Energy Economics 29 (2007) 405 – 427 www.elsevier.com/locate/eneco

Oil and energy price volatility Eva Regnier * Defense Resources Management Institute, Naval Postgraduate School, 699 Dyer Road, Monterey, CA 93943-5201, United States Available online 30 January 2006

Abstract It is commonly believed that since the 1973 oil crisis, oil and energy prices have been more volatile than other commodity prices. This study examines monthly producer prices for thousands of products over the period January 1945 through August 2005. The results show that crude oil, refined petroleum, and natural gas prices are more volatile than prices for about 95% of products sold by domestic producers. Relative to crude commodities, however, crude oil prices are currently more volatile than about 65% of other products, and oil price volatility first exceeded the median for crude commodities following the 1986 drop in oil prices. D 2005 Elsevier B.V. All rights reserved. JEL classification: Q40; E39 Keywords: Oil prices; Energy prices; Commodity prices; Price volatility

1. Introduction The conventional wisdom is that oil prices have been more volatile than prices of most other commodities since the oil crisis in 1973 (Fleming and Ostdiek, 1999; Verleger, 1993).1 This assumption has been used has been used to justify price and allocation controls and energy efficiency subsidies and recently has been the basis for recommendations for national energy policy to diversify energy sources away from oil (Awerbuch, 2003; Humphreys and McClain, 1998; Lovins et al., 2004). The related assumption that prices for other energy products are more volatile than prices for non-energy goods been used to explain microeconomic behavior and in * Tel.: +1 831 656 2912; fax: +1 831 656 2139. E-mail address: [email protected]. 1 Bookout (1990) makes the stronger claim that bafter 1973 volatility in oil prices was greater than in all other commoditiesQ. 0140-9883/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.eneco.2005.11.003

406

E. Regnier / Energy Economics 29 (2007) 405–427

particular the perceived underinvestment in energy conservation technology by consumers and industry. The question whether oil and energy prices are really more volatile than other commodity prices, or whether many commodities’ prices are actually more volatile than oil prices, has both macroeconomic and microeconomic implications. For commodity importing and exporting nations, an understanding of the risks associated with price volatility can allow better targeted policy. At the microeconomic level, the stochastic behavior of prices is relevant to evaluating real investments using modern asset pricing techniques. The relative volatility of various commodities is also a factor in evaluating descriptive models of energy conservation behavior. Whether oil prices are more volatile than prices for other commodities is a complicated question, some facets of which have been tested empirically. Plourde and Watkins (1998) found that crude oil price volatility during the 1985–1994 period was higher than price volatility for nine other commodities and that the differences were significant for six of these. In a broadbased comparison of commodity price volatility, Clem (1985) noted that during the 1975–1984 period, agricultural commodities were the most volatile. Crude petroleum and coal were less volatile than most crude nonfood materials during the three periods he analyzed (1975–1984, 1979–1981, and 1982–1984); in particular, primary metals prices were noticeably more volatile than fuels prices. Pindyck (1999) analyzed oil, coal, and natural gas price changes during the period 1870–1996,2 but did not compare their volatilities, and he calculated a constant volatility measure over a very long period during which volatility is believed to have changed. This study uses the monthly producer price index (PPI) commodity series for the period January 1945 through August 2005 to provide a heterogeneous set of product prices and a long time-horizon for examining the volatility of energy prices in relation to other products, including crude commodities. The results are generally consistent with previous studies, but much more detailed. While oil and energy price volatility increased following the 1973 oil crisis, this was mirrored by an increase in price volatility for all products, which explains why fuels did not stand out in Clem’s (1985) analysis. However, in the late 1970s, price volatility for most products returned to pre1973 levels, while oil price volatility continued to increase. Oil price volatility surpassed most other crude commodities’ price volatility in the mid-1980s and this pattern continues today. Plourde and Watkins (1998) concluded that oil price volatility was bnot clearly beyond the bounds set by other commoditiesQ. This study shows that since 1986, oil prices have been more volatile than about 95% of PPI products; a Mann–Whitney rank-sum test of log differences confirms that oil prices are significantly more volatile than over 93% of other commodities. However, the PPI commodities include intermediate and finished products that would not typically be called commodities and are generally less price volatile than primary (crude) commodities. Among PPI crude commodities, crude oil is currently more price volatile than about 65% of other products, and the difference is significant for about 60% of the crude commodity series. Moreover, crude oil price volatility was below the median volatility for crude commodities until 1986, and it dipped below the median again in 1996–1998. The finding that many crude commodities are at least as price volatile as oil suggests the need to manage input and output price risk for other commodities, through hedging and modern asset pricing methods and perhaps policies. The finding that oil price volatility is not as unusual as commonly believed also suggests that other factors make its price volatility so important: volatility in crude oil prices is largely passed to consumers who are usually somewhat insulated 2

The natural gas price series begins in 1920.

E. Regnier / Energy Economics 29 (2007) 405–427

407

from volatility in production input prices, and fuels simply make up a very large portion of the basket of goods that producers and consumers purchase. Other characteristics of oil price dynamics, such as the rate of mean reversion and the correlation with other prices and macroeconomic variables, are also very important in determining the impact of changes in oil prices. Section 2 gives the background and motivation for the current study and the implications of price volatility. Section 3 describes the data set and methodology. Section 4 gives an overview of volatility trends for energy prices and Section 5 gives both qualitative and statistical comparisons among oil, other energy commodities, and non-energy products. Section 6 discusses the volatility differences by stage of processing and evaluates the volatility of energy commodities relative to crude commodities and Section 7 concludes. 2. Background and motivation The dynamics of commodity prices are of considerable interest because of their macroeconomic and microeconomic impacts, with policy implications at both levels. 2.1. Macroeconomic implications Some economists argue that price volatility is not inherently bad if it is properly planned for, particularly now that markets have provided means to hedge to some extent against price risk. Price fluctuations are nevertheless painful for household consumers, who are not generally subject to highly volatile crude-commodity markets, but who participate in the markets for fuel oils and natural gas. For importing countries and consuming industries, energy commodities are especially important in part because they are a major input requirement to so many economic activities. Approximately 7% of U.S. gross domestic product is spent on energy, and about half that is for petroleum.3 Energy prices are also especially important because about 4% of industrial energy and material expenditures are for energy commodities.4 The U.S. government has taken costly, if not necessarily effective, steps to reduce oil price volatility, including price caps and mandatory allocation of domestic supply (under the Emergency Petroleum Allocation Act of 1973). The U.S. and other International Energy Agency member countries hold strategic petroleum reserves whose stated purpose is to stabilize supplies, not prices. However, because prices and supplies are tightly linked, supply disruptions and price fluctuations are often rhetorically interchanged, and there are frequent calls for the release of supplies from strategic petroleum reserves to mitigate price increases – most recently from members of the U.S. Congress,5 the Air Transport Association,6 and the European Union Energy Commission.7 Both the first President Bush and President Clinton allowed releases from the U.S. reserves to stabilize prices (Bamberger, 2005), as distinct from releases during times of supply disruption, as following Hurricanes Lili, Ivan, and Katrina. 3

