|Salary:||From £16,062 current stipend at UKRI level, tax-free per annum for up to 3 years|
|Placed On:||9th June 2022|
|Closes:||31st August 2022|
Consumer price indices are among the most used official statistics outputs, but assessing their quality is challenging. They are produced using two primary inputs, weights & prices. Both have typically been derived from surveys, although weights data have more recently been post-processed through national accounts balancing which adds a further layer of complexity. Prices & weights may in future come from more comprehensive scanner and web-scraped sources. The complexity of the designs (particularly for price collection, which may involve both probability samples & non-probability sources) makes it challenging to produce quality measures, and there is a need to extend quality measurement to account for the new data sources. Current developments in price indices include calculation of indices relevant for population subgroups & for regions, and there is a need to evaluate their quality to help users to understand and interpret their usefulness.
An initial phase will consider the three basic approaches to calculation of sampling errors, Taylor linearisation methods, replication-based methods, & model-based methods (reviewed in Smith 2021, see also Zhang 2010). The first two approaches generally require separate estimation of the quality measures for the price and weight sources, and this may involve the evaluation of the quality of small area estimates where this kind of estimation is used. A second phase will consider how these separate measures may be combined to produce overall quality measures for price indices, and for changes in price indices. A third phase will examine how these approaches could be applied when more data are available from scanner data, web scraping and other big data sources, or what alternative approaches are needed to generate quality measures for these non-probability data sources. There is a range of further issues in the calculation of price indices such as the use of quality adjustment, hedonic methods and imputation which suggest some possible extensions to this work.
Smith, P.A. (2021) Estimating sampling errors in Consumer Price Indices. International Statistical Review 89 481-504. www.doi.org/10.1111/insr.12438.
Zhang, L.-C. (2010) A model-based approach to variance estimation for fixed-weights & chained price indices. In Official Statistics: Methodology & Applications in Honour of Daniel Thorburn, Eds. M. Carlson, H. Nyquist & M. Villani, Stockholm University & Statistics Sweden: Stockholm, pp. 149–166.
Candidates must have or expect to gain a first or strong upper second class degree, & a Merit at Master’s degree in an appropriate discipline, not necessarily Statistics. The research is, however, methodological in nature, and the successful candidate will have solid statistical theoretical knowledge, as well as programming skills & ability to carry out advanced data analysis & simulation studies.
How to Apply
Applicants should apply for ‘PhD Social Statistics & Demography’ via https://studentrecords.soton.ac.uk/BNNRPROD/bzsksrch.P_Search. Please state in the funding section of the on-line application that you wish to be considered for this Studentship.
Informal enquiries may be made to Prof Paul Smith (email@example.com).
For the latest information on postgraduate opportunities within Social Statistics & Demography, please visit our website at https://www.southampton.ac.uk/demography/postgraduate/research_degrees.page.
Applications will be considered in the order that they are received, and the position will be considered filled when a suitable candidate has been identified.
This project is supported by the Office for National Statistics & University of Southampton. The studentship is funded with a stipend at UKRI level, currently £16,062 tax-free per annum for up to 3 years, with a Research Training Support Grant of £750, together with tuition fees. The studentship is for 3 years.
Type / Role: