Best of Both Worlds?: Can we make convenience samples representative?

Pete Doe and Robert Smith
Nielsen

Introduction/Background

How good are online convenience panels? They are certainly an economical and quick way of getting research data. Many companies have large respondent panels recruited online that can be contacted cheaply and quickly to provide survey insights such as voting intention, consumer product purchase behavior and attitudes towards cable TV programs. The ability to collect data cheaply and quickly is valuable, but are the results accurate and actionable? It is well known that these samples contain the possibility of significant selection bias and this in turn can lead to misleading results. How can such panels be best sampled to minimize bias?

This paper examines different methods to select and weight samples from a convenience panel to create as representative a sample as possible, using a smaller representatively drawn sample as a reference.