National Academies Press: OpenBook

Improving Energy Demand Analysis (1984)

Chapter: 2 The Effects of Price on Demand

« Previous: 1 Formal Modeling and Problem-Oriented Research
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 27
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 28
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 29
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 30
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 31
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 32
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 33
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 34
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 35
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 36
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 37
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 38
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 39
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 40
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 41
Suggested Citation:"2 The Effects of Price on Demand." National Research Council. 1984. Improving Energy Demand Analysis. Washington, DC: The National Academies Press. doi: 10.17226/10457.
×
Page 42

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

- The Effects of Price on Demand There is no doubt that price affects demand, but the way that happens is not yet clearly established. Behav- ioral research gives reason to question some of the most common assumptions about what features of price consumers respond to. The research suggests that price may be a more highly differentiated variable than the one estimated in most formal demand models. This chapter discusses five different dimensions of price stimulus that can plausibly have important effects on individual demand for energy at any given time: (1) average price versus marginal price; (2) real price versus nominal price; (3) price levels versus price changes; (4) price-change thresholds versus linear price effects; and (5) price increases versus price decreases. For the first three dimensions, we examine the empirical evidence, dis- cuss its practical significance, and consider possibili- ties for future research. The issue of the formation of price expectation is covered briefly in the context of responses to price change, but the question of how con- sumers form long-run expectations when making long-run decisions is not covered. The last two dimensions of price are discussed more briefly. Our ability to explore these dimensions of price reflects the types of data available. Prior to the early 1950s, almost all econometric analyses were confined to examining the changes of highly aggregated economic mea- sures (e.g., national gasoline consumption and average retail price) over time. Since many policies are con- cerned with such aggregates, such time-series analyses are potentially useful for forecasting. However, the useful- ness of aggregate data in understanding individual behav- ior is very limited. Since the 1950s, data at the indi- vidual or household level of analysis have become much 27

28 more readily available. These data have usually been cross-sectional, that is, based on measurements at one time. One can hope to infer the causes of individual behavior with such data by comparing individuals presumed to be similar in all respects except for the level of the causal variables in question. Inference is complicated by the fact that not all factors that affect outcomes for individuals can be identified or measured. If there Is any association of unmeasured factors with the variables of interest, the estimates of those variables' effects may be biased. Similar problems occur when unknown functional relationships are misspecified. With the time-of-use electricity pricing experiments of the 1970s, some experi- mental data have become available on consumer demand behavior. However, because of the flaws in the designs of many of those experiments (see Hill et al., 1978 for an evaluation of the designs), very few analyses of these data have estimated treatment effects. Most recently, longitudinal data sets have been devel oped that use households as the unit of measurement and that follow panels of households over time. With these data it is possible to measure change in outcome variables directly and to relate those changes to the changes in antecedent variables. While the resulting estimates of the effects Ray still be biased, this bias may be reduced because the repeated measurement of the same households holds constant the effects of any unmeasured individual differences that are stable over time. For the purposes of this chapter, then, most of the useful evidence is from time-series data; only one or two of these studies, how- ever, are longitudinal. ~ — DO CONSUMERS REACT TO AVERAGE OR MARGINAL PRICES? Although elementary economic analysis may lead one to believe that consumers respond to marginal prices, there is some behavioral evidence that householders, at least, do not seem to notice them. Kempton and Montgomery (1982), for example, conducted detailed interviews with Michigan householders and concluded that the most common units people used to quantify their energy use was dollars per month--a measure of average price. If such awareness shapes behavior, householders will not respond as might be expected to the block-rate schemes common in utility billing; changing the price differential between blocks, for example, would not affect the demand of users whose

