Navigation – Plan du site

AccueilRubriquesEspace, Société, Territoire2000Who’s in? Lifestyle data and geog...

2000
119

Who’s in? Lifestyle data and geographical research

Les données sur le "style de vie" et la recherche géographique
Richard Harris

Résumés

Les bases de données construites par les entreprises de géomarketing donnent des informations mises à jour à l’échelle des individus. Cependant, la recherche géographique les a encore peu utilisées. Ce paradoxe est analysé à la lumière de la géomatique; Les possibles biais de cette information peuvent dissuader les utilisateurs : les personnes qui répondent à des enquêtes de "style de vie" sont auto-choisies, et les jeunes adultes ont une probabilité plus faible d’y figurer que les retraités. Le degré de couverture géographique de l’enquête est variable. Aussi, bien que la source soit potentiellement riche d’une information susceptible de compléter les données du recensement au niveau individuel, un contrôle scientifique est nécessaire avant de pouvoir l’utiliser pour produire des résultats acceptables par les chercheurs.

Haut de page

Texte intégral

Acknowledgements : My sincere thanks to the ‘data-supplying company’, to Paul and to Siân for their on-going help, support and encouragement. Any errors or omissions remain my own. The research is sponsored by ESRC scholarship R00429634059. 1991 Census © The Crown. ESRC/JISC joint purchase. The lifestyles data are © ‘the data-supplying company’. This paper © Richard Harris, 1999.

Introduction

1The rationale behind this paper is described in the following quotation :

more and more attribute data are becoming available in the socioeconomic realm, which may be exploited to devise richer ‘digital personae’ as components of models of human activity patterns if problems of respondent self-selectivity and bias can be successfully accommodated (Longley 1998 : 5; after Birkin 1995).

2Accordingly, Harris (1998b) concludes, "question marks seem to lie over this data, the suspicion being that it is biased towards particular sections of the population … More research is being undertaken to ascertain the nature of response bias." The current paper presents much of this research (see also, Harris, 1999).

3The attribute data which Longley describes frequently originate from marketing. Referred to as ‘lifestyles data’, they quantify, to varying degrees, the behaviours of individual consumers. Openshaw and Turton (1998) have advocated that interest in lifestyles databases should not lie solely with marketers but with geographers too. Lifestyles data are generally up-to-date and are explicitly about people. They are not describing (Census-) areas which have little intrinsic geographical meaning (after Openshaw 1984 : 4). The data are geo-referenced by residential address and may be modelled at the individual or household, theoretically unconstrained by artificial (and geographically meaningless) zoning schemes.

4Why, then, have lifestyles databases not won broader academic credibility ? Most probably it is due to the issues which Longley describes : the problems of respondent self-selectivity and bias which effect the results of lifestyles-based analyses. The digital personae, or ‘lifestyles population’ which constitute the database are a far from random sample of the population, they are not selected by the mathematical principles of ‘traditional’ survey design (see, for example, Moser and Kalton 1985). Indeed, the lifestyles population is self-selecting. The databases are comprised by individuals who have volunteered information. It is a legal obligation to return a Census form. The undertaking of a consumer survey is voluntary. A leap of faith is require to believe that all members of are society are equally likely to solicit lifestyles information. Yet without that faith, or without details of the database composition, it is doubtful whether models of the lifestyles population can be generalised to the national, or other populations. Unfortunately, there is a lack of published information offering such details : who are the lifestyles population? Where do they live? It is these questions which the paper considers. Firstly, however, it is of relevance to consider the general debate surrounding a date-led approach to geographical analysis. The debate is of particular salience to the field of geocomputation.

Geocomputation and data (-led) analysis

5By a recent working definition, Couclelis (1998 : 17) described geocomputation as an "eclectic application of computational methods and techniques ‘to portray spatial properties, to explain geographical phenomena, and to solve geographical problems’" (emphasis added). These methods include automated cluster analysis, probability-based modelling, neural networks and artificial intelligence, and also genetic algorithms. Potential applications are well documented with respect to both social-scientific research and to marketing (Openshaw and Openshaw 1997; Berry and Linoff 1997). Indeed, Goodchild and Longley (1999) comment that the computer is fast becoming no longer an adjunct to the research environment but the research environment itself!

6In contrast to the positivist formulation of the scientific model, which characterised geography’s ‘quantitative revolution’, geocomputation has :

no philosophy (and proud of it!), no set of approved practices to help define the standards of acceptable work, a haphazard collection of techniques developed for all sorts of different purposes in a variety of areas with sometimes unproven intellectual credentials, and, last but no least, no obvious conceptual framework to define some alternative vision of what geography may be about, or where it might go (Couclelis p.22, emphasis added).

