Estimating Trends In Mental Distress Using Survey Data

One way of examining how mental health has changed over time is to use information available in large national surveys. While these survey-based measures are valuable in principle, and have been widely employed in prior research, a number of challenges may limit their practical usefulness or require considerable care in their application. This study uses data on U.S. 18-64 year olds from five major household surveys over the period 1999-2019 to investigate three specific questions. First, do different measures or the same indicator from different data sets consistently show how mental distress has changed over time? Second, how does the performance of single-item measures compare to those of psychometrically validated screeners based on multiple questions? Third, what are the benefits or potential costs of using different functional forms of these measures? The answers to these questions are as follows: 1) the estimated trends vary, often dramatically, across measures and, for the same indicator, across data sets and, as a result, some proxies provide a misleading understanding of how mental health has changed over time; 2) the single-item measures do as well or better than those based on multiple questions; 3) no single functional form demonstrates clear superiority over the others. After demonstrating these points, lessons are elucidated for researchers attempting to use survey data to examine trends in mental health.