# Effect of thermal improvements in housing on residential energy demand.

The residential sector accounts for approximately one-fifth of U.S.
energy consumption. Home heating is the biggest user among energy using
household activities. Although public concern about energy conservation
has abated, there remains a continuing longrun interest in energy
conservation. Households can use two strategies to conserve energy used
in home heating. One is simply to lower the room temperature. This
strategy substitutes warmth for other consumption goods in the short
run. A second strategy is to improve the efficiency of heat production
within the house by making investments such as adding insulation,
weatherstripping, storm windows, or replacing an inefficient furnace.
This latter strategy results in the possibility of achieving a given
room temperature with less energy. Thus, it can be thought of as input
substitution in the production of warmth. It is the substitution of
housing capital for fuel.

Rising prices of energy coupled with favorable tax treatment of residential conservation investments resulted in considerable input substitution during the late 1970s and early 1980s. For example, in 1981 3.4 million households installed roof insulation, 2.0 million households installed wall insulation, and 5.0 million households added storm windows or storm doors (U.S. Bureau of the Census 1982). The impact these conservation investments have in terms of reduced fuel requirements to achieve a given level of warmth is important for consumers to know when deciding whether to make these investments. Engineers have estimated this reduction by observing the impact of energy conserving modifications on model homes (Delene and Gaston 1976; Hutchins and Hirst 1978; Moyers 1971; O'Neal, Corum, and Jones 1981).

However, this technique does not recognize the role conservation investments play in the behavior of households. The introduction of energy conserving modifications to a house serves to reduce the marginal price per unit of warmth. In the face of lower per unit warmth, households may increase their demand for warmth, offsetting the reduction in fuel demand. This implies that energy conserving modifications may impact energy demand less than engineering computations suggest. Similarly in an examination of demand for gasoline and automobile fuel efficiency, both inputs impact the demand for miles driven (Blair, Koserman, and Topel 1984; Mayo and Mathis 1988). These studies provide evidence that some benefits from increased fuel efficiency are taken in more miles driven. Hence, demand for gasoline does not fall as much as predicted.

Economists have estimated residential demand for energy used in home heating taking into account the thermal characteristics of the house (Neels 1981; Scott 1980). But, with a couple of exceptions (Khazzoom 1986), economists have not generally treated energy for home heating as an input demand derived from the demand for home warmth.(1) As a result economists have not specified the demand functions in a way which takes into account what is known by engineers about heat loss.

In this paper the authors treat demand for household energy used in home heating as a derived input demand, using a heat loss function to represent home warmth production. The authors estimate this demand function and compute energy savings and dollar savings, under the 1980-1981 price regime, associated with conservation investment. These savings calculations incorporate changes in the demand for home warmth resulting from reductions in home warmth prices due to the investments themselves.

THEORETICAL CONSIDERATIONS

The economic model used is the standard utility-maximizing model. Household utility, V, is a function of home warmth, W, and other goods, G, and is conditional on the demographic characteristics of the household, D. Home warmth is produced by using fuel for home heating, F, given the thermal characteristics of the house itself, C, and the outside temperature, T. Investments that improve the thermal characteristics of the house improve marginal productivity of fuel in the production of warmth. Given the thermal characteristics of the house, the production function for home warmth, and household resources, the problem is to choose the level of warmth that maximizes the consumer's utility:

V = v(W, G; D) (1)

W = w(F; C, T) (2)

Y = |p.sub.G~G + |p.sub.F~F (3)

where |p.sub.G~ is the price of all other goods, |p.sub.F~ is the price of fuel, and Y is income. Equation (2) is a production function for home warmth. The consumer chooses a level of home warmth by maximizing utility, equation (1), subject to the production relation and the budget constraint. Here the interest is not in the demand for home warmth, but in the derived demand for fuel arising from the consumer's choice of home warmth. An input demand function for fuel can be obtained by solving this problem, and is given by:

F = F(Y, |p.sub.G~, |p.sub.F~ |Delta~W/|Delta~F; D, C, T, W). (4)

The price of home warmth, |p.sub.F~(|Delta~W/|Delta~F), is defined as the marginal cost of fuel, |p.sub.F~, multiplied by the marginal productivity of fuel in the production of home warmth, |Delta~W/|Delta~F. The marginal productivity of fuel in the production of home warmth depends on thermal characteristics of the house. Characteristics of housing that serve to improve the efficiency of fuel in the production of home heating are equivalent to reductions in the price of fuel. Both reduce the price of warmth to the household.(2)

Equation (2), the production function for home warmth, has been the focus of most engineering work. This work has focused on heat loss, that is, the amount of heat that escapes through the shell of a building over a specified period or, equivalently, the amount of heat needed to keep a house at a specified temperature for a given period. If one knows the heat loss properties of a structure, one can calculate the energy requirements for heating the structure to some specified temperature. In this paper the authors incorporate the engineering specifications for the production relation into the estimates of the demand for fuel equation.

Engineering Specification of the Production Relation

Heat loss occurs because of transmission heat loss and infiltration heat loss. Transmission heat loss per square foot, caused by the heat conductivity of building material, is measured by the heat transmission coefficient, U, the number of Btus transmitted through one square foot of surface of a particular material for one degree of temperature difference between one side of the material and the other, T, over one hour. The heat transmission coefficient is a known property of particular materials. The reciprocal of U is R, the heat resistance coefficient. Although engineers use the heat transmission coefficient, the authors express heat loss as the more familiar heat resistance coefficient. Larger R-values indicate materials that are better at stopping heat from escaping. The R-value of an outside surface is the sum of the individual R-values for the building materials, |R.sub.mi~, and the insulation. As insulation prevents heat from escaping, the effect of insulation can be expressed as a reduction in Btus lost. Let the R-value for insulation be measured per inch of insulation, |R.sub.Ii~, and let |I.sub.i~ be the number of inches of insulation. Then transmission heat loss for a surface area i for one year can be approximated by:

TH|L.sub.i~ = |(1/|R.sub.mi~) - (1/|R.sub.Ii~) x |I.sub.i~~ x |A.sub.i~ x HDD x 24 (5)

where TH|L.sub.i~ is transmission heat loss per year for the surface area i; |A.sub.i~ is the surface area i in square feet; HDD is number of heating degree days, a measure of T over a year; and 24 is number of hours in a day. The i outside surface areas of the house include walls, roof, floors, doors, and windows. Increasing the heat savings of the outside surface of the building reduces transmission heat loss. This might be done by adding inches of insulation to walls, ceilings, or floors and storm doors and storm windows.

Infiltration heat loss is caused by cold air entering through holes in the building. The air change method for estimating infiltration heat loss is usually used for houses and small buildings. The air change rate, AC, measures how many times per hour outdoor air exchanges completely with indoor air. The air filtration heat loss over one year, AIHL, is the air change rate multiplied by the indoor volume of a house, VL, and by the energy in Btus needed to warm the cold air over one year:

AIHL = AC x VL x .018 x HDD x 24 (6)

where 0.018 is the number of Btus required to warm one cubic foot of air one degree Fahrenheit; this is multiplied by 24 hours in a day and the number of heating degree days in a year, HDD.