7This conflicts with desires for a "push button tool of academic respectability where all the heavy stuff happens behind the scenes but the results cannot be misinterpreted" (Mckeon, Infoshare : e-mail : 1997; cited by Openshaw 1998). The second from Couclelis’ five-point list of challenges to geocomputation states : "geocomputation must develop scientific standards that will make it fully acceptable to the quantitative side of geography and to their geosciences" (p.27).

8Geocomputation can be fundamentally data-led. Yet, as Longley (1998 : 11) argues, increased quantities of data do not make them any less imperfect, incomplete or error-prone. ‘Data fusion’ - the creation of composite datasets from different and often disparate data sources - does not guarantee better analytical results but may, instead, lead to error propagation. Resultant data models contain error, ambiguity and uncertainty (after Fisher 1999). Referring back to Couclelis’ definition of geocomputation, these models may mis-portray spatial properties, giving fallacious explanations and solutions. The sheer quantity of attribute data available in the socio-economic realm can offer qualitatively richer databases of digital personae but are these profiles actually representative of real-world individuals? Longley (p.7) contends :

platitudes about data-rich environments and information economies cannot disguise the fact that the sources and operations of many biases in data collection and availability are very poorly understood. A good example is provided by the supplementation and partial replacement of (census-based) geodemographic data with lifestyles datasets. Lifestyles data are richer, more detailed and up-to-date and, arguably, predicatively successful, than geodemographic forebears : yet at the same time they are fundamentally more biased and unscientific.

9For an introduction to geodemographics and lifestyles see Sleight, 1997.

Considering lifestyles and Census data

10A definition of lifestyles data might emphasise their salient characteristic : they quantify attributes of individual consumer behaviour, preference and choice. This, a wide definition, is inclusive of data from a variety of sources, for example : consumer questionnaires; sales data; share ownership records; county court judgements. Lifestyles data are also distinctive from the 1991 (UK) Census data. Whereas lifestyles data can be analysed at the non-aggregated, individual or household level, the Census data are aggregated, to the level of the enumeration district at least (containing approximately 150 households). The Census measures the social-economic characteristics of areas (census enumeration districts, EDs) and provides only indirect information upon the consumer practices of individuals.

11The lifestyles dataset to be analysed in Section 4 (below) can be more narrowly defined as being constituted by : data obtained and stored at the non-aggregated level of a named individual and geo-referenced by their address; the data are collated from the return of (consumer) questionnaires mailed directly to the British electorate, their addresses being identified from the Electoral Register.

12The postal questionnaires were distributed between September and October, 1996 to addresses on the Register as of the previous October, 1995. Thus, recipients were at addresses where at least one person had registered with their local authority to vote. However, Register entries indicating more than three different surnames per address were excluded from the survey (these are, in effect, Houses in Multiple Occupation). Individuals registering not to receive such mailings, through the Mail Preference Scheme, were also excluded. Wherever possible the questionnaires were addressed to a female resident. The incentive offered to complete the questionnaire was the same to each recipient (- the chance to win draw entries in the National Lottery). Trial television advertising supported the survey within the West Country broadcasting region only. In total, 20 million questionnaires were distributed across Great Britain of which 2 million were returned : a national response rate of 10% (regional average for the West Country, 16%).

13The sampling fraction is the ratio between the size of a sample and the size of its parent population. For the analysis of Section 4 it shall be assumed that the parent population of the lifestyles database is equivalent to the population base of the 1991 Census statistics (to be called the ‘Census population’). The assumption is made to compare the 1991 Census and the lifestyles data. Two objections against this method are now considered.

14Objection 1 : The Census population is not parent to the lifestyles population. The Census was undertaken 5 years prior to the lifestyles survey.

15Compounded with differences in the specification of variables relating to tenure and property type it is to be admitted that the two surveys are not immediately comparable. However, distributions of the population by age, sex or marital status are unlikely to have changed significantly over the five years, and the housing types (detached, semi-detached, etc.) are somewhat fixed in stone! Throughout Section 4 careful attention is given to the variable definitions, aligning the two datasets. Levels of car ownership, the distribution of dwellings by tenure or the distribution of persons by economic position are potentially more fluid over the period, related to the national economic climate. Nevertheless, large differences between the lifestyles and Census populations are unlikely to be merely spurious.

