estimate a dynamic model of residential natural gas demand as a function of real natural gas prices, real per capita income, and total population. The results indicate that timeinvariant regional effects account for roughly 75% of the total residual variance for residential natural gas demand, with the estimated long-run price and income elasticities of -0.63 and 0.62, respectively. Bloch (1980) investigates the determinants of residential natural gas demand based on data from 151 families in Twin Rivers, New Jersey, for 27 winter Natural gas demand studies have also been examined across countries. See, for example, Verhulst (1950), Berndt and Watkins (1977), Griffin (1979), Pindyck (1980), Estrada and Fugleberg (1989), Eltony (1996), Sorak and Satman (2003), Labandeira et al. (2006), Potocnik et al. (2007), Asche et al. (2008), Yoo et al. (2009), Erdogdu (2010), and Forouzanfan et al. (2010). 1

140

months over the period October 1971 to January 1976. Block (1980) specifies natural gas demand as a function of the real price of natural gas, the number of heating degree days, and a time trend. The results indicated that the coefficient on heating degree days was greater than one, which suggests that household indoor temperature and thermostat settings vary inversely with outside temperatures. The coefficient for real natural gas prices is negative and less than one in the range of -0.666 to -0.583. The coefficient on the time trend is also negative, reflecting the secular decline in natural gas consumption. Beierlen et al. (1981) estimate residential natural gas demand for the northeastern region of the U.S. over the period 1967 to 1977, employing seemingly unrelated regression to find short-run and long-run price elasticity estimates of -0.053 and -3.42, respectively; income was statistically insignificant. The electricity price had a positive impact on residential natural gas demand while the price of fuel oil had a negative impact. Beierlen et al. (1981, p. 406) attribute the negative impact of fuel oil price to the possibility that increases in fuel oil prices accompany increases in all other petroleum products which comprise a substantial portion of household expenditures. In turn the increase in fuel oil prices induces a substantial decrease in real income. Hence, the negative income effect is outweighing the positive substitution effect. Barnes et al. (1982) utilize household data from the 1972-1973 Consumer Expenditure Survey to investigate household natural gas consumption in relation to the stock of gas appliances and demographic profile of the households. Using an instrumental variable technique, Barnes et al. (1982) find an overall short-run own-price elasticity of -0.682 with significant variation across end-use categories. Blattenberger et al. (1983) and Taylor et al. (1984) use pooled time-series cross-section data for all 48 continental U.S. states over the period 1960 to 1974 while recognizing natural gas availability to estimate natural gas demand. The variancecomponents estimation technique of Balestra and Nerlove (1966) shows that, in addition to heating degree days, own-price, electricity price, and income are statistically significant determinants, while fuel oil price is statistically insignificant. The own-price elasticity estimates are -0.049 in the short run and -0.333 in the long run. The income elasticity estimates are 0.113 in the short run and 0.769 in the long run. The electricity-price elasticity estimates are 0.048 in the short run and 0.327 in the long run.

Payne, Loomis, and Wilson

Liu (1983) analyzes U.S. aggregate as well as regional natural gas consumption across sectors over the period 1960 to 1978 using both ordinary least squares and two-stage least squares. Liu finds in the aggregate for residential natural gas that own price, income, and a dummy variable for the oil embargo of the 1970s are statistically significant, while real electricity prices and oil prices are statistically insignificant. The own-price elasticity estimate is -0.490, and the income elasticity estimate is 0.553. The results also reveal significant variation in elasticity estimates across regions. Herbert (1986a) discusses the importance of the appropriate measurement of heating degree days and the use of the lagged dependent variable as a proxy for the fixity of the household stock of equipment using natural gas. 2 Using data for six Northeastern states over the period 1970 to 1982, Herbert provides ordinary least squares estimates of residential natural gas demand. The own price elasticity estimate is -0.22 in the short-run and -0.56 in the long-run. The income elasticity estimate is 0.17 in the short-run and 0.44 in the long-run. The cross price elasticity for real electricity prices is 0.10 in the short-run and 0.26 in the long-run. In addition, heating degree days and the OPEC oil shock dummy variables were also statistically significant. Grady (1986) utilizes pooled data from 1968 to 1978 for all 50 U.S. states, grouped into five categories based on natural gas availability, to estimate residential natural gas demand using an error component model. The results indicate that only own price and heating degree days are statistically significant, while prices for electricity and fuel oil, income, and the ratio of households using natural gas relative to total households are statistically insignificant. Moreover, Grady reports the average short-run and long-run own-price elasticity estimates across the five categories of states as -0.1 and -0.4, respectively. Herbert (1987a) pools data from 1960 to 1982 for seven states in the east central part of the U.S. to estimate residential natural gas demand using generalized least squares. Herbert specifies residential natural gas as a function of own price, heating degree days, income per capita, real electricity price, a dummy variable for the 1973-1974 oil embargo, and a variable measuring residential tax credits for energy conservation and renewable energy