Energy consumption expenditures from Energy Information Agency Annual Energy Review 2002 Table 3.4 and GDP figures from Bureau of Economic Analysis, downloaded July 8, 2004, from http://www.bea.doc.gov/bea/dn/gdplev.xls. 4 Based on inputs to industrial sector from Bureau of Economic Analysis Input–Output Accounts, Standard Use Table, 1998–2002, and Annual Energy Review 2002 Table 2.1 (Industrial Sector Energy Consumption) and Table 3.1 (Fossil Fuel Production Prices). 5 The Christian Science Monitor, May 17, 2004, bWhat the options are for lowering oil pricesQ, by Ron Scherer. 6 PR Newswire, May 19, 2004, bRecord Fuel Prices Demand Change in Strategic Petroleum Reserve PolicyQ. 7 Lloyd’s List, May 25, 2004, p. 1, bDe Palacio urges more oil output to stabilise priceQ.

408

E. Regnier / Energy Economics 29 (2007) 405–427

Recent studies support the belief that price volatility – not just high price levels – adversely affects macroeconomic variables in the U.S.8 First, there is considerable empirical evidence causally linking oil price changes with variables including gross domestic product (Hamilton, 1983), stock returns (Sadorsky, 1999; Papapetrou, 2001), and interest rates (Papapetrou, 2001; Ferderer, 1996). In some cases, the effects have been shown to be asymmetric, i.e. oil price increases depress the economy, but low oil prices do not boost it proportionately (Sadorsky, 1999; Ferderer, 1996; Mork, 1989). This asymmetry is one explanation for the impact of price volatility, over and above the effect of price levels (Mork, 1989; Loungani, 1986). For a survey of similar studies, see Sauter and Awerbuch (2002). For oil-exporting countries and oil-producing firms, the dynamics of oil prices are even more important. The Organization of the Petroleum Exporting Countries (OPEC) lists among its principal aims bto devise ways and means of ensuring the stabilization of prices in international oil markets with a view to eliminating harmful and unnecessary fluctuations. . .Q (OPEC, 2005, p. 11). MacBean (1999) outlines arguments that have been used to justify price stabilization policies, based on the effect of export price instability on developing countries, many of which depend on a single product for more than 50% of their export earnings. Price stabilization policies are generally discredited: for example, many people believe U.S. price and allocation controls increased oil price levels during 1973–1981. However, a number of policy options are still recommended to reduce oil price volatility or mitigate its consequences. These include subsidies for investment in production and storage capacity (Verleger, 1993), international commodity agreements (Labys and Maizels, 1993) and increasing the availability of short-term loans or risk-reduction tools like oil futures to consumers and small businesses. Ferderer (1996) clearly delineated the theories that may explain the macroeconomic consequences of oil price changes. The two mechanisms for the relationship between oil price changes and the macroeconomy that are best supported by his empirical study are also the two that generalize to other commodities. In other words, they imply that price volatility for any commodity that is an input to production will cause negative impacts on the macroeconomy. The first of these mechanisms is sectoral shocks: some labor and capital are dedicated to a particular production process or industry and it is costly to move them to another industry. Therefore price changes – in any input, or, presumably, output – that change the optimal allocation of labor and capital across industries will be costly to the macroeconomy. The second mechanism is the option value of delaying irreversible investment when there is uncertainty about the future level of prices for inputs or outputs. Volatility in the price of any input or output – not just oil – will create a non-negative option-value for irreversible investment; however, if volatility exists, prices will vary such that, over time, the value of delaying is offset by prevailing prices, so it is not clear that the general level of investment will be reduced. Evidence that many other commodity prices are as volatile as oil prices would imply that policies to mitigate the impacts of price volatility for consumers, industrial consumers, and producing firms and countries should be considered. For example, as this study and others show, metals tend to have higher volatility than other commodities, at least in part because supply is relatively inelastic (MacBean, 1999) and metals are also frequently a major component of the exports of many developing nations. Agricultural products are much more price volatile than

8

Awerbuch (2003) argues that bfuel price volatility probably represents a more important aspect of energy security [than fuel supply disruption] Q.

E. Regnier / Energy Economics 29 (2007) 405–427

409

most other commodities, due partly to their seasonality. Agriculture policies might be updated to respond to price volatility, or move toward market-based means of mitigating the price risk for producers using derivatives. 2.2. Microeconomic implications At the microeconomic level, the volatility of energy prices has been used in both descriptive and prescriptive models for energy conservation behavior. The befficiency gapQ – a term used in the energy policy literature to refer to an apparent discrepancy between the optimal and observed levels energy conservation – has been documented in industry (DeCanio, 1998; Sanstad et al., 1995) and among consumers (Koomey and Sanstad, 1994). Generally, this discrepancy is explained as a result of market failures. However, one explanation for the discrepancy takes the high volatility of energy prices as a given, and argues that the implicit discount rates observed in energy conservation behavior are much higher than the cost of capital because they are risk-adjusted to account for the volatility of the cost savings associated with energy conservation. A related argument is the option value argument: the implicit discount rates for energy conservation investments are higher than for other investments because energy price volatility drives up the option value of delaying a conservation investment (Hassett and Metcalf, 1993).9 Modern asset pricing methods that require an understanding of price dynamics are increasingly recommended for evaluating real-investment decisions (Brealey and Myers, 2000), and are being put into practice particularly in the petroleum industry (Smith and McCardle, 1999; Bradley, 1998; Laughton, 1998). These methods, which include option value and portfolio methods, generally originated in financial markets, and have been adapted to real investment decisions. Volatility parameters figure in both portfolio-based methods to optimize hedging of risks, and methods such as option value that depend on the stochastic modeling of price evolution, as explained in Schwartz (1997). Commodity price volatility is relevant to pricing real investments that affect production inputs and outputs, such as investments in capacity and reductions in energy and material intensity. Evidence that volatility of non-oil commodities is comparable to oil price volatility indicates that industries that consume or produce other commodities should take advantage of the modern asset pricing methods already in use in the oil industry. For example, agricultural-commodity prices are highly volatile, which could imply that biomass-based inputs tend to increase profit volatility and decrease shareholder value. Also in the energy-policy literature, a number of researchers have argued that because oil prices and, more recently, natural gas prices are more volatile than prices for renewables,10 oil or gas-linked cash-flows should be adjusted to account for price risk, making renewable-energy sources and energy conservation appear more profitable.11

9

The reverse argument has also been advanced in an expected utility framework. Thompson (1997) argues that because the volatile prices are linked to future costs, not benefits, in a conservation investment, reducing these costs through fixed investment is more desirable than a net present value or internal rate of return analysis would indicate. 10 Capital-intensive energy sources – in particular solar and hydropower – can be considered less price volatile because their operating costs are so low. 11 There is also a stream of research that takes essentially a portfolio approach and argues that prices for renewable energy are uncorrelated or have low correlation with oil prices, and therefore diversifying energy sources away from oil decreases overall risk (Humphreys and McClain, 1998; Awerbuch, 2003).