16Objection 2 : The lifestyle database providers are not interested in the entire population, only the ‘mail responsive’.

17Initially in agreement with this contention, Lawson (1998 : 22-23) writes that "in the past, individual information was almost exclusively used to make up [!] mailing lists in the direct marketing industry and this is still by far the largest application for it." He continues,

but as geographic coverage and UK penetration of the surveys has increased, new data products have arrived that challenge the traditional dominance of Census-based geodemographic data in marketing applications that have a geographic element such as site location, penetration analysis, door drops, and media planning.

18A promotional brochure for the ‘PRIZM Micromarketing’ product implies the data-providers, Claritas UK, are offering an alternative to the Census, one which is explicitly consumer-based and up-to-date :

  • "At PRIZM’s core lies the Lifestyle Census";

  • "Over 75% of data utilised by PRISM is less than three years old … Census data was last collected in 1991";

  • "The lifestyle and demographic data on which PRIZM is built has been collected specifically for marketing purposes … Census data is collected specifically for local services planning purposes."

19Claritas claim up to 300 pieces of information for 75% of UK households are available to the user of PRIZM.

Comparing the lifestyles database and the 1991 Census

Gender (of ‘prime respondent’)

Table 1 - The lifestyles and Census populations by gender

sampling fraction (by households) = 0.09
* prime respondents
** GB residents aged 18 or above

20The national lifestyles dataset contains entries for 1,951,975 ‘prime respondents’ across Britain of whom gender is known. The prime respondent is the person to whom the questionnaire is addressed, or the first name on the returned form. The prime respondent completes the ‘You’ answers on the survey form. Consistent with the guidelines for completing the questionnaire, other answers, attributed to a ‘partner’, may in fact be completed : by a partner; by the prime respondent on behalf of their partner; by another member of the household; by the prime respondent on behalf of other members of the household; or left blank. Hence, whilst the prime respondent corresponds to an individual, the ‘partner’ may not.

21Table 1 shows the lifestyles database to contain 1,951,975 prime respondents. The relative bias of 21% towards females (and thus also against males) is to be understood as the deliberate targeting of females as prime respondents (see Section 3). The 1991 Census gives a count of 21,802,787 households (with residents and in permanent buildings across Britain). Assuming one prime respondent per household and treating the Census as the ‘parent’ to the lifestyles population, then an estimated sampling fraction (by households) can be derived, as in equation 1.

221,951,975 ÷ 21,802,787 ≈ 0.09 [1]

23This fraction of 9% is of a similar magnitude to the estimated response rate of 10%. Some Census data are derived from a similar sized sample, 1-in-10 of the Census forms (namely the 10% tables of the Small area Statistics). The Census, however, takes a random sample. The lifestyles sample is self-selecting.

Age (of prime and their partner)

Table 2 - The lifestyles and Census populations by age

24Table 2 shows the proportions of the lifestyles and Census populations within each of six age bands. Summing over these bands gives a total lifestyles population of 2,702,102 who recorded their date of birth on the questionnaire (5% of the prime respondents shown in Table 1 chose not to reveal their age). A difference of (minus) 9% between the lifestyles and Census populations is shown for those aged 18-24. The lifestyles survey enumerates (‘penetrates’ into) only 2% of British adults aged under 25 years (i.e. 100,531 × 100 / 5,614,045), whereas the mean penetration across the six age bands is 6%. The Pearson correlation between the absolute Census data and lifestyles data (the second and fourth columns of Table 2) suggests the extent of the linear relationship between the two datasets to be 71%. Notably, the extent reaches 83% when the 18-24 age group is excluded.

25Similar results were obtained by the author for a regional sample of the lifestyles database comprised by an area inclusive of and surrounding Bristol, England (Census population, 636,366 adults). In this study thirteen ‘lifestage groups’ were considered, as shown in Figure 1 and broadly numbered by ascending age (see Harris 1998a for precise definitions of the groups). Hence, the youngest lifestage groups are 1 and 2 (aged 16-24, without and with children in the household, respectively). Lifestage 1, especially, identifies that young adults form a relatively small proportion of the lifestyles population when compared with the Census population. As Rae (1998 : 6) has commented, referring to the results of the regional study, "groups that are equally common in the population [-at-large] have radically different representation in lifestyle data" - compare lifestage groups 4 and 11, for example (aged 25-34, children aged 0-4 years; and aged 55-64, working or retired).

