Thursday, July 19, 2007

The Fed still hasn’t recovered the plot: core vs. headline inflation again

Oh boy…

Let me quote from the Testimony of Chairman Ben S. Bernanke, Semiannual Monetary Policy Report to the Congress, before the Committee on Financial Services, U.S. House of Representatives July 18, 2007

Sizable increases in food and energy prices have boosted overall inflation and eroded real incomes in recent months--both unwelcome developments. As measured by changes in the price index for personal consumption expenditures (PCE inflation), inflation ran at an annual rate of 4.4 percent over the first five months of this year, a rate that, if maintained, would clearly be inconsistent with the objective of price stability (footnote in original: Despite the recent surge, total PCE inflation is 2.3 percent over the past twelve months). Because monetary policy works with a lag, however, policymakers must focus on the economic outlook. Food and energy prices tend to be quite volatile, so that, looking forward, core inflation (which excludes food and energy prices) may be a better gauge than overall inflation of underlying inflation trends. Core inflation has moderated slightly over the past few months, with core PCE inflation coming in at an annual rate of about 2 percent so far this year.”

Here we have the Chairman of the Fed reassuring us that inflation is under control because, when you strip out the items in the price index with the highest inflation rates, the inflation rate of the remainder is only slightly above target.

It is indeed true that food and energy prices tend to be quite volatile. Chairman Bernanke, however, makes at least two logical errors in going from that defensible empirical assertion to the conclusion that core inflation may be a better gauge than overall inflation (headline inflation) of underlying inflation trends. First, volatile food and energy prices don’t necessarily imply volatile food and energy price inflation. If successive values of highly volatile (logarithms of ) core price levels were negatively correlated, core inflation rates could have low volatility. No former academic economist should miss an opportunity to remind (or explain to) the great unwashed the key distinction between levels and changes, between high, rising and accelerating. It is disappointing to see someone with Chairman Bernanke’s background waste an opportunity to strike a blow for conceptual hygiene. In this case, only general harm is done – a didactic opportunity was squandered - but no specific harm: food and energy price inflation has indeed historically been more volatile than core inflation. Food and energy price inflation has also, historically, been less persistent (more ‘mean-reverting’) than core inflation.

Chairman Bernanke’s second logical error is the mental leap from the correct statement that food and energy price inflation have historically been quite volatile (and more volatile than core inflation) to the conclusion that core inflation may be a better gauge of underlying inflation trends than overall inflation. Let’s also grant the Chairman the greater persistence, historically, of core inflation compared to food and energy price inflation.

This second logical error involves two ‘sub-errors’. The first reason Chairman Bernanke’s conclusion does not logically follow from his empirical observation is that he has to slip in the assumption that the future is going to be like the past he based his empirical characterisation on. In an earlier blog on this issue (The Folly of the Fed or: Why is the Fed so Hardcore? ), I have pointed out that even the (recent) past (2002-2007) has not been like the past that Chairman Bernanke appears to base his empirical judgments on (probably something like 1987 – 2001) , and that the future is more likely to be like the recent past than the more distant past.

The key feature of the recent past has been the sustained, persistent and continuing increase in the relative price of food and energy to core goods and services, driven by globalisation and specifically by the entry into the global economy of China, India and other emerging giants as demanders of non-core goods and suppliers of core goods. There are reasons to believe that this pattern of a rising relative price of non-core goods to core goods will persist for a fair number of years, and that as a result, the expected inflation rate of food and energy will be systematically above the expected inflation rate of core goods and services.

The second sub-error is that even if the future were to be just like the past Bernanke appears to have in mind, and even if therefore food and energy inflation were to continue to be more volatile and less persistent than core inflation, it does not follow that core inflation “may be a better gauge than overall inflation of underlying inflation trends”. I interpret underlying inflation trends as the medium-term rate of headline inflation, where the medium term is the period over which current monetary policy decisions have most of their full effect on inflation- somewhere between one and three years probably in the US.

Given all this, would it be better to forecast future underlying inflation using just current and past headline inflation or just current and past core inflation? First, note that this is a silly question. It’s obvious that no rational human being would do either. You would in general use both current and past core inflation and current and past non-core inflation (or, equivalently, current and past core inflation and current and past headline inflation). It is wildly unlikely that there is no predictive content for future headline inflation in current and past food and energy inflation, even after we condition on current and past core inflation (see my earlier discussion of Granger causality tests in The Folly of the Fed or: Why is the Fed so Hardcore? In fact, you would use any information, in addition to current and past core and headline inflation, that may have a sufficiently stable statistical relationship with future medium-term headline inflation. The craving for a single number, be it core inflation or headline inflation, that can serve as a sufficient statistic for all information relevant to the prediction of future inflation trend is very unhealthy and has, not surprisingly, led to a bad case of monetary indigestion.

