Learning from Inflation Experiences

From Ulrike Malmendier and Stefan Nagel:

How do individuals form expectations about future inflation? We propose that personal experiences play an important role. Individuals adapt their forecasts to new data but overweight inflation realized during their lifetimes. Young individuals update their expectations more strongly in the direction of recent surprises than older individuals since recent experiences make up a larger part of their lives so far. We find support for these predictions using 57 years of microdata on inflation expectations from the Reuters/Michigan Survey of Consumers. Differences in lifetime experiences strongly predict differences in subjective inflation expectations. Learning from experience explains the substantial dis- agreement between young and old individuals in periods of high surprise inflation, such as the 1970s. It also explains household borrowing and lending behavior, including the choice of fixed versus variable-rate investments and mortgages. The loss of distant memory implied by learning from experience provides a natural microfoundation for models of perpetual learning, such as constant-gain learning models.

They conclude:

The learning-from-experience framework differs from more conventional representative- agent applications of learning in that it generates heterogeneity in inflation expectations. Nevertheless, its implications for the average level of inflation expectations are similar to those resulting from representative-agent constant-gain learning algorithms that are popular in macroeconomics (see, e.g., Orphanides and Williams (2005a); Milani (2007)). There are, however, two important differences.

First, the learning-from-experience theory points to a different and complementary reason why data in the distant past are down-weighted and learning dynamics are perpetual. While standard implementations of constant-gain learning motivate a gradual loss of memory with structural shifts and parameter drift, learning from experience implies that memory of macroeconomic history is lost as new generations emerge whose subjective beliefs are shaped by relatively recent experience.

Second, the learning-from-experience framework allows us to estimate individuals’ reaction to inflation surprises (their gain parameter) from heterogeneity between cohorts. This opens up a new dimension of data as a source of identification. Remarkably, even though we rely purely on cross-sectional variation in survey expectations in estimating the gain, the implications of our estimates for time-variation in the beliefs of the average person are quantitatively similar to those obtained in earlier work in macroeconomics that estimated the gain to fit macroeconomic time-series and aggregate survey expectations.

The expectations heterogeneity generated by learning from experience has the potential to produce interesting macroeconomic effects. For example, Piazzesi and Schneider (2012) show that the disagreement about future inflation and real interest rates between younger and older households in the late 1970s helps understand aggregate quantities of household borrowing and lending and the prices of real assets at the time. Our findings help understand the reasons for this dispersion in inflation expectations.