The Economic Effects of Energy Price Shocks

The paper states that “it is widely accepted that energy prices in general and crude oil prices in particular have been endogenous with respect to U.S. macroeconomic conditions dating back to the early 1970s,” because global supply and demand for energy both depend on global macroeconomic aggregates such as global real activity and interest rates. This matters because a correlation between energy prices and macroeconomic outcomes “does not necessarily imply causation,” which motivates the search for measures or models that isolate the exogenous component of energy price changes.

The paper argues that the instruments used to validate the net-oil-price-increase measure’s exogeneity are weak, reporting that estimated gmin statistics “are far below the least conservative critical value of about 4, suggesting that a weak-instrument problem cannot be ruled out,” so that “we have no credible evidence that the responses to net oil price increases correspond to those of truly causal models.” The paper adds that measures of exogenous political oil-supply shocks “fail to explain much of the observed fluctuations in crude oil prices” – at most one fourth of the 1973/74 increase – and that applying the same nonlinear transformation to other industrial commodity prices “generates spurious evidence of exogenous price shocks,” so “oil price series must be treated as endogenous whether they have been transformed to net oil price increases or not.”

The paper substitutes a weaker, “more defensible assumption than strict exogeneity”: that energy price innovations are predetermined with respect to U.S. macroeconomic aggregates, meaning energy prices respond to domestic conditions “only with a delay,” which licenses recursively identified VARs with energy prices ordered first. The paper notes this assumption is “typically inappropriate when working with annual data, but may provide a good approximation when working with quarterly and in particular with monthly data,” which is why the paper’s own empirical work, and much of the literature it reviews, uses monthly VAR models.

The paper states that “oil price shocks by definition cannot be interpreted as productivity shocks for real GDP,” because under standard assumptions imported oil enters the production function of domestic gross output but not of domestic value added, so that, holding capital and labor fixed, “oil price shocks do not move value added”; moreover, to the extent oil does affect output through a cost channel, the effect “should be bounded by the cost share of oil in gross domestic production, which is known to be very small.” The paper reviews three modifications proposed to rescue a supply-side channel – time-varying markups (Rotemberg and Woodford 1996), capital-energy complementarities in a putty-clay model (Atkeson and Kehoe 1999), and reduced capacity utilization (Finn 2000) – but states there is “no consensus which, if any, of these models has empirical support.”

The paper argues that “the primary channel of transmission is on the demand side of the economy,” operating through four mechanisms: a discretionary-income effect bounded by the energy expenditure share, an uncertainty effect that delays purchases of durables, a precautionary-saving effect tied to fears of future unemployment, and an operating-cost effect that reduces demand specifically for energy-using durables such as automobiles. The paper further argues that an “indirect” reallocation effect – the shifting of capital and labor away from energy-intensive sectors such as automobiles, amplified by labor- and capital-market frictions – “could be much larger than the direct effects listed earlier, and is considered by many economists to be the primary channel.”

The paper reports an estimated elasticity of total energy consumption to energy prices of -0.45 (95-percent bounds of -0.27 to -0.66), with gasoline demand reaching an elasticity of -0.48 after one year and heating oil and coal showing a much larger response of -1.47, which the paper attributes to households’ ability to store heating oil. For non-energy consumption, the overall elasticity is -0.15, “driven mainly by a reduction in vehicle purchases,” which has an estimated elasticity of -0.84; the paper computes that for a typical household spending $4,000 per month, a 25-cent-per-gallon unanticipated gasoline price increase cuts expenditure by about $41 one year later based on full-sample estimates, illustrating that “it takes repeated surprise increases in gasoline prices to generate large effects on household consumption.”

The paper reports “no compelling statistical evidence against the symmetry hypothesis” for consumption or investment responses, noting that a symmetric model implies real consumption fell 1.92 percent in 1979 and rose 2.02 percent in 1986 in response to purchasing-power changes – “nearly symmetric” and broadly consistent with the actual growth rates of -2.20 percent and +1.44 percent observed in those years. The paper attributes the genuinely asymmetric behavior of real GDP growth in 1979 versus 1986 not to an intrinsic nonlinearity in the response to energy prices, but to a confounding event – the 1986 Tax Reform Act’s repeal of the investment tax credit, which the paper argues was “mainly responsible for the low rate of real GDP growth in 1986” via its effect on nonresidential fixed investment, separate from the effect of falling oil prices that year.

The paper documents that the response of total real consumption to an unanticipated 1-percent energy price increase fell from -0.30 percent (after 18 months) in 1970.2-1987.12 to -0.08 percent in 1988.1-2006.7, with comparable declines for durables, vehicles, residential investment and unemployment, and attributes this mainly to structural change in the U.S. automobile industry. The paper rules out a declining energy expenditure share as the explanation, since the estimates already control for purchasing-power changes, and rules out declining shock variability, since purchasing-power innovation variability “increased” across the two subsamples; instead, it argues that “by the late 1980s and 1990s the differences between domestic and foreign auto producers had been greatly reduced,” so oil shocks no longer disproportionately depress demand for (formerly fuel-inefficient) domestic cars the way they did in the 1970s.

The paper summarizes a structural VAR of the global crude oil market (world oil production, global real activity, and the real price of oil) identified under the assumptions that oil production “does not respond within the month to demand shocks in the crude oil market” and that oil-market-specific demand shocks do not affect global real activity within the month, decomposing oil price changes into oil supply shocks, global aggregate demand shocks, and oil-market-specific (“precautionary”) demand shocks. This distinction matters because the three shock types “have different implications for the timing, magnitude and persistence of the path of oil prices,” and because demand shocks, unlike supply shocks, carry an additional direct stimulative effect on oil-importing economies through higher demand for their exports – so the net effect of a given oil price increase on U.S. output and stock prices depends on which shock produced it.

The paper states “there is no evidence that this build-up has been driven by production decisions by OPEC or other crude oil supply shocks,” and that “the bulk of the increase in oil prices since 2002 has been associated with increasing global demand for crude oil, along with other industrial commodities,” reflecting demand from emerging economies in Asia. The paper argues this composition helps explain why the price run-up “did not induce a sharp recession or stock market correction, as traditional models of oil price shocks would have suggested,” because the direct stimulative effect of strong foreign demand on U.S. exports initially offset the drag from higher oil prices, with the negative effect dominating only with a delay.

The paper reports that an oil supply disruption produces a temporary, statistically significant decline in the level of real GDP lasting about seven quarters with little effect on the CPI, while an oil-market-specific demand shock produces a larger, more persistent decline in real GDP (peaking after about three years) together with “a large and even more statistically significant increase in the price level,” and an aggregate demand shock produces an initial (statistically insignificant) rise in GDP followed by a delayed, significant decline as the drag from higher commodity prices eventually dominates the initial direct stimulus. For stock returns, the paper reports that “on average, 22 percent of the variation in aggregate stock returns can be attributed to the shocks that drive the crude oil market,” but that the conventional view that higher oil prices necessarily depress stock prices “is seen to apply only to oil-market specific demand shocks,” since aggregate-demand-driven oil price increases are instead associated with higher stock prices in the first year.

The paper is explicit that it is “not a comprehensive survey,” and flags specific open questions, including that the choice between competing nonlinear specifications of energy price shocks “remains an active area of research,” that the identifying assumptions of the predetermined-energy-price VAR approach are “not testable within the VAR framework,” and that the relative importance of endogenous monetary-policy responses to oil shocks “depends to a large extent on the definition of the counterfactual and on the modeling assumptions.” The paper also stresses that expectations of future oil-supply shortfalls driving precautionary demand “need not be realized in the observed sample period,” drawing an explicit analogy to the peso problem in exchange-rate economics, which the paper offers as a caution against reading its own demand/supply decomposition as a complete, mechanical account of any single historical episode.