Indirect effect of income on Survey completition

#1
HI Guys,

I conducting an analysis on competition of a survey: completed surveys vs not completed.This would be my DV. I have two modes of administrating my survey: the old form (paper) and the new one - via Ipad (Isurvey). My findings suggest that clients who completed the survey using IPAD are 6 times more likely to complete the survey (compared to paper). When I introduced my income control, my odd ratio increase from 6 to 12. I know that income is affecting the relationship between mode and competition of survey. but I dont know how explain this. What is going on here? How can I explain this?

Thank you in advance.
Marvin
 
#2
Actually, you don't know that income is affecting the relationship between the method of the survey and whether or not it was completed. Correlation does not equal causation.

I would start with the facts that iPads are expensive and electronic forms are perceived by many to be easier to complete. You might hypothesize that richer people have more access to the easier method. However, I think it would be very important to examine your definition of "completed." I don't think you're actually looking at "completed" surveys, you're looking at completed AND SUBMITTED surveys, right? You don't actually know how many people "completed" a paper survey but didn't bother to return it. (Which is why many people prefer electronic forms -- eliminates the hassle of returning the hard copy.)
 
#3
Well let me explain; we fund all these agencies. these agencies are supposed to administer our survey to all their clients in a paper or ipad format. Complete survey is operationally defined as answering all the questions of the survey (26 questions). Clients can skip the questions in the Ipad version of the survey. I just want to understand this interaction effect.
 
#4
You don't have enough information to determine the interaction between income and survey completion. All you know is that there seems to be a difference. You can't say why.

Now if the agencies either hand you an iPad or a paper survey, have you fill it out right there, and then return the completed survey -- maybe then you have something. But a "please don't forget to take our survey" approach introduces way too many variables that you can't control for. And if you're relying on people to use their own iPad, then that would really tweak the results (IMHO).