Heating degree days, HDD, in equations (5) and (6) is defined as the difference between the outside temperature and 65 degrees Fahrenheit. In the 1930s engineers established that, in the long term, heat gains from solar and internal sources and heat losses offset each other in the average house at an external temperature of 65 degrees. The heating degree days in equations (5) and (6) account for the difference between inside and outside temperatures. Taken together equations (5) and (6) provide an expression for heat loss, HL, for a house:

|Mathematical Expression Omitted~

where i denotes the outside surface area of the house including wails, floors, roof, doors, and windows.

This heat loss equation can now be used to compute the energy requirements for a particular house.(3) The equation assumes no energy usage when both the external and internal temperatures are 65 degrees. To compute energy requirements, E, the heat loss equation (7) is adjusted for the efficiency of the heating system, k, which is a fraction between zero and one indicating the portion of energy lost in the process of heat production and distribution, and for the heat value of the fuel in question, v:

E = ||C.sub.d~/kv~ x HL (8)

|C.sub.d~ is an engineering adjustment made in recognition that insulation and other characteristics of the average house have changed since the 1930s when heating degree days was first defined.(4)

Specification of Demand for Fuel

This discussion of the heat loss equation describes the engineering calculation of the relationship between the physical characteristics of the house and the amount of energy required to raise the temperature inside the house a known, specific amount. However, the input demand function described in equation (4) requires that the household choose the inside temperature. In particular, equation (8) multiplied by the difference between the chosen temperature and 65 degrees yields the energy required to heat the house. The demand for warmth in the model is endogenous, chosen by the household to maximize utility. It depends on the price of warmth, which, in turn, is a function of the heating integrity of the house. To adapt the engineering formulation of the production function for warmth the choice the household has over warmth is incorporated. To do this households choose some level other than 65 |degrees~ as the temperature they demand. The difference between outside temperature and this level defines warmth, W. The demand for W is the demand for warmth arising out of the utility-maximizing model shown. Ideally, the demand for warmth would be estimated, as well as the derived input demand function for fuel given in equation (4). As the data set does not include an observation of the warmth choice, only a reduced form equation for fuel demand can be specified and estimated, using the variables that affect demand for warmth, including income and demographic characteristics, multiplied by the production function for home heating. To the extent that the fuel in question is used for reasons other than home heating, the demand for the fuel must take this into account.

Equation (9) shows the full reduced form equation for the demand for fuel. It is derived by substituting equation (7) into equation (8), then substituting expressions for THL, equation (5), and AIHL, equation (6), into the result. This yields a complete expression for the technical production function, in terms of the energy requirements for maintaining a 65 |degrees~ internal temperature. Finally, multiply the entire production function by W, the chosen level of warmth, to determine the energy requirements for maintaining the chosen temperature. Note that W, the chosen level of warmth, is a function of income and demographics. This is not reflected in the specification shown as equation (9).

|Mathematical Expression Omitted~

Where the subscript for the individual household observation has been suppressed for clarity, and the subscript i denotes the five distinct outside surfaces considered in this paper; walls, roof, floors, doors, and windows. The next to last term in equation (9) refers to the j appliances owned by the household. A|P.sub.j~ is the energy used by the jth appliance per unit time. This is multiplied by the demand for the jth appliance use, |D.sub.j~, which is a function of income, price of use, and household demographics.(5)

Data and Estimation

The Energy Information Administration, U.S. Department of Energy, collected these data in 1980-1981. The data are from the Residential Energy Consumption Survey, 1980-1981, one of a series of surveys(6) that collected information on housing units, demographic and economic characteristics of the households and matched consumption and expenditure data obtained from records maintained by the households' fuel suppliers. The data set included 5,979 households chosen from all 50 states and the District of Columbia using a multistage stratified random sample design developed for this survey.

These data include information on thermal characteristics of the house, demographic characteristics of the household, energy using appliances, and energy saving investments, all collected by household interview. Heating and cooling degree days and energy consumption and expenditures by fuel types were collected monthly or by billing period from the households' fuel suppliers. To maintain anonymity, location data were suppressed and heating and cooling degree days have been injected with random errors.(7)

As the production function for warmth incorporated into equation (9) is specifically for detached units, only households in single-family detached houses were used. The estimated rates of return for investments in energy conserving modifications apply only to homeowners so the authors further restricted the sample to owners. Observations that exhibit rare characteristics, as heated swimming pools or part-year residences, were deleted. Also excluded were observations with missing data on fuel consumption and households not paying their own fuel bills. After these deletions the sample sizes were 1,028 for natural gas, 253 for electricity, and 217 for fuel oil. Table 1 contains a summary of the variables and their definitions. Table 2 provides descriptive statistics for each of these variables.

In principle, equation (9) is determined except for whatever randomness arises because of consumer errors in demand for warmth or appliance use. However, there is randomness associated with approximation in the engineering equations. For example, |C.sub.d~ is an empirically TABULAR DATA OMITTED estimated correction factor needed because of the approximation of 65 degrees as the temperature at which heat gain and heat loss are equal. Heating system efficiency, k, is not fully accounted for in the data set used. Heating system efficiency is likely to vary across systems using different types of fuel (Neels 1981). Compared to these differences in efficiency, variations across systems using the same fuel and of different ages are probably small. Because of this, equation (9) is estimated separately for natural gas, electricity, and fuel oil. Before estimating this equation some modifications are necessary. The data set does not provide details about wall insulation, although it does provide inches of floor and roof insulation. Wall insulation is measured with a binary variable equal to one if surface i is insulated and zero otherwise. Neither the R-value of the insulation nor the R-value of the basic building materials is known. Thus estimated coefficients accompanying each of the surface areas of the house contain the inverse of the R-value for the house surface, and the impact of energy saving due to insulation is captured as part of the estimated coefficients accompanying the insulation terms. The estimated coefficients are negative if insulation results in lower energy use.

The data do not provide details about appliance use. Hence the authors define A|P.sub.j~ to be a binary variable equal to one if the household owns the jth appliance using the fuel in question and zero otherwise. Appliances considered are air conditioners, ranges and ovens, clothes dryers, water heaters, refrigerators, and freezers. As weather conditions affect the use of air conditioners and water heaters, the binary variable for each of these appliances is interacted with cooling degree days or heating degree days. Use of some appliances, particularly clothes dryers, cooking units, water heaters, and air conditioners, depends on the demographic characteristics of the household. The binary variable for each of these appliances is interacted with the appropriate demographic characteristics.

