|
Article Excerpt
I. INTRODUCTION
Past research has documented a strong correlation between oil prices and economic activity, both in individual sectors and at the aggregate level. Davis and Haltiwanger (2001) find that oil shocks explain much of the variation in manufacturing employment from 1972 to 1988. Rotemberg and Woodford (1996) find that a 1% increase in the price of oil caused gross domestic product (GDP) to decline by 0.25%. (1) Although some authors have argued that the actual effect may be smaller (e.g., Bernanke, Gertler, and Watson 1997), it appears that oil shocks played an important role in business cycles from World War II at least until the 1980s. (2)
This result constitutes a puzzle. A simple neoclassical model of the U.S. economy would suggest that the elasticity of GDP to the price of oil should be similar to the cost share of oil, which was about 0.04 (e.g., Rotemberg and Woodford). Previous research has found a much larger elasticity.
Economists have proposed many explanations for the large effect, but there is little direct evidence. (3) Davis and Haltiwanger argue that adjustment costs amplify the effect of a price shock and reduce capital utilization. For example, the 1970s oil shocks increased the demand for small automobiles. Most manufacturers used equipment designed to produce large vehicles, and could not adjust the types of inputs they used, which caused capital utilization to decline. Bresnahan and Ramey (1993) support this argument with data from the automobile industry, but there is little evidence from other industries.
I provide an alternative explanation for the large effect of oil prices on value added and present empirical evidence from across the manufacturing sector. I use a simple model, similar to Horvath (1998, 2000), to argue that demand and supply linkages between industries can amplify an energy price shock. (4) I use data from the Census of Manufactures (CM) to show that these mechanisms significantly increase the effect of a shock on industry value added. In the remainder of the analysis, I use plant data to investigate how energy prices affect individual plants and generate the large aggregate effects, that is, I ask whether energy prices affect entry, exit, or average production per plant. I find that an energy price increase caused a large decline in average plant production and labor demand but had a small effect on entry and exit; these results are consistent with the utilization argument of Davis and Haltiwanger.
More specifically, I focus on three mechanisms by which the price of energy can affect industry-level value added. First, the direct effect is the change in value added when an energy price increase raises energy costs. The direct effect is proportional to the cost share of energy and should be small for the average industry.
Second, if an industry uses energy-intensive materials, an increase in the price of energy may cause materials prices to rise. For example, because paper is energy intensive, an energy price increase causes the price of paper to rise. Publishing uses a lot of paper, so the supply effect would cause the publishing industry to contract as well. (5) I refer to the supply effect as the change in value added for an industry due to an energy price--induced change in materials prices. For a given industry, the supply effect increases with the energy intensity of its materials. The effect would be large if many industries rely on energy-intensive inputs.
It is also possible that demand linkages amplify a price shock. If the output of an industry is used by energy-intensive industries, the demand for its product may fall as the energy-intensive industries contract. The demand effect increases with the average energy intensity of the other industries that use the industry's product. The demand effect would be large if energy-intensive industries use materials produced by other industries.
This study estimates the importance of demand and supply linkages both at the industry and at the plant levels. I first document the strength of the direct, supply, and demand effects in explaining variation in industry value added. Energy costs, materials costs, and demand are potentially endogenous, so I identify the three effects using cross industry variation in input requirements and time series variation in the price of oil. For example, there is considerable variation in the energy intensity of industries' materials, which allows me to estimate the magnitude of the supply effect.
Using CM data from 1963 to 1982, I find that a 1% increase in the price of oil caused value added to decline by 0.07% for the average industry. This elasticity is about four times as large as the energy cost share in manufacturing, which is consistent with previous research. The supply effect accounts for about one half of the response of value added to a price shock. In other words, when the price of energy rises, industries that use energy-intensive inputs experience a large decrease in value added. The demand effect is much smaller, which is consistent with supply and demand relationships across industries; energy-intensive industries are important input suppliers to other industries but use few manufacturing materials themselves.
There is very little evidence about how energy prices affect individual plants, in particular, whether energy prices affect entry, exit, or production by plants that remain in operation. In the second part of the analysis, I use plant-level data from the CM to investigate these possibilities. It appears that oil shocks mainly affected continuing plants. The direct and supply effects caused similar changes in value added per plant as in value added per industry. By comparison, a price increase caused a small and precisely estimated decrease in entry and had no effect on exit.
A reduction in value added should be associated with a reduction in inputs: labor, materials, or energy. I further investigate the plant-level response to an energy shock by estimating how input demands respond to the direct, supply, and demand effects. I find that the direct and supply effects caused large reductions in production worker hours and wages, implying that the labor demand curve shifted toward the origin.
This paper expands on the work of Lee and Ni (2002). They define an oil-intensive industry as an industry that uses oil directly or that uses materials with a large oil cost share. For example, the petroleum-refining industry, which uses oil directly, and the industrial chemicals industry, which uses oil-intensive materials, are both considered to be oil intensive. By estimating a vector autoregression (VAR) for roughly 14 two-digit manufacturing industries, they characterize the response over time to an oil shock. They find that a positive oil shock causes a contraction of oil-intensive industries and they find large demand effects for some industries, such as automobiles.