For a detailed discussion of data measurement issues related to residential natural gas, see Herbert (1986b, 1987b, 1988a,b), Herbert and Barber (1988), Herbert and Kott (1988), and Herbert and Kreil (1988, 1989).

2

Residential Natural Gas Demand in Illinois

investment expenditures. The results are quite robust, with statistically significant short-run elasticity estimates of -0.30 for own price, 0.59 for heating degree days, 0.10 for electricity prices, and .058 for income. The coefficients for the oil embargo and residential tax credits are also statistically significant and negative. Lin et al. (1987) examine the demand for residential natural gas for the four U.S. census regions and the nine sub-census regions over the 1960 to 1983 period. Within a reduced-form dynamic model of residential natural gas demand, Lin et al. employ the error components seemingly unrelated regression procedure to find that, based on the average elasticity estimates across regions, the own-price elasticity estimate is -0.154 in the short-run and -1.215 in the long run. The average income elasticity estimate is 0.105 in the short run and 0.574 in the long run. Furthermore, the respective own price, income, and cross price elasticity estimates and adjustment speeds to the desired level of residential natural gas consumption vary across regions. Maddala et al. (1997) use several shrinkage estimators to address issues with pooling data in order to obtain reasonable elasticity estimates of residential natural gas demand for each of the 49 U.S. states using data over the 1970 to 1990 period. Residential natural gas demand is modeled as a function of real personal income per capita, real residential natural gas prices, real residential electricity prices, and heating and cooling degree days. Own-price elasticity estimates range from -0.092 to -0.177 in the short run and from -0.289 to -1.358 in the long run. The income elasticity estimates range from 0.048 to 0.307 in the short run and -0.425 to 0.491 in the long run.

3. Data, methodology, and results Annual data for the period 1970 to 2007 was obtained from the U.S. Energy Information Administration (EIA), the U.S. Bureau of Labor Statistics (BLS), and the U.S. Bureau of Economic Analysis (BEA). The units of measurement for each variable are as follows: natural gas consumed by (delivered to) the residential sector measured in millions of cubic feet; natural gas prices in the residential sector measured in dollars per million Btu; residential electricity prices measured in dollars per million Btu, and fuel oil prices measured per million Btu. As indicated by the EIA, the measure for heating degree days is computed for each day by subtracting the

141

average of the day’s high and low temperatures from the base temperature (65 degrees), with negative values set equal to zero. Each day’s heating degree days are summed to create a heating degree day measure for a specified reference period. The following are variable names and definitions related to the state of Illinois:

ngc rngcp rpdi rep rfop hdd

residential natural gas consumption per capita real residential natural gas price real personal disposable income per capita real residential electricity price real fuel oil price heating degree days.

Real values for the above-mentioned variables were derived using the Midwest consumer price index. The inclusion of these variables parallels the model specification set forth by Beierlen et al. (1981). As noted by Bohi and Zimmerman (1984, pp. 112-113), among others, an issue in the estimation of natural gas demand is the use of marginal, average, or inframarginal natural gas prices. Though marginal prices are considered pertinent to consumption decisions, one can argue that most residential customers have limited insight on marginal rates within block schedule pricing. Thus, customers are more likely to respond to perceived average price. We begin by examining the time series properties of the respective variables using the augmented Dickey-Fuller (ADF, 1979) and Phillips-Perron (PP, 1988) unit root tests. In the case of both tests, the null hypothesis is a unit root whereas the alternative hypothesis is stationarity in levels. Table 1 presents the results of the unit root tests. Residential natural consumption per capita (ngc) and heating degree days (hdd) appear to be integrated of order zero based on both the ADF and PP unit root tests. The ADF and PP unit root tests suggest that real residential electricity prices (rep) and real fuel oil price (rfop) are both integrated of order one. On the other hand, the results from the unit root tests are mixed with respect to real residential natural gas prices (rngcp) and real personal disposable income per capita (rpdi). Unlike the Engle-Granger (1987) and JohansenJuselius (1990) cointegration procedures, which require the respective time series be integrated of order one, the ARDL approach to cointegration does