To introduce the function for the demand for warmth into equation (9), multiply the economic and demographic variables that determine the demand for warmth, W, by the thermal characteristics of the house. Because of the large number of interactions this introduces into the equation, the collinearity among independent variables is unacceptably high. One solution to this problem is to interact the demographic and economic characteristics with the thermal characteristics of the house as a whole, rather than with each component as equation (9) indicates. Doing this yields:

|Mathematical Expression Omitted~

where the |X.sub.k~'s are the k housing characteristics, the |Z.sub.j~'s are the j appliance variables and other demographic variables, and |a.sub.k~ and |c.sub.j~ are estimated coefficients and |b.sub.1~, |b.sub.2~, and |b.sub.3~ are estimated exponents. The k housing characteristics are the heated area, building age, presence of caulking and weatherstripping, number of storm doors, inches of floor insulation, inches of roof insulation, number of doors, number of windows, number of storm windows, and presence of wall insulation.

Taking the natural logarithm of both sides of equation (10) yields the form of the equation used for the estimates:(8)

|Mathematical Expression Omitted~

Estimates of the Demand for Home Heating Energy

Table 3 shows the estimates of the demand for home heating energy, equation (11). The estimated equation for the demand for natural gas performs best in that the coefficients are generally of the expected signs and carry t-statistics over 2.00. The estimated equation for electricity does not perform as well, perhaps because of the small sample size. A demand equation for fuel oil was estimated but not reported here because the estimation demonstrated so little precision.(9)

The price elasticity of demand for electricity is 0.44 and for natural gas is 0.58. The income elasticities estimated in these equations are .079 for electricity and .039 for natural gas. The income and price elasticities have t-statistics greater than 2.00 in both the electricity and natural gas equations.

Several estimates of shortrun price and income elasticities for various types of energy have been made. These estimates vary considerably, depending on the data used, estimation technique, functional specification, and time period of the data. Shortrun own-price elasticities for electricity estimated from reduced form models range TABULAR DATA OMITTED from -.03 to -.54. There are not as many estimates of price elasticities for natural gas, and some of the demand work on natural gas combines residential and commercial customers. The shortrun own-price elasticity estimates for natural gas using reduced form models range from -.03 to -.50 (Bohi 1981). This study's price elasticities fall within the range of these estimates.

Estimates for income elasticities for these fuels vary even more than estimates of price elasticities. Estimates of shortrun income elasticities for electricity range from 0.02 to 2.00, although most are less than 1.00 (Bohi 1981). Estimates of shortrun income elasticities for natural gas range from 0.03 to 0.96 (Al-Sahlawi 1989). Again, estimates from this study are within this range.

Using these estimated equations, the marginal effect of energy conserving modifications to the house can be calculated. These marginal energy savings are calculated by differentiating equation (11) with respect to the energy conserving item in question, when that item is measured continuously, i.e., roof insulation, floor insulation, storm doors, or storm windows, then multiplying the result by the predicted energy use. Calculate the marginal energy savings for wall insulation, a binary variable, by subtracting energy used with wall insulation from energy used without wall insulation. Marginal energy savings vary by the same characteristics that cause predicted energy use to vary. Of particular interest is the variation in energy saving across regions, house sizes, and incomes. Tables 4 and 5 show these predicted savings for observations in the sample. Both annual Btu savings and dollar savings, calculated using average prices actually paid by sample members in 1980-1981, are shown in the tables.(10) Only TABULAR DATA OMITTED TABULAR DATA OMITTED those energy conserving investments that yield positive savings are shown.

As seen, marginal energy savings from these energy conserving investments are not large when increases in the demand for warmth due to lower warmth prices are considered. These energy savings vary considerably depending on region, income, and size of house. The largest energy savings should accrue to high income households, because they consume more energy and hence have a larger base against which to save. High income households, those with incomes greater than $30,000 who live in the Northeast, have a predicted $13.52 annual savings for one inch of roof insulation if they use electricity for heating, and $18.18 annually if they use natural gas. Low income households, those with incomes less than $18,000 who live in the South save a predicted $3.32 annually if they use electricity and $1.37 annually if they use natural gas. Savings also vary with house size, with those in larger houses saving more. Those using electricity for heating, living in the Northeast, and having a house with more than 2,500 square feet should save $14.78 annually. Using natural gas is predicted to save $29.45 annually. Those living in the South in a house of less than 1,500 square feet save $2.25 annually if they use electricity and $1.08 annually if they use natural gas for heating. Similar variation by income, region, and house size is seen in savings accruing to other types of energy conserving investments.

These savings can be compared with those estimated by Delene and Gaston (1976) who use an 1,800 square foot model home and the engineering method to estimate energy savings from various conservation strategies for 13 U.S. cities. Table 6 shows the Delene and Gaston estimates for three conservation strategies compared to estimates for the same strategies using the equations in this paper. The strategies are six inches of roof insulation versus no roof insulation, 3-1/2 inches of wall insulation versus no wall insulation, and a complete set of storm windows versus no storm windows.

In the model presented here the marginal energy savings for one inch of insulation (or one storm window) are smaller than the average savings. Thus the energy savings presented in Table 6 are calculated TABULAR DATA OMITTED by subtracting the predicted energy use with the energy conservation investment from the predicted energy use without the investment.

The energy savings estimated by Delene and Gaston are larger for all types of energy conserving investments than those estimated using our model. For roof insulation they are two to eight times as large, depending on region and type of fuel. For wall insulation using natural gas, Delene and Gaston estimate savings twice as large, and for storm windows their estimates are a third or more larger.

The differences between the Delene and Gaston estimates and the estimates which use the model presented in this paper are striking. The techniques used in making these two estimates are quite different. The model home method uses the production function presented in equation (8) together with heating degree days assuming inside temperatures set at 68 degrees. The model postulated in this paper allows households to respond to the reduction in the price of warmth by increasing their demand for warmth. This behavioral response to energy conserving investments offsets some of the financial and energy savings associated with the investment. Households still benefit from the investment, but they take some of these benefits in the form of increased satisfaction from the warmer house.

Implications of this work for private investment in energy saving modifications to the house are limited. Households faced with information about cost savings to energy saving investments calculated from the model home technique will make investment decisions on the basis of this information. Once the investment is made they will discover that the cost savings are not as great as predicted, because they have chosen to take some of these savings in the form of greater satisfaction from the warmer house.

Policymakers concerned with actual energy savings for the reasons of price stability, balance of payments, and for political reasons should be more concerned about the implications of this work. Any estimates of reduced residential energy demand achieved through energy conservation investment (i.e., improved insulation) are likely to be greatly exaggerated if these estimates are made using the model home techniques. The estimates provided here are more likely to represent the possible energy savings from energy conserving investment.

1 There is some recognition of this effect when looking at appliance efficiency regulation and labeling. See Hausman and Joskow (1982) and Khazzoom (1980).

2 Of course households choose thermal characteristics of their housing. This choice depends on the cost of the characteristics as well as the impact it has on the price of warmth through marginal productivity. Formulation of the model presented here focuses on the shortrun impact on warmth and, hence, fuel demand. A longrun model would consider simultaneous determination of the thermal characteristics of the house.

