Measuring Social Policy Attitude Stability with Generalizability Theory

• In Event: Poster Session: Public Opinion & Ideas, Knowledge and Politics
  In Poster Session: Poster Session: Public Opinion

Abstract

Existing research in attitude stability focuses only on reducing measurement error of individual responses by averaging multiple survey items. Nonetheless individual responses can only be part of the sources of measurement error, interpretation of survey items, sampling interval between repeated surveys can also be important sources of measurement error. Therefore, existing method (classical test theory) that political scientists use was unable to differentiate whether the error variances are caused by individual responses, survey items, sampling intervals, or the interactions between them. This study introduces Generalizability Theory, a psychometric method, to further decompose the error variances into the variances that generated by items and sampling intervals. Moreover, this study uses Intergroup Conflict Panel Data, which consist of six waves of panel data that conducted among UCLA undergraduate students from 1996 to 2001. This study finds that 1) white students tend to have higher individual response variances, while Asians and Latinos tend to have relatively smaller individual response variances. 2) In sharp contrast, white students have relatively smaller variances caused by items, while Asian and Latino students have large variances that caused by items. Therefore, in this study I argue that due to different socialization and cultural traits, different racial groups tend to different attitude stability that caused by differing interpretations of survey questions.

Author

• Bang Quan Zheng, UCLA