NSF SAI: A Linked Longitudinal American Community Survey Sub-Sample to Understand Commuting Behavior

A Linked Longitudinal American Community Survey Sub-Sample to Understand Commuting Behavior Travel behavior research provides the foundational knowledge necessary to understand how and why people travel and provides the basis for making decisions about transportation infrastructure investments. The resulting knowledge informs the development of the travel demand models used forecast the costs and benefits of proposed transportation projects. It also provides transportation planners and decision- makers with a more general understanding of the types of investments that are likely to be effective at advancing their goals. Planners are usually concerned with understanding the effect of a change to the transportation system, but our data typically do not observe those changes. Instead, the most important data sets in travel behavior research are cross-sectional household travel surveys. These surveys collect a rich inventory of the daily travel patterns for people in a sample of households and provide the data necessary to estimate econometric models of travel behavior. However, there is a risk that the correlations estimated cross-sectionally may be spurious. For example, data show that people who live in more urban, walkable neighborhoods tend to drive less, but that could occur either because the residential built environment causes people to drive less, or because people who wish to drive less choose to live in more urban, walkable neighborhoods. Research into this residential self-selection problem has used statistical controls to separate the effect of the built-environment from self-selection while acknowledging the use of longitudinal data as a superior means of answering the question. Specifically, when the same household is observed at different times in two different locations, or when the household stays in the same location and the built environment around it changes, researchers can more readily separate the effect of the built environment itself. Unfortunately, longitudinal surveys of travel behavior remain rare in the United States (US), largely due to the cost and difficulty of reaching the same respondents repeatedly. Where such surveys exist, they tend to feature small sample sizes or focus on specific neighborhoods. The limited availability of panel surveys of travel behavior severely hampers our ability to answer some of the most fundamental questions in transportation research, and to generalize the results of studies that do. Fortunately, a data set already exists that provides unique opportunity to address this longitudinal data gap for perhaps the most important travel segment—work commutes. The American Community Survey (ACS) collects data from about 2 million households annually, including the home location, work location, number of vehicles owned, usual commute mode, and usual commute time. The ACS is currently available from 2005-2021. For this study, we will use the ACS restricted-use microdata housed in the Kentucky Research Data Center (KRDC), which one of 33 Federal Statistical Research Data Centers. In contrast to its public-use version, the ACS restricted-use data include both the detailed location data essential for transportation research and a Protected Identification Key (PIK) that is unique to each person, analogous to a Social Security Number. While it is not designed as a longitudinal survey, the sampling rates suggest that about 160,000 addresses sampled each year will also have been sampled in a previous wave of the ACS—an order of magnitude larger than most household travel surveys. In this study, we propose to use the PIK to identify individuals who responded to the ACS in multiple years and link those records to create a longitudinal subset of the ACS. We will further link the ACS records to the restricted-use version of the Longitudinal Employer–Household Dynamics (LEHD) data, which use unemployment insurance records to determine the home location, employment location and earnings information for 95% of American workers. The LEHD are available annually with data available in most states for the years 2002-2019. The result of our linkages will be a large and nationally representative longitudinal sample of commuting behavior over a period of 15+ years. We will make these data available to other researchers through the network of Federal Statistical Research Data Centers, enabling a potentially transformative advance in our research capabilities. Further, we propose to use these data to study three important, but vexing questions in transportation research: 1. By how much can changes to the residential built environment reduce car commuting? 2. Does increasing access to jobs lead to higher worker earnings? 3. When low-income workers are displaced from transit-rich areas, how are their commutes and their earnings affected? Our results will provide new evidence on how to effectively design a sustainable and equitable transportation system. We will leverage our existing relationships to share these results directly with those stakeholders who evaluate and make key decisions about transportation infrastructure. In doing so, we will engage and train students from under-represented backgrounds in evidence-based policy making, with the goal of showing them that they belong in the room as these important decisions are made. Together, these proposed activities will serve to strengthen American transportation infrastructure at a critical moment.