410

E. Regnier / Energy Economics 29 (2007) 405–427

Like airlines, which have in the last few years begun to hedge against fuel-price risk by buying oil futures, many firms face higher volatility in their input prices than output prices because volatility tends to decrease at later stages of processing (Clem, 1985; Klemmer and Kelley, 1998; MacBean, 1999). If output prices cannot be adjusted to compensate for fluctuations in input prices, a firm’s overall profits are vulnerable to input price volatility. The more modern asset pricing methods are accepted, the more important commodity price volatility will become. 3. Data and methodology The data for this study are drawn from the producer price index (PPI) commodity price series compiled by the U.S. Department of Labor’s Bureau of Labor Statistics (BLS). PPI commodities data have been used previously in studies of oil prices (Sadorsky, 1999). The PPI series for crude oil is plotted in Fig. 1, which shows that although the PPI index is an aggregation of prices from multiple domestic producers for different types of crude petroleum, the series closely tracks changes in the spot price of West Texas Intermediate oil. This is consistent with statistical results in Klemmer and Kelley (1998) showing that the PPI energy series followed price series from other sources quite closely. The PPI commodities series measure average monthly prices received by U.S. producers for their output, and are meant to reflect prices changes for a bconstant set of goods and services which represent the total output [of a standard item] of an industryQ, averaged over many producing firms (BLS, 1997). They do not reflect other components of actual contracts, such as volume discounts, and are expressed as an index such that percent

Fig. 1. Monthly prices for West Texas Intermediate Oil in Cushing, Oklahoma (Source: Energy Information Agency, http://www.eia.doe.gov/neic/historic/historic.htm) and the PPI series WPU0561, for Crude petroleum (domestic production) not seasonally adjusted.

E. Regnier / Energy Economics 29 (2007) 405–427

411

changes in the index reflect percent changes in average prices. The PPI series are designed to measure average changes in prices paid by consumers and received by producers for exchanges of the commodity, instead of the spot (marginal) price of the commodity on a market.12 A further advantage of the PPI data is that they cover a broad range of commodities, including many whose prices cannot be observed in open markets. The PPI commodities series are divided into fifteen major commodity groups, each identified by a two-digit code. The Appendix describes the coding system and the aggregation and disaggregation into groups. The items priced for the PPI commodities series include all products of the bgoods-producing sectors of the American economyQ, i.e. excluding services and imports (BLS, 1997). The prices are for domestic sales in the United States, and do not necessarily reflect world prices during the period of price caps (1973–1981) for domestic producers. The PPI commodity series (see the Appendix for more detail) are categorized as crude, intermediate, and finished materials (BLS, 1997), and include many products that would not commonly be called commodities. The PPI crude commodities are so designated because they are not sold directly to producers. Therefore, this category contains mostly primary materials such as agricultural products, lumber and metals that are more likely to be traditional commodities. The purpose of this study is to compare oil and energy price volatility with volatility of a broad range of other products to answer the question whether oil and energy are really more price volatile. For reasons discussed in the Appendix, statistical comparisons treating a set of series as a population – for example, a hypothesis test of whether oil price volatility exceeds the mean price volatility of a set of other commodities – may be suggestive but are not strictly appropriate. Therefore, the results of this study are presented as more general summaries. For example, the average volatility for all fuel and energy series was about 7% and crude oil’s volatility was higher than the volatility of 93% of other products for the period January 2001–August 2005. On the other hand, statistical comparisons between two series – for example, a hypothesis test of whether oil price volatility is significantly higher than another specific price series – may be appropriate, and are reported. Standard deviation of price differences is commonly used as a measure of volatility of commodity prices (Slade, 1991; Fleming and Ostdiek, 1999; Ferderer, 1996). Here, volatility is measured as the standard deviation over a 5-year period of log differences in monthly PPI price series, deflated using the BLS Consumer Price Index. Generally, studies fitting models to commodity price data, including oil, find that commodity prices are highly autocorrelated and mean reverting with stochastic volatility (Schwartz and Smith, 2000; Deaton and Laroque, 1992; Schwartz, 1997). This data set also shows stochastic volatility. A Levene test for constancy of variance for the log differences over 5-year periods showed significantly ( p-value = 0.05) non-constant variance over 5-year periods for about 14% of all series. However, the selection of a best model – and therefore measure – of stochastic volatility remains an open question, and might differ across commodities. Therefore standard deviation of log price differences is the best general measure of volatility and an indicator of changes in volatility over time. 12

According to Verleger (1993, p. 51), market prices may not reflect prices of delivered oil during the 1980s, but more recently these have converged.

412

E. Regnier / Energy Economics 29 (2007) 405–427

4. Energy price volatility Various dates for the increase in oil price volatility have been cited, from the 1970s (Fleming and Ostdiek, 1999) to the 1980s (Plourde and Watkins, 1998; Ferderer, 1996) to the 1990s (Earley, 2001). In particular, the OPEC oil crisis beginning in 1973 and deregulation in 1981 are commonly believed to have triggered dramatic increases in oil price volatility. A plot of overlapping 5-year volatilities in Fig. 2 shows that oil price volatility began increasing a few years before the 1973 oil crisis, and then jumped with the dramatic price hikes in 1973. The 5-year volatility level declined in 1979, when the 1973 price increases drop out of the 5-year standard deviation, but the volatility never again approached its pre-1970s level. Deregulation in 1981, with the removal of price and allocation controls on domestic producers, shows up as a small increase in price volatility, but it is dramatically overshadowed by the huge volatility rise due to the 1986 crash in oil prices. The 1991 Gulf War caused the biggest volatility increases seen to date, but the volatility drops back once the 1991 spike drops out of the rolling volatility measure. The 5-year smoothing effect of the volatility measure disguises short-term

Fig. 2. Crude oil 5-year price volatility, measured as standard deviation of log price differences and plotted monthly. Volatility measurement periods are overlapping.

