Theoretical and empirical work modeling the demand for electricity, natural gas, and petroleum products has been given less research attention recently than in the 1970s and early 1980s. Yet econometric, data mining, and other statistical techniques for modeling economic relationships have improved markedly since that time.
Many of the existing models provide estimates of the short-run and long-run price-elasticity and income-elasticity of demand or elasticity of demand with respect to other measures of economic activity. However, relatively few of these models have been estimated so as to differentiate between behavioral changes in use of energy services and technological changes in the energy-using capital equipment. Yet it is likely that investment in energy efficient capital equipment will lead to permanent reductions in energy use, while purely behavioral changes in use of energy services are likely to lead to only transient changes.
We believe that it is important to understand in depth both the permanent and the transient reductions, since many policy choices depend on that understanding. For example, the advantage of real-time pricing of electricity is that it provides incentives for transient changes in electricity usage, generally motivating reductions in the use of energy services at times of the day when the marginal cost of electricity generation is high, shifting use to those times of the day when marginal costs of electricity generation are relatively low. Non-time-differentiated increases in the price of electricity may motivate both reductions in the average use of energy services and improvements in the efficiency of energy-using capital equipment. Reductions in the use of energy services may occur very quickly, over the course of minutes, hours, or days, in response to changing conditions, while changes in the energy using capital equipment occur only much slower. And policy interventions which lead to long-term reductions in the use of energy services may reduce economic welfare while policies that promote the adoption of energy-efficient equipment can be expected to increase economic welfare.
Another promising approach is to estimate energy demand models whose functional form allows different responses to energy price changes depending upon whether reductions in energy use are due to behavioral or technical factors. In a study of more than 70 different countries, Gately and Huntington (2002) demonstrated that energy demand responded much more strongly to more permanent energy price changes that exceeded previous maximum levels than to price changes that fluctuated over a range previously experienced. Presumably, these larger effects were due to fundamental technical changes in energy-using equipment that occurred during the 1970s and early 1980s. Asymmetric responses to energy price changes (both increases and decreases) are an important research methodology that could be fruitfully applied at a more disaggregated level to explore these issues.
We plan several initial research areas in energy demand modeling. First, we plan on conducting in-depth surveys of the energy demand literature, focusing on the distinctions discussed above, particularly on energy demand responses associated with characteristics of energy-using equipment vs. energy demand responses associated with changes in the use of energy services.
Second, we plan on using the California experience as a case study, more fully quantifying the underlying reasons why California per-capita electricity use has remained virtually unchanged since the mid-1970s while U.S. per-capita electricity use has continued to grow. As Bernstein et al (2002) show, differences in states' energy intensity are due not only to energy prices and government energy programs but also to new construction, capacity utilization, population, climate, and technical innovations. But the California-U.S. difference has not been fully explained.
Third, we plan on empirically estimating a set of energy demand models, designed to quantify the distinctions discussed above. The first such project will be further development of the econometric modeling of light-duty-vehicle energy demand, now being conducted by Lawrence Goulder, and briefly described above.
Analysis of barriers to energy efficiency can be improved by a clear way of measuring whether a particular option is consistent with current best available practice or the degree to which the option falls below the best available. One approach is to establish efficiency frontiers for different energy-use applications (buildings, industrial plants, etc.). Efficiency frontiers are an objective methodology for evaluating whether any particular energy-efficient option lies along the surface representing 'best available practice' where all inputs are being used efficiently.
The approach is relatively simple but powerful. First, the analyst derives the frontier of best practice from a comprehensive set of building, plants or other observations of interest. The appropriate technique can be either statistical analysis of production frontiers or linear programming solutions to the frontier, both techniques that are used widely in economics and management sciences. Second, each building or plant that does not lie on this frontier uses too much energy or other inputs and is therefore technically inefficient. The approach allows one to estimate how much technical inefficiency should be associated with each observation. The approach has been used widely to evaluate the efficiency of different organizations, such as hospitals, electric power plants and other enterprises.
An efficiency frontier has several main advantages towards understanding the energy efficiency problem. First and foremost, it promotes dialogue between engineers, economists and policy makers. It focuses attention away from conceptual debates about whether inefficiency is possible and towards the empirical issue of how much inefficiency can be assigned to a particular observation. This approach allows researchers and decision-makers to agree and understand that inefficiency may be more applicable to certain energy-use applications than to others. Second, the approach allows one to compare the inefficiency in some applications (lighting in buildings) with others (temperature control in buildings). Such comparisons may be particularly important for helping policymakers to understand policy tradeoffs. And third, it helps to identify the conditions that may improve technical efficiency. For example, some researchers have found that steel plants tend to be more efficient when they use newer capital stock vintages or when they are operating closer to full capacity.
Hillard Huntington (1994) summarizes the approach and provides references to the few studies that have tried to apply it to understanding energy efficiency.