Naïvely Predicting Inflation

Keyword: 
Inflation
Topic: 
Financial Economics

Standard economic models are populated by agents forming rational expectations. A growing number of economists, yet, believes this to be too extreme a view. Some argue that individuals are not optimizers; they are adaptive and only behave “as if” they optimize (e.g., Kirman, 2006); others suggest that forecasters unconsciously “combine many insights” and that predictions result from both optimal and sub-optimal procedures (e.g., Fuster et al., 2010).  

From the empirical standpoint, evidence shows that models making standard assumptions have trouble explaining the behavior of economic variables as observed in the data. A point in case is the “twin puzzles” found in the recent US inflationary dynamic: the higher-than-expected inflation despite economic slack from 2009 to 2011 and the weakening inflation despite economic recovery starting in 2012 (Friedrich, 2016).  

A naïve expectations formation mechanism (EFM) aimed to explain how the median – hence lay – forecaster predicts the one-year-ahead inflation may help to clarify these and other episodes in US inflation (Bovi, 2019). This said, the external validity of the EFM is self-evident.    

Involved Variables. The EFM is a linear combination of long-run expectations, current inflation and uncertainty. Considering the current inflation is natural. Inserting the long-run expected inflation allows contemplating the possibility that agents’ short-run expectations are anchored to some reference level of inflation. Uncertainty enters the EFM because higher uncertainty may induce higher forecasting errors.   

Updating frequency. As a fixed-income (wage or pension) earner, the median US consumer has an aversion toward underestimation that increases with the inflation level. In the last decades, inflation dynamics witnessed tremendous changes, so that the EFM allows frequent updating.  

Parameter calibration. Despite overestimations may be costly, lay forecasters perform simple informal calibrations mainly aimed to avoid costlier underestimations. In addressing boundedly rational lay forecasters simplicity is strength. Separating out the four possible combinations of inflation and uncertainty may clarify the theorized modus operandi.   

Case A: Low inflation, Low uncertainty. In this ideal environment, the reference (trend) inflation is likely to be not smaller than the current inflation and, hence, the long run expected inflation is a suitable benchmark for agents with a propension to overestimate. Instead, agents have less need to consider current inflation and uncertainty to form their short-term expectations.

Case B: High inflation, Low uncertainty. With respect to the Case A, the median consumer places larger weights on current inflation and uncertainty. When inflation is high, in fact, the parametrization of the Case A would lead to underestimations. Since inflationary environments are particularly dangerous for fixed-income agents, then, even relatively low uncertainty might affect their predictions.     

Case C: High inflation, High uncertainty. In this dramatic stance, the occurrence of costly underestimations is more likely than in any other environment. Naïve but cautious forecasters react increasing the weight put on current inflation and uncertainty. Because of this re-balance, the reference level turns out to have the smallest weight. 

Case D: Low inflation, High uncertainty. As in the Case A, low inflation shifts the emphasis toward the reference inflation in preference to the current inflation. Despite the probability to incur in underestimations is small, large uncertainty leads wary forecasters to consider it. 

Optimality. In naively “combining insights” consumers may occasionally, albeit unintentionally, end up with predictions congruent with optimal forecasting rules. The proposed EFM is compatible with agents that forecast “as if” they optimize. 

Inflation expectations data gathered by the University of Michigan’s Survey of Consumers sustain the argued behavior and contribute to interpret several historical episodes.  

Amid exceptionally challenging periods – such as, e.g., the Great Inflation/Disinflation of the “80s – lay forecasters combine insights such that they often underestimate and their expectations are suboptimal (indeed, even professionals’ forecasts were poor in those periods).  

As expected, when inflation level and uncertainty shrink – as, e.g., in the Great Moderation (GM) – agents opt for an EFM disregarding uncertainty. Possibly because of the favorable environment, then, agents’ informally chosen weighting scheme is compatible with optimally unbiased predictions. This is congruent with the so-called random walk model (Atkeson and Ohanian, 2001): when inflation is low and stable, the Phillips curve flattens, and short-run expectations are only based on inflation levels.

When inflation is low and uncertainty is high – as, e.g., in the last two decades – lay forecasters’ instinctive tuning end up with optimally biased predictions. Optimality stems from the low inflation, biasedness from the high uncertainty.

Predictions grounded on trend inflation (higher than current inflation) and uncertainty (which was not small) may explain the 2009-2011 missing disinflation. The reduction in uncertainty coupled with the persistent reference to the long-run expected inflation may then explain the lack of reflation expected after 2011.  

Finally yet importantly, since mid-2016 US short-run expectations seem based eminently on longrun expectations. An interpretation of this “new normal” is that the current low inflation rate besides the FED’s inflation targeting strategy may have eventually convinced the median, fixed-income, consumer to strengthen the link between short-run and long run expectations. This EFM is historically rare and intriguing with respect to future developments. To the extent that the actual inflation depends on expected inflation, e.g., this kind of EFM may support the current phase of low price dynamics. 


References: 

Atkeson, A. and L.E. Ohanian (2001) “Are Phillips Curves Useful for Forecasting Inflation?” Federal Reserve Bank of Minneapolis Quarterly Review 25(1): 2–11. 

Bovi, M. (2019) “A Time Varying Expectations Formation Mechanism”, Econ Polit. https://doi.org/10.1007/s40888-019-00171-7. 

Friedrich, C. (2016) “Global inflation dynamics in the post-crisis period: What explains the puzzles?” Economics Letters, 142(2):31–34 

Fuster, A., D. Laibson, and M. Brock (2010) “Natural Expectations and Macroeconomic Fluctuations.” Journal of Economic Perspectives 24 (4): 67-84. 

Kirman, A.P. (2006), “Demand Theory and General Equilibrium: From Explanation to Introspection, a Journey Down the Wrong Road,” History of Political Economy, 38, 246-280.

Share this