Relative and absolute bias - Likert-type data

#1
I'm working on a data set which was collected through a questionnaire by a certain regulatory institution. 50 institutions (whole population of that type of institutions) answered a series of questions (100 questions). The answers were collected via a Likert-type form (5-items, anchored). The institutions were asked for their opinion on relevance (first 50 questions) and usefulness (last 50 questions) of certain legal requirements. There were no reversed questions.

It is obvious that there could have been a significant bias because the institutions might have been eager to answer in the way which would, in their opinion, please the regulator. Furthermore, it can be assumed that the bias, for all the institutions could have the same direction, although intensity might vary. However, the data is intrinsically highly valuable (it almost certainly could not be obtained by a non-regulatory questionnaire) – hence I would like to get the most of it. I’m planning to perform some simple correlation analyses (interpreting data as ordinal).

I have a hunch that although there probably is an absolute bias
  • (e.g. an institution might have answered more agreeably to all the posed questions therefore “compressing” the answers to one side of the scale),
the order/ranking might preserve true opinions.
  • (e.g. if the institution thinks that legal requirement A is not useful, and that legal regal requirement B is somewhat useful, it might answer that A is somewhat useful and that B is very useful).

Are you aware of any research that would corroborate (or dispute) this hunch? Do you have any other ideas on how to deal with this (potential) bias? I do not expect the data to be normally distributed.

Help will be highly appreciated!
 
#2
I see that there were many readers, but no replies to the thread. Do I need to explain the environment in more details? Or something else?