Abstract Item response rates frequently serve as indicators of data quality and potential nonresponse bias. However, key variables from surveys, such as total household income or net worth, are often composite variables constructed from several underlying components. Because such composite variables do not have clearly identifiable response rates, inference on the data quality of these key measures is more difficult. This article proposes three new methods for aggregating data on response rates across questions to create a measure of item response for composite variables. To compare the three methods and illustrate how they can be used (both individually and collectively) to investigate data quality, I analyze item response for net worth in the Survey of Income and Program Participation (SIPP) and the Survey of Consumer Finances (SCF). These new measures provide detailed information about net worth estimates that would be difficult to assess without an item response aggregation method. Overall, these new item response rate methods provide a new way of describing data quality for key measures in surveys and for analyzing changes in data quality over time.

中文翻译：

---

复合变量的项目响应率

摘要项目响应率通常是数据质量和潜在的无响应偏差的指标。但是，来自调查的关键变量，例如家庭总收入或净资产，通常是由几个基本要素构成的综合变量。由于此类复合变量没有清晰可辨的响应率，因此很难推断出这些关键指标的数据质量。本文提出了三种新方法，可汇总问题之间的响应率数据，以创建针对复合变量的项目响应的度量。为了比较这三种方法并说明如何使用它们（分别和集体）调查数据质量，我在收入和计划参与度调查（SIPP）和消费者财务调查（SCF）中分析了净值的项目响应。这些新措施提供了有关净值估计的详细信息，如果没有项目响应汇总方法，这些信息将很难评估。总体而言，这些新的项目响应率方法提供了一种描述调查中关键指标的数据质量以及分析数据质量随时间变化的新方法。