How do you assess for mode bias and scale effects when designing survey questionnaires? We typically educate clients on the potential bias of one mode over another (online versus paper surveys for example) and the bias a scale can have when designing questionnaires.  A client recently asks me how we could test for this bias and I am looking for ideas? 
 A: If you want to assess scale or mode effects in surveys you have to set up an experiment. Split your sample in two groups, and interview each group using a given mode. For such an experiment to be valid, you have to be able to control your environment as much as possible. At the end you have to be sure that possible differences in the outcomes can be attributed to the interview mode. This can be achieved if you assign the interview modes randomly. In the experiment, the choice of the mode should not depend on internal or external factors. Internal factors, are mainly the profiles of the interviewees and external factors are related to the environment the interviewees live in.
As @rolando2 has pointed out, there are papers on mode effects, and it could be worthwhile to have a look at that literature. 
A: For testing, the best method would be to include both modes in one project and compare the results. This is commonly done when switching from  phone to online methodology, when sufficient funds are available. 
You can also test sequentially, which is cheaper, but this always leaves the question of whether any differences are due to methodology or an actual change. 
