Inflation expectations quantification with fuzzy control system

Abstract

The inflation expectations of economic agents are one of the most important variables for central banks and monetary policy conduct, but such expectations are not directly observable. Consumers’ expectations are examined in qualitative surveys. The key task is to transform consumers’ survey responses into quantitative proxies of expectations. In this examination, we investigate an alternative method to quantify consumers’ inflation expectations using fuzzy rule-based systems. In order to generate rules in the learning phase, we compare two methods: the Wang-Mendel method based on space partition and the subtractive clustering method. The learning data are information about past inflation and consumers’ opinions as expressed in surveys. The system is built and tested on data from January 2002 to June 2019 for non-euro member states of the EU. The fuzzy rules-based system returns results that outperform a benchmark—the standard quantification procedure—in terms of correlation with a raw approximation of consumer expectations, which is survey balance statistics. The results are robust regarding the benchmark choice. We find the method more promising for economies with lower stability, as it gives more room for the training phase of the algorithm application.