What do Thirteen Million Price Records have to Say about Consumer Price Rigidity?*
| Date | 01 April 2007 |
| Published date | 01 April 2007 |
| DOI | http://doi.org/10.1111/j.1468-0084.2007.00473.x |
| Author | Patrick Sevestre,Laurent Baudry,Sylvie Tarrieu,Hervé Le Bihan |
139
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford, 2007. Published by Blackwell Publishing Ltd,
9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 69, 2 (2007) 0305-9049
doi: 10.1111/j.1468-0084.2007.00473.x
What do Thirteen Million Price Records have to
Say about Consumer Price Rigidity?*
Laurent Baudry†, Herv´
e Le Bihan†, Patrick Sevestre†‡
and Sylvie Tarrieu†
†Direction de la Recherche, Banque de France, Paris, France
(e-mail: laurent.baudry@banque-france.fr; herve.lebihan@banque-france.fr;
sylvie.tarrieu@banque-france.fr)
‡UniversitéParis I-Panthéon Sorbonne, Paris School of Economics, Paris, France
(e-mail: sevestre@univ-paris1.fr)
Abstract
Based on the analysis of 13 million price records underlying the computation of
the French consumer price index, we provide a detailed assessment of consumer
price rigidity. Our main results are as follows. The average duration of prices is
around 8 months. Price durations and the patterns of price-setting strongly differ
across sectors. Price cuts are almost as frequent as increases, suggesting no speci-
fic downward nominal rigidity. Price changes typically have large absolute sizes.
Time variation in the frequency of price changes and in their size both contribute
to inflation fluctuations. Overall there is evidence of both time- and state-dependent
price-setting.
*This study has been developed in the context of the Eurosystem Inflation Persistence Network. It was
first circulated with the title ‘Price Rigidity: Evidence from the French CPI Micro-data’. We are grateful to
the INSEE (Direction des Statistiques D´emographiques et Sociales) for providing access to the data, and in
particular to Dominique Gu´ed`es for discussion and advice. We are grateful to Jonathan Temple (the editor)
and two anonymous referees for helpful comments and suggestions. We also thank an anonymous referee
of the ECB Working Paper series, as well as many colleagues and participants at the NBER Summer Insti-
tute 2004, the 2004 AFSE Conference in Paris, and seminars at the Banque de France, the European Central
Bank, the INSEE and the French Ministry of the Economy, Finance and Industry. In particular we are grateful
to L. Alvarez, I. Angeloni, E. Dhyne, D. Foug`ere, J. Gal´ı, I. Hernando, A. Kashyap, H. Kempf, A. Levin,
V. Mignon, P. Neves, D. Romer, P. Vermeulen for helpful remarks and discussions. Any remaining errors
are our own. The views expressed in this paper do not necessarily reflect the opinion of the Banque de
France.
JEL Classification numbers: E31, D43, L11.
140 Bulletin
I. Introduction
Price stickiness is a major issue when assessing the potential impact of various kinds
of shocks on the economy. Indeed, the response of output, inflation and employment
to a shock to, e.g. interest rates or energy prices, is highly dependent on the flexi-
bility of prices (and wages). However, while a number of microeconomic theo-
retical models of price stickiness have been developed (e.g. Taylor,1980; Rotemberg,
1982; Calvo, 1983) and used in empirical and theoretical macroeconomic models,
their empirical assessment at the microeconomic level has remained relatively lim-
ited, as can be inferred from the surveys by Weiss (1993) or Wolman (2000). This
lack of microeconomic empirical evidence reflects the scarcity of the available sta-
tistical information on prices at the microeconomic level. Indeed, until very recently,
most existing micro-studies focused on specific products or markets (e.g. the seminal
contributions by Cecchetti, 1986, on magazine prices; by Lach and Tsiddon, 1992,
on food product prices; by Kashyap, 1995, on goods sold through catalogues; or the
more recent one by Genesove, 2003, on apartment rents). The purpose of the present
paper is to contribute to the sparse although growing literature that documents price
rigidity based on larger, comprehensive data sets of consumer prices.