As is clear from Chart 2, core inflation and headline inflation have, historically, been negatively correlated. This is comforting for those who believe that, in the long run, inflation is a monetary phenomenon (i.e. is made by central banks) rather than a relative price phenomenon, although it does not prove that belief to be correct This negative correlation means that, if it is possible to predict core inflation and non-core inflation (and if expected core inflation is not always expected to equal expected non-core inflation), one should use both core and non-core inflation to predict future headline inflation. The same logic also implies that, if future core and non-core inflation are sufficiently negatively correlated, you could do a better job predicting future headline inflation using just current and past headline inflation than you could using just core inflation, even if the volatility of non-core inflation is higher than the volatility of core inflation. Of course, as pointed out earlier, you would be crazy to use just one predictor if more are readily available.

Does the Fed not believe that, in the long run, inflation is a monetary phenomenon?

The basic confusion in the Fed appears to be the view that monetary policy determines underlying or trend core inflation and that they “they” (global oil markets, the weather, the Munchkins of OZ) determine the deviations of food and energy price inflation from core inflation. The shocks driving these deviations have a high variance, and the deviations are transitory. Food and energy price inflation converges to core inflation, with the core inflation being the dog and food and energy price inflation the tail. Non-core inflation is beyond the ability of the Fed to influence and is effectively treated as exogenous to the monetary policy decision.

This is the only theory that allows me to make sense of the utterances coming out of DC, other than the indefensible theory that the Fed believes that the best predictor of tomorrow’s relative price of core and non-core goods is always today’s relative price. If this Canine Theory of Inflation is what the Fed believes, it is bad monetary economics. If one has to have dogs and tails in the monetary transmission mechanism story, the better story is the following. Monetary policy (the Fed in the USA) determines the underlying inflationary trend for headline inflation. This is the dog. Both core inflation and non-core inflation are ‘tails’ in the following sense.

“They” (globalisation, Opec, the weather, the Munchkins of OZ) determine the equilibrium changes in the relative price of non-core and core goods. The actual, real-time behaviour of the non-core and core components of the price index does not match what their behaviour would be under full nominal price flexibility for both core and non-core goods, that is, with instantaneous full adjustment to the equilibrium relative price of core and non-core goods. Instead, the prices of core goods and services are subject to nominal rigidities and tend to adjust only gradually in response to differences between their actual and equilibrium values. Non-core prices are fully flexible and therefore can and do jump around a lot more than core prices. Since core goods prices (and indeed core goods inflation rates) are sticky, non-core goods and prices (and indeed non-core goods inflation rates) will often overshoot their equilibrium values in the short run. In the absence of further relative price changes between core and non-core goods and other shocks, both core and non-core inflation rates will converge to the underlying trend headline inflation rate determined by the central bank.

While better, the second story is also not fully satisfactory. In a well-articulated model of a monetary economy, what “they” do to the equilibrium relative price of core and non-core goods and what the Fed does to determine the underlying inflationary trend for the headline price index are jointly endogenously determined. But if you want a simple story, the second one is much less dangerous than the first.

It is possible that there will be no further increases in the relative price of non-core to core goods in the next few years. In that case, predicting headline inflation is predicting core inflation (or predicting food and energy inflation). In that case, and in that case only, no harm is done by the Fed’s defective analytical and predictive framework.

It is possible, but not likely. Headline inflation has been well above any reasonable notion of price stability (say, a point target of an annual rate of PCE inflation of 1.5%, corresponding to an annual CPI inflation rate of around 2.0%), ever since the third quarter of 2002 (see Chart 1) . So for the past 5 years, the Fed has either systematically under-predicted the underlying inflationary pressures or failed to raise interest rates by enough and fast enough to keep headline inflation near the target. From the statement of Chairman Bernanke, it looks as though there is considerable persistence in this prediction error/policy implementation error.

Chart 1

Source: Bureau of Economic Analysis, Quarterly data, SA.
Chart 2


Waltraud Schelkle said...

Very good piece, Willem. The positive correlation between the level and the relative change in headline inflation (Chart 2) is quite impressive. Could that be shown for the different years (say, the data points from 2002 onwards in another coulour) so as to highlight that there is indeed a structural break in the recent past which you claim the Fed misses?
Waltraud Schelkle, European Institute, LSE.

Willem H. Buiter said...
This comment has been removed by the author.
Ivan Kitov said...

I had a different representation which illustartes the issue in more detail. You have to plot coreCPI- CPI against time. The difference has three distinct periods 1960-1980, 1980-1999, 2003 - present which are charaterized by linear dependence of the difference on time. effectively, core CPI and CPI are equivalent as targets - one gets a precise value of CPI is targets core CPI

See my posts for figures: