Innovative Approaches for Small Area Estimation Methods

• In Event: Small Area & Geographic Estimation with New Data and New Methods

Abstract

Introduction

All politics are local. The same argument could be made for every intervention. However, tailoring interventions to address specific situations depends on the availability of localized metrics and measures. Often, these measures are not granular enough to capture the nuances and the contextual realities of the deficits these interventions are meant to address. The present work addresses this shortcoming. Our approach addresses critical gaps in localized data necessary for informed decision-making, particularly at the neighborhood or sub-city level where many interventions are typically implemented. The innovative approaches we will be demonstrating in this talk are domain agnostic and could be applied to multiple uses - healthcare access, travel mode choices, residential energy consumption, and in proactively planning for impactful policy prescriptions particularly for relatively nascent services such as EV Charging or On-Demand delivery services.

Approach

Our innovation employs administrative and proprietary data and a variety of publicly available microdata files such as the US Public Use Microdata Sample datafiles; the Pulse Survey, regional and national sector specific household surveys and an approach that integrates iterative proportional fitting (IPF), data imputation methods and machine learning, to create a replicable and reproducible approach for generating synthetic population. Detailed profiles of the synthetic population will be generated by creating composite variables that capture the cohort or segment of population of interest. A significant benefit of this approach is that it is not data intensive, and it could easily be tailored to address niche populations. The talk will also include a trend analysis by presenting a multi-year assessment of these profiles.

Results’ Demo Site

We showcased our approach using a demonstration webpage https://seattle-travel.replit.app/ - that provides an overview of how these small area estimates are generated using Seattle as a case study. The page presents estimated travel mode choices, destinations, and volume of trips generated in a given period at the Census Block Group (CBG) level for Seattle, thus providing crucial insights that could be localized to specific geographical areas, and for designing targeted policy responses. Though this demo site is on transportation, the approach could easily be extended to multiple domain areas.

Conclusion

By borrowing methodologies from related domains and integrating auxiliary data sources, such as publicly available microdata sample files, this research offers more reliable and actionable insights. The refined level of estimation provided is crucial for city planners, municipal governments, and community-based organizations to track and respond to local trends effectively, ultimately fostering better-targeted interventions.

Authors

• TAYO FABUSUYI, University of Michigan Ann Arbor
  Presenting Author

• Yangyang Wang, University of Michigan Ann Arbor
  Non-Presenting Co-Author