A Study of Two-Phase Mail Survey Data Collection Methods

Abstract

Over the past two decades, landline random digit dial (RDD) telephone surveys have faced declining coverage and response rates. The availability of sampling frames based on U.S. Postal Service (USPS) files has led survey researchers to turn to address based sampling (ABS) as an alternative to RDD. However, ABS does not have a prescribed data collection methodology, as the ABS sampling frame offers a variety of mode options, including the possibility of using mixed modes. As survey researchers have developed procedures for administering surveys selected by ABS, several data collection models have emerged. We describe and compare two of these models, including one that was used in a very large-scale field test for the National Household Education Surveys Program (NHES) in 2011. This field test offered the opportunity for methodological research to assess the specific approaches to be used, with an aim at developing “best practices.” To that end, several experiments were embedded in the NHES:2011 Field Test. We describe those experiments and present some key findings. Relative to landline RDD, we have found the two-phase mail-based ABS approach described here yields much higher coverage and response rates at about the same cost per complete.