Oil, Automobiles, and the U.S. Economy: How Much Have Things Really Changed?
2011; University of Chicago Press; Volume: 25; Issue: 1 Linguagem: Inglês
10.1086/657541
ISSN1537-2642
AutoresValerie Ramey, Daniel J. Vine,
Tópico(s)Global Energy and Sustainability Research
ResumoPrevious articleNext article FreeOil, Automobiles, and the U.S. Economy: How Much Have Things Really Changed?Valerie A. Ramey and Daniel J. VineValerie A. RameyUniversity of California, San Diego, and NBER Search for more articles by this author and Daniel J. VineBoard of Governors of the Federal Reserve System Search for more articles by this author University of California, San Diego, and NBERBoard of Governors of the Federal Reserve SystemPDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinkedInRedditEmailPrint SectionsMoreI. IntroductionBetween 2002 and mid-2008, the average real price of gasoline in the United States increased more than twofold after having risen only modestly in the preceding 15 years. Not surprisingly, this run-up led to renewed interest in the effects of oil shocks on the U.S. aggregate economy. Hamilton's (1983) seminal paper documented the negative effects of oil shocks on the aggregate economy, and numerous papers since that time have extended or questioned the strength of these effects.1 Most recently, several authors have argued that the effects of oil price shocks on U.S. aggregate activity have declined since the mid-1980s (e.g., Blanchard and Riggi 2009; Edelstein and Kilian 2009; Herrera and Pesavento 2009; Blanchard and Galí 2010). These papers have variously attributed the decline to improved monetary policy, a smaller share of oil in production, or more flexible labor markets. Empirical work has also shown that a more muted response in the consumption of motor vehicles to energy price shocks has played a large role in obtaining these results (Edelstein and Kilian 2009).This paper reexamines the extent to which the impact of oil shocks on the aggregate economy—and on the motor vehicle industry in particular—has changed over time. We first discuss the array of energy cost measures that authors in the literature have used to define oil price shocks, and then we survey the theoretical contributions from a number of dynamic stochastic general equilibrium (DSGE) macro models that include various roles for oil in the economy. Using these models, there are a number of structural parameters that reasonably could have changed over time and reduced the potency with which oil price fluctuations depress aggregate output.However, all of these macro models assume that the price of oil reflects the true cost of energy for firms and consumers. While much of the recent empirical work uses published measures of oil or gasoline prices as an indicator of the strength of the oil price shocks, we find that these measures neglect the impact of the shortages that occurred in the critical 1973–74 and 1979 oil shock episodes, due to price controls.Using two oil shock measures that include the effects of both price and nonprice rationing, we reexamine the evidence from vector autoregressions (VARs) that oil disturbances have had less impact on the real economy in the past 20 years than in the preceding decades. The results show that the responses of motor vehicle consumption and aggregate output to shortage-adjusted oil price shocks appear just as great during the last couple of decades as they were in the 1970s and early 1980s. However, even the new measures imply that the impact on nominal variables has become noticeably muted.Why has there been so little change over time in the response of motor vehicle consumption to oil price changes? We find that, despite the many innovations in the way the U.S. economy produces and uses motor vehicles that have occurred over the past 40 years, the primary channels through which oil prices directly affect motor vehicles have not changed much over time. Namely, we present evidence that the recent increases in gasoline prices have caused just as much anxiety in consumers now as was observed 40 years ago, and the shifts in demand across vehicle size classes have also been as disruptive to motor vehicle capacity utilization since 2000 as they were in the 1970s and early 1980s.The paper proceeds as follows. Section II reviews the data available on various measures of oil prices and discusses how modern DSGE models accommodate the role of oil in the economy. It also presents evidence that the presence of energy price controls and gasoline shortages in the 1970s may cause problems in empirical work because published prices in that era do not reflect the true cost of energy. Using measures that include the cost of shortages, we find no evidence of weaker effects of oil shocks on the