Web Survey Bibliography

Title Access pools as a solution to the nonresponse problem?
Author Stoop, I.
Year 2004
Access date 13.09.2004
Abstract As obtaining high response rates is increasingly difficult, an increasing number of organisations revert to so-called access pools as a means to ensure high response rates. Access pools are groups of people who have agreed to regularly partici­pate in future sur­veys run by a specific organiza­tions, generally a market research organization. For a sur­vey among an access pool, a either a sample is drawn from a pool of cooperative individuals or households that is similar to the gen­eral population on socio-demographic character­istics, or a specific sample is selected, for in­stance parents of small children in rural areas. The latter possibility to select specific subgroups, makes it a very attractive in­strument. Access pools are nowadays more and more web access pools. Access pools can be based on probability sampling and on nonprobability sampling. The former are also called pre-recruitment panels, the latter volunteer panels. In pre-recruitment panels re­cruitment is passive. At the end of each sur­vey run by a market research organization, for in­stance, respondents are asked if they want to be an access pool member, and after each survey the pool gets larger. If the re­cruitment is from ran­dom sample surveys, and information on initial inclusion probabili­ties and response rates are available (which is rarely the case), a survey among (a random sample from) an access pool can theoreti­cally be considered as based on probability sampling. If inclusion prob­abilities and nonre­sponse in earlier phased are unknown, and if there is a large attrition in earlier phases, the prob­ability character of access pools is easily lost. In vol­unteer pools recruitment is active: respon­dents can apply for membership of an access pool. If respondents can self-select, it is definitely a non-probability sample, and actually quite similar to a quota sample (provided there are controls on pool membership or con­trols on selecting pool members for a survey). Access pools have a number of advantages. Firstly, nonprobability samples generally have higher response rates than probability samples and particularly ac­cess pools have high response rates. Market research organizations even tout customers for their ac­cess pools with the recommen­dation that high response rates are guaranteed, because the re­spondents (here: members of the access pool) have stated that they are will­ing to participate in surveys. A second advantage is that both quota samples and access pools can be made ‘represen­tative’ a priori and that weighting for under-represented groups (because of sampling errors and nonre­sponse) is not neces­sary any more. This is related to the third advantage, namely that spe­cial groups (homosexuals, parents of young children) can easily be selected from access pools as many characteristics of the pool members are known, or because screening is relatively inexpen­sive. The fourth advantage is that the turnaround time of quota samples and access pools is gen­erally shorter than for random sam­ples. In the latter case, many calls have to be made to contact a sam­ple household (in face-to-face surveys), mail reminders have to be sent, and initial refusers have to be converted. When using quota sampling, not much time has to be spent on including a par­ticular sample person, as any other with the same characteristics will do. In access pools, re­spon­dents are interested and willing to participate, so obtaining response goes fast too. The final ad­vantage is related to the previous ones: they are often less expensive than random samples. This may partly be due to the fact that ac­cess pools generally use web or mail surveys. They are presumably also less expensive, because pool members have promised to regularly partake in surveys, and therefore less time has to be devoted to persuading respondents to participate. On the other hand, maintaining a pool requires time and money too. The paper will discuss the assumptions behind access panels, the meaning of representativeness, the possibility of statistical inference from access panels, comparisons between access pools and random samples and the question whether access pools can solve the nonresponse problem.
Access/Direct link Homepage - conference (abstract)
Year of publication2004
Bibliographic typeConferences, workshops, tutorials, presentations
Print