International Journal of Market Research – summaries of articles from current issue

2010, volume 52 issue 2

The Past Matters: Eliminating the Pro-Labour Bias in British Opinion Polls, John Curtice (University of Strathclyde) & Nick Sparrow (ICM)

The first paper is the latest to be published from last year’s Call for Papers: ‘Researching Voting Intentions’. The authors have a long history of research in this field, and publishing papers in IJMR. Their paper describes the persistent tendency since 1983 to over estimate the Labour vote in UK opinion polls, particularly noticeable since 1997, and they discuss the methods adopted to try to reduce this effect. As the authors state, the MRS investigation into the 1992 failure of the polls to predict a Conservative victory in that year’s UK General Election uncovered inadequacies in the processes used by polling companies to select demographically representative samples. However, Curtice & Sparrow argue that a more fundamental need is to explore whether demographic representativeness leads to political representativeness, contending that the link between the two has weakened over time. They also argue that due to declining levels in voter turnout, using last vote as a weighting factor to reduce the pro Labour effect, a method adopted by many polling companies, might also have weaknesses. In the paper they revisit the findings from research published by Himmelweit et al in 1978 who concluded firstly that people were less likely to accurately recall having voted for the Liberal party, or abstaining, than recalling having voted Conservative or Labour, and secondly, when recall was in error, people were more likely to claim to have voted for the party they currently support than would be randomly expected. Curtice & Sparrow discovered from their own analysis of ICM and British Election Panel study data that whilst this picture remains true today, the effect may have been exaggerated. The authors conclude that in designing weighting methods, pollsters need to take more account of variables related to likely behaviour than demographics. They also believe that whilst past vote based weighting remains an essential tool in helping to correct for pro Labour bias, greater care needs to be taken when deriving targets - pollsters need to weight by past turnout in addition to past vote choice.

Researching behavioural differences among ethnic minority groups: the case for inferring ethnicity on the basis of people’s names, Richard Webber (Kings College, University of London)

As with the authors of the first paper, the author of the second paper will also be very familiar to many readers of IJMR. Webber developed the first geodemographic coding system in the UK (ACORN) in the 1970’s, followed by the worldwide family of MOSAIC classifications. In this paper, Webber argues that peoples’ personal and family names provide a proxy for inferring ethnicity, religion and cultural origin and can provide a more effective method for researching behaviour than using data drawn from direct questioning. The author discusses the difficulties in using ethnicity as a demographic classification system and identifying true ethnicity within surveys, for example identifying the different generations of immigrants who may have very different cultural backgrounds (also recently explored by Sekhon & Szmigin in ‘The bicultural value system: undertaking research among ethnic audiences’, published in IJMR Vol. 51 Issue 6) and relating groups to a given geographic area. Webber also discusses the context in which ethnicity is used and argues that even if research surveys covered the topic in more depth, larger data sets are necessary in order to provide effective sample sizes for analysing and targeting ethnic groups. The author describes the development of techniques for inferring ethnicity based on names, leading to a file of 700m.records covering the USA and W. Europe that identified 5m.different family names and 1m.personal names used in these regions. Webber agues that automated methods represent the only effective analysis method, using rule-based algorithms, describing the OriginInfo method that he has applied to analysing 1.5m.commonly occurring family names and 0.5m.personal names; how the method was validated and issues of granularity and accuracy. The method is illustrated by two case studies in the UK covering British National Party supporters and the London borough of Tower Hamlets Primary Healthcare Trust. The author contends that these methods provide consistency, actionability similar to that available through geodemographics and insights that could not be provided through survey data. Webber concludes that this is a field of growing importance to commercial companies and public sector bodies in an increasingly multi-cultural society in order to understand needs and target products and services more accurately. However, the author expresses concerns that the market research industry will, as with geodemographics, fail to take the lead in developing and applying methodologies in this field and that the expertise will be vested elsewhere.