3 See Hsueh (1984) for further details on derivation of the heat loss and energy use equations.

4 |C.sub.d~ is an empirically determined correction to the heating degree days calculation. Engineers have determined that heating degree days overestimates heat loss associated with a house's surface. This overestimate varies with the number of heating degree days. It is smaller for HDD over 5,000, but larger for HDD under 5,000. In other words, using HDD overestimates heating energy requirements more for colder areas.

5 This expression allows for the introduction of j separate appliances. However, in the empirical work, only those appliances that consume a significant amount of fuel are used.

6 Earlier surveys include the Lifestyle and Household Energy Use, 1973 and 1975 Surveys, the National Interim Energy Consumption Survey, Residential, 1978, and the Residential Energy Consumption Survey, Household Screener Survey, 1979-1980.

7 The injection of random errors into the heating and cooling degree days was done by the Department of Energy to assure that demographic characteristics together with consumption data and degree day data do not identify specific individuals. This presents an unusual type of errors-in-variables problem. The procedure for introducing this random error is to multiply HDD and CDD for each observation by a random number drawn from a normal distribution, mean 1.00, variance .0225. In addition, the minimum error is constrained to be one percent and the maximum is 60 percent. This procedure results in OLS estimators that are symptomatically inconsistent. Hwang (1984) has developed a consistent estimator for this case. The model used this estimator. Although the absolute value of estimated coefficients for variables related to HDD and CDD are increased and those of socioeconomic variables and regional dummy variables are decreased, the magnitude of change is small when the sample size is large. The details of the estimator are reported in Hwang (1984) and Hsueh (1984). The results reported here are those using the consistent estimator.

8 This form of the demand equation eliminates the multicollinearity found in equation (10), but at the cost of the linear relationship between energy consumption and inches of insulation. A linear form of the equation was estimated for comparison. The signs of the estimated coefficients were generally similar between the two forms, and the log-linear form eliminates the heteroscedasticity found in the linear form. Thus the log-linear form is preferable.

9 Estimates for the fuel oil equation are available on request.

10 Although the prices used for these calculations reflect actual prices paid by the surveyed households, these prices are not radically different from current prices. The fuel oil and other household fuel commodities component of the CPI was 86.1 in 1980 and 81.7 in 1989 (1982-1984 = 100). The gas (piped) and electricity component of the CPI was 71.4 in 1980 and 107.5 in 1989. The Annual Energy Review (1989) reports that average prices paid for natural gas in 1981 were $4.29/thousand cubic feet and $5.65/thousand cubic feet in 1989. For electricity the average price paid in 1981 was $0.62/kilowatt hour and $0.0764/kilowatt hour in 1989.

REFERENCES

Al-Sahlawi, Mohammed A. (1989), "The Demand for Natural Gas: A Survey of Price and Income Elasticities," The Energy Journal (January): 77-90.

Blair, Roger D., David L. Koserman, and Richard C. Topel (1984), "The Impact of Improved Mileage on Gasoline Consumption," Economic Inquiry, 22(April): 209-216.

Bohi, Douglas R. (1981), Analyzing Demand Behavior: A Study of Energy Elasticities, Baltimore, MD: Johns Hopkins University Press.

Delene, J. G. and J. B. Gaston (1976), "A Regional Comparison of Savings from Various Residential Energy Conservation Strategies," Oak Ridge National Laboratory, ORNL/TM-5146 (February).

Energy Information Administration, U.S. Department of Energy (1983), Residential Energy Consumption Survey, Housing Characteristics 1980, and Consumption and Expenditure, April 1980 Through March 1981, Public Use Tape Technical Documentation, PB 83-199547, Washington, DC: National Technical Information Service (January).

Energy Information Administration (1987), Residential Energy Consumption Survey: Trends in Consumption and Expenditures, 1978-1984, DOE/EIA-0482, Washington, DC.

Energy Information Administration (1989), Annual Energy Review, Washington, DC: U.S. Government Printing Office.

Hausman, Jerry A. and Paul L. Joskow (1982), "Evaluating the Costs and Benefits of Appliance Efficiency Standards," American Economic Review, 72(2, May): 220-225.

Hsueh, Li-Min (1984), "A Model of Home Heating and Calculation of Rates of Return to Household Energy Conservation Investment," unpublished doctoral dissertation, Cornell University, Ithaca, NY.

Hutchins, Paul F., Jr. and Eric Hirst (1978), "Engineering-Economic Analysis of Single-Family Dwelling Thermal Performance," Oak Ridge National Laboratory, ORNL/CON-35.

Hwang, Jiunn T. (1984), "Multiplicative Errors-in-variables Model with Applications to the Recent Data Released by USDOE," Technical Report No. 84-1, Cornell University Statistical Center.

Khazzoom, J. Daniel (1980), "The Economic Implications of Mandated Efficiency Standards for Household Appliances," The Energy Journal (October): 21-40.

Khazzoom, J. Daniel (1986), An Econometric Model Integrating Conservation Measures in the Residential Demand for Electricity, London: JAI Press.

Latta, Robert B. (1981), The National Interim Energy Consumption Survey: Exploring the Variability in Energy Consumption: A Supplement, DOE/EIA-0272/S, Washington, DC: U.S. Department of Energy, Energy Information Administration, Office of Energy Markets and End Use, Energy End Use Division.

Mayo, John W. and John E. Mathis (1988), "The Effectiveness of Mandatory Fuel Efficiency Standards in Reducing the Demand for Gasoline," Applied Economics, 20: 211-219.

Moyers, John C. (1971), "The Value of Thermal Insulation in Residential Construction: Economics and Conservation of Energy," Oak Ridge National Laboratory, ORNL-NSF-EP-9.

Neels, Kevin (1981), "Family, Houses, and Demand for Energy: Housing Assistance Supply Experiment," Rand Note, N-1242-HUD(PB82-125733), Santa Monica, CA: The Rand Corporation (February).

O'Neal, D. L., K. R. Corum, and J. L. Jones (1981), "Estimate of Consumer Discount Rate Implicit in Single-Family Housing Construction Practices," Oak Ridge National Laboratory, ORNL/CON-62.

Scott, Alex (1980), "The Economics of House Heating," Energy Economics, 2(3, July): 130-141.

U.S.Bureau of the Census (1982), Statistical Abstract of the United States, 1982-1983, 103rd Edition, Washington, DC.

Li-Min Hsueh is Associate Research Fellow, Chung-Hua Institution for Economic Research, Taipei, Taiwan; and Jennifer L. Gerner is Associate Professor, Department of Consumer Economics and Housing, Cornell University, Ithaca, NY.

Rising prices of energy coupled with favorable tax treatment of residential conservation investments resulted in considerable input substitution during the late 1970s and early 1980s. For example, in 1981 3.4 million households installed roof insulation, 2.0 million households installed wall insulation, and 5.0 million households added storm windows or storm doors (U.S. Bureau of the Census 1982). The impact these conservation investments have in terms of reduced fuel requirements to achieve a given level of warmth is important for consumers to know when deciding whether to make these investments. Engineers have estimated this reduction by observing the impact of energy conserving modifications on model homes (Delene and Gaston 1976; Hutchins and Hirst 1978; Moyers 1971; O'Neal, Corum, and Jones 1981).