E. Regnier / Energy Economics 29 (2007) 405–427

413

events, but the steady increase in this rolling measure during the late 1990s is a sign of not a one-time event, but a sustained increase in volatility due to both positive and negative price changes, although the average oil price levels were relatively low – about the same as during the late 1980s. More recently, the high oil price levels of 2004 and 2005 that have attracted much attention have actually been associated with declining volatility. Previous studies have also found that price volatility differs across fuels. In the past, both coal and natural gas have been considered less price volatile than oil. Pindyck (1999) examined oil, coal, and natural gas over a long horizon (75 years for natural gas and 125 years for oil and coal), and found that oil was more volatile than natural gas which was more volatile than coal. This general ordering is preserved in this study, except that gas price volatility has spiked since the end of Pindyck’s data set. Table 1 shows the 5-year volatilities for series in the fuels and energy group, averaged by subgroup.13 Within the fuels and energy group, there are important differences. The relative stability of coal prices is a recent phenomenon: coal was actually more volatile than crude oil during 1950–1970, a period of oligopoly control of domestic oil prices. The crude oil and refined petroleum series were the most volatile of the fuels and energy series in the 1980s and early 1990s, though more recently gas fuels, which include natural gas as well as propane and butane, have become more volatile. Price volatility for most energy products more than doubled from the 1951–1970 period into the 1971–1975 period. Refined petroleum is the exception because its price volatility was already high. Prices and price volatility for non-petroleum energy products, including gas and coal, increased right along with oil, which is not surprising as they are partial substitutes for petroleum. However, as discussed in the next section, price volatility for other commodities also spiked at the same time, rendering energy series’ price volatility relatively typical through the early 1980s, and even through the 1981 oil deregulation. 5. Volatility of non-energy commodities Factors that contribute to volatility in commodity prices include the physical characteristics of the commodity, the market structure, output elasticity and the availability of substitutes. There are a number of reasons to expect oil prices to be more volatile than other commodity prices. The holding costs for oil and other fuels are relatively high and oil is not a completely standardized commodity. The geographic concentration of oil supply is often cited as another contributor to volatility. In addition, demand is weather-sensitive and seasonal. The most commonly cited causes of the 1970s’ increase in oil price volatility are powerful market players, specifically OPEC, supply disruptions, deregulation (Fleming and Ostdiek, 1999), a decline in storage capacity and idle production capacity, and the introduction of derivatives markets, in the 1980s for oil and in the early 1990s for natural gas (Fleming and Ostdiek, 1999).14

13

All price series for which observations were available for at least 30 of the 60 months in each period were included in this average. 14 Whether derivatives markets would stabilize or destabilize commodity prices is a subject of debate (Fleming and Ostdiek, 1999).

414

Table 1 Volatility for all fuels and energy series (averaged by subgroup) and for all series (averaged by group) Average volatility 1946–1950 1951–1955 1956–1960 1961–65 1966–1970 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995 1996–2000 2001–2005 0.0222 0.0201 – – – 0.0320 0.0485 –

0.0293 0.0112 – – – 0.0157 0.0298 –

0.0208 0.0145 0.0579 0.0036 – 0.0156 0.0302 –

0.0193 0.0070 0.0942 0.0048 – 0.0028 0.0228 –

0.0248 0.0243 0.0386 0.0061 – 0.0098 0.0193 –

0.0516 0.0376 0.0673 0.0184 – 0.0361 0.0409 –

0.0110 0.0135 0.0390 0.0275 – 0.0225 0.0235 –

0.0146 0.0195 0.0379 0.0294 – 0.0294 0.0303 –

0.0176 0.0154 0.1034 0.0312 – 0.1160 0.0825 0.0739

0.0182 0.0116 0.0702 0.0246 0.0238 0.0681 0.0709 0.1076

0.0189 0.0186 0.1020 0.0189 0.0458 0.1106 0.0751 0.0514

0.0158 – 0.1251 0.0197 0.0785 0.0976 0.0753 0.0461

Group Farm products Foods and feeds Textiles Leather Fuels and energy Chemicals Rubber and plastic Lumber Pulp and paper Metals Machinery Furniture Minerals Transportation Miscellaneous

0.0999 0.0510 0.0265 0.0487 0.0392 0.0360 0.0326 0.0330 0.0306 0.0265 0.0193 0.0151 0.0148 0.0143 0.0168

0.1009 0.0478 0.0184 0.0473 0.0249 0.0258 0.0248 0.0165 0.0229 0.0218 0.0148 0.0104 0.0119 0.0081 0.0116

0.0899 0.0400 0.0115 0.0409 0.0252 0.0197 0.0282 0.0156 0.0268 0.0217 0.0136 0.0102 0.0101 0.0093 0.0102

0.0941 0.0353 0.0085 0.0321 0.0201 0.0188 0.0114 0.0202 0.0126 0.0164 0.0089 0.0070 0.0075 0.0030 0.0079

0.0878 0.0325 0.0107 0.0417 0.0199 0.0203 0.0149 0.0251 0.0135 0.0181 0.0120 0.0091 0.0117 0.0080 0.0117

0.1270 0.0588 0.0200 0.0504 0.0338 0.0375 0.0250 0.0456 0.0320 0.0318 0.0194 0.0127 0.0182 0.0131 0.0148

0.1133 0.0446 0.0175 0.0574 0.0230 0.0276 0.0183 0.0343 0.0221 0.0287 0.0152 0.0132 0.0166 0.0180 0.0161

0.1096 0.0290 0.0132 0.0262 0.0267 0.0263 0.0170 0.0279 0.0162 0.0242 0.0136 0.0132 0.0139 0.0120 0.0164

0.1339 0.0280 0.0106 0.0167 0.0554 0.0235 0.0137 0.0181 0.0171 0.0197 0.0099 0.0109 0.0094 0.0092 0.0101

0.1750 0.0273 0.0089 0.0120 0.0520 0.0193 0.0104 0.0263 0.0324 0.0150 0.0084 0.0075 0.0087 0.0075 0.0102

0.1731 0.0327 0.0088 0.0101 0.0551 0.0196 0.0075 0.0190 0.0308 0.0131 0.0075 0.0074 0.0089 0.0070 0.0101

0.1730 0.0345 0.0096 0.0101 0.0655 0.0178 0.0097 0.0158 0.0247 0.0162 0.0082 0.0078 0.0089 0.0082 0.0093

All series

0.0340

0.0280

0.0246

0.0200

0.0209

0.0333

0.0274

0.0231

0.0209

0.0220

0.0215

0.0269

The number of series included in the average for each period and subgroup is given in Table 5 in the Appendix.

E. Regnier / Energy Economics 29 (2007) 405–427

Subgroup Coal Coke Gas fuels Electric power Utility gas Crude oil Refined petroleum Other petroleum

E. Regnier / Energy Economics 29 (2007) 405–427

415

The goal of this study is to test the hypothesis that oil price volatility and volatility of other energy products are unusual among commodities – irrespective of the causes. Therefore, this study addresses the heterogeneity among commodities by comparing oil and

Fig. 3. Energy series’ 5-year price volatility (median within each subgroup) as a percentile of all commodities. Volatility measurement periods are overlapping.