Media placement versus advertising execution, Edward Malthouse & Bobby Calder, Northwestern University, USA)

In the third paper, Malthouse and Calder describe their research investigating the relationship between reader engagement with a media vehicle and the effectiveness of advertising placed in that vehicle. Building on a literature review of existing research in the fields of engagement linked to media experiences and advertising effectiveness, the authors tested one hypothesis: ‘The average level of engagement with a media vehicle increases advertising effectiveness’. The authors contend that the focus at respondent level in earlier research ignores the fact that media buyers buy a vehicle to reach their target market, not individuals. Their research examined 72 magazines across 15 genres and over 5,000 advertising executions using the VISTA Print Effectiveness Rating Service, described by the authors as the industry standard in the USA. The authors describe their research process and the scales developed for their study plus the model derived to explain the relationship between engagement and advertising effectiveness. The authors conclude that their findings extend those from earlier research by showing that the relationship between the engagement of individual readers and advertising effectiveness is also true at vehicle (magazine) level: ‘the engagement of the audience as a whole with the media vehicle is reflected in the effectiveness of an ad for the entire audience’. The authors discuss the impact of their findings on media space pricing. They argue that it is in the interests of advertisers, and media owners, to factor engagement into decision making. Media owners could add value by measuring and reporting overall levels of engagement, but measures would need to be standardised to enable comparisons across and between vehicles.

Consumer-generated websites versus marketer-generated websites in consumer decision making, Fred Bronner (University of Amsterdam) & Robert de Hoog, (University of Twente)

Two potentially counter forces are available on the internet to help buyers make effective buying decisions. Firstly, there are the sites managed by product manufacturers and retailers, and of increasing importance, sites based on the experiences of consumers. Bronner and de Hoog have investigated the levels of trust, differing roles and usage of these two types of sites, focussing on vacation planning amongst a sample of consumers in the Netherlands. The authors explored four issues in their research, comparing the roles of consumer versus commercial sites. Firstly, usage; secondly, levels of trust; thirdly, the nature of contributions to consumer sites; fourthly, the sub-decisions influenced by the two types of sites. The authors’ research was based on a sample of 700 respondents drawn from the continuous vacation panel of consumers, extensively used by the travel industry in the Netherlands. The results were also compared with those from a nationwide study of online information usage. The results showed that the internet is an important source of information, with both types of sites used by many people. Both types of sites are trusted, but in different ways. Whilst relatively few contributed views to a consumer site, in the main the comments were favourable. Finally, usage of consumer sites often related to post purchase experience, such as accommodation location and service quality in the case of a vacation. The authors advise researchers and marketers to monitor consumer sites in order to rapidly respond to issues being raised and use eWOM as a source of insight alongside traditional market research methods.

Research into questionnaire design: a summary of the literature, Petra Lietz (Jacobs University, Bremen, Germany)

The fourth paper should be of interest to many readers. The author has contributed an extensive review of the literature covering the cognitive and communications aspects of question design. The paper is primarily divided into three parts, each being divided into key topic related subsections. The first part covers the design of questions, with eight sub sections covering topics such as question length, social desirability, ambiguity (‘double-barrelled’ questions), negative wording. The second part covers response related issues, such as the ‘don’t know’ option, and opinion filtering. The final part covers scale options. Based on the review, the author provides readers with thirteen key conclusions, many of which whilst being familiar and common sense based, are not always reflected in real life examples of questionnaires. Finally, the author points out that whilst some aspects of design have been extensively researched, other areas have been less widely investigated and could benefit from empirical research. As this paper is a literature review, readers will also find it a very useful source of references where they can find more detailed guidance from a wide range of research in the field of question design.

Forum

Ethics in practice: using compliance techniques to boost telephone response rates, David Bednall, (Deakin University, Victoria, Australia)

Declining response rates is a concern across the market research sector. The author of the Forum article discusses this issue and the potential for using compliance techniques as a way to encourage more respondents to participate in telephone research. Bednall describes seven forms of compliance technique, some of which he argues rely on deception or half-truths, which the author feels research agencies under extreme pressure to achieve interview numbers might be tempted to use. The author discusses how these techniques fit with the Australian market research industry’s Code of Conduct, which is similar to other national codes and the international ESOMAR Code of Conduct – in particular the principles that respondent co-operation is voluntary and they must not be misled – and concludes that compliance methods can violate these Codes if they are deliberately deceptive or provide misinformation. Bednall undertook eight in depth interviews with telephone research managers followed by a quantitative survey of 347 agencies in Australia to assess attitudes towards compliance techniques, and whether they are used in practice. Based on the findings, the author identified four categories of compliance methods: those rejected as being unethical; those rejected as being counter-productive; those deemed acceptable but with limited application, and finally, ‘Liking’ which often formed the basis for selecting and training interviewers. The author concludes that some methods would tend to increase the problem of non response, rather than provide a solution and recommends researchers in other countries investigate local attitudes in order to compile an international perspective.

Back to IJMR page

What's New - Membership - Company Partner Service - Members' Area - Code/Guidelines - Qualifications - Training - Awards
Events - Networking - Publications - Media Info - Market Research - Search - Site Map - A-Z Directory - Contact Us - Home