real economy. Because we find that the motor vehicle industry plays a central role in the propagation of the oil shocks, the remainder of the paper studies this industry in detail. Section III examines the role of the motor vehicle industry in the overall economy. Section IV discusses how gas prices affect vehicle demand, and Section V shows the ways in which these shocks affect production. Section VI concludes.II. Oil Shocks and the U.S. EconomyWe begin by reviewing the behavior of several key measures of oil prices over the past few decades. After describing how macro DSGE models have been used to understand the role of energy costs in the economy, we present evidence that price controls may have led to a wedge between the published price of oil and the true cost of oil during the large oil price shocks in the 1970s. Using VARs that are similar to those estimated by Edelstein and Kilian (2009) and Blanchard and Galí (2010), we then show how mismeasurement of the true cost of oil in the 1970s may have caused the appearance of structural instability in the impulse-response functions of real output to oil price shocks. Using cost measures that account for shortages, we find that the impulse-response functions have not changed much over time.A. Overview of Oil PricesFigure 1 displays three oil price measures: the producer price index for crude petroleum (PPI-oil), the refiner acquisition cost of imported oil (RAQ), and the consumer price index for gasoline (CPI-gas). Hamilton (2003, 2009) typically uses the PPI-oil measure, and Mork (1989) and Barsky and Kilian (2002) use versions of the RAQ measure.2 Unfortunately, the RAQ measure starts only in 1974.3 We include the CPI-gas measure because several authors have shown that gasoline is a large share of U.S. petroleum consumption, and gasoline prices are also the most relevant energy price measure for the automobile sector.4Fig. 1. Petroleum prices, January 1967 through March 2010. A, Nominal price indexes; B, real price indexes. Data come from the Bureau of Labor Statistics. The real indexes normalize the changes in petroleum prices by the changes in headline consumer price inflation. Refiners acquisition cost data begin in 1974. For each series x, the log index is calculated as .View Large ImageDownload PowerPointOil and gas prices—displayed in log current dollars in figure 1A and in log real index points in figure 1B—have risen notably at several points in history. Four episodes stand out in particular: first, the real price of gasoline rose 27% between October 1973 and May 1974, the result of an even larger rise in the price of crude oil after the Yom Kippur War. After falling back a bit over the next 4 years, the price of crude oil began to rise again at the end of 1978. By the spring of 1980, the Iranian Revolution and the Iran-Iraq War led to losses in crude oil production that pushed up the price of imported oil 71% and the price of gasoline 46%. Between 1982 and 1985, the nominal price of gasoline grew only modestly until Saudi Arabia abandoned production quotas and the price of crude oil plunged.Real gasoline prices continued to trend lower after 1985, and, by the end of the 1990s, real gasoline prices had receded to record low levels. This pattern changed abruptly at the beginning of 1999, when the Organization of Petroleum Exporting Countries (OPEC) members phased in several cuts to production quotas. The real price of gasoline surged 43% by the summer of 2001 before the weakening world economy put downward pressure on crude oil prices. The relief was short lived, however. Gas prices began to rise again in early 2002, when political turmoil in Venezuela shut down much of the country's crude oil production; real crude prices climbed 588% and gasoline prices climbed 127% by summer 2008, and then they collapsed when the financial crisis spread from the housing sector to the rest of the economy and interrupted aggregate demand.B. Oil Shocks in Macro DSGE ModelsEconomists take a keen interest in oil prices because these episodes of steep increases in prices were often followed by recessions. The literature has introduced into macro models four principle channels through which oil or energy shocks can lead to recessions: (i) energy serves as an important input to production; (ii) energy is an important consumption good; (iii) changes in energy prices lead to costly shifts in demand across sectors; and (iv) the policy response to oil price shocks includes monetary tightening, a move that depresses output. Often layered on top of these channels are forces that multiply and propagate the effects of oil price shocks on aggregate output, such as real wage rigidities (e.g., Bruno and Sachs 1982; Blanchard and Riggi 2009; Blanchard and Galí 