142

Payne, Loomis, and Wilson

Table 1. Unit root tests. Variable

ngc

∆ngc rngcp ∆rngcp rpdi ∆rpdi rep ∆rep rfop ∆rfop hdd ∆hdd

ADF -3.799b

PP -3.746b

-------

-------

-3.479c

-1.879

-------

-6.727a

-5.579a

-2.722

-------

-5.864a

-1.570

-1.596

-4.076b

-4.078b

-1.795

-1.849

-6.144a

-6.145a

-4.298a

-4.812a

-------

-------

logarithms. The null hypothesis of no cointegration ( H 0 : π 1 = π 2 = π 3 = π 4 = π 5 = π 6 = 0) is tested against the alternative hypothesis of cointegration

( H 0 : π 1 ≠ 0, π 2 ≠ 0, π 3 ≠ 0, π 4 ≠ 0, π 5 ≠ 0, π 6 ≠ 0)

Notes: ADF and PP unit root tests include an intercept and trend. Critical values for the ADF and PP unit root tests are as follows: 1%(a) -4.235, 5%(b) -3.540, and 10%(c) -3.202.

not require the variables to be integrated of the same order (Pesaran et al., 1997, and Peasran and Shin, This attribute of the ARDL approach is 2001). 3 quite useful given the relatively low power of unit root tests. The ARDL procedure allows for the simultaneous modeling of the short-run and long-run dynamics, providing short-run and long-run elasticity estimates associated with residential natural gas demand. This procedure begins by estimating a conditional ARDL-ECM model as follows: 4

∆ngc t = α 0 + ∑ip=1 γ i ∆ngc t −i + ∑ip=1 δ i ∆rngcp t −i + ∑ip=1 φi ∆rpdit −i + ∑ip=1θ i ∆rep t −i + ∑ip=1 λi ∆rfopt −i + ∑ip=1ψ i ∆hdd t −i

(1)

+ π 1ngc t −1 + π 2 rngcp t −1 + π 3 rpdit −1 + π 4 rep t −1 + π 5 rfop t −1 + π 6 hdd t −1 + ε t where α 0 is the drift component and ε t are the random errors. All variables are in natural 3 Madlener (1996) provides a survey of the econometric approaches of cointegration analysis in the estimation of residential energy demand. See, for example, Pindyck (1979), Bohi (1981), Bohi and Zimmeran (1984), and Al-Sahlawi (1989) for comprehensive examinations of energy demand, including residential.

Given that natural gas was heavily regulated prior to the Natural Gas Policy Act of 1978 which deregulated the natural gas industry, preliminary analysis incorporating a dummy variable for the period prior to 1978 (the 1970 to 1977 period) did not indicate statistical significance. 4

via a partial F-test. However, as discussed by Pesaran et al. (2001), the asymptotic distribution of the F-statistic is non-standard, regardless of whether the variables are I(0) or I(1). Pesaran et al (2001) provide lower and upper bound critical values, where the lower bound critical values assume all variables are I(0) while the upper bound critical values assume all variables are I(1). If the calculated F-statistic is above the upper critical value, the null hypothesis of no cointegration can be rejected irrespective of the orders of integration of the variables. If the calculated F-statistic is below the lower critical value, the null hypothesis of no cointegration cannot be rejected. However, if the calculated F-statistic falls between the lower and upper critical values, the result is inconclusive. This process is carried out with each variable as the dependent variable in order to infer which variables are long-run forcing variables in the determination of the long-run equilibrium relationship. Table 2 reports the partial F-statistics to test the null hypothesis of a long-run relationship among the variables along with the asymptotic critical values of the bounds testing procedure. As shown in Table 2, the F-statistic exceeds the upper critical bound at the 10% significance level with respect to residential natural gas consumption. However, for the remaining variables the F-statistic either falls below the lower critical bound or in the inconclusive region. Thus, the remaining variables are considered longrun forcing variables in the long-run equilibrium relationship. Table 2. ARDL bounds testing procedure tests of long-run relationships. F(ngc|rngcp, rpdi, rep, rfop, hdd) F(rngcp|ngc, rpdi, rep, rfop, hdd) F(rpdi|ngc, rngcp, rep, rfop, hdd) F(rep|ngc, rngcp, rpdi, rfop, hdd) F(rfop|ngc, rngcp, rpdi, rep, hdd) F(hdd|ngc, rngcp, rpdi, rep, rfop)