Evidence on price stickiness based on such data has recently been provided for
the US by Bils and Klenow (2004) and Klenow and Kryvtsov (2005). Evidence for
several European countries has been produced in parallel to the present study by
Alvarez and Hernando (2006), Aucremanne and Dhyne (2004), Dias, Dias and Neves
(2004), among others (see Dhyne et al., 2006, for a survey of these findings). Here we
use a large and comprehensive data set containing more than 13 million observations
consisting of price records collected in order to compute the French consumer price
index (CPI). Those data cover a large part of the economy and allow the construction
of indicators of price rigidity that are representative of the whole non-farm business
economy. The data set is of very large size in the cross-sectional dimension (with
more than 750,000 individual products identified at the outlet level) and fairly large
in the time dimension (the sample of monthly prices spans July 1994 to February
2003).
These data are used to answer several questions of interest for macroeconomic
modelling and policy: How long is the average duration between two price changes?
Is there a nominal downward price rigidity? What is the size of a typical individual
price change? Do the data suggest state- or time-dependent price-setting by retailers?
A particular effort is devoted here to assessing the duration of prices, an indicator of
the degree of price flexibility and a key parameter in most new Keynesian macroeco-
nomic models (e.g. see Gal´ı and Gertler, 1999; Taylor, 1999).
The main findings we obtain are the following. Consumer prices are rather sticky
(lasting for nearly three quarters on average) but not particularly subject to down-
ward nominal rigidity (because four price changes out of 10 are price decreases). The
average size of a price change is large (around 10%) but small changes are not rare
(the median price increase/decrease is around 5%). In addition, as noted in several
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford 2007
What do thirteen million price records have to say about consumer price rigidity? 141
recent studies and illustrated here by estimating a multinomial logit model, hetero-
geneity across outlet types and sectors is pervasive. The model estimates also point to
a significant impact of various shocks (ValueAdded Tax rate changes, the euro cash
changeover) and of seasonality on the occurrence of price increases and decreases.
The overall picture is that the low inflation observed over the period results, in terms of
individual price changes, from a combination of infrequent price increases and price
decreases of a rather large absolute magnitude, with price increases being slightly
more frequent than price decreases. Following Klenow and Kryvtsov (2005), we
compute the contributions of the variations over time in the frequency and in the
magnitude of price changes to the variance of inflation, and extend this decompo-
sition by explicitly distinguishing between price increases and decreases. Both the
fluctuations in the size of prices increases and decreases and in the share of price
decreases in price changes contributed to the variation in inflation over the period.
The paper is organized as follows. Adescription of the data is provided in section
II. Section III presents direct evidence on price duration, emphasizing measurement
issues. Section IV deals with the frequency of price changes and provides our pre-
ferred estimates of duration. Sections V and VI deal with the determinants of the
probability of a price change, focusing on heterogeneity and duration-dependence,
respectively.Section VII presents evidence on the sign and size of variations in prices,
and their time-series relation. Several robustness checks are presented in section VIII.
Finally, section IX concludes.
II. The data: over 13 million consumer price quotes
2.1. Overview
The data are taken from a longitudinal data set of monthly price quotes collected by
the Institut National de la Statistique et des Etudes Economiques (INSEE) in order to
compute the French CPI. The methodology of data collection is described in INSEE
(1998), and also in Lequiller (1997).
The sample contains CPI records from 1994:7 to 2003:2, each record relating
to a precisely defined product sold in a particular outlet. With each individual price
quote (i.e. the exact price level of the product), the following additional information
is recorded: the year and month of the record, an individual product identification
number, a qualitative ‘type of record’ code and (when relevant) the quantity sold.1By
‘individual product’, we mean a particular product, of a particular brand and quality,
sold in a particular outlet. The individual product identification number allows us to
follow the price of a product through time, and to recover information on the type of
outlet, the category of product and the regional area. The sequence of records corres-
ponding to one individual product is referred to as the price trajectory. Importantly,
if in a given outlet a given product is definitively replaced by a similar product of
1When relevant, we divide prices by the indicator of quantity sold in order to recover a consistent price per
unit.
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford 2007
To continue reading
Request your trial