This application proposes to integrate insights from psychology and neuroscience into choice models estimated from survey data in the population sciences. The central insight is that response time (RT) data that is either currently available or could be collected in surveys can be used to better model behavioral outcomes. While survey researchers tend to rely only on survey answers to model individual choice and outcomes, cognitive psychologists and neuroscientists rely both on answers as well as the process of answering questions. RT data plays a central role. For example, there is a well-developed literature in psychology that shows that RT is related to measures of general intelligence and to attention to a task and is related to whether a respondent is being truthful in his or her answer. Our work investigates whether models of human cognition in psychology and neuroscience that use RT as well as subject responses can better estimate behavioral models within a survey context. Our work recognizes that the survey context differs significantly from the laboratory context and hence factors about the survey context may also influence a subject's answers and RT. We propose to study the usefulness of RT data and the application of models from cognitive psychology for analyzing survey data with two case studies. The first will ask whether RT data is useful in assessing the truthfulness of answers to questions where a respondent's answer might meet with social stigma. We specifically are interested in underreporting of pregnancy and abortion and its effects on estimates of the impacts of teenage childbearing on women's socioeconomic outcomes. Second, we ask whether and/or with what modifications can well developed methods in cognitive psychology that combine a subject's answer to test questions and RT to those questions to better measure cognitive ability within a survey as well as the effort a respondent puts into the cognitive test. Should we be able demonstrate the value of RT data to population scientists, this will open new avenues of understanding behavioral relationships with data that already exists in almost all surveys that use computer assisted data collection methods. In addition, it would open existing population science datasets for use in cognitive psychology and neuroscience.
This project will use data on outcomes and response times (RT) to evaluate if RT data within surveys can better model behavioral outcomes. We focus on the use of RT data to better measure sensitive issues, such as abortion, within surveys and better measure cognitive and non-cognitive skills in test taking.
