An Empirical Assessment of New and Established Benchmarking Methods in Official Statistics

• Activity Number: 169
• Type: Topic Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Jennifer Davies* and Duncan Elliott and Homesh Sayal and John Aston
• Companies: Office for National Statistics and Office for National Statistics and University of Cambridge and University of Cambridge
• Keywords: wavelets ; benchmarking ; national accounts ; time series ; official statistics

Abstract

Benchmarking is a well established method used widely in official statistics to ensure consistency between time series of differing frequencies. For example, in the production of UK National Accounts a number of areas constrain quarterly time series to sum over the year to a related annual time series from a different source. The UK Office for National Statistics (ONS) currently use the Cholette-Dagum regression based method for benchmarking.

This paper compares two well established benchmarking methods (Denton and Cholette-Dagum) to a novel wavelet based approach to benchmarking. Wavelet benchmarking can be useful for handling more complicated benchmarking problems due to the time and frequency localisation properties of wavelets. These different benchmarking methods are compared using both real ONS time series and simulated data through a range of quality measures relevant to both users and producers of official statistics.