29 total bills remained unchanged; neither would inverting the rate structures. When a price change is restricted to the marginal block of energy used, the demand of con- sumers who response to average price will be less elastic than if they responded to marginal price; and base price increases, such as for service, might decrease energy use even though they have no effect on marginal prices. Of the questions discussed in this chapter, the ques- tion of whether consumers react to average or marginal prices is the most thoroughly researched by economists. Most formulations of demand theory assume that consumers are able to purchase various goods at a constant per-unit price. When this is true, average and marginal prices are identical. But because there are large fixed components to the costs of supplying energy, marginal production costs are generally lower than average costs. As a result, some (often complicated) price schedules have evolved. The declining block-rate schedule has, until recently, been the predominant form of pricing electricity and nat- ural gas in the United States. The price per unit for consumption of these energy goods is a declining function of the level of consumption (usually calculated on a monthly basis). For a set minimum level of consumption, customers are charged a constant per-unit rate. If a customer consumes more than the minimum, the additional amount is billed at a lower marginal rate, which is in effect for consumption between the minimum and a set higher level of consumption. Beyond that level, a still lower rate is charged. While most utilities only distin- guish two or three such rate blocks, some have more. In any event, such a pricing schedule results in differences between marginal and average prices, and those differences can be quite large. When marginal and average prices are different, the question arises as to which is more important in deter- mining consumer behavior. The answer that can be derived from demand theory is both; theory also predicts that consumer demand will also be affected by the structure of lover the last decade many utilities have instituted "life-line" rates and other forms of inverted block pricing schedules. The fact that new electric generating capacity now costs more than the average cost of elec- tricity has contributed to this change.

30 the entire rate schedule. 2 Economists explain this through an analysis of the declining block-rate structure. They define the marginal price as that charged in the consumption block corresponding to the highest levels of energy use; the rest of the energy cost--consumption in lower blocks times the price differential between those blocks and the highest block--is called a tax. This "tax n is sometimes called an "income component" (see, e.~., Taylor, 1978; Taylor, Blattenberger, and Verleger, 1980; Nordin, 1976) because its effect is to decrease consumers' real incomes by an amount that may be affected little or not at all by marginal energy use. 3 The theory is used to argue that changes in average prices when the marginal price is constant have an impact equal in magnitude to the effect on demand of an ordinary change in income. This analysis implies that demand is a function of both mar- ginal price and of the "tax" on lower blocks of consump- tion, which affects energy use through the budget con- straint. The size of the "tax" is a function both of average price and of the starting point of each consump- tion block. In theory, this combination of factors deter- mines the shape and position of the demand function and therefore the equilibrium level of demand for energy priced under a given rate schedule. Empirical evidence also suggests that both average and marginal price play a role in determining behavior. Invariably, however, the effect of average price has been found to be larger than that of an ordinary change in income--in one case (Hill et al., 1978), by a full order of magnitude. The data suggest, then, that consumers respond to average price as distinct from marginal price and also that they are more responsive to average prices than to changes in income. However, considerable care must be exercised in interpreting this consistent finding from the various models. While the per-dollar effect of changed cost in lower consumption blocks is, in most 2 Indeed, Blattenburger (1977) has shown that, in theory, some changes in consumption can be induced by changing rate schedules, even though average and marginal prices remain unchanged. 3This n income component" is also referred to in the eco- nomics literature as the "rate schedule premium" (Barnes, Gillingham, and Hagemann, 1982), "intra-marginal expendi- tures" (Taylor, 1978), and "implicit income" (Hill, Ott, Taylor, and Walker, 1983).

