Douglas R. Bohi was the principal investigator in a recent project that reviewed the literature concerning estimates of price elasticities of energy demand. The manuscript was prepared for the Electric Power Research Institute and will be submitted to RFF's Publications Committee as a proposed book.
The use of price elasticities of demand as a policy tool is not always understood, and the many ways of estimating them need to be evaluated. What follows is an amended version of the introductory chapter of the study, which discusses these problems.
The choice of strategies to implement an energy policy often hinges on how consumption is expected to change in response to a price change. Answers to questions about the wisdom of oil price controls, or the need for oil import controls, or the effect of a gasoline tax, depend on what is expected to happen to consumption when the price of oil changes. The production plans and investment decisions of private firms also depend on assumptions about how energy consumption will respond to a change in price. Electric utilities, for example, make investment decisions today in order to supply customers many years in the future, and one determinant of expected future consumption is the price of electricity. Similarly, auto manufacturers take into consideration the effect of increasing gasoline prices on auto sales, including the mix of fuel-efficient autos. The rapid increases in energy prices since 1974, and the prospects for further increases in the future, will continue to alter consumption patterns for energy products and the goods and services which use energy. These changes in consumption patterns will have important implications for future industrial investment, regional development, and lifestyles.
Yet, there is no consensus about how price changes affect energy consumption. The notion that the level of consumption of a product will vary inversely with a change in its price is a well-accepted principle of economics. What economics cannot tell us, however, and what is not so well-accepted, is how much consumption will be affected by a change in price, or how much a change in one price will affect the consumption of related goods and services. These are matters left to statistical evaluation based on past experience.
Unfortunately, the statistical research on this subject may only confuse the decisionmaker. For any given energy product used by any given consuming group, one can find a range of statistical results wide enough to support virtually any predisposition about the importance of the price effect. Statistical estimates of the price elasticity of demand, a measure of the responsiveness of consumption to a change in price, are close to zero in some studies and very large in others. These measures also vary by product, by consuming group, by region, by season, and by time period. Some of these differences are not disturbing to the experienced forecaster, because they may be explained, but others are not easily reconciled, particularly those which arise from virtually identical samples using different estimation methods.
One consequence of this confusion is a certain skepticism about using the price elasticity of demand in forecasting exercises. This skepticism is useful when it leads to cautious and judicious applications of the concept. It is misplaced when it leads to disdain for the importance of a price effect or the value of information contained in available research studies.
The concept of a price elasticity of demand is disarmingly simple and straightforward. It is defined as the percentage change in quantity consumed that is induced by a given percentage change in price. When the percentage change in price results in a larger percentage change in quantity demanded, the demand for the product is said to be elastic. Conversely, when the percentage change in price results in a smaller percentage change in quantity demanded, the product is said to be inelastic. But there are serious problems in accurately inferring a causal relationship between price and consumption from past experience, and in using that information appropriately in forecasting future events.
Simplicity and convenience serve to encourage the widespread use of price elasticities in policy analysis models and in projections of fuel demands. It is a parameter that, when multiplied by an expected change in price, gives an estimate of the expected change in consumption, or, alternatively, when divided by an expected change in supply, gives an estimate of the expected change in price. Thus, it is clear that estimates of price elasticities of demand can play an important role in forecasting the consequences of pricing and supply decisions. Some examples are straightforward, such as the implications of price decontrol on energy consumption, or the impact of oil import controls on domestic energy prices.
Other applications are more subtle. For example, price elasticities may be useful in designing a tax on oil consumption by providing some indication of how the tax burden will be shared. A tax may be imposed on crude oil or on different oil products, but the tax burden will vary according to relative demand elasticities. A tax paid by refiners on crude oil will not be spread evenly over all petroleum products, but will likely be distributed more toward products with the lowest price elasticities in order to minimize the effect on total sales. Similarly, taxes imposed on less price-elastic products will have less of an effect on sales compared to taxes imposed on more price-elastic products. It follows that the burden of a tax on inelastic products will affect consumers more than a tax on elastic products, but the conservation effect is greater for elastic products.
Price elasticities are also useful in connection with decisions on plant utilization and capacity expansion. Relative prices affect market shares and sales volume, altering short-run requirements for purchasing labor and materials and long-run plans for changes in the scale of production. The more successful firms are those that gauge the market correctly and respond accordingly. Elasticities constitute valuable information about the character of the market and, whether explicitly calculated or implicitly perceived through experience, can be used to advantage.
While estimates of price elasticities may be usefully applied in forecasting exercises, there are a number of dangers involved. These dangers arise from essentially two sources. The first involves the problems associated with inferring from available evidence an accurate measure of the price elasticity of demand.
The second involves the problems of using past experience to forecast future events. The second problem will arise because the importance of energy price changes may differ in the future from the past. This may happen for any one of a number of reasons. Changes in consumer tastes, preferences, and lifestyles constitute a large number of possibilities. Changes in technology alter the range of energy consumption choices and the ways energy products are used in production and consumption. General economic and political conditions may alter from the past to the future in ways that alter consumption behavior. In addition, institutional changes in the economy, such as environmental protection regulations or mandatory energy consumption rules, may alter the importance of energy prices in future consumption behavior. In cases where these changes are important, estimates derived from past experience may be misleading when applied to the future.
The second problem will also arise because the experience we draw from is not relevant to the event we are forecasting. Past experience is related to a given set of prices, incomes, and other determinants of demand. Future price changes, or changes in other determinants of demand, outside the range of past experience may require a change in the appropriate measure of the price elasticity. In particular, elasticities derived from a period of falling prices may not be appropriate for use during a period of rising prices, and elasticities derived from a period of small price changes may not be relevant for forecasting demand in a period of large price changes. An indication of the stability of elasticities may be judged from a comparison of estimates derived from periods that differ in these respects.
Any statistical measure is by its very nature subject to error and even under ideal circumstances should not be accepted uncritically. Departures from ideal conditions are inevitable in all practical applications; they increase the probability of errors and reduce the reliability of estimators as approximations of their true values. This makes it difficult to choose from a wide range of values that are all subject to error, and all the more difficult if different studies are subject to different errors and there is no basis for choosing among them.
The literature on energy demand is varied enough to provide some information on this score. A comparison of estimates of the same parameter derived from a variety of estimation methods, model specifications, and sample data provides, first, some indication of the robustness of the parameter in different circumstances. This information is valuable in judging the sensitivity of estimates to the particular way in which they are derived. It is also of value in judging the reliability of the estimates in forecasting exercises, as future conditions will undoubtedly differ from past conditions. Second, comparisons of different results obtained in different ways can be useful in ascertaining the empirical importance of some estimation problems. While these problems may lead to estimation errors, it is often impossible to determine their importance except by a comparison of empirical results.
In addition, a comparison of research studies provides valuable information for guiding future research efforts and for evaluating their results. The literature contains useful information about which approaches seem to work and which do not in various contexts. It provides indications of the nature of problems that will be encountered and the state of the art in dealing with them. Past results also provide a basis of comparison with new findings, where agreements and disagreements may both require some explanation. In general, the existing literature provides a basis of information to guide future researchers in designing their analysis and in evaluating their findings.
A version of this article appeared in print in the June 1980 issue of Resources magazine.
