Among the disappointments in the 2015:Q1 GDP figures was weak consumption growth, which was a little surprising given the extra cash most consumers have on hand as a result of lower energy prices. I wanted to take a look at how the recent consumer behavior compares with what we’ve seen historically.
The graph below plots the price of energy goods and services relative to the overall price consumers pay for other purchases. Real energy prices have fallen about 20% from where they had been last summer.
Figure 1. Ratio of implicit price deflator for energy goods and services to overall PCE deflator, monthly 1959:M1 to 2015:M3.
Many consumers buy the same number of gallons of gasoline each week regardless of whether the price goes up or down. Such behavior would mean that someone who used to spend 5% of their budget on energy would only need to spend about 4% if energy prices fell 20%. And indeed we see in the data that purchases of energy goods and services now account for only 4.4% of total consumer spending, down from 5.6% a year ago.
Figure 2. Consumer purchases of energy goods and services as a percentage of total consumption spending, monthly 1959:M1 to 2015:M3.
In 2007, Paul Edelstein and Lutz Kilian studied the relation between energy prices and consumer spending using simple forecasting relations known as a vector autoregression. They summarized the effects on consumer budgets of energy prices by multiplying the monthly percent change in the relative price plotted in Figure 1 by the energy share in Figure 2– call this variable xt. Thus for example if the energy share is 5% and the price increases 20%, xt would equal 1, corresponding to the 1% loss of spending power described above. They then related xt to the monthly growth rate of real consumption spending. They used data from 1970:M7 to 2006:M7 to estimate a pair of forecasting equations to predict xt and real consumption growth based on observed values of the two variables over the previous 6 months. The black line in the graph below shows how an n-month ahead forecast of real consumption in that system would change when xt goes up by 1, a graph that economists describe as an “impulse-response function”. The green lines indicate 95% confidence intervals. Historically, an energy price increase that reduces consumer spending power by 1% would on average be followed within a few months by a 1% decline in real consumption spending and by closer to a 2% decline by the end of 6 months. One interpretation of why the latter effect is bigger than 1% is that it could reflect second-round macroeconomic multiplier effects of the lower consumer spending.