2010), imperfect competition (Rotemberg and Woodford 1996), variable utilization rates (Finn 2000), vintage capital effects (Atkeson and Kehoe 1999; Wei 2009), and multiplier effects created by externalities across firms (Aguiar-Conraria and Wen 2007). We will briefly discuss each of these channels and point out which parameters in these DSGE models are suspected to have changed over time.1. Energy as an Input to ProductionBerndt and Wood (1975), Bruno and Sachs (1982), and Pindyck and Rotemberg (1983) were among the first to study energy price shocks in a framework that accommodates energy as an input to production. The strength of this channel is limited, however, by the small share of energy in total production costs, even in the 1970s. Finn (1991, 2000) modifies the standard model to reflect the notion that the energy requirements of installed capital are often fixed, and thus energy must be used in fixed proportions to capital use. This feature makes output more sensitive to increases in energy prices.In many of the models mentioned above, a decrease in the amount of oil required to produce a unit of output would reduce the effect of oil shocks on the aggregate economy. This result suggests that increases over time in the fuel efficiency of many types of production technology may have weakened the relationship between oil price and real output. In addition, structural parameters not directly related to the use of energy can also affect the transmission of energy price shocks in DSGE models. For example, Blanchard and Galí (2010) show that a decline in the rigidity of wages in these types of models reduces the effects of oil price shocks on output.2. Energy as a Consumption GoodIn addition to the consequences of reduced output in general equilibrium, increases in oil prices also have direct effects on demand. First, oil shocks can lead to declines in demand for goods for which consumption is complementary with purchases of oil. Hamilton (1988) and Wei (2009) use models of demand for motor vehicles to show this effect. Second, oil shocks introduce uncertainty into the outlook for future energy prices, and increases in uncertainty can dampen demand for goods if purchases are costly to reverse (Bernanke 1983). Third, for energy-consuming capital goods, increases in the price of energy change the desired characteristics of the capital in use. Because the energy efficiency of the existing stock of consumer durables available in the short run is largely fixed, demand for new goods can shift between products in an exaggerated fashion and reflects the widening differential in the relative cost of ownership between different types of goods. For motor vehicles, smaller and more fuel-efficient models naturally become more desirable.In these types of models, one parameter that has likely changed over time is the energy efficiency of consumer durable goods, including motor vehicles and other appliances. When energy efficiency rises or the share of these types of goods in total consumption falls, then we would expect the impact of oil shocks on output to diminish.3. Sectoral Shifts and Costly Factor MobilitySeveral papers have investigated sectoral shifts as a way in which oil price shocks affect the aggregate output. Davis (1987) and Hamilton (1988) both suggest that oil price shocks have a bigger effect on output if the shocks induce sectoral shifts and factor adjustment is costly. Bresnahan and Ramey (1993) argue that oil shocks can lead to disruptive sectoral shifts, even within narrowly defined industries. They present empirical evidence that shifts in demand between size classes of automobiles disrupted output in the U.S. automobile industry during the 1970s.In the context of the multisector models, it is not clear that the structure of the economy has changed in a way that would weaken the transmission of oil price shocks through the sectoral-shifts channel. We find evidence in the motor vehicle industry that this channel remains quite potent.4. Monetary Policy Reaction FunctionsBernanke, Gertler, and Watson (1997) argue that the endogenous response of monetary policy to an increase in oil prices is an important part of the outsized declines observed in output, a result they showed by using a structural VAR and counterfactual experiments with different monetary policy rules.5 Using a calibrated DSGE model, Leduc and Sill (2004) find that 40% of the decline in output that follows a positive shock to oil prices reflects the systematic component of monetary policy.To look for changes in policy parameters over time and assess whether these changes may have reduced the impact of oil shocks on output, a number of papers have either simulated monetary DSGE models or estimated monetary structural VARs. Blanchard and