However, this technique does not recognize the role conservation investments play in the behavior of households. The introduction of energy conserving modifications to a house serves to reduce the marginal price per unit of warmth. In the face of lower per unit warmth, households may increase their demand for warmth, offsetting the reduction in fuel demand. This implies that energy conserving modifications may impact energy demand less than engineering computations suggest. Similarly in an examination of demand for gasoline and automobile fuel efficiency, both inputs impact the demand for miles driven (Blair, Koserman, and Topel 1984; Mayo and Mathis 1988). These studies provide evidence that some benefits from increased fuel efficiency are taken in more miles driven. Hence, demand for gasoline does not fall as much as predicted.

Economists have estimated residential demand for energy used in home heating taking into account the thermal characteristics of the house (Neels 1981; Scott 1980). But, with a couple of exceptions (Khazzoom 1986), economists have not generally treated energy for home heating as an input demand derived from the demand for home warmth.(1) As a result economists have not specified the demand functions in a way which takes into account what is known by engineers about heat loss.

In this paper the authors treat demand for household energy used in home heating as a derived input demand, using a heat loss function to represent home warmth production. The authors estimate this demand function and compute energy savings and dollar savings, under the 1980-1981 price regime, associated with conservation investment. These savings calculations incorporate changes in the demand for home warmth resulting from reductions in home warmth prices due to the investments themselves.

THEORETICAL CONSIDERATIONS

The economic model used is the standard utility-maximizing model. Household utility, V, is a function of home warmth, W, and other goods, G, and is conditional on the demographic characteristics of the household, D. Home warmth is produced by using fuel for home heating, F, given the thermal characteristics of the house itself, C, and the outside temperature, T. Investments that improve the thermal characteristics of the house improve marginal productivity of fuel in the production of warmth. Given the thermal characteristics of the house, the production function for home warmth, and household resources, the problem is to choose the level of warmth that maximizes the consumer's utility:

V = v(W, G; D) (1)

W = w(F; C, T) (2)

Y = |p.sub.G~G + |p.sub.F~F (3)

where |p.sub.G~ is the price of all other goods, |p.sub.F~ is the price of fuel, and Y is income. Equation (2) is a production function for home warmth. The consumer chooses a level of home warmth by maximizing utility, equation (1), subject to the production relation and the budget constraint. Here the interest is not in the demand for home warmth, but in the derived demand for fuel arising from the consumer's choice of home warmth. An input demand function for fuel can be obtained by solving this problem, and is given by:

F = F(Y, |p.sub.G~, |p.sub.F~ |Delta~W/|Delta~F; D, C, T, W). (4)

The price of home warmth, |p.sub.F~(|Delta~W/|Delta~F), is defined as the marginal cost of fuel, |p.sub.F~, multiplied by the marginal productivity of fuel in the production of home warmth, |Delta~W/|Delta~F. The marginal productivity of fuel in the production of home warmth depends on thermal characteristics of the house. Characteristics of housing that serve to improve the efficiency of fuel in the production of home heating are equivalent to reductions in the price of fuel. Both reduce the price of warmth to the household.(2)

Equation (2), the production function for home warmth, has been the focus of most engineering work. This work has focused on heat loss, that is, the amount of heat that escapes through the shell of a building over a specified period or, equivalently, the amount of heat needed to keep a house at a specified temperature for a given period. If one knows the heat loss properties of a structure, one can calculate the energy requirements for heating the structure to some specified temperature. In this paper the authors incorporate the engineering specifications for the production relation into the estimates of the demand for fuel equation.

Engineering Specification of the Production Relation

Heat loss occurs because of transmission heat loss and infiltration heat loss. Transmission heat loss per square foot, caused by the heat conductivity of building material, is measured by the heat transmission coefficient, U, the number of Btus transmitted through one square foot of surface of a particular material for one degree of temperature difference between one side of the material and the other, T, over one hour. The heat transmission coefficient is a known property of particular materials. The reciprocal of U is R, the heat resistance coefficient. Although engineers use the heat transmission coefficient, the authors express heat loss as the more familiar heat resistance coefficient. Larger R-values indicate materials that are better at stopping heat from escaping. The R-value of an outside surface is the sum of the individual R-values for the building materials, |R.sub.mi~, and the insulation. As insulation prevents heat from escaping, the effect of insulation can be expressed as a reduction in Btus lost. Let the R-value for insulation be measured per inch of insulation, |R.sub.Ii~, and let |I.sub.i~ be the number of inches of insulation. Then transmission heat loss for a surface area i for one year can be approximated by:

TH|L.sub.i~ = |(1/|R.sub.mi~) - (1/|R.sub.Ii~) x |I.sub.i~~ x |A.sub.i~ x HDD x 24 (5)

where TH|L.sub.i~ is transmission heat loss per year for the surface area i; |A.sub.i~ is the surface area i in square feet; HDD is number of heating degree days, a measure of T over a year; and 24 is number of hours in a day. The i outside surface areas of the house include walls, roof, floors, doors, and windows. Increasing the heat savings of the outside surface of the building reduces transmission heat loss. This might be done by adding inches of insulation to walls, ceilings, or floors and storm doors and storm windows.

Infiltration heat loss is caused by cold air entering through holes in the building. The air change method for estimating infiltration heat loss is usually used for houses and small buildings. The air change rate, AC, measures how many times per hour outdoor air exchanges completely with indoor air. The air filtration heat loss over one year, AIHL, is the air change rate multiplied by the indoor volume of a house, VL, and by the energy in Btus needed to warm the cold air over one year:

AIHL = AC x VL x .018 x HDD x 24 (6)

where 0.018 is the number of Btus required to warm one cubic foot of air one degree Fahrenheit; this is multiplied by 24 hours in a day and the number of heating degree days in a year, HDD.

Heating degree days, HDD, in equations (5) and (6) is defined as the difference between the outside temperature and 65 degrees Fahrenheit. In the 1930s engineers established that, in the long term, heat gains from solar and internal sources and heat losses offset each other in the average house at an external temperature of 65 degrees. The heating degree days in equations (5) and (6) account for the difference between inside and outside temperatures. Taken together equations (5) and (6) provide an expression for heat loss, HL, for a house:

|Mathematical Expression Omitted~

where i denotes the outside surface area of the house including wails, floors, roof, doors, and windows.

