Respondent Characteristics That Are Predictive of Financial Item Nonresponse
Flicker, L., Ruppenkamp, J. W., Bibb, B. S., Riley, S., & Lindblad, M. (2009, May). Respondent Characteristics That Are Predictive of Financial Item Nonresponse. Presented at AAPOR 2009, .
Item nonresponse is a concern for researchers making inferences about target populations. Statistical imputation techniques assume that data are missing at random, yet that assumption is often violated. Item nonresponse has been shown to vary by subgroup for survey content related to sexual orientation (Gruskin, et al., 2001), sexual behavior (Kupek, 1998), drug use (Wolfe, 2003), health (Blaum and Liang, 1996; Slymen, et al., 1994), and income (Gruskin, et al., 2001; Turrell, 2000; Dengler, et al., 1997; Schräpler, 2004).
Moreover, some of the same respondent characteristics that affect unit-level nonresponse also contribute to item nonresponse. That may compound problems for data quality, especially for income estimation (Turrell, 2000). Education, age, and gender have been shown to independently contribute to item nonresponse (e.g., Gruskin, et al., 2001; Slymen, et al., 1994; Dengler, et al., 1997; Blaum and Liang, 1996). Elsewhere, gender has been shown to interact with other characteristics to affect item nonresponse (Turrell, 2000) or to have no independent or interactive effects on item nonresponse (Wolfe, et al., 2008). Race in some cases contributes to nonresponse after controlling for socioeconomic status (Gruskin, et al., 2001; Kupek, 1998).
The goal of the present research is to determine whether specific respondent characteristics contribute to item nonresponse for sensitive data. For this analysis, the missingness of financial data is investigated.