416

E. Regnier / Energy Economics 29 (2007) 405–427

other fuels with the most comprehensive set of commodities available. The volatility of commodity prices is summarized by group in Table 1, which shows the average volatility for each group in each 5-year period. During the period from 1951 to 1970, most commodity prices were less volatile than during any other period. The volatility increase for oil that first appeared in the 1971–1975 period was mirrored in all 15 commodity groups. The highest volatility periods for most groups are 1946– 1950 and 1971–1975 or 1976–1980. The only exceptions are fuels and energy products and farm products. During the 1980s, most commodities’ volatility began to decline to pre-1970 levels. Fuels and energy and farm products’ price-volatilities never returned to their pre-1970 levels, and instead rose further in the late 1980s to mid-1990s. Fig. 3 shows the relative position of fuels among all series: this is the percentile of the fuels series’ price volatility, among price volatility for all commodities, with volatility measured over 5-year periods. The price volatility is a rolling measure, and the percentile is plotted for every month. Since the mid1980s, all the major fuels’ commodity prices have been more volatile than most other commodity prices. The only exception is the farm products group, whose price volatility has kept pace with oil’s. Refined petroleum and natural gas have been more volatile than most other series throughout the 1946–2005 period. Oil used to be less price volatile than most other products, during the oligopolistic market of the 1960s. Electricity price volatility is also rising, although the PPI index – a monthly average – does not reflect intra-day spikes associated with recent utility deregulation. As discussed above, statistical comparisons treating any set of series as a population are problematic. However, pairwise statistical comparisons between series are possible, as the log differences are not significantly auto-correlated. The magnitude of the absolute value of log differences for each series in the data set was compared with crude oil. Because log differences were highly non-normal,15 a Mann–Whitney non-parametric (rank-based) test was used (Conover, 1999). Table 2 summarizes the results, giving the percentage of series that was significantly ( p = 0.05) more (less) volatile than crude oil. These are perhaps the most striking results. They confirm the significance of differences in average volatility summarized in Table 1. Oil prices were significantly less volatile than about a third of all other products’ prices before 1970, and significantly more volatile than none. On the other hand, since 1986, oil prices have been significantly more volatile than at least 93% of all other product prices. 6. Stage of processing As discussed earlier, this study examines the behavior of prices for a very broad set of products, of many types and in various stages of processing. Some are consumer items; others are primary materials that are only purchased by industrial users. In general, prices of crude materials are more volatile than prices for commodities that are more advanced in the production process (Clem, 1985; Klemmer and Kelley, 1998; MacBean, 1999). Labys and

15

Jarque–Bera normality tests of each series in each 5-year period indicate that log differences were significantly ( p = 0.05) non-normal in 85% of the cases.

Table 2 Results of pairwise Mann–Whitney rank-sum tests comparing absolute log differences of crude oil price series with other commodity series ( p = 0.05) Percentage of all series significantly less (more) price volatile than crude oil

Farm products Foods and feeds Textiles Leather Fuels and energy Chemicals Rubber and plastic Lumber Pulp and paper Metals Machinery Furniture Minerals Transportation Miscellaneous

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

All series

0 (24)

1946–1950 1951–1955 1956–1960 1961–1965 1966–1970 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995 1996–2000 2001–2005 (91) (57) (15) (62) (44) (21) (13) (55) (10) (12) (0) (0) (0) (0) (0)

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

(99) (69) (40) (69) (44) (18) (19) (57) (12) (25) (8) (40) (21) (67) (2)

0 (34)

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

(98) (66) (31) (67) (40) (12) (14) (61) (2) (23) (10) (26) (9) (33) (2)

0 (31)

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

(95) (73) (33) (63) (45) (14) (29) (70) (19) (35) (24) (39) (31) (50) (14)

0 (38)

0 0 0 0 0 0 0 0 0 0 0 0 2 0 2

(92) (67) (23) (67) (42) (12) (36) (68) (27) (25) (23) (12) (17) (13) (5)

0 (31)

0 (91) 2 (42) 6 (16) 8 (33) 2 (2) 7 (5) 5 (15) 0 (44) 2 (14) 4 (12) 8 (0) 18 (0) 16 (3) 15 (0) 15 (0) 6 (14)

0 2 0 0 2 0 0 0 0 1 2 4 5 0 4

(84) (48) (9) (59) (46) (19) (9) (46) (20) (21) (3) (1) (8) (0) (0)

1 (20)

0 22 36 44 2 21 35 16 36 38 62 55 47 47 60

(83) (35) (2) (26) (52) (19) (7) (49) (4) (22) (1) (2) (5) (0) (4)

39 (17)

20 89 100 100 62 96 100 96 98 93 100 100 100 100 100

(29) (2) (0) (0) (2) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0)

93 (1)

22 93 100 100 62 97 100 91 89 97 100 100 100 100 100

(31) (3) (0) (0) (2) (0) (0) (0) (1) (0) (0) (0) (0) (0) (0)

94 (2)

49 95 100 100 85 99 100 100 93 100 100 100 100 100 100

(26) (3) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0)

97 (1)

40 92 100 100 48 97 100 100 97 98 100 100 100 100 100

(28) (4) (0) (0) (3) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0)

E. Regnier / Energy Economics 29 (2007) 405–427

Group

93 (2)

417

418

Group

Average volatility 1946–1950 1951–1955 1956–1960 1961–1965 1966–1970 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995 1996–2000 2001–2005

Farm products Foods and feeds Leather Fuels and energy Chemicals Lumber Pulp and paper Metals Minerals

0.1104 0.0527 – 0.0275 0.0212 – 0.1705 0.0411 0.0081

0.1091 0.1971 – 0.0213 0.0122 – 0.1717 0.0711 0.0053

0.0976 0.1598 – 0.0218 0.0080 – 0.0950 0.0691 0.0047

0.0980 0.1525 – 0.0124 0.0120 – 0.0679 0.0380 0.0031

0.0988 0.1261 – 0.0134 0.0397 – 0.0497 0.0522 0.0053

0.1419 0.1401 – 0.0330 0.1034 – 0.1108 0.0856 0.0076

0.1212 0.1678 – 0.0238 0.0343 – 0.0772 0.0800 0.0098

0.1148 0.2873 – 0.0195 0.0174 0.0218 0.0618 0.0663 0.0070

0.1289 0.1684 – 0.0496 0.0275 0.0176 0.0641 0.0535 0.0067

0.1792 0.1749 – 0.0429 0.0226 0.0280 0.1687 0.0413 0.0056

0.1754 0.1862 – 0.0696 0.0237 0.0228 0.1283 0.0404 0.0044

0.1768 0.2253 0.0418 0.0581 0.0214 0.0249 0.1034 0.0580 0.0067

All crude series

0.0962

0.1011

0.0878

0.0823

0.0824

0.1226

0.1039

0.0832

0.0844

0.1279

0.1251

0.1299

The number of series included in the average for each period and group is given in Table 6 in the Appendix.

E. Regnier / Energy Economics 29 (2007) 405–427

Table 3 Price volatility for all crude commodities series, averaged by group

E. Regnier / Energy Economics 29 (2007) 405–427

419

Fig. 4. Crude energy series’ 5-year price volatility (median within each subgroup) as a percentile of all crude commodities. Volatility measurement periods are overlapping. Refined petroleum and fuels and energy, average are not shown because they are not crude commodities.

Maizels (1993) explain that primary commodities’ (which they refer to as commodities, as distinct from consumer-marketed items) prices respond more quickly to supply and demand and precede changes in price levels generally. This data set shows the same pattern. Table 3 gives average volatility of crude commodities in each group. This table is analogous to Table 1 except that commodities in other stages of processing are excluded.16 In the most recent period, 2001–2005, the average volatility over all crude series is about 13%, nearly five times the average for all series in Table 1. Fig. 4 shows the percentiles of crude energy commodities among all crude commodities; this is identical to Fig. 3, except that non-crude commodities are excluded.

16

See the Appendix for a detailed explanation of the assignment of series to stages of processing.

420

Group

Percentage of all crude series significantly less (more) volatile than crude oil

Farm products Foods and feeds Leather Fuels and energy Chemicals Lumber Pulp and paper Metals Minerals

0 (95) 0 (100) – 0 (0) 0 (0) – 0 (100) 0 (100) 0 (0)

0 (98) 0 (100) – 0 (33) 0 (0) – 0 (100) 0 (75) 0 (67)

0 (97) 0 (100) – 0 (43) 0 (33) – 0 (50) 0 (63) 0 (0)

0 (96) 0 (100) – 0 (33) 0 (0) – 0 (100) 0 (89) 0 (33)

0 (94) 0 (100) – 0 (25) 0 (33) – 0 (100) 0 (67) 0 (0)

0 (92) 0 (100) – 0 (0) 0 (0) – 0 (100) 0 (56) 100 (0)

0 (84) 0 (67) – 0 (50) 0 (50) – 0 (100) 0 (100) 33 (0)

0 (85) 0 (100) – 0 (9) 0 (25) 25 (50) 0 (0) 14 (57) 57 (0)

26 (18) 33 (33) – 60 (0) 100 (0) 100 (0) 81 (0) 64 (0) 100 (0)

25 (32) 38 (63) – 40 (0) 100 (0) 100 (0) 19 (6) 82 (0) 100 (0)

54 (27) 42 (42) – 50 (0) 100 (0) 100 (0) 63 (0) 100 (0) 100 (0)

All crude series

0 (80)

0 (88)

0 (84)

0 (85)

0 (81)

4 (78)

1 (78)

7 (59)

54 (9)

45 (20)

67 (17)

1946–1950 1951–1955 1956–1960 1961–1965 1966–1970 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995 1996–2000 2001–2005 44 33 100 67 100 100 85 71 100

(31) (67) (0) (17) (0) (0) (0) (0) (0)

60 (21)

E. Regnier / Energy Economics 29 (2007) 405–427

Table 4 Results of pairwise Mann–Whitney rank-sum tests comparing absolute log differences of crude oil price series with other crude-commodity series ( p = 0.05)

E. Regnier / Energy Economics 29 (2007) 405–427

421

Tables 3 and 4 and Fig. 4 paint a very different picture about how unusual oil price volatility is. Oil prices and natural gas prices are more volatile than about 65% of crude commodities—a substantial fraction, but this by no means indicates that oil and natural gas are extreme outliers in commodity price volatility.17 Coal volatility and average volatility for all crude fuel series remained well below the median among crude commodities throughout the period of the study. The pairwise Mann–Whitney tests, summarized in Table 4 indicate that 60% of crude series are significantly less price volatile than oil, but 21% are significantly more price volatile. The series that are more price volatile than oil are other fuels and energy series, farm products, and foods and feeds. 7. Conclusions Energy price volatility is not as unusual as commonly perceived. Oil prices are highly volatile compared with all products manufactured in the United States, but among crude commodities, oil prices are not so unusual, and were in fact less volatile than most crude commodity prices until 1986. Instead, oil and energy volatility are important in part because of the volume of demand. Using the sectoral shock principle, Hamilton (1988) showed in a general equilibrium model that fluctuations in the supply of primary commodities can increase unemployment and decrease aggregate output, and smaller fluctuations were required to cause these effects for commodities that represent a large share of the economy – hence by this model, importer economies are more vulnerable to fluctuations in oil prices than to similar volatility in other commodity prices. Oil prices were very stable during the 1960s, with a volatility of well under 1%, lower than 80 to 90% of other products. Although other commodity price-volatilities spiked in 1973, the oil price increases affected consumers and most industries. One reason crude oil price volatility is perceived as unusually high is consumer-product prices are usually much less volatile than commodity prices, but consumers are exposed to crude oil price fluctuations and farm-product price fluctuations. Price changes are transmitted more readily from farm products to consumer food items and from crude oil to refined petroleum products than is price volatility for most raw materials. In addition, disruption caused by high prices is often attributed to volatility when levels and volatility are related but not perfectly correlated. The late 1990s was a period of relatively low price levels, comparable to the late 1980s, but high volatility due to both positive and negative changes. By contrast, oil price levels have been increasing fairly steadily since 2001, but that steadiness has meant low volatility. Arguments justifying interventions to reduce oil price volatility should therefore be based on the volume of oil consumed, and its role as an input to so many activities and a major expense for consumers, not on its volatility. Acknowledgement This research was supported in part by NSF Award 0411930. 17

Conservative assumptions were made, as discussed in the Appendix, such that the fraction of crude commodities less volatile than crude oil is likely to be less than 65%.

