Estimating completion and breakoff functions in a business web survey

A business web survey should be of an appropriate length. On the one hand it should include all the questions which are important to the researcher, but on the other hand, it should not be too long as the breakoff rate in this case tends to be high, resulting in a low response rate. In consequence, the researcher is forced to invest more time and money in order to reach a sample size which would enable an appropriately performed statistical analysis. In this paper, completion and breakoff times are observed and compared across different questionnaire and respondent characteristics. A regression modelling approach has been adopted to estimate the completion and breakoff functions to help a researcher determine which respondents completed a questionnaire and which broke it off too quickly or too slowly. By omitting such respondents, a researcher is able to obtain the relevant estimates more efficiently. In addition, the completion and breakoff functions offer a better insight into the completion and breakoff development rates, allowing the researcher to make a betterinformed decision as to whether the survey requires any modifications or not.