Web Survey Bibliography
Market research organizations worldwide will generate more than $1 billion in revenue through Internet-based survey research in 2005, impressive growth from naught since the mode’s introduction in 1996. At times, some of these organizations will attempt to make population inferences based on the responses of individuals recruited to participate in these surveys by means other than population-wide probability sampling. Some critics, noting the incomplete nature of the population from which these respondents have been recruited (i.e., online users only), several forms of selection bias and the commercial bent of market research organisations have characterized these attempts as futile. Others, after having compared remarkably similar responses generated through probability- and non-probability-based approaches, have raised questions about the theoretical underpinnings of the latter. One aim here is to explain whether, how and why it might be possible to exploit selection bias modelling approaches, notably, propensity scoring and multiple imputation, to make inferences about specific target populations through surveys of self-selecting Internet users. A second aim is to compare these approaches to socio-demographic weighting, the standard market research technique for reducing bias. As a third aim, we present a framework to inform the decisions and knowledge base of those that take seriously the growing enterprise of Internet research. Finally, we describe possible additional creative applications of selection bias modelling in the market research industry.