A framework for large scale use of scanner data in the Dutch CPI

A framework for large scale use of scanner data in the Dutch CPI

De Haan, J.

Architecture and The Built Environment

OTB

2015-12-31

Statistics Netherlands is planning to use scanner data on a large scale for the compilation of the CPI, covering supermarkets, department stores, do-it-yourself stores, etc. Ideally, to make the production process as efficient as possible, a limited number of fully or semi-automated methods would be used. The purpose of this paper is to propose a framework supporting these plans. Our basic aggregation formula is what we refer to as a “quality-adjusted unit value index”, which is equal to the value index divided by a quantity index that is defined as the ratio of quality-adjusted or standardized quantities. Time dummy regression models play an important role in the estimation of the quality-adjustment (standardization) factors. There are two extreme cases. If information on all relevant item characteristics is available, then the use of time dummy hedonic models is preferred. When characteristics information is lacking, the use of time-product dummy (fixed effects) models is proposed. We also discuss a number of issues that need to be resolved before our ideas can be implemented in CPI production, such as the definition of homogeneity, treatment of revisions and choice of estimation window.

fixed effects
hedonic regression
quality adjustment
unit values

http://resolver.tudelft.nl/uuid:c3c8191b-a6cb-462f-b73c-559f8b080d97

Ottawa Group

Report from Ottawa Group 14th meeting, International Working Group on Price Indices, Tokyo (Japan), May, 2015

Institutional Repository

conference paper

(c) 2015 De Haan, J.