Most discussions of monetary policy take for granted that if the Federal Reserve has an implicit inflation target, it’ll follow some variant of the Consumer Price Index.
But the CPI isn’t the only price index: there’s also the GDP deflator (the price index for final output produced in the US) and the Producer Price Index (the price index for costs faced by producers). It isn’t obvious, a priori, that we should be tracking one index or the other.
Mankiw and Reis provide an elegant framework for thinking about this question in their 2003 paper, What Measure of Inflation Should a Central Bank Target? They consider a central bank attempting to construct a “stability price index”: the price index that, if targeted, leads to the greatest possible stability in economic activity. The criteria they derive are pleasantly intuitive, and they include:
- The more responsive a sector is to the business cycle, the more weight that sector’s price should receive in the stability price index.
- The greater the magnitude of idiosyncratic shocks in a sector, the less weight that sector’s price should receive in the stability price index.
- The more ﬂexible a sector’s price, the less weight that sector’s price should receive in the stability price index.
- The more important a price is in the consumer price index, the less weight that sector’s price should receive in the stability price index.
There’s a tremendous amount of insight here. Prices are partly responsible for their signaling value: if a sector is more responsive to the business cycle, its price conveys more information about the macroeconomy and deserves a larger place in the index. By the same token, if a sector is constantly hit by shocks unrelated to macroeconomic fundamentals, it’s tough to extract a meaningful signal about where the economy is headed.
Point #3 is a classic: all else being equal, you should target less flexible prices. Aoki (2001) proves a special case of this result, showing that in an economy where one sector has sticky prices and the other has perfectly flexible prices, the optimal policy is to target sticky-price inflation. Why? From a stabilization perspective, monetary policy doesn’t matter for goods with flexible prices: if the prices are free to move around, they’ll adjust to their equilibrium values no matter what the Fed does.
But the real shocker is #4: all else being equal, prices that are more important to the consumer price index should be less important to the stability price index. In other words, the optimal index for stabilizing macroeconomic activity is in some sense the opposite of the optimal index for measuring the cost of living. Mankiw and Reis explain the intuition as follows:
What is the intuition behind this surprising result? Under inﬂation targeting, undesirable ﬂuctuations in output arise when there are shocks to equilibrium prices, which the central bank has to oﬀset with monetary policy. The eﬀect of a shock in sector k depends on its consumption weight. The greater is the consumption weight, the more the shock feeds into other prices in the economy, and the more disruptive it is. Thus, to minimize the disruptive eﬀect of a shock, a central bank should accommodate shocks to large sectors. Under inﬂation targeting, such accommodation is possible by reducing the weight of the sector in the target index. Hence, holding all the other parameters constant, sectors with a larger weight in the consumption index should receive a smaller weight in the target index.
It’s instructive to consider how energy prices fare under these criteria. They do well under #1: they’re very responsive to changes in the business cycle. Unfortunately, they’re horrible on #2: energy prices (and natural resource prices more broadly) are subject to more idiosyncratic shocks than prices in virtually any other sector. Unlike most prices in the US, they reflect the highly volatile and unpredictable shifts of a global market. And if we consider #1 and #2 jointly, thinking about the signal/noise ratio of energy prices, it’s clear that we have much better ways to gather information about the domestic economy.
Meanwhile, energy prices are the canonical example of flexible prices: the gas station is possibly the only place that a typical consumer sees price updates over the course of a day. Certainly, then, energy prices don’t merit much emphasis under criterion #3 either.
The lessons here, of course, apply to questions much broader than the inclusion of energy prices. We typically construct our price indices based on some kind of basket of goods: either a basket of consumption goods (the CPI), a basket of intermediate goods used in production (the PPI), or a basket of final goods produced (the GDP deflator). These baskets are supposed to accurately represent the relative weight of each good in consumption or production. But this goal has absolutely nothing to do with the goal of designing indices that provide effective targets for stabilization. And if we use an index that wasn’t designed for monetary policy, we shouldn’t be surprised when it doesn’t work so well.