This heat loss equation can now be used to compute the energy requirements for a particular house.(3) The equation assumes no energy usage when both the external and internal temperatures are 65 degrees. To compute energy requirements, E, the heat loss equation (7) is adjusted for the efficiency of the heating system, k, which is a fraction between zero and one indicating the portion of energy lost in the process of heat production and distribution, and for the heat value of the fuel in question, v:

E = ||C.sub.d~/kv~ x HL (8)

|C.sub.d~ is an engineering adjustment made in recognition that insulation and other characteristics of the average house have changed since the 1930s when heating degree days was first defined.(4)

Specification of Demand for Fuel

This discussion of the heat loss equation describes the engineering calculation of the relationship between the physical characteristics of the house and the amount of energy required to raise the temperature inside the house a known, specific amount. However, the input demand function described in equation (4) requires that the household choose the inside temperature. In particular, equation (8) multiplied by the difference between the chosen temperature and 65 degrees yields the energy required to heat the house. The demand for warmth in the model is endogenous, chosen by the household to maximize utility. It depends on the price of warmth, which, in turn, is a function of the heating integrity of the house. To adapt the engineering formulation of the production function for warmth the choice the household has over warmth is incorporated. To do this households choose some level other than 65 |degrees~ as the temperature they demand. The difference between outside temperature and this level defines warmth, W. The demand for W is the demand for warmth arising out of the utility-maximizing model shown. Ideally, the demand for warmth would be estimated, as well as the derived input demand function for fuel given in equation (4). As the data set does not include an observation of the warmth choice, only a reduced form equation for fuel demand can be specified and estimated, using the variables that affect demand for warmth, including income and demographic characteristics, multiplied by the production function for home heating. To the extent that the fuel in question is used for reasons other than home heating, the demand for the fuel must take this into account.

Equation (9) shows the full reduced form equation for the demand for fuel. It is derived by substituting equation (7) into equation (8), then substituting expressions for THL, equation (5), and AIHL, equation (6), into the result. This yields a complete expression for the technical production function, in terms of the energy requirements for maintaining a 65 |degrees~ internal temperature. Finally, multiply the entire production function by W, the chosen level of warmth, to determine the energy requirements for maintaining the chosen temperature. Note that W, the chosen level of warmth, is a function of income and demographics. This is not reflected in the specification shown as equation (9).

|Mathematical Expression Omitted~

Where the subscript for the individual household observation has been suppressed for clarity, and the subscript i denotes the five distinct outside surfaces considered in this paper; walls, roof, floors, doors, and windows. The next to last term in equation (9) refers to the j appliances owned by the household. A|P.sub.j~ is the energy used by the jth appliance per unit time. This is multiplied by the demand for the jth appliance use, |D.sub.j~, which is a function of income, price of use, and household demographics.(5)

Data and Estimation

The Energy Information Administration, U.S. Department of Energy, collected these data in 1980-1981. The data are from the Residential Energy Consumption Survey, 1980-1981, one of a series of surveys(6) that collected information on housing units, demographic and economic characteristics of the households and matched consumption and expenditure data obtained from records maintained by the households' fuel suppliers. The data set included 5,979 households chosen from all 50 states and the District of Columbia using a multistage stratified random sample design developed for this survey.

These data include information on thermal characteristics of the house, demographic characteristics of the household, energy using appliances, and energy saving investments, all collected by household interview. Heating and cooling degree days and energy consumption and expenditures by fuel types were collected monthly or by billing period from the households' fuel suppliers. To maintain anonymity, location data were suppressed and heating and cooling degree days have been injected with random errors.(7)

As the production function for warmth incorporated into equation (9) is specifically for detached units, only households in single-family detached houses were used. The estimated rates of return for investments in energy conserving modifications apply only to homeowners so the authors further restricted the sample to owners. Observations that exhibit rare characteristics, as heated swimming pools or part-year residences, were deleted. Also excluded were observations with missing data on fuel consumption and households not paying their own fuel bills. After these deletions the sample sizes were 1,028 for natural gas, 253 for electricity, and 217 for fuel oil. Table 1 contains a summary of the variables and their definitions. Table 2 provides descriptive statistics for each of these variables.

In principle, equation (9) is determined except for whatever randomness arises because of consumer errors in demand for warmth or appliance use. However, there is randomness associated with approximation in the engineering equations. For example, |C.sub.d~ is an empirically TABULAR DATA OMITTED estimated correction factor needed because of the approximation of 65 degrees as the temperature at which heat gain and heat loss are equal. Heating system efficiency, k, is not fully accounted for in the data set used. Heating system efficiency is likely to vary across systems using different types of fuel (Neels 1981). Compared to these differences in efficiency, variations across systems using the same fuel and of different ages are probably small. Because of this, equation (9) is estimated separately for natural gas, electricity, and fuel oil. Before estimating this equation some modifications are necessary. The data set does not provide details about wall insulation, although it does provide inches of floor and roof insulation. Wall insulation is measured with a binary variable equal to one if surface i is insulated and zero otherwise. Neither the R-value of the insulation nor the R-value of the basic building materials is known. Thus estimated coefficients accompanying each of the surface areas of the house contain the inverse of the R-value for the house surface, and the impact of energy saving due to insulation is captured as part of the estimated coefficients accompanying the insulation terms. The estimated coefficients are negative if insulation results in lower energy use.

TABLE 2 Descriptive Statistics for the Variables (standard deviation in parenthesis) Variable Electricity Natural Gas BTU (|10.sup.4~) 7,335.41 11,562.44 (2,478.42) (5,587.19) AVEPRICE 0.1273 0.0374 (0.046) (0.009) INCOME 23,219.37 23,630.84 (13,921.00) (15,219.80) AREA 1,793.91 1,878.51 (808.38) (1,039.90) NOHOME 0.905 0.968 (0.88) (0.91) ROOF 6.49 4.85 (3.62) (3.39) FLOOR 1.23 0.22 (2.12) (1.05) WINDOW 28.47 33.44 (12.03) (15.97) STORMWIN 19.63 20.27 (16.99) (20.62) DOORS 3.19 2.86 (1.25) (0.99) STORMDOOR 1.75 1.53 (1.42) (1.16) BLDAGE 12.49 25.81 (10.72) (12.38) HDD 4,281.03 5,019.84 (1,890.38) (2,122.00) CDD 1,353.75 1,148.74 (1,054.90) (842.00) REF 1.92 -- (0.81)

The data do not provide details about appliance use. Hence the authors define A|P.sub.j~ to be a binary variable equal to one if the household owns the jth appliance using the fuel in question and zero otherwise. Appliances considered are air conditioners, ranges and ovens, clothes dryers, water heaters, refrigerators, and freezers. As weather conditions affect the use of air conditioners and water heaters, the binary variable for each of these appliances is interacted with cooling degree days or heating degree days. Use of some appliances, particularly clothes dryers, cooking units, water heaters, and air conditioners, depends on the demographic characteristics of the household. The binary variable for each of these appliances is interacted with the appropriate demographic characteristics.