F-statistic 3.5715c 1.3615 0.8506 3.0819 1.9063 2.9889

Notes: Critical value bounds for the partial F-statistics with intercept and no trend for the number of variables, k = 6: 99% I(0) 3.267 I(1) 4.540 95% I(0) 2.476 I(1) 3.646 90% I(0) 2.141 I(1) 3.250 10 percent significance level denoted by “c”.

Residential Natural Gas Demand in Illinois

143

Panel A of Table 3 displays the long-run coefficients, which are long-run elasticity estimates given the logged variables, from the models based on the Akaike information criterion for the ARDL procedure as outlined in the Appendix. As expected, the coefficient for real residential natural gas prices is negative, -0.263, and significant at the 1% level. The coefficient for real personal disposable income per capita, though positive, is statistically insignificant. Beierlen et al. (1981), Grady (1986), and Herbert (1986a) also find income to be statistically

insignificant and provide several explanations. First, there is little variability in income defined in constant dollars at the state level, and income is frequently correlated with other variables that do exhibit variability. Second, with an increase in real income, consumers may turn up their thermostats, but they may also spend less time at home and increase dining at restaurants, entertainment, and vacations. Such activities would result in a decrease in the use of gas appliances.

Table 3. Long-run estimates and ARDL error correction model. Panel A: Long-Run Estimates

ngct = −2.0889 − 0.26297 rngcpt + 0.02401rpdit + 0.12239rept + 0.06726rfopt + 0.62487hdd t (-0.89)

(-2.79)a

(0.16)

(1.79)c

(1.43)

(2.88)a

Notes: t-statistics are in parentheses. Asymptotic standard errors based on Bewley (1979), see Appendix. Significance levels: a(1%) and c(10%).

Panel B: ARDL Error Correction Model

∆ngct = −0.18450∆rngcpt + 0.01684∆rpdit − 0.16102∆rept + 0.04719∆rfopt (-2.88)a

(0.16)

(-1.36)

(1.51)

+ 0.71705∆hdd t + 0.16983∆hdd t −1 − 0.70159ectt −1 (8.39)a

(1.91)c

Adj. R2 = 0.814 F = 23.17

(-4.61)a

[0.00]a

Notes: t-statistics are in parentheses and probability values in brackets. Significance level: a(1%) and c(10%).

3.8

3.7

3.6

3.5

3.4

1972

1981

1990 LNGC

Figure 2. Plot of actual and fitted values of LNGC.

1999 Fitted

2007

144

With respect to substitutes, the coefficient for real residential electricity prices is positive, 0.122, and significant at the 10% level, while the coefficient for real residential fuel oil prices is positive but statistically insignificant. The statistical insignificance of fuel oil prices has been addressed by Herbert (1986a) as well. As pointed out by Herbert, virtually all of the increase in the percentage of gas space-heating customers occurred in the 1970s and early 1980s as a result of switching from oil furnaces to gas furnaces in response to the OPEC oil price increases. Furthermore, Herbert also points out that all energy prices are highly correlated, making the coefficient estimates less precise with the inclusion of multiple energy prices. The coefficient for heating degree days is positive, 0.626 and statistically significant at the 1% level. Thus, the long-run results indicate that residential natural gas consumption in Illinois is influenced by real residential natural gas prices, real residential electricity prices, and heating degree days. In the long-run residential natural gas consumption is price inelastic, as a 1% increase in real residential natural gas prices decreases residential natural gas consumption by 0.264%. A 1% increase in real residential electricity prices increases residential natural gas consumption by 0.123%, whereas a 1% increase in heating degree days increases residential natural gas consumption by 0.626%. The relatively low elasticity estimates are perhaps indicative of limited availability of substitutes, inadequate public knowledge of available substitutes, high economic costs to participate in consumer switching, or some combination thereof. Panel B of Table 3 reports the corresponding ARDL error correction model which will highlight the short-run elasticity estimates. In the short run only real residential natural gas prices and heating degree days have a statistically significant impact on residential natural gas consumption. The short-run elasticity estimate for real residential natural gas prices is -0.185 while the short-run elasticity estimate for heating degree days is 0.888. It is not surprising that in the short run real personal disposable income per capita and the price of substitutes (real residential electricity prices and real fuel oil prices) are statistically insignificant given the rigidity against switching heating sources. The error correction term is negative (-0.703) and significant at the 1% level,

