HS&B

ABSTRACT: For its recent field test for the First Follow-up of the High School and Beyond Longitudinal Study of 2022 (HS&B:22), NCES experimented with different data collection strategies for parent surveys. In the Base Year (2022–23 school year), parent response rates were the lowest among the 5 different respondent groups for the study. That is, they were lower than those of high school students, their math teachers, school counselors and school principals. Because parents provide important demographic data about their high schooler sample members, eliminating the parent questionnaire entirely would leave data gaps in key variables. A frequent reason for survey refusal reported to interviewers is that parents are "too busy" or "don't have the time" to participate. While an abbreviated questionnaire is often offered toward the end of the data collection window for NCES sample surveys as a refusal conversion technique, few studies have tried drastically shortening the questionnaire at the outset of data collection. Considering this, NCES explored options for dramatically minimizing the time burden requested to test whether reducing the burden might encourage more parents to respond.

To test procedures to boost response rates and/or minimize missingness for key variables, NCES experimented with data collection utilizing micro-questionnaires in paper-and-pencil interview (PAPI), computer-assisted telephone interview (CATI) and web-based interview modes. Results are discussed in terms of response rates, data quality and potential for bias. Implications for further implementation of micro-surveying sum up the discussion.

ABSTRACT: The National Center for Education Statistics (NCES) has been working in recent years to improve declining response rates in its sample surveys. Across multiple longitudinal and cross-sectional surveys, several methods of adaptive design have been tested. Using models to determine likelihood to introduce bias should sample members not respond, NCES has effectively discerned which sample members to target for special intervention, in order to boost response rates. However, it is clear that sample members' propensity to respond is not always strongly correlated with their likelihood to introduce bias if they do not respond, necessitating more research on tailored ways to encourage response. Incentives and other treatments thus become an important area of investigation. This presentation describes the adaptive design approach that NCES used in the High School Longitudinal Study of 2009 (HSLS:09) Second Follow-up Field Test, conducted in spring and summer, 2015. This field test examined the effectiveness of several types of incentive treatments to encourage sample member response. The comparative effectiveness of four treatment methods will be discussed: 1) prepaid incentives, 2) response-contingent incentives, 3) incentive boosts, and 4) an abbreviated interview. The paper will describe the results from the adaptive design experiment and the plans for implementation of the findings in full-scale data collection in 2016.

ABSTRACT: This presentation highlights some of the adaptive design strategies used in recent NCES longitudinal surveys. We review model selection, treatments, and results from the 2012 follow-up of the Education Longitudinal Study of 2002 (ELS:2002), the 2008-12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12), and the 2012–14 Beginning Postsecondary Students Longitudinal Study (BPS:12/14). We discuss the potential for using adaptive design to reduce survey nonresponse bias and note considerations for implementation.

ABSTRACT: Elise Christopher and Ted Socha of the National Center for Education Statistics will be discussing how responsive design (RD) models have been implemented in two of the Center's nationally-representative longitudinal survey data collections: the Education Longitudinal Study of 2002 (ELS:2002) and the 2008–12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12). Topics will include the following: (1) choice of RD models, (2) how each RD model was applied, (3) implications of using RD for monetary and non-monetary incentives, (4) substantive and paradata model variables, (5) dealing with missingness in substantive variables, and (6) what their preliminary results mean for nonresponse bias reduction.