To introduce the function for the demand for warmth into equation (9), multiply the economic and demographic variables that determine the demand for warmth, W, by the thermal characteristics of the house. Because of the large number of interactions this introduces into the equation, the collinearity among independent variables is unacceptably high. One solution to this problem is to interact the demographic and economic characteristics with the thermal characteristics of the house as a whole, rather than with each component as equation (9) indicates. Doing this yields:

|Mathematical Expression Omitted~

where the |X.sub.k~'s are the k housing characteristics, the |Z.sub.j~'s are the j appliance variables and other demographic variables, and |a.sub.k~ and |c.sub.j~ are estimated coefficients and |b.sub.1~, |b.sub.2~, and |b.sub.3~ are estimated exponents. The k housing characteristics are the heated area, building age, presence of caulking and weatherstripping, number of storm doors, inches of floor insulation, inches of roof insulation, number of doors, number of windows, number of storm windows, and presence of wall insulation.

Taking the natural logarithm of both sides of equation (10) yields the form of the equation used for the estimates:(8)

|Mathematical Expression Omitted~

Estimates of the Demand for Home Heating Energy

Table 3 shows the estimates of the demand for home heating energy, equation (11). The estimated equation for the demand for natural gas performs best in that the coefficients are generally of the expected signs and carry t-statistics over 2.00. The estimated equation for electricity does not perform as well, perhaps because of the small sample size. A demand equation for fuel oil was estimated but not reported here because the estimation demonstrated so little precision.(9)

The price elasticity of demand for electricity is 0.44 and for natural gas is 0.58. The income elasticities estimated in these equations are .079 for electricity and .039 for natural gas. The income and price elasticities have t-statistics greater than 2.00 in both the electricity and natural gas equations.

Several estimates of shortrun price and income elasticities for various types of energy have been made. These estimates vary considerably, depending on the data used, estimation technique, functional specification, and time period of the data. Shortrun own-price elasticities for electricity estimated from reduced form models range TABULAR DATA OMITTED from -.03 to -.54. There are not as many estimates of price elasticities for natural gas, and some of the demand work on natural gas combines residential and commercial customers. The shortrun own-price elasticity estimates for natural gas using reduced form models range from -.03 to -.50 (Bohi 1981). This study's price elasticities fall within the range of these estimates.

Estimates for income elasticities for these fuels vary even more than estimates of price elasticities. Estimates of shortrun income elasticities for electricity range from 0.02 to 2.00, although most are less than 1.00 (Bohi 1981). Estimates of shortrun income elasticities for natural gas range from 0.03 to 0.96 (Al-Sahlawi 1989). Again, estimates from this study are within this range.

Using these estimated equations, the marginal effect of energy conserving modifications to the house can be calculated. These marginal energy savings are calculated by differentiating equation (11) with respect to the energy conserving item in question, when that item is measured continuously, i.e., roof insulation, floor insulation, storm doors, or storm windows, then multiplying the result by the predicted energy use. Calculate the marginal energy savings for wall insulation, a binary variable, by subtracting energy used with wall insulation from energy used without wall insulation. Marginal energy savings vary by the same characteristics that cause predicted energy use to vary. Of particular interest is the variation in energy saving across regions, house sizes, and incomes. Tables 4 and 5 show these predicted savings for observations in the sample. Both annual Btu savings and dollar savings, calculated using average prices actually paid by sample members in 1980-1981, are shown in the tables.(10) Only TABULAR DATA OMITTED TABULAR DATA OMITTED those energy conserving investments that yield positive savings are shown.

As seen, marginal energy savings from these energy conserving investments are not large when increases in the demand for warmth due to lower warmth prices are considered. These energy savings vary considerably depending on region, income, and size of house. The largest energy savings should accrue to high income households, because they consume more energy and hence have a larger base against which to save. High income households, those with incomes greater than $30,000 who live in the Northeast, have a predicted $13.52 annual savings for one inch of roof insulation if they use electricity for heating, and $18.18 annually if they use natural gas. Low income households, those with incomes less than $18,000 who live in the South save a predicted $3.32 annually if they use electricity and $1.37 annually if they use natural gas. Savings also vary with house size, with those in larger houses saving more. Those using electricity for heating, living in the Northeast, and having a house with more than 2,500 square feet should save $14.78 annually. Using natural gas is predicted to save $29.45 annually. Those living in the South in a house of less than 1,500 square feet save $2.25 annually if they use electricity and $1.08 annually if they use natural gas for heating. Similar variation by income, region, and house size is seen in savings accruing to other types of energy conserving investments.

These savings can be compared with those estimated by Delene and Gaston (1976) who use an 1,800 square foot model home and the engineering method to estimate energy savings from various conservation strategies for 13 U.S. cities. Table 6 shows the Delene and Gaston estimates for three conservation strategies compared to estimates for the same strategies using the equations in this paper. The strategies are six inches of roof insulation versus no roof insulation, 3-1/2 inches of wall insulation versus no wall insulation, and a complete set of storm windows versus no storm windows.

In the model presented here the marginal energy savings for one inch of insulation (or one storm window) are smaller than the average savings. Thus the energy savings presented in Table 6 are calculated TABULAR DATA OMITTED by subtracting the predicted energy use with the energy conservation investment from the predicted energy use without the investment.

The energy savings estimated by Delene and Gaston are larger for all types of energy conserving investments than those estimated using our model. For roof insulation they are two to eight times as large, depending on region and type of fuel. For wall insulation using natural gas, Delene and Gaston estimate savings twice as large, and for storm windows their estimates are a third or more larger.

The differences between the Delene and Gaston estimates and the estimates which use the model presented in this paper are striking. The techniques used in making these two estimates are quite different. The model home method uses the production function presented in equation (8) together with heating degree days assuming inside temperatures set at 68 degrees. The model postulated in this paper allows households to respond to the reduction in the price of warmth by increasing their demand for warmth. This behavioral response to energy conserving investments offsets some of the financial and energy savings associated with the investment. Households still benefit from the investment, but they take some of these benefits in the form of increased satisfaction from the warmer house.

Implications of this work for private investment in energy saving modifications to the house are limited. Households faced with information about cost savings to energy saving investments calculated from the model home technique will make investment decisions on the basis of this information. Once the investment is made they will discover that the cost savings are not as great as predicted, because they have chosen to take some of these savings in the form of greater satisfaction from the warmer house.

Policymakers concerned with actual energy savings for the reasons of price stability, balance of payments, and for political reasons should be more concerned about the implications of this work. Any estimates of reduced residential energy demand achieved through energy conservation investment (i.e., improved insulation) are likely to be greatly exaggerated if these estimates are made using the model home techniques. The estimates provided here are more likely to represent the possible energy savings from energy conserving investment.

1 There is some recognition of this effect when looking at appliance efficiency regulation and labeling. See Hausman and Joskow (1982) and Khazzoom (1980).

2 Of course households choose thermal characteristics of their housing. This choice depends on the cost of the characteristics as well as the impact it has on the price of warmth through marginal productivity. Formulation of the model presented here focuses on the shortrun impact on warmth and, hence, fuel demand. A longrun model would consider simultaneous determination of the thermal characteristics of the house.