Payne, Loomis, and Wilson

indicating that adjustment toward long-run equilibrium takes roughly 1.42 years. 5 Figure 2 displays the ability of the error correction model to replicate the actual behavior of residential natural gas consumption per capita in Illinois over the period 1970 to 2007. The model adequately captures the downward trend in natural gas consumption as well as the turning points over time.

6. Concluding Remarks Residential natural gas consumption in the U.S. has trended downward over time, which can be attributed to several factors: (1) improvement in the energy efficiency in new home construction and heating equipment; (2) an increase in the number of natural gas customers who use alternative energy sources for primary heating; (3) population migration to warmer climates; and (4) an increase in natural gas prices due to the deregulation of the natural gas industry with the Natural Gas Policy Act of 1978. While a number of studies have examined residential natural gas demand in the U.S. over the period of the 1960s through early 1990s, the literature also lacks more recent studies on residential natural gas demand at the state level in general. Furthermore, a survey of the published literature shows the absence of any published studies specific to the state of Illinois. This study attempts to rectify this omission in the recent literature by estimating the demand for residential natural gas in Illinois using an autoregressive distributed lag (ARDL) model to do the following: (1) simultaneously model the short-run and long-run residential natural gas demand dynamics; (2) provide short-run and longrun income, own price, and cross price elasticity estimates; (3) determine the impact of heating degree days on residential natural gas demand; and (4) evaluate the residential natural gas demand model’s performance in tracking actual residential natural gas consumption behavior. Using annual data from 1970 to 2007, the ARDL bounds testing approach reveals a long-run equilibrium relationship between residential natural gas consumption per capita and real residential natural gas prices, real personal disposable income per capita, real residential electricity prices, real fuel oil prices, and heating degree days. More specifically, the coefficients of the long-run equilibrium relationship

Speed of adjustment is computed as the reciprocal of the absolute value of the coefficient on the error correction term.

5

Residential Natural Gas Demand in Illinois

measuring the long-run elasticity estimates show that only real residential natural gas prices, real residential electricity prices, and heating degree days are statistically significant at least at the 10% level. The own-price long-run elasticity estimate is -0.264, the elasticity estimate with respect to real residential electricity prices is 0.123, and the elasticity estimate with respect to heating degree days is 0.626. The results from the corresponding error correction model indicate that only real residential natural gas prices and heating degree days are statistically significant, yielding short-run elasticity estimates of -0.185 and 0.888, respectively. The error correction term suggests that the speed of adjustment towards the long-run equilibrium is relatively fast at 1.42 years. In addition, the estimated ARDL model performs well in tracking the actual residential natural gas consumption behavior in Illinois. Future research could extend the ARDL modeling platform to the analysis of natural gas demand across various sectors and for other states in order to differentiate the regional variation in elasticity estimates.