31 studies, much larger than that of normal income, the cost changes ranged from $2 to $20 per month, while incomes ranged from $200 to $4,000 or more per month. Thus, regu- lar income is a far more important determinant of energy consumption. Similarly, while it is generally found that costs in lower blocks (as measured by standardized regres- sion coefficients, t-ratios, marginal R2s, and so forth) are more strongly predictive of demand than is marginal price, it must be recalled that these measures are designed to capture all of the n income effect" of the rate schedule and that they are definitionally related to con- sumption. Marginal prices are, in the common analytic construction, not definitionally related to consumption, the entire motivation of the exercise generally being to isolate the true impact of marginal prices. Even with these caveats in mind, it seems safe to con- clude that consumers react more vigorously to rate- schedule components associated with average prices than economic theory suggests. That is, consumers seem to react both to average and marginal price, although it is not yet possible to estimate the relative magnitude of these effects. What practical difference is the effect of average price likely to make. and what can be done to improve policy analysis? Even though cost changes in lower con- sumption blocks are small and their effects on energy consumption are likely also to be small in comparison with income effects, consumer reaction to average price may have potentially important ramifications for policy models. Rate schedules are currently undergoing major restructuring: declining block-rate schedules are rapidly being replaced by life-line rates, n inverted or increas- 1ng D10CK rates, and even tlme-ot-day seasonally adjusted flat rates. The divergence of average from marginal prices that may arise from these new structures may be much greater that that observed in the past for declining block rates, and responses to average prices may gain cor- respondingly in analytical importance. There are ways to assess this possibility, and the empirical literature on the effects of differences between average and marginal prices contains an important key. When the issue of responses to average price first arose, economists were forced to base their estimates on the average electricity prices average consumers paid. With considerable effort, analysts progressed to using industry-compiled information for "typical electric bills at 100 kWh/month, 250 kWh/month, and 750 kWh/month for . .

32 consumers in a given state. Finally, with even more effort, some researchers have recently used information about the actual rate structure individual consumers face to explain their behavior. Understanding has grown in proportion to the level of detail and disaggregation of the data available for analysis. The fact that rate structures are presently changing rapidly means that an important natural experiment is taking place. Energy analysts are not monitoring it in any systematic fashion. Following even a small number of energy consumers through the course of this natural experiment could be one of the best possible investments in energy research. DO CONSUMERS REACT TO REAL OR NOMINAL PRICES ? . By the definition of economically rational choice, that choice is a function of real price. But according to self-reports, residential energy users mainly pay atten- tion to nominal prices (Kempton and Montgomery, 1982). This possibility, that people suffer what economists call a money illusion, is consistent with a body of literature in cognitive psychology that shows that people use cog- nitive heuristics to simplify complex choices for them- selves (Rahneman, Slovic, and Tversky, 1982) and that people's choices sometimes depend not only on the expected values of the alternatives but on the way the alternatives are described (e.g., Tversky and Kahneman, 1981). Thus, it is plausible that in times of inflation people may operate on a rule of thumb that "all prices are going up, n and thus fail to respond to a real price increase for a particular commodity. Or a decline in real income may be taken as a signal to cut all expenditures, even for items whose real prices are declining faster than income. Economists have explored such possibilities under the rubric of the homogeneity of demand functions. If con- sumers base their expenditures on real prices, proportion- ally equal increases or decreases in all nominal prices and incomes should leave the demand for all goods unaltered. Relationships that exhibit this property are termed "homogeneous of degree zero. n When demand func- tions are homogeneous, the demands for all commodities add up to total income. The exact form of this restriction depends on how preferences, and thereby the demand func- tions, are specified. To test the implications of homo- geneity, it is necessary to analyze the demand for all

33 goods at once. Although this is not feasible for all goods taken individually, it has been accomplished in some models for broad classes of goods (such as food, housing, entertainment, and so forth) as well as for some subsets of goods, such as peak and off-peak electricity. Almost without exception, in studies of demand for meat, consumer durables, and electricity, the restrictions on behavior implied by homogeneity have been rejected (see., e.g., Atkinson, 1977; Byron, 1970; Caves and Christensen, 1978; Court, 1967; Diewert, 1974; Theil, 1965). However, in all of the studies the hypothesis rejected was actually a joint hypothesis: that demands are homogeneous and that the particular specification of the demand function used is appropriate. The rejection of this joint hypothesis does not allow one to determine which part is incorrect. There is some evidence that it is the specifications-- which are usually highly simplified--that are incorrect. The residential time-of-day electricity demand systems used by such authors as Atkinson (1977) or Caves and Christensen (1978) provide a good example. These models use single-valued demand functions (i.e., functions allow- ing only one level of demand for each price) and ignore the fact that intervening between electricity consumption at various hours of the day and consumers' utility is a If the complicated set of home production relationships. production of commodities at home were a smooth and con- tinuous function of peak and off-peak electricity usage and if the law of diminishing returns also applied, there would be no problem. But some home-produced commodities, such as hot water, are produced primarily in batches and stored from one period to the next, to take advantage of economies of scale: householders consume from the result- ing inventories in accordance with classic inventory adjustment processes (Hill, 1978). This means that for some uses electricity in one period is substitutable for that in another and, therefore, that for certain combina- tions of peak and off-peak prices the derived demand functions are multivalued, nondifferentiable, and even discontinuous. While some possible aggregations of activities might tend to smooth out the production sur- face, it remains true that home production is complex and its outputs may not be representable by a smooth function of peak and off-peak electricity demand. Thus, it would be fortuitous if a single-valued continuous demand speci- fication would yield results consistent with homogeneity (or any condition implied by demand theory except negativ-

34 ity), even if preferences were perfectly well-behaved when defined in terms of home-produced commodities such as clean clothes and cooked meals. Further indirect evidence that it is the specification of demand functions that is in error rather than the assumption of homogeneity is provided by nonparametric tests based on the axioms of revealed preference. 4 In general, the restrictions of demand theory (for example, negatively sloped demand functions) are found by those tests to hold for aggregate data. Unfortunately, however, the homogeneity hypothesis is not directly testable with these axioms (see Deaton, 1983). In conclusion, the empirical evidence contained in the literature on demand systems does not rule out the possibility that consumers react to nominal rather than to real prices. Two other sources of evidence in economics may be per- tinent to the question of responses to nominal price: macroeconomics and the study of equilibrium in single markets. In the macroeconomic literature, the issue of real versus nominal prices is generally cast in terms of differential perception of price and income changes. If consumers perceive only price increases, inflation should increase saving because the price increases would dampen consumption; the reverse type of money illusion (i.e., consumers' perceiving only income increases) would reduce saving.s The evidence has suggested that money illusion is a relatively short-term phenomenon. Until quite recently, savings rates had remained fairly constant for 200 years while price levels had increased manyfold. How- ever, differential perceptions of and responses to price and income changes may be important in determining indi- vidual behavior in the short run. Evidence from the study of single markets (e.g., studies of gasoline demand) is hard to consolidate because there are so many markets. Most of the studies use only real (i.e., deflated) prices and therefore do not inves- 4The axioms of revealed preference were originally developed by Samuelson (1970) and have been extended and operationally defined by Afriat (1973), Diewert (1973), and Varian (1982). Much empirical evidence in the psycho- logical literature, however, raises serious questions about the adequacy of the axioms (e.g., Tversky and Kahneman, 1981; Kahneman, Slovic, and Tversky, 1982). sHowever, if consumers expect continuing (or escalating) inflation, different behavior would be expected.

35 Ligate the question of whether it is real or nominal prices that best explain behavior. When comparisons are attempted, they are hampered by the fact that nominal prices for particular commodities and general price indices tend to covary. In cross-sectional analyses, there is only one value for a commodity price and one for the general price index, and so it is impossible to exam- ine the issue. Dividing the variable nominal prices by the constant price inflator may change the nonstandardized price coefficient (because it changes the scale of mea- surement), but it will not alter any goodness-of-fit mea- sure of the estimating algorithm and therefore provides no means of judging which form of price best explains the data. With longitudinal or time-series data, as well as with cross-sectional data containing local price-level information, the problems in analyzing the data are less statistical and more conceptual. If the observation period is long enough and varied enough, there may be sufficient independent variation in nominal prices and price-level indices for statistical methods to distinguish which price concept reveals more about behavior, but choosing the weights upon which the index is based requires care and can dominate the results. Thus, the empirical evidence on whether consumers respond to real or nominal prices is inconclusive. How- ever, economists seem to agree that while the speed of reaction may well depend more on nominal prices, the over- all reaction is constrained by real prices. The impor- tance of the issue of consumer reaction to real versus nominal prices depends on the level and stability of the overall rate of price change. When inflation is low and constant, as in 1955-1965, it does not really matter if consumers are reacting to nominal or real changes since the two are nearly equal and move in the same direction. During periods of higher inflation and, especially, vari- able inflation, the issue may be more important. The last 10 years have been such a period. Recent price instabil- ity offered a good opportunity to collect relevant data on price response, but energy analysts did not do so. There may be other chances in the future, and analysts should be ready to take advantage of them. DO CONSUMERS REACT TO PRICE LEVELS OR PRICE CHANGES? Although demand analysts usually express equilibrium con- ditions in terms of price levels, economists have hypothe-

36 sized for years that the speed of consumer adjustment to new prices is proportional to the rate of change of prices. Such ~ ~ ~ ~ ~ ~ ~ a hypothesis is consistent with an extensive empirical literature in the psychology of perception (e.g., Helson, 1964). The most common form of the hypoth- esis in economics is the adaptive expectations model in which consumers are assumed to allocate expenditures on the basis of "expected" long-term (or normal) prices rather than actual current prices. Expectations are assumed to be adapted by a constant fraction of the dis- crepancy between most recent expected and current actual prices. That is, Pt - Pt_1 = (1 - r)(Pt ~ Pt-l) ~ where Pt is the price expected in period t for period t + 1; Pt_1 is the price expected in period t - 1 for period t; Pt is the actual price in period t; and r is a positive number less than one. If prices have been stable until the current period, it is easy to show that the current expected price equals the last period's actual price plus a fraction of the change in actual price between the two periods: that is, Pt = Pt-1 + (1 - r)(Pt ~ Pt_l) when Pt_1 = Pt_1 . Thus, if consumers do base their decisions on expected prices and if they base their expectations on deviations of actual from expected prices, then both the level and the rate of change of prices should affect consumer reac- tion. The adaptive expectations model has been used numerous times in energy demand studies (see, e.g., Dahl, 1979; Hopthakker, Verleger, and Sheehan, 1974; Kwast, 1980; Taylor, 1978). The data are almost always found to be consistent with the hypothesized model. The trouble is that in most cases, the hypothesis is not incorporated into the model directly but via the Koyck transformation (Koyck, 1954). This technique makes it nearly impossible to distinguish whether the data are being generated by adaptive expectations, by partial adjustment to current prices, or even by simple serial correlation of the errors (see Griliches, 1967) because all three hypotheses imply the same things for the estimated coefficients in the model.6 6 In theory one should be able to tell which mechanism is operating because the adaptive expectations model

37 The work of Archibald and Gillingham (1978a, 1978b) is notable in that it allows alternative hypotheses to be distinguished. In their longitudinal study of 1972-1973 consumer expenditure data, they include both current prices and quarterly rates of change in price in a regres- sion model of gasoline demand. They find that demand decreases as price levels increase but, holding price levels constant, is higher when the price increases have been more rapid. This model fits the data better than a related alternative that includes price level and its square. While Archibald and Gillingham do not specifi- cally discuss the adaptive expectations hypothesis, this pattern of a negative effect of current price and a posi- tive effect of price change is exactly what one would expect to find during a period of rapid price changes under an adaptive expectations hypothesis. It is as if people do not expect that recent rapid price increases will be sustained into the future. In terms of current- period actual prices (the variable Archibald and Gillingham used), the current expected price can be expressed as Pt = Pt ~ r(Pt ~ Pt_l) when Pt-1 = Pt-1 . Thus, the effects of price level and price change should have opposite signs and the absolute value of the latter should be smaller that that of the former. Archibald and Gillingham found that the coefficient for (the natural logarithm of) gasoline price level was -.549 while that for price change was +.142. These coefficients imply a level of r of roughly 25 percent, which means that con- sumers increase their expectations of prices by about 75 percent of the discrepancy between last quarter's expected prices and current prices. Before concluding that the evidence supports the idea that consumers react to both price levels and price changes, we should note that Archibald and Gillingham's findings are subject to alternative interpretations. Their sample period ended in the first quarter of 1974, exactly the period in which the effects of the 1973 oil implies serial correlation of the errors and the partial adjustment mechanism does not. In practice it is difficult to tell if any observed serial correlation is due to the incorporation of the Koyck transformation or is inherent to the original data.

