Assessing the term structure of expected inflation using treasury inflation—protected securities: near real-time measures for those who need quick estimates

Business Economics, Jan, 2003 by Albert E. DePrince, Jr.

At the same time, relative volatility of the TIPS-based forecast might reflect a fundamental difference in the expectations formation process behind the two measures of expected inflation. This may not be an unreasonable assertion. The SPF forecasts are the average of a limited number of forecasters. The TIPS-based estimates are derived from market-wide data. The informational input is thus far broader for the TIPS-based forecasts than the SPF forecasts. Another difference, of course, is that SPF forecasters may or may not be participants in Treasury securities markets.

As a point of departure, the study sought to assess the extent to which an adaptive process may be behind either or both of the inflation estimates. Turning first to the SPF ten-year inflation forecasts, a simple regression showed a close relationship between the SPF expected inflation measures and a three-year lag on observed quarterly CPI inflation. More important, the sum of the coefficients on the lagged inflation rates summed virtually to unity; and the constant term did not differ from zero, after accounting for systematic information in the residuals. These characteristics suggest that SPF inflation forecasts are adaptive. Therefore, the relatively smooth movement is not surprising, and a weighted average of past inflation measures would be about as accurate as SPF forecasts.

This characteristic may stem from two factors: the nature of forecasting models and judgment used in the production of the individual forecasts accumulated in the SPF survey. Many models rely on lags in the adjustment of economic variables to shifting equilibrium conditions; forecasts produced by these models would be related to recent trends in the forecast variable. Many forecasters rely heavily on judgment when making forecasts or adjusting forecast equations, and such judgment may be heavily influenced by recent developments.

Turning to the TIPS-based inflation measures, fewer observations are available for the TIPS-based ten-year inflation forecast, and results might be suspect. Nevertheless, the same exercise was repeated. The same regression form (but using monthly frequency) showed virtually no relation between recent inflation measures and the TIPS-based inflation forecasts. This outcome suggests that current market-based information, as embodied in both the conventional and the TIPS, dominates the expectation process, and past inflation had minimal influence.

Thus, it should not be surprising that the TIPS-based estimates do not follow as smooth a pattern as do the SPF results. Expectations embodied in the TIPS-based spreads adjust rapidly to incoming information with two distinctive characteristics: (1) they are more volatile than those portraying adaptive characteristics, and (2) they bear minimal relation to recent movements in actual inflation. Therefore, TIPS-based inflation forecasts have responded more rapidly to new information than surveys and should continue to do so in the future.

The Accuracy of TIPS-Based Inflation Forecasts