1. Introduction
In recent years, methodological aspects of inflation forecasting have increasingly
attracted attention in empirical economics. One strand of the literature focuses on
the determinants of future inflation such as monetary aggregates (e.g. Carstensen
2007) or the output gap (e.g. Stock and Watson 1999). The other concentrates on
finding an adequate measure of the trend component of inflation which is expected
to be a better predictor for overall inflation. A prominent role in this second group
of papers plays core inflation, which is understood as the permanent part of infla-
tion that is not being influenced by random short term fluctuations. A quite famil-
iar concept for measuring core inflation is to exclude prices for energy and un-
processed food from the recorded basket of goods, since their prices tend to be
very volatile without any trend (Gordon 1975). Alternatively, median based indi-
cators or trimmed mean measures have been proposed (Bryan et al. 1994), as well
as smoothed versions of these indicators (Rich and Steindel 2005). Moreover,
another suggestion is to incorporate co-integration restrictions (Smith 2004).
Though having some aspects in common with median-based measures of core
inflation, our paper focuses on another aspect of price trends. We construct an
indicator capturing whether a given inflation rate is the result of similar price
increases for many items in the goods basket, or whether it results from price
hikes for a few relatively important goods (e.g. furniture or cars). Since the con-
crete way an overall increase in the price level is coming about will influence how
it is perceived by consumers and firms, we label our indicator inflation sentiment.
The same inflation rate may have different consequences for the future, depending
on the distribution of price increases of individual items. When many prices are on
the rise, inflation climate may have changed. Producers might be inclined to pass
through higher costs because everybody does it, and workers might feel inflation
to be more severe and struggle more fiercely for higher wages. As a consequence
inflation will tend to increase further in the future.
To quantify this fundamental idea of inflation sentiments one might alternatively
follow the concept developed by Brachinger (2006). He constructs his index of
perceived inflation by re-weighting the components of German CPI according to
their frequency of purchases. This approach requires information going beyond
the data regularly provided by the CPI statistics, such as on the expenditures on all
components of the CPI and the prices per unit purchased. Therefore it might be
rather difficult to construct such indices for many countries for purposes of inter-
national comparison or to provide a long time series.
We propose a more simple methodology. It makes use of the fact that inflation
measurement is standardized to some extent around the world. Most statistical
offices collect price data in sufficient detail to provide price indices for the prod-