Riggi (2009), Herrera and Pesavento (2009), and Blanchard and Galí (2010) all find evidence that oil price shocks have had less impact in recent decades, in part, because of the changes in monetary policy.To summarize, the theoretical literature has suggested a variety of ways in which oil shocks affect the economy. Some of these effects could be weaker now, while other effects could easily be as strong.C. The Importance of Nonprice Rationing in the 1970sIn the models described above, it is assumed that the price of oil reflects the true acquisition cost of energy for firms and households. While the literature on the effects of oil shocks has debated the merits of various measures of oil prices and whether the effects are nonlinear, much of it has missed a potentially important change in the degree to which oil prices reflect oil disruptions.6 In particular, other than Mork (1989), macroeconomists have not paid much attention to the embargoes, price controls, and shortages that marked the oil price disturbances in the 1970s. Helbling and Turley (1975) document that price controls were first imposed on the U.S. domestic oil industry in August 1971 as part of the general imposition of price controls. The controls on other sectors of the economy were phased out, but the controls were made more stringent on the domestic oil industry in response to the OPEC embargo of October 1973. These complex controls, which imposed a price ceiling on "old" oil that was lower than the one imposed on "new" oil, led to significant disruptions in the production of domestic oil and held the average domestic price of crude oil below the world price. Most of the effects of these controls were felt in the markets for gasoline and diesel fuel. According to some estimates, 20% of the gasoline stations ran out of gas during the height of the crisis (Frum 2000).Pisarski and de Terra (1975) detail the policy responses to the embargo in various European countries. While most European countries did not impose the types of price controls imposed in the United States, they responded with other sorts of controls, such as bans on Sunday driving (Germany, Italy, Netherlands, Switzerland) and limits on gas purchases (Great Britain, Netherlands, Sweden, Switzerland). Almost all countries imposed lower speed limits.Multiple oil and gas price controls also helped produce shortages after the Iranian Revolution of 1979. In April 1979, President Carter announced gradual decontrol of oil prices but proposed a windfall profits tax. In January 1981, President Reagan signed an order leading to the complete deregulation of oil and gas prices.To quantify the additional cost imposed on consumers by nonprice rationing in the 1970s, Frech and Lee (1987) use data on urban and rural traffic patterns in California and estimates of the price elasticity of demand for gasoline from Lee (1980). They estimate that the time cost of the queues added between 13% and 84% to the price of a gallon of gasoline between December 1973 and March 1974; the additional time cost implicitly paid by consumers between May 1979 and July 1979 varied from 6% to 33%. Thus, the price index for gasoline shown in figure 1 potentially understates the true cost of gasoline quite severely in periods affected by the two oil shocks of the 1970s.The PPI-oil measure suffers from the same problem because of the price controls on domestic crude oil. The refiners' acquisition cost of imported oil used by Barsky and Kilian (2002) comes closer to measuring the world price of oil. However, this measure still does not capture all of the additional costs imposed on the U.S. economy by distortions caused by price controls and the entitlement system. The reason is that price controls cause inefficiencies and deadweight loss that are larger than the gap between actual prices and market-clearing prices.7In order to capture the true cost of gasoline during these episodes, we propose two new variables: the first variable augments published gas prices with estimates of the additional time cost during the periods of gasoline lines. In particular, we use the average of the rural and urban estimates from table 1 in Frech and Lee (1987), which compares the time costs per gallon to the published price per gallon of gasoline for the months of December 1973 through March 1974 and May 1979 through July 1979. All told, rationing is estimated to have added between 8% (in July 1979) and 67% (in March 1974) to the shadow price of a gallon of gasoline.8 Using these estimates, we construct a shortage-adjusted index for the real price of gasoline, which is shown as the dashed line in figure 2.Fig. 2. Real gasoline prices, January 1967 through March 2010. Dashed line represents the CPI (consumer price index) for