422

Table 5 Number of series included in summaries in Tables 1 and 2 for each period and group Number of series 1946–1950 1951–1955 1956–1960 1961–1965 1966–1970 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995 1996–2000 2001–2005 3 6 – – – 4 21 –

3 6 – – – 4 21 –

7 6 2 2 – 4 22 –

7 6 2 2 – 6 26 –

7 6 2 2 – 5 28 –

5 7 2 18 – 1 13 –

11 6 6 18 – 1 23 –

13 1 6 18 – 1 13 –

11 1 6 18 – 1 18 3

9 1 6 4 5 1 16 3

8 1 4 4 5 1 14 2

4 – 4 4 4 1 10 2

Group Farm products Foods and feeds Textiles Leather Fuels and energy Chemicals Rubber and plastic Lumber Pulp and paper Metals Machinery Furniture Minerals Transportation Miscellaneous

81 141 138 39 34 166 30 58 49 163 252 50 33 3 57

83 153 140 39 34 168 31 58 51 170 253 58 33 3 57

90 161 146 39 43 169 35 62 54 243 264 66 35 3 61

96 177 159 40 49 243 35 64 64 347 335 85 39 4 72

93 205 141 36 50 271 36 72 62 377 444 101 41 8 65

77 209 96 36 46 176 66 70 50 370 621 88 37 13 62

77 188 85 29 65 231 33 65 50 393 650 67 40 11 57

63 213 84 34 52 155 43 49 114 314 433 86 57 30 50

69 185 143 35 58 170 38 67 153 313 459 131 63 40 90

72 208 123 25 45 157 37 65 143 332 429 115 68 41 104

68 173 107 22 39 131 38 67 104 287 387 105 69 49 89

67 141 47 12 29 72 26 35 98 168 210 55 48 25 66

1294

1331

1471

1809

2002

2017

2041

1777

2014

1964

1735

1099

All series

E. Regnier / Energy Economics 29 (2007) 405–427

Subgroup Coal Coke Gas fuels Electric power Utility gas Crude oil Refined petroleum Other petroleum

Group

Number of series 1946–1950 1951–1955 1956–1960 1961–1965 1966–1970 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995 1996–2000 2001–2005

Farm products 58 Foods and feeds 2 Leather 0 Fuels and energy 6 Chemicals 3 Lumber 0 Pulp and paper 3 Metals 1 Minerals 3

60 6 0 6 3 0 3 8 3

67 6 0 7 3 0 2 8 3

70 6 0 9 3 0 3 9 3

62 6 0 8 3 0 3 9 3

53 6 0 4 3 0 3 9 3

55 6 0 6 4 0 3 9 3

48 2 0 11 4 4 3 7 7

50 3 0 5 3 5 16 25 9

63 8 0 5 2 5 16 22 10

59 12 0 4 1 6 16 19 10

59 9 1 6 1 4 13 14 10

All crude series

89

96

103

94

81

86

86

116

131

127

117

76

E. Regnier / Energy Economics 29 (2007) 405–427

Table 6 Number of series included in summaries in Tables 3 and 4 for each period and group

423

424

E. Regnier / Energy Economics 29 (2007) 405–427

Appendix A. Details on data set The PPI data were downloaded September 18, 2005, from ftp://ftp.bls.gov/pub/time.series/ wp/. They include final figures through April 2005, and preliminary figures through August 2005. Tables 5 and 6 give the number of series observed in the data set for each commodity group and each 5-year period. A.1. Inclusion of commodities series in the analysis Each PPI price series has a five- to eleven-digit series identifier. The last two to eight digits designate the item priced. Items with an eight-digit code are individual items, which are grouped into six-digit subproduct classes, which are in turn grouped into four-digit product classes, three-digit subgroups, and finally into fifteen major commodity groups, each identified by a two-digit code (BLS, 1997).18 An extract of the hierarchy is given in Table 7. Each two-digit series is a weighted average of the three-digit subgroup series within the two-digit group, with each component series weighted according to a relative importance measure that reflects the dollar value of sales by domestic producers (BLS, 1997). In general, each three-digit series is a weighted average of its component four-digit series, which are weighted averages of their component six-digit series, which are weighted averages of their component eight-digit individual item series. However, for some product classes and subproduct classes, no individual items are defined. In these cases, only the four- or six-digit series is defined. Therefore, in this study, the set of all series is defined to include the lowestlevel index available, in many cases the individual item (eight-digit) series, subject to the constraint that all lineages are used. Specifically, for each 5-year period, the lowest-level series in a given lineage for which at least half of the observations are available is used. For example, if a four-digit series is defined in a given 5-year period, either it is included, or one or more six- or eight-digit series that are in that four-digit class is included in the set of all series. The sampling method and construction of each PPI index are adjusted over time as products and industries change. Usually the changes are slight, but major redefinitions have occurred as well. Most important, between 1975 and 1986, the sampling method was transitioned from a commodity-based sampling of individual products to a statistical sampling of firms and products within an industry—referred to as the industry-based sampling method.19 As the industry-based method was phased in, imports were removed from all PPI indices. To the extent that international commodities prices are linked, the qualitative results in this study will apply to international commodity markets. The redefinition led to some arbitrary repetition: for example, in some cases when a new series replaced an old series, the new series was back-filled using the old series data, and both series are reported for early years, leading to duplication of data. The BLS economists take pains to adjust the index so that its level reflects price changes, rather than quality or other types of changes. In addition, as products change, indices may drop out of the PPI and new indices may be created. These changes are most common at the eight-digit level. For this reason, statistical comparisons treating a set of series as a population – for example, a 18

Unlike the BLS industry classifications, these groupings do not correspond to the Standard Industrial Classification (SIC) groupings. 19 Personal communication, Joseph Kowal, Bureau of Labor Statistics, Producer Price Index, Index Analysis and Public Information, August 1, 2002.

E. Regnier / Energy Economics 29 (2007) 405–427

425

Table 7 Selected PPI commodity series codes and item names Group (two digits)

Subgroup (three digits)

Product class (four digits)

Subproduct class (six digits)

Individual item (eight digits)

05 051 052 053 054 055 056 057 0571 0572 0573 0574 0575 0576 057601 05760101 05760102 05760103 05760104 05760106 05760111

Item name Fuels and related products and power Coal Coke oven products Gas fuels Electric power Utility natural gas Crude petroleum (domestic production) Petroleum products, refined Gasoline Kerosene and jet fuels Light fuel oils Residual fuels Lubricating oil materials Finished lubricants Automotive oil Automotive motor oil, retail Other automotive oil, retail Automotive motor oil, commercial Other automotive oil, commercial Industrial oils gal. Petroleum grease lb

hypothesis test of whether oil price volatility exceeds the mean price volatility of a set of commodities – may be suggestive but are not strictly appropriate. Therefore, the results of this study are presented as more general summaries, such as percentage of a set of commodities whose price volatility are higher than oil’s, and average price volatility for a group or class of commodities. Statistical comparisons between any two series may be appropriate, and are reported. A.2. Stage of processing In some contexts, the only products that would be called commodities are those for which there are derivatives such as futures traded on a financial market. A somewhat broader meaning would include any product that is well-standardized and can be traded in large volume, as in auctions or other clearinghouses. The PPI commodities include intermediate and finished products that would not typically be called commodities and are generally less price volatile than primary (crude) commodities. At the six-digit level, the PPI commodity series are categorized by stage of processing into crude, intermediate, and finished materials (BLS, 1997; Gaddie and Zoller, 1988). Because the stage of processing is only identified at the subproduct class (six-digit) level, and because each subproduct class can be assigned zero to three stages of processing, the identification of two-, three-, four- and eight-digit series with a stage of processing was ambiguous. Therefore, any individual item whose six-digit series was assigned to the crude stage of processing, even if not uniquely, was included as a crude item, and any two-, three-, and four-digit series that contained any six-digit subseries assigned to the crude stage of processing was also assigned to crude stage of processing. This rule tends to bias results towards the inclusion of non-crude

