Saturday, December 15, 2012

Are we ready for a single-frame cell phone survey design?

It feels as if in this rapidly changing social and technological environment, surveys methods are merely trying to catch up with changes. It took a while for RDD telephone surveys to incorporate cell phone numbers. It took more time for researchers to use sample optimization formulas for allocation of sample across frames.

What if we try to look at least a step ahead. If one plots the landline and cell phone service for adults in the US from Blumberg and Luke (2012) based on NHIS data, there is a clear linear trend. More importantly, there are two clear trends: increasing cell phone service, and just as importantly, decreasing landline telephone service. Another key observation that can be made from the graph below is that we are rapidly approaching almost complete coverage of the telephone households through cell phones.

If we use an allocation optimization formula, we would fail to detect a situation in which the optimum design is not a particular dual-frame allocation, but rather, a single-frame cell phone sample design.

This is what I did with Ben Neely and presented at JSM earlier this year (to appear in Public Opinion Quarterly soon), projecting what the population may look like in 2013 and used simulation on collected survey data to evaluate different dual-frame and single-frame sample designs, constrained to the same interviewing cost. Indeed, for some estimates the optimum next year, in terms of mean square error, would be a single-frame cell phone. It depends on many factors, many of which need to be explored further, but a single-frame cell phone design should be considered for future surveys.

Why could this be the case? The two main factors are the increasing coverage of the telephone population through cell phones (in the above figure) and the loss in efficiency in the dual-frame estimators (if done well). There is a tremendous loss in precision due to weighting in the dual-frame RDD survey design, even when the allocation of sample to each sampling frame is almost proportional to the telephone service split in the population, as shown in our results below.