Figure 1 - The lifestyles and Census regional populations, by lifestage group (after Harris 1998a)

Economic position (of prime and their partner)

Table 3 - The lifestyles and Census populations by economic position

* GB residents aged 18 or above

26Table 3 shows the lifestyles and Census populations distributed by economic position. It also provides an estimate of the total lifestyles population (prime and partners) : 2,761,532 (presumed) individuals. Assuming this value, and knowing that the number of prime respondents equals 1,951,975 (Table 1), then the number of partners is 809,557, a ratio of 71 :29 (prime :partner). The Census counts 21,802,787 households across Britain (with residents and in permanent buildings), of which 6,738,654 are comprised by one adult only (with or without dependent children). Hence, 15,064,133 (69%) of households have (at least) a second adult resident - there is a ‘partner’ by the definition of the lifestyles questionnaire. Applying this percentage to the lifestyles database, then 69% of the 1,951,975 prime respondents are expected to have partners. The expected number of partners becomes 1,346,863 : 1.7 times greater than the observed value, calculated as 809,557 above. There are two possibilities causing this shortfall :

  • the lifestyles survey method tends to enumerate one-adult rather than multiple-adult households;

  • the partner’s details are often absent from the returned questionnaire.

27In Table 3 the Census data refer to GB residents aged 18 or over. Effectively, the lifestyles data are also of those aged 18 and above, that is, those who are eligible to vote and who are on the Electoral Register. The ‘employed’ group of the lifestyles population are those who have an occupation indicated as a director, a manager, a skilled worker or trader, a manual worker, an office worker, a shop worker, a public sector worker, or a professional. The student group of the Census population considers only economically inactive students. Table 3 shows that the retired group constitute a larger proportion of the lifestyles population than they do the Census population, the difference being +7%. Intuition suggests that people spending greater time at home are increasingly likely to complete the lifestyles questionnaire. The survey takes about twenty minutes to complete in full.

Tenure and property type (of addresses with prime respondents)

Table 4 - The lifestyles and Census populations by tenure and by property type

* includes properties rented with a job or business
** may be either detached [a] or semi-detached [b]

28The lifestyles questionnaire permits no direct response to indicate tenancy of a new town or Scottish home. It is assumed that these tenants would therefore indicate a council or housing association tenancy. A comparable Census variable can then be derived by amalgamating the new town and Scottish home properties into a broader council and housing association group. It is further assumed that those indicating ownership of properties on the lifestyles questionnaire are equivalent to the Census’ owner-occupied group which includes mortgagees. Under the privately rented group, the total number of household spaces across GB varies depending upon the inclusion or exclusion of properties rented with a job or business. The inclusive total (1,969,571) and corresponding calculations are shown in parentheses in Table 4. There is no notable difference between the distributions of the lifestyles and Census populations by tenure, other than a slight, dis-proportionate ‘over-sampling’ of the owned group by the lifestyles survey, at the expense of the council and housing association group.

29Table 4 indicates that it is permissible to indicate a property as being a bungalow on the lifestyles questionnaire but that no equivalent category exists for the Census. Therefore, the bungalows are firstly treated as being all of a detached type (column [a]) and secondly as all being semi-detached (column [b]). In either instance it is shown that lifestyles survey tends to enumerate persons living in semi-detached properties rather than those living in flats or terraced properties. This tallies with the relative under-enumeration of the 18-24 age group (Table 1). Hypothesising that young adults tend to reside near the centres of cities, are also where Houses in Multiple Occupation are located (recalling that these properties are excluded from the survey), then the ‘sampling fraction by enumeration district’ should exhibit a geographical pattern. The sampling fraction by ED is the total lifestyles population for an ED (prime respondents plus partners, where applicable) divided by the total Census-population resident.

30An interpolated surface model of the sampling fractions per ED is illustrated by Figure 2. The area shown is that of the Bristol study region for which Figure 1 (above) was derived. Areas coloured blue generally exhibited a cold ‘response’ to the survey, whereas areas shaded red gave a warm response. Evidently the city centre is dominated by a low sampling fraction, as anticipated. The drift of the area of low response beyond the city border is due to EDs being absent from the lifestyles dataset (hence, a zero sampling fraction). ‘Edge effects’ from the interpolation method are also apparent.

Figure 2 - Interpolated surface model estimating the proportion of the Census population enumerated by the lifestyles survey

Conclusion : lifestyles data and geographical research

31This paper was intended to be critical of a lifestyles database in the evaluative sense only. There has been a tendency, perhaps, to treat the Census as the benchmark by which lifestyles should be judged when in reality the Census is not perfect or well suited to every analytical task. This imbalance is partly redressed by Figure 3 and the following discussion. However, see Openshaw (1995) for a full consideration of the analytical and spatial aspects of the Census data.

