We’ve seen over the last few years that there are some examples of successful use of paradata to gain efficiency in the survey data collection, and to alert us to errors that can happen. There was a time when we used these data and didn’t call them paradata, and they served that function back then as well. But we have seen that we still face serious challenges in using paradata in an analytical way. We also still see challenges with the tailored collection of paradata, and in particular the transfer across surveys and survey organizations. There is great opportunity here for funding agencies like the National Science Foundation (NSF) to help. So far we have failed to use the wide variety of paradata that are out there, or could be out there, and have mostly focused on non-response and measurement error. The U.S. Census Bureau has made efforts recently to combine cost information and paradata, and increase the modeling of paradata. A lot more could be done in that area too.