This paper differs in several respects. First, I characterize the cross-sectional response to an oil shock, rather than the dynamic response. Second, I quantify the importance of linkages between industries across the entire manufacturing sector. Third, I use plant data to investigate whether the aggregate effect is due to entry, exit, or average production per plant.
II. ESTIMATING THE EFFECT OF SUPPLY AND DEMAND LINKAGES
A. The Price of Energy, Linkages, and Value Added
The model in this section highlights the importance of linkages between industries. The economy is composed of a set of industries, and each industry has a number of identical plants. There is one time period.
I partition the industries into two subsets. The first subset, I, consists of industries that produce intermediate goods used by other industries. The intermediate goods industries do not use other materials themselves. The second set, J, includes industries that produce final goods. These industries demand the products from the first set of industries. Horvath (1998) uses a similar framework to demonstrate the importance of interindustry linkages in amplifying a productivity shock.
I derive the supply curve of plants in final goods industry j. The industry has a fixed number of price-taking plants; the price of output is exogenous. (6) Each plant, n, produces output, [Y.sub.jn], from one unit of capital, [E.sub.jn] units of energy, and [M.sub.ijn] units of intermediate materials from industry i. The production function is Cobb-Douglas:
(1) ln [Y.sub.jn] = [s.sup.E.sub.j]ln ([E.sub.jn]) + [[summation over (i[member of]I)] [s.sup.M.sub.ij]ln ([M.sub.ijn]),
where [s.sup.E.sub.i] and [s.sup.M.sub.ij] are parameters. I assume that all firms in the industry have the same production function and that [s.sup.E.sub.i] + [[summation].sub.i[member of]I][s.sup.M.sub.ij] < 1. I do not include labor m the production function for simplicity; the results are similar if labor is modeled as a variable input. In Equation (1), capital is fixed, but the plant can smoothly adjust energy and materials. I relax these, and the Cobb-Douglas, assumptions in the empirical analysis.
Plants take prices as given and maximize profits by purchasing energy and materials:
(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where [p.sub.j] is the output price for industry j (i.e., the price of good j), [p.sup.E] is the price of energy, and R is the cost of capital. The first-order conditions for input demands are the standard Cobb-Douglas equations, and the parameters [s.sup.E.sub.j] and [s.sup.M.sub.ij] are the cost shares of energy and materials. Using the first-order conditions and Equation (1), I obtain the following supply curve for each plant:
(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],
where [r.sub.j] = 1 - [s.sup.E.sub.j] - [[summation].sub.i[member of]I] [s.sup.M.sub.ij]; [P.sup.E] is the real price of energy for industry j, [p.sup.E]/[p.sub.j]; [P.sub.i] is the real price of material i, [p.sub.i]/[p.sub.j]; and [P.sub.i] is a constant that depends on parameters. Each plant's output is decreasing in real input prices. The elasticity of output with respect to the real price of energy is proportional to the cost share of energy, [s.sup.E.sub.j]. The elasticity of output with respect to the real price of material i is proportional to the cost share of material i, [s.sup.M.sub.ij].
Free entry in the intermediate goods industries determines the price of good i. Because these industries do not use goods produced by other industries, the price of intermediate good i is given by:
(4) ln [p.sub.i] = [s.sup.E.sub.i] ln([p.sup.E]) + [k.sub.i].
The elasticity of the price of good i to the price of energy is equal to the industry's cost share of energy, [s.sup.E.sub.i].
Combining Equations (3) and (4) gives the supply curve for final good industry j, in terms of the real price of energy. To simplify the notation, I define the supply elasticity for industry j, [S.sub.j], as [S.sub.j] = [[summation].sub.i][s.sup.M.sub.ij][s.sup.E.sub.ij]. The variable is the inner product of the materials cost share for good i and the energy intensity of the material. The supply curve is given by:
(5) 1n [Y.sub.jn] = - 1/[r.sub.j][s.sup.E.sub.j]ln ([P.sup.E]) - l/[r.sub.j][S.sub.j]ln ([P.sup.E]) + [k.sub.j],
where the constant, [k.sub.j], has absorbed additional parameters. The equation shows two effects of the price of energy on output. The first term is the direct effect. Plants in energy-intensive industries have lower output. The second term is the supply effect. Plants that use large amounts of energy-intensive materials have a large value of [S.sub.j]. These plants have higher materials costs, reducing output.
Equation (5) shows the effect of the price of energy on gross output per plant. I use Equation (5) and the first-order conditions for energy and materials to obtain an expression for value added, which excludes energy and materials:
(6) in [V.sub.jn] = [B.sub.1][s.sup.E.j]ln ([P.sup.E]) + [B.sub.2][S.sub.j]ln ([P.sup.E]) + [k.sub.j],
where [B.sub.1] and [B.sub.2] are nonlinear functions of the cost shares and are constant. This equation is similar to the output equation and shows two effects on value added: the direct effect, which is proportional to the industry's energy cost share, and the supply effect, which is proportional to the (weighted) average energy intensity of the industry's materials.
B. Estimation of the Direct, Supply, and Demand Effects
Equation (6) characterizes value added per plant and is the basis for the estimating equation. Because the number of plants is fixed in the model, industry output is a constant multiple of output per plant and...
|
