In survey research those that are less engaged are less likely to participate in the survey process and those who do but fail to provide quality responses drive measurement error.
In this analysis the metadata from 1100 diverse studies emanating from a great variety of online sources as well as covering an abundance of sources was examined to determine the driving factors of respondent engagement.  
There has been considerable discussion as to the influence of varying structural designs within a questionnaire that might improve respondent engagement.   Surely, instruments replete with grids that drag on over forty minutes beg the questions as to the quality of responses that we are eliciting.
In this work, we find that the source of the sample has great influence, only exceeded by the affinity that the respondent has toward the subject matter under study.   The impact of these two variables is so dominating as to swamp out structural changes within the questionnaire as a factor.
Interestingly, open ended questions appear to discourage respondents more than might have been previously thought.   Perhaps it is the change in pace that an unstructured question might represent, from an endless sea of grids, but it is the demand for placing answers in their own language that appears to discourage respondents most.
Here we treat a large number of possible variables and present our effort to create a predictive model of engagement.   It would be useful to know the impact that various drivers of engagement have on our research results and how structural changes that we control might create a more engaged respondent.