Core blimey, Governor
This post recaps on today's tweetstorm about core inflation, which follows Paul Krugman's blog on the topic. It's late on Friday, productivity levels are falling, and I can defend not doing my bit for my co-authors on my revise and resubmit.
1. In sticky price models, theory tells us that the optimal inflation rate is one weighted by the stickiness of prices. So if it could be proven that food and energy are more flexible than other prices, that would legitimate the Fed dropping these prices from the core rate.
2. By the same token, equal prominence should be given to nominal wages in formulating policy. If prices were entirely flexible, and nominal wages sticky, the central bank should be given a nominal wage inflation, not a price inflation target.
3. Another defensible argument is that some shocks to inflation may come and go more quickly than monetary policy can respond, given, ahem, the long and variable lags between interest rate changes and changes in inflation. Looking through such shocks may often wind up looking like excluding some prices from the headline index.
4. It's also defensible to formulate a core index on grounds of measurement error. There are two kinds we can envisage. One is that our statistics agency has the correct concept of price, but simply measures it with random noise induced by sampling. It has long been known [see for example work by Svensson and Woodford] that optimal policy indicators should weight component variables by how well they are measured. A similar argument was made by Bryan and Cecchetti in the early 1990s when they argued for the 'trimmed mean' inflation rate. A second kind of measurement error is that the statistics agency simply has a concept of price that is not appropriate for monetary policy purposes. (This doesn't mean that the price is necessarily 'wrong', since it may do perfectly well for achieving some other policy goal.) Responding to this kind of measurement error would hopefully amount to surgically removing the offending component index, and replacing with one more in conformity with what theory or common sense tells you should go in. And this process might resemble constructing a core inflation index. A classic example of this genre is that the theory consistent measure looks for an aggregate of nondurable goods. Yet CPI weights the prices of many goods which provide services over a long period. Ideally we would figure out the service flow and substitute this in place of the price of the computer, or car, or whatever. [Although my Windows 8 computers struggle to provide a service flow over multiple periods].
5. These arguments have to be weighed against the costs of the central bank, or authority which defines the mandate, appearing to choose and change the goalposts to suit itself. If there is a strong feeling that something is inflation in the populace, it may be costly for the authorities to entirely disregard controlling that, since it may undermine the credibility of their actions in this and other spheres of public policy, and may unhinge expectations of the true measure of inflation, aggravating the problem of controlling it. Lorcan Roche Kelly responded to my tweetstorm by pointing to 'CPI flat', a joke index of the prices which, when aggregated, amount to the same number each month. This is not such a joke. It reminds me of a long period in the early 2000s when the ECB were continually pointing to above target inflation as being due to 'price level shocks', as though these were mysterious things that had nothing to do with the ECBs own actions. 'Price level shock' madness haunted the corridors of the BoE too, in the early days of post-crisis high inflation, before the MPC made more conscious and honest reference to an explicit choice to tolerate higher inflation on account of moderating the recession.