Figure 3 - A comparison of lifestyles and Census data for geographical research

32As Figure 3 implies, the appeal of lifestyles is the availability of up-to-date information specifically about people themselves and not just the areas where they live. It is no longer necessary to infer individual attributes from indirect, surrogate data sources if the individual is well represented by the lifestyles data. Alternatively, lifestyles databases provide extensive micro-data, adding another string to the analytical bow (after Lawson 1998 : 23). Lawson quotes Wade from the Data Consultancy company :

The Census has excellent coverage and little ambiguity, while some lifestyle surveys have dozens of questions which can be ambiguous or simplistic and are often returned incomplete. But they are up-to-date and identify individual households. They are different to each other but not better … perhaps best seen as complementary.

33Are they different but complementary, or just different? Winters (1997 : 8) writes, "more information is needed on both types of data [Census-based geodemographics and lifestyles] to decide when to use one or the other, either individually or in combination."

34Wade continues : "a lot of people are now using lifestyle data aggregated to postcode sector for site location." Lawson (p.24) also cites Sleight : "Lifestyle data is usually aggregated to postcode sector for any kind of geographic analysis. It’s simply overkill at the individual level." Figure 4 shows the proportions of households living within different property types and tenures, estimated using both the lifestyles and Census data for an inner-city postcode sector (Bristol, ‘BS6’). The lifestyles dataset has a notably exaggerated proportion of semi-detached properties (compare with Table 4).

Figure 4 - The percentage of ‘BS6’ households in particular property types and tenures, estimated using lifestyles data and by the Census.

35Rae from CACI Limited argues that aggregation is possible "but you have to take care" (Lawson, op. cit.). CACI’s ‘LifestylesUK’ product adds weighting to a lifestyles dataset, to reflect a UK distribution of the population. The dataset is, in fact, derived from a recent successor to the database analysed in Section 4 and results from substantively similar postal-survey method (with additional Census-based data and other indicators culled from the Electoral Register, Sleight 1998 : 8). It is therefore expected that LifestylesUK gives higher weight to young adults and lower weight to pensioners. Post-stratification weighting is, however, an imperfect solution to compensate for sampling bias. Consider the 18-24 age group, for example. Those of this age group who complete a lifestyles questionnaire are the exception not the rule (they comprise an estimated 2% of all adults of this age, Table 2). It is questionable indeed to assume that their lifestyles are representative of others within the group. Yet, weighting the lifestyles data to reflect the national age distribution gives these respondents increased weight, not less!

36LifestylesUK fills the holes in the lifestyles database by modelling the characteristics of non-respondents to the consumer survey (but who are on the Electoral Register) form the characteristics of respondents and assigning their resulting digital persona a probability score on each of 300 variables. CACI’s new product, ‘PeopleUK’ goes a stage further, feeding the probability scores into a clustering algorithm and classifying each member of the UK electorate into one of 46 different types. Who, though is being classified, real-world individuals or their discrepant digital counterparts? Sleight (op. cit.) quotes Rae’s assessment of PeopleUK : "it’s not going to be 100 per cent correct on all dimensions, naturally, because of the clustering process [and the underlying data!], although we would expect it to be broadly correct on most" (emphasis added).

37To some applications this level of uncertainty could be acceptable. A lifestyles product might be usefully applied to an exploratory data analysis, for example. Nevertheless, the inability to establish firm (statistical) confidence in analytical results undermines lifestyles’ credentials for academic research or local authority and service planning. In this context, and by way of conclusion, Couclelis’ challenge to geocomputation (Section 1) can be re-issued to the providers and analysts of lifestyles databases : to develop scientific standards that will make application of the data fully acceptable to the quantitative side of geography

Haut de page

Bibliographie

Berry MJA and G Linoff (1997) Data Mining Techniques : for marketing, sales, and customer support, New York : John Wiley & Sons Ltd.

