David Greenlaw | New York
There have certainly been bigger market movers in recent years, but Friday’s CPI report was one of the most shocking data releases in memory. The reason -- unlike employment numbers or retail sales data -- the CPI figures tend to exhibit very little month-to-month volatility. In fact, a forecast miss of 0.2 percentage points on the core CPI is about a two standard deviation event. To put it another way, over the past 10 years, the core CPI outcome has been 0.2 percentage points higher or lower than the consensus on only 8 occasions (or 6.7% of the time). Most importantly, when such a surprise does occur, it is almost always traceable to a big move in a single volatile component -- such as tobacco or hotel rates. In this case, there was no such sole special factor responsible. And, to top it all off, the big downside surprise in November followed on the heels of a notable -- although not quite as large -- downside surprise in October.
Friday’s report was particularly shocking from another standpoint. Mathematically, it is virtually impossible to get a 0.0% result for the core when the shelter category, which accounts for 41.7% of the core, is up 0.4%. Yet that is exactly what happened in November. As seen in the accompanying figure, the core CPI excluding shelter was -0.2% in November -- the lowest reading in the 40-year history of the data.
From our standpoint there are three possible explanations for the sharp deceleration seen in the core CPI over the past two months.
1) The data are correct and should be taken at face value. Core inflation experienced a significant run-up in the first nine months of the year (rising from around +2.0% to a +2.9% yr/yr rate in September) and we are now simply seeing a rapid unwind, reflecting the pullback in energy prices and a weaker economy. Of course, the problem with this story is that the transmission from energy prices to consumer prices is hardly instantaneous. It takes at least a few months -- if not a few quarters -- for this chain of events to play out. Moreover, while economic growth has slowed, labor markets remain very tight and cost pressures -- even after taking into account the latest revisions -- continue to edge gradually higher. We assign about a 20% probability to this scenario.
2) The October and November data reflect statistical quirks that will be unwound in relatively short order. While there was no single special factor responsible for the much lower than expected core CPI results over the past couple of months, some of the categories that played important roles simply do not seem to square with reality. Two obvious such items are motor vehicles and air fares. Automakers have pared production dramatically over the course of 2006 so that they could discount less -- not more. Indeed, vehicle inventories at the end of November were at their lowest level for that particular month in the past five years -- hardly a recipe for a stepped-up pace of price cuts. Meanwhile, airline industry load factors remain quite elevated and industry pricing data simply do not support the notion that there have been sizeable fare reductions of late. It’s certainly conceivable that we will see a sharp rebound in vehicle prices and air fares along with a flattening out of apparel prices and a continued escalation in OER over the course of coming months. This could put us right back at a +2.9% yr/yr rate by February. We assign about a 35% probability to this scenario.
3) Finally, it’s quite possible that the October and November data merely reflect an unwind of some quirks which had temporarily elevated the core CPI readings in the first three quarters of the year. In other words, both the prior up moves and the down moves of late have merely reflected statistical noise. Core inflation has actually been holding fairly steady all along. One possible culprit in this scenario is inadequate seasonal adjustment. Interestingly, in both 2004 and 2005, the core CPI experienced a run-up in the early part of the year followed by significant deceleration later on. While this seems to us to be the most likely scenario -- we assign it a 45% probability -- there are still plenty of unanswered question. Specifically, while a seasonal bias may be evident in the data over the past few years, the swings in both 2004 and 2005 are almost entirely attributable to big moves in a single volatile category -- hotel rates. And, there does not appear to be any sign of such a seasonal bias in the core CPI for the 10 years or so prior to 2004.
In the end, only time will help tell us which one of these scenarios best explains the swings in the core CPI over the course of 2006. In the meantime, it seems reasonable to assume that the inflation picture is not as scary as previously feared. However, with labor markets still tight, with productivity showing signs of some modest cyclical slowing, and with energy prices remaining quite elevated relative to a few years ago, it would be wrong to assume that inflation risk has disappeared entirely.