University of Michigan Professor Lutz Kilian, whose research we’ve often highlighted here (, ), and Michigan Ph.D. candidate Paul Edelstein have an interesting new paper on how energy price changes affect the economy.
Edelstein and Kilian explored one implication of the popular idea that energy expenditures are relatively price-inelastic. If you go on purchasing the same amount of energy even though the price goes up, you have to cut back on other discretionary spending. The rough size of this effect would be given by the dollar magnitude of the loss in purchasing power. The authors sought to quantify this effect by calculating exactly how much consumers would have to decrease (or be able to increase) discretionary spending each month given how much energy prices went up (or down) that month, if they wanted to keep the quantity of energy used constant. In a typical month, energy prices might change by 1.5%, which would amount to a little less than 0.1% of total consumption spending. The biggest monthly shock in their sample (1970-2006) was associated with the aftermath of Hurricane Katrina, which reduced consumer purchasing power by -0.66%.
The authors then looked at the correlations between these measures of potential changes in discretionary spending and subsequent changes in various categories of actual consumption spending. Most previous studies had focused on output measures like GDP and had found asymmetric responses– energy price increases were often followed by slower GDP growth, whereas energy price decreases seemed to have little positive benefit. By contrast, Edelstein and Kilian found energy price declines are followed by higher consumption spending, just as price increases lead to consumption cutbacks.
However, like previous studies, Edelstein and Kilian found that the overall size of these effects, either up or down, is much bigger than can be explained by the discretionary spending story alone. The diagram below shows the average historical changes following an energy price increase that reduces purchasing power by about 0.1%. The horizontal axis measures the time (in months) after the initial energy price change, and the vertical axis measures the percentage change in the indicated category of real consumption spending. Ninety percent confidence intervals are indicated by dashed lines. The upper left panel indicates that even though potential spending has only decreased by 0.1%, consumption spending on average actually dropped by 0.23%.
The other panels give a pretty clear signal of where the biggest effects are coming from– motor vehicle purchases change sharply and dramatically. Domestic car sales (which tend to be concentrated in the less fuel efficient models) were hit harder than foreign,
and light trucks (which includes SUVs) much harder than autos:
Consumers seem to be responding in a very direct and tangible way to the gasoline prices themselves.
Another strong effect that the authors found was on consumer sentiment. Even though a 1.5% increase in energy prices typically meant only a 0.1% drop in purchasing power, on average it was followed by a 1.6% drop in the University of Michigan index of consumer sentiment:
All of this could be reconciled with rational behavior if one believed that car sales are important for aggregate economic activity. Because car sales indeed fall when gas prices go up, consumers rationally have increased worries about future job security when gas prices rise, and these insecurities are an additional factor in and of themselves inducing them to cut back on spending by even more than the direct impact on their wallets would require. Here’s what Edelstein and Kilian concluded:
Our results suggest that Hamilton (1988) was right that expenditures on consumer durables that
are complementary in use to energy (such as cars) are sensitive to even small energy price fluctuations.
We showed that indeed there is a strong decline in the real consumption of motor vehicles in
response to unanticipated purchasing power losses. This decline accounts for much of the anomalous
response of consumer durables and generates increased aggregate unemployment. However, there
is no evidence in the real consumption or consumer expectations data that changes in the demand
for vehicles cause a sectoral reallocation effect that amplifies the effect of energy price increases and
cushions the effect of energy price decreases, as postulated in Hamilton’s model.
That interpretation also led Edelstein and Kilian to draw some conclusions about why the economy may be less sensitive today to energy price increases than it had been historically. First, they noted that domestic automakers today have a better mix of large and small cars, so that some segments can continue to do well regardless of which way gasoline prices lurch. Second, with the increasing market share of imported cars and decreasing importance of manufacturing overall, the domestic auto sector matters less for the overall economy than it used to.
Nonetheless, as Econbrowser readers are well aware, that doesn’t stop me from worrying.