References Al-Sahlawi, M.A. 1989. The Demand for Natural Gas: A Survey of Price and Income Elasticities. Energy Journal 10:77-87. Asche, F., O.B. Nilsen, and R. Tveteras. 2008. Natural Gas Demand in the European Household Sector. Energy Journal 29:27-46. Balestra, P., and M. Nerlove. 1966. Pooling Cross Section and Time Series data in the Estimation of a Dynamic Model: The Demand for Natural Gas. Econometrica 34:585- 612. Barnes, R., R. Gillingham, and R. Hagemann. 1982. The Short-Run Residential Demand for Natural Gas. Energy Journal 3:59-72. Beierlen, J.G., J.W. Dunn, and J.C. McConnon. 1981. The Demand for Electricity and Natural Gas in the Northeastern United States. Review of Economics and Statistics 63:403-408. Berndt, E.R., and G.C. Watkins. 1977. Demand for Natural Gas: Residential and Commercial Markets in Ontario and British Columbia. Canadian Journal of Economics 10:97-111. Bewley, R. 1979. The Direct Estimation of the Equilibrium Response in a Linear Dynamic Model. Economics Letters 55:251-276. Blattenberger, G.R., L.D. Taylor, and R.K. Rennhack. 1983. Natural Gas Availability and the Residential Demand for Energy. Energy Journal 4:23-45.

145

Bloch, F.E. 1979. Residential Demand for Natural Gas. Journal of Urban Economics 7:371-83. Bohi, D.R. 1981. Analyzing Demand Behavior: A Study of Energy Elasticities, Baltimore, MD: Johns Hopkins University Press and Resources for the Future. Bohi, D.R., and M.B. Zimmerman. 1984. An Update on Econometric Studies of Energy Demand Behavior. Annual Reviews of Energy 9:105-154. Dickey, D.A., and W.A. Fuller. 1979. Distribution of the Estimate for Autoregressive Time Series with a Unit Root. Journal of the American Statistical Association 74:427-431. Eltony, M.N. 1996. Demand for Natural Gas in Kuwait: An Empirical Analysis Using Two Econometric Models. International Journal of Energy Research 20:957-963. Energy Information Administration. 2010. Trends in U.S. Residential Natural Gas Consumption, June, 1-12. Engle, R.F., and C.W.J. Granger. 1987. Co-integration and Error Correction: Representation, Estimating and Testing. Econometrica 55:251-276. Erdogdu, E. 2010. Natural Gas Demand in Turkey. Applied Energy 87:211-219. Estrada, J., and O. Fugleberg. 1989. Price Elasticities of Natural Gas Demand in France and Germany. Energy Journal 10:77-90. Fourouzanfar, M., A. Doustmohammadi, M.B. Menhaj, and S. Husanzadeh. 2010. Modeling and Estimation of the Natural Gas Consumption for Residential and Commercial Sectors in Iran. Applied Energy 87:268-274. Grady, S.T. 1986. Regional Demand for Natural Gas in the Residential Sector. Review of Regional Studies 16:19-28. Griffin, J.M. 1979. Energy Consumption in the OECD: 1880-2000, Cambridge, MA: Ballinger Publishing Company. Herbert, J.H. 1986a. Data Analysis, Specification, and Estimation of An Aggregate Relationship for Sales of Natural Gas Per Customer. Journal of Economic and Social Measurement 14:165-174. Herbert, J.H. 1986b. Data Analysis of Sales of Natural Gas to Households in the United States. Journal of Applied Statistics 13:199-211. Herbert, J.H. 1987a. Demand for Natural Gas at the State Level: Twenty Years of Effort. Review of Regional Studies 17:79-87. Herbert, J.H. 1987b. Data Matters: Specification and Estimation of Natural Gas Demand Per Customer in the Northeastern United States. Computation Statistics and Data Analysis 5:67-78.