426

E. Regnier / Energy Economics 29 (2007) 405–427

items and therefore toward lower price volatility of the crude commodities. This means the result showing that crude oil is more price volatile than 65% of crude commodities is conservative, and it is probably more price volatile than a smaller percentage of crude commodities. References Awerbuch, S., 2003. A note: The implications of fossil price volatility and other market risks on electricity generating costs and energy security. Renewable Energy World 6 (2), 53 – 61. Bamberger, R., 2005. Strategic Petroleum Reserve. Congressional Research Service Issue Brief for Congress, updated September 7, 2005. (http://www.ndu.edu/library/docs/crs/crs_ib87050_07sep05.pdf, accessed October 30, 2005). Bookout, J.F., 1990. Managing Volatility in the Oil Industry. National Academy Press, Washington, DC. Ch. 3. Bradley, P.G., 1998. On the use of modern asset pricing for comparing alternative royalty systems for petroleum development projects. The Energy Journal 19 (1), 47 – 81. Brealey, R.A., Myers, S.C., 2000. Principles of Corporate Finance, 6th ed. McGraw Hill, New York. Clem, A., 1985. Commodity price volatility: Trends during 1975–1984. Monthly Labor Review 108 (6), 17 – 21. Conover, W., 1999. Practical Nonparametric Statistics, 3rd ed. John Wiley and Sons, New York. Deaton, A., Laroque, G., 1992. On the behaviour of commodity prices. Review of Economic Studies 59 (1), 1 – 23. DeCanio, S.J., 1998. The efficiency paradox: Bureaucratic and organizational barriers to profitable energy-saving investments. Energy Policy 26 (5), 441 – 454. Earley, R., 2001. Energy price volatility: Trends and consequences. International Energy Agency World Energy Outlook. (http://www.worldenergyoutlook.org/weo/papers/Slt200148.pdf, accessed June 24, 2004). Ferderer, J.P., 1996. Oil price volatility and the macroeconomy. Journal of Macroeconomics 18 (1), 1 – 26. Fleming, J., Ostdiek, B., 1999. The impact of energy derivatives on the crude oil market. Energy Economics 21 (2), 135 – 167. Gaddie, R., Zoller, M., 1988. New stage of process price system developed for the producer price index. Monthly Labor Review 111 (4), 3 – 16 (http://www.bls.gov/opub/mlr/1988/04/art1full.pdf). Hamilton, J.D., 1983. Oil and the macroeconomy since World War II. Journal of Political Economy 91 (2), 228 – 248. Hamilton, J.D., 1988. A neoclassical model of unemployment and the business cycle. Journal of Political Economy 96 (3), 593 – 617. Hassett, K.A., Metcalf, G.E., 1993. Energy conservation investment: Do consumers discount the future correctly? Energy Policy 21 (6), 710 – 716. Humphreys, H.B., McClain, K.T., 1998. Reducing the impacts of energy price volatility through dynamic portfolio selection. The Energy Journal 19 (3), 107 – 130. Klemmer, K.A., Kelley, J.L., 1998. Comparing PPI energy indexes to alternative data sources. Monthly Labor Review 121 (12), 33 – 41 (http://www.bls.gov/opub/mlr/1998/12/art2full.pdf). Koomey, J.G., Sanstad, A.H., 1994. Technical evidence for assessing the performance of markets affecting energy efficiency. Energy Policy 22 (10), 826 – 832. Labys, W.C., Maizels, A., 1993. Commodity price fluctuations and macroeconomic adjustments in the developed economies. Journal of Policy Modeling 15 (3), 335 – 352. Laughton, D.G., 1998. The potential for use of modern asset pricing methods for upstream petroleum project evaluation: Concluding remarks. The Energy Journal 19 (1), 149 – 153. Loungani, P., 1986. Oil price shocks and the dispersion hypothesis. The Review of Economics and Statistics 68 (3), 536 – 539. Lovins, A.B., Datta, E.K., Bustnes, O.-E., Koomey, J.G., Glasgow, N.J., 2004. Winning the Oil Endgame. Rocky Mountain Institute, Snowmass, CO. MacBean, A.I., 1999. The prima facie case. In: Greenaway, D., Morgan, C. (Eds.), The Economics of Commodity Markets. Edward Elgar, London. Mork, K.A., 1989. Oil and the macroeconomy when prices go up and down: An extension of Hamilton’s results. Journal of Political Economy 97 (3), 740 – 744. Organization of the Petroleum Exporting Countries, 2005. General information booklet. (http://www.opec.org/library/ General%20Information/pdf/geninfo.pdf, accessed October 31, 2005). Papapetrou, E., 2001. Oil price shocks, stock market, economic activity and employment in Greece. Energy Economics 23 (5), 511 – 532. Pindyck, R.S., 1999. The long-run evolution of energy prices. The Energy Journal 20 (2), 1 – 27.

E. Regnier / Energy Economics 29 (2007) 405–427

427

Plourde, A., Watkins, G., 1998. Crude oil prices between 1985 and 1994: How volatile in relation to other commodities? Resource and Energy Economics 20 (3), 245 – 262. Sadorsky, P., 1999. Oil price shocks and stock market activity. Energy Economics 21 (5), 449 – 469. Sanstad, A.H., Blumstein, C., Stoft, S.E., 1995. How high are option values in energy-efficiency investments? Energy Policy 23 (9), 739 – 743. Sauter, R., Awerbuch, S., 2002. Oil price volatility and economic activity: A survey and literature review. Research Paper. International Energy Agency, Paris. Schwartz, E.S., 1997. The stochastic behavior of commodity prices: Implications for valuation and hedging. The Journal of Finance 52 (3), 923 – 973. Schwartz, E., Smith, J.E., 2000. Short-term variations and long-term dynamics in commodity prices. Management Science 46 (7), 893 – 911. Slade, M.E., 1991. Market structure, marketing method, and price instability. The Quarterly Journal of Economics 106 (4), 1309 – 1340. Smith, J.E., McCardle, K.F., 1999. Options in the real world: Lessons learned in evaluating oil and gas investments. Operations Research 47 (1), 1 – 15. Thompson, P.B., 1997. Evaluating energy efficiency investments: Accounting for risk in the discounting process. Energy Policy 25 (12), 989 – 996. U.S. Department of Labor, Bureau of Labor Statistics, 1997. BLS Handbook of Methods. (http://www.bls.gov/opub/hom/ homtoc_pdf.htm). Verleger, P.K., 1993. Adjusting to Volatile Energy Prices. Institute for International Economics, Washington, DC.