However, if the latest consumer expectations for inflation are correct, inflation would revive from its slumber over the next 12 months. The latest Reuters/University of Michigan survey predicted that consumer prices will rise 3.0% over the next year.
This latest prediction offers a good opportunity to take a closer look at inflation predictions. Based on the 12-month inflation predictions made in the University of Michigan survey cited above from May 1978 through May 2008, one finds:
• A median predicted inflation rate was 3.1% vs. the median inflation rate that prevailed of 3.2%
• The predictions were tightly clustered in the 2.5%-4.9% range. Nearly 81% of predictions fell in that range. Inflation during the prediction period fell in that range just under 60% of the time. The predictions typically avoided the long-tails (changes in consumer prices of less than 0% and those of 10% or more). Just 1.4% of predictions fell in that area. Actual inflation outcomes fell in those areas 9.6% of the time.
• There was a mean forecast error of 1.2% with a standard deviation of that error of 1.2%. In other words, the magnitude of the forecast error was quite volatile. The smallest mean forecasting errors occured in the 2.5%-4.9% forecast range. Then, the mean error was 0.9%. That conditon is not too surprising given that inflation fell in that range about 60% of the time, while forecasts called for such inflation around 80% of the time. However, when the inflation approached or reached the tails, the mean forecast error rose. Given the present macroeconomic environment, it should be noted that when consumer prices fell, the mean forecast error was 4.7% and when inflation was in the 0%-2.4% range, the average forecast error was 1.2%. Increased forecasting errors also occured during times of elevated inflation.
• In general, the higher the predicted inflation rate, the greater the error. The exception concerned predictions of 10% or greater inflation. A plausible reason for that outcome is that consumers were very reluctant to forecast 10% or greater inflation until such inflation had become established. Consumers forecast 10% or greater inflation just 1.4% of the time. Consumers were also reluctant to expect a continuation of such inflation. 10% or greater inflation occurred 8.2% of the time. The 12-month inflation rate reached 10% in March 1979. Consumers made their first forecast for the 12-month inflation rate to average 10% or above in November 1979. Their last such forecast was made in April 1980. The last double-digit 12-month inflation figure was recorded 6 months after consumers last expected it.
What All This Means:
Comparisons of 12-month inflation predictions and 12-month inflation outcomes provide some general lessons concerning human forecasting. Those lessons include: (1) Forecasts focus in a tight range, namely ranges that are familiar to those making such predictions. That is why tail risk-type events such as major crises are typically foreseen by only a few; (2) During times abnormally high inflation or abnormally low inflation/deflation, forecasting errors are large; and, (3) People view economic phenomena as more stable than such phenomena actually are.
Those general lessons also have specific ramifications for the 3.0% inflation forecast. Adding the prevailing economic dynamics (household deleveraging, an rise in saving as a share of disposable personal income, a weak economy, tighter credit markets, and the loss of the earlier “wealth effect” from high stock and real estate prices that helped encourage consumption) to the general lessons from the inflation prediction experience, odds favor an inflation rate of less than 3% for the 12-month period ending in July 2010. Given the average error associated with various inflation outcomes, it is probably more likely that inflation will amount to 1%-2% over the next 12 months rather than 3%, barring some significant geopolitical shock that disrupts the energy markets for an extended period of time and/or much more robust economic growth than expected develops.
&&
No comments:
Post a Comment