38 embargo were being felt. This timing, combined with the fact that the investigators pooled four quarterly obser- vations per household to form their sample, means that the change measure they used for gasoline prices coincided with an oil embargo during the measurement period. If the price-change variable is picking up the effects of the embargo, the implication might be not that consumers respond to price changes as well as levels, but that they are unwilling to reduce consumption when the price level is due to a (perhaps temporary) supply disruption. 7 The weight of the evidence is that consumer response is partly determined by the rate of change of prices, not just by price levels. The issue can be very important in periods of volatile prices. Much more work needs to be done to estimate the magnitude of consumer response to the rate of change of prices, and better data are needed for that work. IS THERE A THRESHOLD FOR THE RESPONSE TO PRICE? Voluminous psychological literature--mostly from labora- tory studies but also from field studies of organizational behavior--supports the hypothesis that energy users often act as problem avoiders (Stern and Aronson, 1984) and the related hypothesis that responses to price may not occur until a threshold has been crossed. The notion of prob- lem avoidance also suggests that once a threshold is crossed and the consumer believes that a response is needed, the size of the short-run adjustment may not be strongly affected by the size of the price change. The 7 There are further complications with the Archibald and Gillingham study that cloud interpretation. Cross- sectional demand analyses are usually interpreted as long- run in nature, but Archibald and Gillingham interpret them as short run. They justify their interpretation on the basis of controlling for a vehicle stock measure and some locational measures through which the long run is supposed to operate. They cannot, however, control for all factors associated with the long-run response, and notably missing from their control variables is distance of residence from work. Other analyses have shown that work distance is a major determinant of the long-run gasoline demand response; thus it is not clear how to interpret Archibald and Gillingham's model.

39 findings of an experiment with time-of-use electricity pricing are consistent with this view (Heberlein and Warriner, 1982): households shifted electricity use to off-peak periods in response to price differentials rang- ing from 2:1 to 8:1. Within that range, however, atti- tudinal variables accounted for much more of the variance in energy use than prices did. It is possible to conduct studies to formally investi- gate the threshold hypothesis with respect to energy demand. Experiments with alternative electricity rates, such as the time-of-use pricing experiments, could be used as tests if properly designed rate schedules were included among the experimental conditions. It might also be pos- sible to illuminate the issue through analyses of panel data on electricity consumption because electricity rates often change in different proportions in different utility service areas. DO PRICE INCREASES AND DECREASES PRODUCE DIFFERENT RESPONSES ? There is evidence from psychological research that people do more to avoid a loss than to reap an equal gain (Kahneman and Tversky, 1979); this effect is sometimes true for energy-related behavior as well (Yates, 1982). This loss avoidance may mean that energy users will respond more vigorously to a Drice increase than to comparable decrease. This possibility might be addressed by analyses of panel data in the face of price increases and decreases (as there have recently been for gasoline and home heating oil). CONCLUSIONS Most energy policy models, including several used by the Department of Energy's Office of Policy, Planning, and is, use one or two energy price measures that are ~ _ _ . ~ allowed to affect predicted energy demand via some simple lag structure. In contrast, our examination of the price variable suggests that at least three dimensions of price may have important effects on energy demand. While in no case is the empirical or theoretical evidence sufficient to suggest concrete changes to existing policy models, some general conclusions do seem justified. i

40 Models dealing with energy sources that are sold under complex rate schedules should make allowance for differ- ential response to average and marginal prices since both seem to affect demand. Similarly, if policy models are to be used to assess the short-term impact of pricing policy (as opposed to the long-run equilibrium effects), they must be reformulated so as to reflect the complex nature of response to both price levels and to rates of change in prices. Simple geometric lag structures cannot be expected to capture the dynamic aspect of real-life consumer behavior. Similarly, the dynamics of consumer response are also likely to be affected differentially if real energy price changes are the result of changes in nominal energy prices or of changes in the nominal prices of other goods. We should note that all the hypotheses discussed in this chapter can easily be represented in terms usable in formal demand models. In fact, given sufficiently detailed data, these hypotheses can also be tested by modeling techniques. Thus, both problem-oriented research and general data collection efforts can feed into the model-building process in addressing the questions raised in this chapter. (Chapter 3 raises some questions about the nature of consumer response to financial stimuli that are not so easily incorporated into existing models.) Although the several price variables discussed in this chapter can readily be modeled, there is not yet enough empirical evidence to offer much guidance on what changes to make in the models. This state of affairs can only be remedied by further analysis of existing data, by col- lecting new data, and by developing the theory. On the question of average versus marginal prices, further analysis of existing data may yield enough infor- mation to reformulate current policy models in a more realistic manner. Any data base sufficiently rich to allow analysts to adjust incomes to account for the n income component" of rate schedules is also sufficiently rich to use equations that do not force the effects of this component to equal that of ordinary income. Very few demand analysts have reported the results of such model specifications; more should do so. Existing data are not rich enough, however, to change models to account for the different dynamics that may result from changes in real versus nominal price or changes in price levels versus rates of change in price levels. There has been insufficient independent varia- tion in these dimensions of price to allow precise esti-

41 mation of their respective effects. In most data for any particular economy, real prices have been highly cor- related with nominal prices. What variation does exist is confounded by the fact that effects that result in inflation in a particular period may have independent effects on energy demands. Furthermore, most analysts have access only to national cost-of-living indices and thus cannot capitalize on what cross-sectional variation there is in price levels. Similarly, there is very little cross-sectional variation in price changes for several important fuels (e.g., gasoline and fuel oil), and past price levels are negatively correlated by definition with rates of change in prices. Because different features of price have covaried in the past, simple assumptions, such as that consumers respond only to real prices, may be reasonably adequate for modeling existing data--even if the assumptions are in error. But to the extent that future events do not follow past patterns, more differentiated knowledge about how price affects demand will have practical importance. Some relatively inexpensive improvements to existing data sets might help analysts in addressing the price issues. The Bureau of Labor Statistics computes, but until recently did not release, a consumer price index (CPI) for each of 23 Standard Metropolitan Statistical Areas. If detailed geographic area measures were made available, the full set of these CPI measures could be used to compute price indices that are comparable both across area and over time. There may be enough indepen- dent variation in these data to determine the relative effects of real and nominal prices on energy demand. Furthermore, with data on the individual commodity prices upon which the indices are constructed, repeated measures of consumption over the same units of time--such as is available in a crude form from the Panel Study of Income Dynamics and may in the future be available from the Residential Energy Consumption Survey (RECS)--might lead to real progress in identifying the dynamics of consumer response to price levels and changes. New and continued data collection will be necessary to provide satisfactory evidence about which dimensions of price affect consumer behavior. The past decade has presented analysts with a number of natural experiments involving energy price changes. Had the decade begun with a measuring instrument such as RECS in place, far more detailed information would be available for modeling energy demand behavior. If detailed measurement had been

42 done before the change in economic conditions, each house- hold could have acted as its own control for the analysis of the effects of the change in conditions. The lesson for the future is that it is necessary to have a data collection capability in place well before it is of any current interest. Furthermore, once a sample is created, it is valuable to maintain it by contacting sampled house- holds on a regular basis. In short, to develop an ade- quate understanding of how price changes affect consumer behavior, it is important to conduct RECS and similar studies even during rather "dull" periods in which energy supplies are plentiful and prices stable.

Next: 3 The Effects of Financial Incentives on Energy-Efficient Investments in Residential Buildings »
Improving Energy Demand Analysis Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!