146

Herbert, J.H. 1988a. Institutional and Economic Factors in Residential Gas Markets in the Northeastern United States between 1960-1984. Energy 13:211-216. Herbert, J.H. 1988b. A Data Analysis of Residential Demand for Natural Gas Using Step-Down Multiple Comparison Tests. Applied Stochastic Models and Data Analysis 4:149-157. Herbert, J.H., and L.J. Barber. 1988. Regional Residential Natural Gas Demand: Some Comments. Resources and Energy 10:387-391. Herbert, J.H., and P.S. Kott. 1988. An Empirical Note on Regressions with and without a Poorly Measured Variable Using Gas Demand as a Case Study. The Statistician 37:293-298. Herbert, J.H., and E. Kreil. 1988. Supplement to ‘An Analysis of Monthly Sales of Natural Gas to Residential Customers in the United States’. Energy Systems and Policy 12:43-46. Herbert, J.H., and E. Kreil. 1989. Specifying and Evaluating Aggregate Monthly Natural Gas Demand by Households. Applied Economics 21:13691381. Johansen, S., and K. Juselius. 1990. Maximum Likelihood Estimation and Inference on Cointegration with Applications to the Demand for Money. Oxford Bulletin of Economics and Statistics 52:169-210. Kliesen, K.L. 2006. Rising Natural Gas Prices and Real Economic Activity. Federal Reserve Bank of St. Louis Review Nov/Dec:511-526. Krichene, N. 2002. World Crude Oil and Natural Gas: A Demand and Supply Model. Energy Economics 24:557-576. Labandeira, X., J.M. Labeaga, and M. Rodriguez. 2006. A Residential Energy Demand System for Spain. Energy Journal 27:87-111. Lin, W.T., Y.H. Chen, and R. Chatov. 1987. The Demand for Natural Gas, Electricity and Heating Oil in the United States. Resources and Energy 9:233-258. Liu, B.-C. 1983. Natural Gas Price Elasticities: Variations by Region and by Sector in the USA. Energy Economics 5:195-201. Maddala, G.S., R. P. Trost, H. Li, and F. Joutz. 1997. Estimation of Short- Run and Long-Run Elasticities of Energy Demand from Panel Data Using Shrinkage Estimators. Journal of Business & Economic Statistics 15:90-100. Madlener, R. 1996. Econometric Analysis of Residential Energy Demand: A Survey. Journal of Energy Literature 2:3-32.

Payne, Loomis, and Wilson

Pesaran, M.H., and Y. Shin. 1997. An Autoregressive Distributed Lag Modelling Approach to Cointegration Analysis. In Centennial Volume of Ragnar Frisch edited by S. Strom, A. Holly and P. Diamond, Econometric Society Monograph, Cambridge University Press: Cambridge, UK. Pesaran, M.H., Y. Shin, and R.J. Smith. 2001. Bounds Testing Approaches to the Analysis of Level Relationships. Journal of Applied Econometrics 16:289326. Phillips, P.C.B. and P. Perron. 1988. Testing for a Unit Root in Time Series Regression. Biometrika 75:335-346. Pindyck, R.S. 1979. The Structure of World Energy Demand, Cambridge, MA: MIT Press. Pindyck, R.S. 1980. International Comparisons of the Residential Demand for Energy. European Economic Review 13:1-24. Potocnik, P., M. Thaler, E. Gorekar, I. Grabec, and A. Poredos. 2007. Forecasting Risk of Natural Gas Consumption in Slovenia. Energy Policy 35:42714282. Sorak, H., and A. Satman. 2003. The Degree-Day Method to Estimate the Residential Heating Natural Gas Consumption in Turkey: A Case Study. Energy 28:929-939. Taylor, L.D., G.R. Blattenberger, and R.K. Rennhack. 1984. Residential Energy Demand in the United States: Empirical Results for Natural Gas. Advances in the Economics of Energy and Resources 5:129-140. Yoo, S.-H., H.-J. Lin, and S.-J. Kwak. 2009. Estimating the Residential Demand Function for Natural Gas in Seoul with Correction for Sample Selection Bias. Applied Energy 86:460-465. Verhulst, M.J. 1950. The Theory of Demand Applied to French Natural Gas Industry. Econometrica 18:45-55.

Residential Natural Gas Demand in Illinois

147

Appendix In order to obtain the long-run relationship between the respective variables, the autoregressive distributed lag (ARDL) model is specified in general form as follows:

φ ( L, p ) xt = β1 ( L, q1 ) y t + u t

(A1)

2 p where L is the lag operator, i.e., φ ( L, p ) = 1 − φ1 L − φ2 L − ... − φ p L , and

β i ( L, qi ) = β 0i + β1i L + ... + β qi Lq .

Selection of the order of the ARDL model for the computation of the long-run coefficients is based on Akaike information lag length selection criteria up to two lags. The long-run coefficient estimates are obtained as follows:

θˆi =

βˆi (1, qˆ i ) βˆ0i + βˆ1i + ... + βˆ qˆ ,i , = φ(1, pˆ ) 1 − φˆ1 − φˆ2 − ...φˆp

i = 1,2,..., k

(A2)

The estimates of the asymptotic standard errors of the long-run coefficients are computed by the regression approach advanced by Bewley (1979).