Landmark paper commemorates the work of Paul Harris
03 January 2018

Read the latest Landmark Paper:
‘The effect of clustering on costs and sampling errors of random samples’, Paul Harris (then NOP) - JMRS Vol 39 No 1, January 1997

On the day of this year’s MRS Awards dinner (December 4^th), was also held the funeral of Paul Harris. Paul was a leading market and social research statistician of his generation, and in addition to those who learned their trade in his long career with NOP (now GfK), Paul trained countless others in specialist courses or as modules in the MRS Winter and Summer Schools where many of today’s leaders learnt their craft.

Linking back to the Awards dinner, this Landmark paper won Paul the MRS Silver Medal in 1979. It was initially published in JMRS Vol. 19, No. 3 (July 1977), and then reprinted in one of the two special issues of JMRS celebrating MRS’s 50^th anniversary. And along with Bill Blyth and Ken Baker, Paul is one of only three authors to have two papers published in those special issues, Paul’s other paper being ‘Data fusion: an appraisal and experimental evaluation’, with co-authors Ken Baker and John O’Brien (originally published in JMRS Vol 31 No 2, April 1989 as an MRS Conference Proceedings paper from the 1989 MRS Annual Conference). This latter paper was based on research funded by the Market Research Development Fund. Paul was no stranger to developing and applying innovative new methods.

Not surprisingly, Paul was awarded the MRS Gold Medal in 1996 for his overall contribution to the research sector in so many ways, over such a long period of time.

This Landmark paper also links back to my last Landmark selection, back in October, as it discusses issues to do with conducting research based on random probability samples (RPS). In particular, the paper explores cost issues when using this ‘gold standard’ sampling method, making it very timely in today’s world where costs and time pressures are leading to innovations that can address these challenges without reducing data quality, as described by the speakers at the November IJMR Lecture (see Editorial in 60/1, to be published in early February, and the Editorial in 59/5 covering the SRA Summer event).

Back in 1977, RPS was not simply the ‘gold standard’, it was still routinely used across the market research sector, as Harris described in the paper. As the title of the paper indicates, Harris examines the role that clustering can play in RPS based survey design, a summary of clustering theory being included to set the scene. This includes a detailed discussion of the intra-class correlation coefficient (rho) within clusters in sample design. Paul investigates the impact of rho values obtained from two by-election surveys (Hull North in 1966 and Workington in 1976), and town surveys conducted in Scotland (1976), as there was no national level data available. 

These findings are then applied to a hypothetical national survey based on clustered samples, and the data used to calculate the comparative cost implications based on 100 sampling points with 20 interviews spread across a ward and 200 sampling points with 10 interviews contained within a small cluster of 260 consecutive names on the electoral roll, with the control a normal omnibus sample design. Some decrease in sampling errors are obtained, but at survey cost increases estimated as +20% and +35%, with the 200 sampling point design being the more efficient of the two options.

Paul concluded with a discussion of further experimental work being planned at NOP to create a databank for designing a national sample. As was always the case with Paul’s work, the topic provides a perspective on a topical issue; it describes an innovative solution; the paper is logically structured and easy to read; it is fully evidenced based. Unlike too many submissions to IJMR today, it is also succinct.

Numerous tributes to Paul have been collected by Corrine Moy (GfK), one of his former pupils and who plays a similar role today within GfK as Paul did in his time, her celebration of his life having been published recently in Research-live. 

In addition to being a contributing author for JMRS, I recruited Paul onto the Editorial Advisory Board of IJMR shortly after I was appointed Editor in Chief and he provided me with a steady stream of high quality reviews until he decided to retire from that role.

Ruwix is the best place to learn about the Worlds best selling puzzle toys. Go and check it out here!