Strategies to Retain Hispanic Sample Members in a Longitudinal Study and Effects on Survey Responses
*Bryan Bell Rhodes, RTI International
Keywords: Hispanic, locating, longitudinal, non-response bias, survey
Both anecdotal and empirical evidence show challenges associated with retaining certain racial/ethnic groups in longitudinal studies. Researchers speculate that factors such as living in a cell-only household, mobility, and citizenship status may affect retention of Hispanics. This paper will (1) examine the effectiveness of methods used to locate Hispanic study participants in a longitudinal study and (2) determine whether high-cost methods effectively reduce survey bias by measuring differences in demographic characteristics and key survey measures by method of location.
Data will come from a baseline and follow-up telephone survey. The former was completed about 4 years ago with more than 2,700 mothers of newborns, representing a cross-sectional sample drawn from birth records in Oklahoma (African Americans, American Indians, and Hispanics were oversampled). Several rounds of panel maintenance were conducted, and a follow-up telephone survey with the same respondents will be completed in June 2011. Data from the surveys will be used to evaluate whether having assets—in this instance, savings for the newborn’s college education—has an impact on child and family-well being.
To locate Hispanic sample members, this study used a variety of low- and high-cost methods (e.g., mail, telephone, proprietary data bases, other named contacts, in-person). The paper will first examine which methods were most effective in reaching Hispanic study participants relative to other racial/ethnic groups. The paper will then determine whether responses to key survey questions from Hispanic sample members located through more expensive methods (such as visits to the sample member’s last known address(es)) differ from those located through less expensive methods (such as outbound calls)—and determine whether factors such as country of birth, language spoken in the home, and employment status act as mediating factors.