This graph of inflation versus expected inflation and the WSJ article accompanying it have set off a discussion over what it means for monetary policy now that inflation expectations are so much more stable than in the past. For example, the existence and source of this stability is one part of the debate I am having with James Galbraith over monetary policy and the value of inflation targeting. But are expectations more stable? (the picture is from Brad DeLong's sidebar):
This comparison is between one year inflation rates, the noisier of the two lines, and the average expected inflation rate over the next five years, very different quantities (averages ought to be more stable - more on this below).
Not too long ago, the Cleveland Fed took a look at inflation expectations. The authors of the article, Michael Bryan and Linsey Molloy, say:
Keeping inflation expectations “contained” seems to be an important preoccupation for central bankers. This is because the expectation of higher inflation induces changes in economic behavior that impose costs on the economy... For example, when people anticipate an increase in inflation—and the corresponding decline in the purchasing power of their money—they are more likely to invest their wealth in real assets, such as land or commodities. This reallocation is a less efficient use of resources than what may have occured if people didn’t have to seek this inflation-protected form of saving. Rising inflation expectations may also help to perpetuate an otherwise temporary rise in prices, making the job of maintaining price stability more difficult to achieve.
To gauge inflation expectations, economists turn to a number of sources, including surveys of consumers, financial market data, and economists' predictions. A recent look at each of these sources shows inflation expectations are running anywhere from 2 to 3-1/2 percent, depending on the source and the period over which inflation expectations are projected (1 to 10 years).
Here are a few graphs from their
article. First, here are data from the University of Michigan’s Survey of
Consumers as in the first graph above:
*Mean expected change as measured by the
University of Michigan’s Survey of Consumers.
Source: University of Michigan.
The one-year ahead expected inflation series shows quite a bit more variation than the five to ten year ahead series. In discussing this graph the authors note:
Consumers’ year-ahead inflation expectations were elevated in the wake of Hurricane Katrina and geopolitical issues that drove up the price of oil through last summer but have since tumbled downward.
Thus, as they indicate, one year ahead expectations are sensitive to new information.
The second graph shows more variation in five to ten year ahead expected inflation than is shown in the first graph from the WSJ partly because it starts from two years earlier, in 1995, when expectations were higher. Starting the graph in 1997 gives a more stable picture.
The graphs above are based upon survey evidence and there are questions as to the accuracy and reliability of expectations derived from survey data. What do financial markets say? Here are the inflation expectations derived from TIPS (inflation protected government bonds), also from the Cleveland Fed article (these should also be interpreted cautiously):
*Derived from the yield spread between the 10-year
Treasury note and Treasury inflation-protected securities.
a. Ten-year TIPS-derived expected inflation, adjusted for the liquidity premium on the market for the 10-year Treasury note. Sources: Federal Reserve Bank of Cleveland; and Bloomberg Financial Information Services.
*Blue Chip panel of economists.
Source: Blue Chip Economic Indicators,
January 10, 2007.
Given the degree of volatility in, say, the third graph showing forecasts of inflation from the TIPS market for inflation averages over long time periods, I am not seeing the same degree of stability in inflation expectations that others are seeing (the corrected series, which is slightly less volatile that the uncorrected series, is the best one to use -- but with close to a 2% swing within a 0% to 3.5% range it is still shows a lot of variation). Thus, I would be very reluctant, given this evidence, to test whether a sustained increase in inflation to bring down unemployment would in fact cause a corresponding increase in expected inflation. That's particularly true if expectations ratchet upward easier than they fall.
There's one more reason to be careful with this evidence. The expectations are for the average inflation rate over a five to ten year period, so that is what they should be compared to. To round out the picture, then, here is average expected inflation over the next ten years from TIPS data along with actual inflation over one and five year time periods (annualized in the latter case). Notice that the estimate of average inflation (expected average inflation) is more stable than actual one year inflation rates, but is quite a bit less stable than inflation over longer time periods, e.g. the five year time period shown in the diagram (ten year inflation rates are also very stable). From this perspective, expectations of average inflation may overreact to new information as reflected in the year-ahead inflation rate:
Click for larger version
[Note: the averages are over the last one to five years while the forecasts are for inflation rates in the future, so the series do not strictly align, though even so the red line (ten year average expected inflation) appears to be a fairly good estimate of the dotted blue line (actual inflation averaged over five years) conditioned on one year rates. If actual data is aligned with the forecasts, the averages over the next ten years would end in 1997 since the average would extend to 2007 (it would end in 2002 for a five year ahead average) and the actual and expected inflation series would not even overlap. So I chose to present the data in this way to illustrate the point that forecasts of average inflation are actually noisier than average inflation itself, and I chose five year averages for actual inflation so the series could also be compared, roughly, to the expected inflation series shown in the first graph which is for a five to ten year time period.]
I'd be happy to say let's not worry about inflation anymore, let's do whatever is necessary to maintain full employment without regard to the inflationary impact and longer run consequences of the policy, or at least to worry less than we used to worry about the long-run inflationary consequences of stabilization policy. The graphs do indicate that the variability of inflation expectations has diminished somewhat in recent years, but not enough in my opinion to conclude that there's been any fundamental shift in the relationship over and above what would be expected with growing confidence in the Fed's commitment to anti-inflation policies. But as I said, I'd be happy to follow the recommendations of others and worry less about inflation and, with any luck I've overlooked something obvious. So, please, convince me that I'm wrong.
Update: There is quite a bit more discussion as James Galbraith responds and I reply, and then he responds and I reply again, in comments.