The Use of Subjective Information in Statistical Models
研究了将自我报告的主观信息作为预测变量时出现的问题，指出其社会建构性导致内生性，需在结构模型中明确建模，并以健康自评为例说明群体差异的影响。
This paper examines problems that arise in using self-reports of subjective information as predictors in mathematical models. The analysis demonstrates that the social construction of such subjective information makes its use in comparative analysis problematic since both its accuracy and the outcome for which it is employed as a predictor are influenced by the respondent's culture and social location. We argue that subjective information is socially and culturally constructed and, by definition, endogenous in models in which self-reports are used as predictors. Methodologically, this requires that the endogenous status of subjective information be explicitly modeled when it is used as a predictor in structural models. As an illustration of the substantive consequences of group-specific response patterns, we examine the use of self-reported health status in comparing the health levels of Mexican-Americans, blacks, and non-Hispanic whites.