3 See Hsueh (1984) for further details on derivation of the heat loss and energy use equations.

4 |C.sub.d~ is an empirically determined correction to the heating degree days calculation. Engineers have determined that heating degree days overestimates heat loss associated with a house's surface. This overestimate varies with the number of heating degree days. It is smaller for HDD over 5,000, but larger for HDD under 5,000. In other words, using HDD overestimates heating energy requirements more for colder areas.

5 This expression allows for the introduction of j separate appliances. However, in the empirical work, only those appliances that consume a significant amount of fuel are used.

6 Earlier surveys include the Lifestyle and Household Energy Use, 1973 and 1975 Surveys, the National Interim Energy Consumption Survey, Residential, 1978, and the Residential Energy Consumption Survey, Household Screener Survey, 1979-1980.

7 The injection of random errors into the heating and cooling degree days was done by the Department of Energy to assure that demographic characteristics together with consumption data and degree day data do not identify specific individuals. This presents an unusual type of errors-in-variables problem. The procedure for introducing this random error is to multiply HDD and CDD for each observation by a random number drawn from a normal distribution, mean 1.00, variance .0225. In addition, the minimum error is constrained to be one percent and the maximum is 60 percent. This procedure results in OLS estimators that are symptomatically inconsistent. Hwang (1984) has developed a consistent estimator for this case. The model used this estimator. Although the absolute value of estimated coefficients for variables related to HDD and CDD are increased and those of socioeconomic variables and regional dummy variables are decreased, the magnitude of change is small when the sample size is large. The details of the estimator are reported in Hwang (1984) and Hsueh (1984). The results reported here are those using the consistent estimator.

8 This form of the demand equation eliminates the multicollinearity found in equation (10), but at the cost of the linear relationship between energy consumption and inches of insulation. A linear form of the equation was estimated for comparison. The signs of the estimated coefficients were generally similar between the two forms, and the log-linear form eliminates the heteroscedasticity found in the linear form. Thus the log-linear form is preferable.

9 Estimates for the fuel oil equation are available on request.

10 Although the prices used for these calculations reflect actual prices paid by the surveyed households, these prices are not radically different from current prices. The fuel oil and other household fuel commodities component of the CPI was 86.1 in 1980 and 81.7 in 1989 (1982-1984 = 100). The gas (piped) and electricity component of the CPI was 71.4 in 1980 and 107.5 in 1989. The Annual Energy Review (1989) reports that average prices paid for natural gas in 1981 were $4.29/thousand cubic feet and $5.65/thousand cubic feet in 1989. For electricity the average price paid in 1981 was $0.62/kilowatt hour and $0.0764/kilowatt hour in 1989.

REFERENCES

Al-Sahlawi, Mohammed A. (1989), "The Demand for Natural Gas: A Survey of Price and Income Elasticities," The Energy Journal (January): 77-90.

Blair, Roger D., David L. Koserman, and Richard C. Topel (1984), "The Impact of Improved Mileage on Gasoline Consumption," Economic Inquiry, 22(April): 209-216.

Bohi, Douglas R. (1981), Analyzing Demand Behavior: A Study of Energy Elasticities, Baltimore, MD: Johns Hopkins University Press.

Delene, J. G. and J. B. Gaston (1976), "A Regional Comparison of Savings from Various Residential Energy Conservation Strategies," Oak Ridge National Laboratory, ORNL/TM-5146 (February).

Energy Information Administration, U.S. Department of Energy (1983), Residential Energy Consumption Survey, Housing Characteristics 1980, and Consumption and Expenditure, April 1980 Through March 1981, Public Use Tape Technical Documentation, PB 83-199547, Washington, DC: National Technical Information Service (January).

Energy Information Administration (1987), Residential Energy Consumption Survey: Trends in Consumption and Expenditures, 1978-1984, DOE/EIA-0482, Washington, DC.

Energy Information Administration (1989), Annual Energy Review, Washington, DC: U.S. Government Printing Office.

Hausman, Jerry A. and Paul L. Joskow (1982), "Evaluating the Costs and Benefits of Appliance Efficiency Standards," American Economic Review, 72(2, May): 220-225.

Hsueh, Li-Min (1984), "A Model of Home Heating and Calculation of Rates of Return to Household Energy Conservation Investment," unpublished doctoral dissertation, Cornell University, Ithaca, NY.

Hutchins, Paul F., Jr. and Eric Hirst (1978), "Engineering-Economic Analysis of Single-Family Dwelling Thermal Performance," Oak Ridge National Laboratory, ORNL/CON-35.

Hwang, Jiunn T. (1984), "Multiplicative Errors-in-variables Model with Applications to the Recent Data Released by USDOE," Technical Report No. 84-1, Cornell University Statistical Center.

Khazzoom, J. Daniel (1980), "The Economic Implications of Mandated Efficiency Standards for Household Appliances," The Energy Journal (October): 21-40.

Khazzoom, J. Daniel (1986), An Econometric Model Integrating Conservation Measures in the Residential Demand for Electricity, London: JAI Press.

Latta, Robert B. (1981), The National Interim Energy Consumption Survey: Exploring the Variability in Energy Consumption: A Supplement, DOE/EIA-0272/S, Washington, DC: U.S. Department of Energy, Energy Information Administration, Office of Energy Markets and End Use, Energy End Use Division.

Mayo, John W. and John E. Mathis (1988), "The Effectiveness of Mandatory Fuel Efficiency Standards in Reducing the Demand for Gasoline," Applied Economics, 20: 211-219.

Moyers, John C. (1971), "The Value of Thermal Insulation in Residential Construction: Economics and Conservation of Energy," Oak Ridge National Laboratory, ORNL-NSF-EP-9.

Neels, Kevin (1981), "Family, Houses, and Demand for Energy: Housing Assistance Supply Experiment," Rand Note, N-1242-HUD(PB82-125733), Santa Monica, CA: The Rand Corporation (February).

O'Neal, D. L., K. R. Corum, and J. L. Jones (1981), "Estimate of Consumer Discount Rate Implicit in Single-Family Housing Construction Practices," Oak Ridge National Laboratory, ORNL/CON-62.

Scott, Alex (1980), "The Economics of House Heating," Energy Economics, 2(3, July): 130-141.

U.S.Bureau of the Census (1982), Statistical Abstract of the United States, 1982-1983, 103rd Edition, Washington, DC.

Li-Min Hsueh is Associate Research Fellow, Chung-Hua Institution for Economic Research, Taipei, Taiwan; and Jennifer L. Gerner is Associate Professor, Department of Consumer Economics and Housing, Cornell University, Ithaca, NY.

Printer friendly Cite/link Email Feedback | |

Author: | Hsueh, Li-Min; Gerner, Jennifer L. |
---|---|

Publication: | Journal of Consumer Affairs |

Date: | Jun 22, 1993 |

Words: | 5543 |

Previous Article: | Probabilities of small claims judgment satisfaction and factors influencing success. |

Next Article: | Perceived time pressure and recommended dietary practices: the moderating effect of knowledge of nutrition. |

Topics: |