Birkin M (1995) Customer targeting, geodemographics and lifestyle approaches. In P Longley and G Clarke (eds.) GIS for Business and Service Planning, Cambridge : GeoInformation International, 104-49

Couclelis H (1998) Geocomputation in Context. In PA Longley, SM Brooks, R McDonnell, B Macmillan (eds.) Geocomputation : a primer, Chichester : John Wiley & Sons Ltd., 17-29

Fisher PF (1999) Models of uncertainty in spatial data. In PA Longley, MF Goodchild, DJ Maguire, DW Rhind (eds.) Geographical information systems : principles, techniques, management and applications, New York : John Wiley & Sons Ltd., 191-204

Gittus E (1964) The structure of urban areas : a new approach. Town Planning Review, 35, 5-20

Goodchild MF and PA Longley (1999) The future of GIS and spatial analysis. In PA Longley, MF Goodchild, DJ Maguire, DW Rhind (eds.) Geographical Information Systems : principles, techniques, management and applications, New York : John Wiley & Sons Ltd., 567-80

HARRIS R (1999) Geodemographics and the analysis of urban lifestyles. Bristol University : Ph.D. thesis.

Harris R (1998a) Considering (mis-)representation in geodemographics and lifestyles. In RJ Abrahart (Production Co-Ordinator) Proceedings of the 3rd International Conference on GeoComputation, University of Bristol, CD-ROM (ISBN 0-9533477-0-2). Also at http ://www.bitsmart.com/rich.harris/

Harris R (1998b) Is there an 'I' in GDIS? The problem of representation in geodemographic and lifestyles systems. Cybergeo, 63, 1-22. At http://www.cybergeo.presse.fr/

Lawson J (1998) Once in a lifestyle. Mapping Awareness, 12, 22-25. Published by Adams Business Media, e-mail : ma@mail.aip.com

Longley PA (1998) Foundations. In PA Longley, SM Brooks, R McDonnell, B Macmillan (eds.) Geocomputation : a primer, Chichester : John Wiley & Sons Ltd., 3-15

Moser CA and G Kalton (1985) Survey Methods in Social Investigation, Aldershot, Hampshire : Gower Publishing

Openshaw S (1984) The Modifiable Areal Unit Problem (CATMOG 38), Norwich : Geo Books

Openshaw S (1995) Census user's handbook, Cambridge : GeoInformation International

Openshaw S (1998) Building Automated Geographical Analysis and Explanation Machines. In PA Longley, SM Brooks, R McDonnell, B Macmillan (eds.) Geocomputation : a primer, Chichester : John Wiley & Sons Ltd., 95-115

Openshaw S and C Openshaw (1997) Artificial intelligence in geography, Chichester : John Wiley & Sons Ltd.

Openshaw S and I Turton (1998) Geographical research using lifestyles databases. Unpublished paper, presented at the RGS-IBG annual conference, Kingston. At http://www.geog.leeds.ac.uk/staff/ s.openshaw/ibg98/P001.html

Girault F., Orhan J.M. (1999).La qualité des données de géomarkéting. Cybergeo, n°104, 12/07/99 http://www.cybergeo.presse.fr/essoct/texte/orhan.htm

Rae J (1998) Why do you have to mix Lifestyle and Census data. Marketing Systems Today, 13, 6-7. Published by CACI Ltd., e-mail : marketing@caci.co.uk

Sleight P (1997) Targeting Customers : how to use geodemographic and lifestyle data in your business, 2nd edition, Henley-on-Thames : NTC Publications

Sleight P (1998) Opinion. Database Marketing, 1, p. 8. Blue Sky Publishing, e-mail : dmarket@ndirect.co.uk

Winters P (1997) Best of both worlds. New Perspectives, 7, 8-9. (New Perspectives is no longer published)

Haut de page

Notes

1 The sampling fraction is different from the response rat. To calculate the true response it would be necessary to know both the number of questionnaires sent to addresses per ED and number returned.

Haut de page

Pour citer cet article

Référence électronique

Richard Harris, « Who’s in? Lifestyle data and geographical research », Cybergeo : European Journal of Geography [En ligne], Espace, Société, Territoire, document 119, mis en ligne le 03 février 2000, consulté le 27 janvier 2021. URL : http://journals.openedition.org/cybergeo/4343 ; DOI : https://doi.org/10.4000/cybergeo.4343

Haut de page

Auteur

Richard Harris

School of Geographical Sciences, University of Bristol, University Rd., Bristol, United Kingdom. BS8 1SS Telephone +44 (0)117 928 9000 ext. 3851 Fax +44 (0)117 928 78 78
R.J.Harris@bris.ac.uk

Articles du même auteur

Haut de page

Droits d’auteur

Licence Creative Commons
La revue Cybergeo est mise à disposition selon les termes de la Licence Creative Commons Attribution - Pas d'Utilisation Commerciale - Pas de Modification 3.0 non transposé.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search