gasoline augmented with the shadow cost of waiting time in gas lines in 1973, 1974, and 1979 as estimated by Frech and Lee (1987). The log index is calculated as in figure 1.View Large ImageDownload PowerPointBecause the rationing-by-queue cost estimates likely capture the effect of shortages imperfectly, we also consider a second measure—the special question posed by Thomson Reuters and the University of Michigan in the Survey of Consumer Sentiment. Respondents to the monthly survey are asked several questions related to car-buying conditions. The survey tracks the portion of respondents who cite the price of gasoline or possible fuel shortages as a reason that car-buying conditions are poor.9 This measure is shown in figure 3. The portion of consumers that expressed anxiety over fuel prices ramped up sharply at the time of the oil price shocks in the 1970s and early 1980s. Although the rise in real gas prices was much greater in the 2000s than in the 1970s and early 1980s, the consumer sentiment variable hit similar peaks in both periods. A CNN opinion research poll conducted in June 2008 found that consumers were more concerned about long gas lines than about high prices (CNNMoney.com, June 10, 2008). This may explain why the run-up in gas prices in the 2000s, although bigger than the run-ups in earlier episodes, did not cause a larger effect on consumer sentiment.Fig. 3. Consumer sentiment toward gasoline, January 1970 through April 2010. Share of respondents to the Thomson Reuters and University of Michigan Survey of Consumer Sentiment who cite high gasoline prices or shortages of gasoline as reasons that car-buying conditions are poor. Gasoline price question was asked on a quarterly basis before January 1978, and the series was extrapolated to a monthly frequency by the authors.View Large ImageDownload PowerPointD. The Responses of Output, Consumption, and Prices to Oil Price ShocksUsing the published measures of energy prices and the two variables that account for nonprice rationing in the 1970s, we now revisit the evidence used to suggest that aggregate activity has been responding less to energy price shocks in recent years than it used to in the past. The energy price measures we consider are as follows: (1) CPI-gas, (2) Hamilton's (2003, 2009) "net oil price increases," (3) CPI-gas that has been augmented with the time cost of rationing by queue, and (4) the measure of consumer attitudes toward gasoline prices and fuel shortages.10 Our strategy is as follows: first, we show that the impulse-response functions from VARs estimated by Edelstein and Kilian (2009) and Blanchard and Galí (2010) do not change much if published gasoline prices are used in place of the authors' original energy price measures. Second, we show that the impulse-response functions based on gasoline price measures that account for the effects of shortages present a different story.We begin by estimating a VAR that is similar to the one used by Blanchard and Galí (2010). The VAR system we estimate isIn the VAR estimated by Blanchard and Galí with quarterly data, includes the nominal price of oil, the CPI, the gross domestic product (GDP) deflator, nominal nonfarm compensation, real GDP, and nonfarm business hours. In other specifications, they also included the federal funds rate. In our version of their analysis, is built from monthly observations of the following variables (in order): (i) a selected version of one of the oil shock variables, (ii) the CPI, (iii) nominal wages of private production workers, (iv) industrial production, (v) civilian hours, and (vi) the federal funds rate. Function A(L) is a matrix of polynomials in the lag operator L, and U is a vector of disturbances. All variables except the sentiment measure and the federal funds rate are in logs. The shock to oil prices is identified using a standard Cholesky decomposition. We include a linear time trend and six lags of the variables. The data are monthly and span 1967:1–2009:12.Blanchard and Galí (2010) compare samples that are split between 1983 and 1984, which is the typical split for studies of the Great Moderation. Edelstein and Kilian (2009) study samples split between 1987 and 1988. We choose a split between 1985 and 1986, as this date is between the dates used by these authors, and it also coincides with the rather dramatic change in the nature of the oil market that occurred in 1986.We summarize the results in Table 1. As an alternative to showing dozens of impulse-response functions from various permutations of oil price measures and estimation periods, Table 1 shows the peak response of key macro variables to a shock in each oil price indicator in each period.Table 1. The Peak Effects of Oil Shocks on U.S. Variables Estimation
Referência(s)