Future Work and Limitations

The scope of this project warrants us to consider the limitations of our approaches and outcomes. It requires us to constantly evaluate our methods in the context of the public good outcomes we’re trying to accomplish. This leaves room for improvements on this ongoing project as we hand it off.

• There is no clear ground truth to serve as a benchmark for our results. We can compare the naive definition to the Census, and use Census information for other approximate comparisons, but there’s no way to know whether we get any households “right.”
• We have more information on some types of households than others, depending on which people interact more with these government agencies. For example, families receiving assistance from DSHS anti-poverty programs may be disproportionately represented, while people who don’t vote or have driver’s licenses may be missing.
• We have limited information on children, so our household sizes likely undercount the actual number of people in many cases. For example, some of our one-person households (which we overcount according to the Census) are single parents. And, we haven’t leveraged what little information we do have about children, in DSHS assistance units and birth certificate records, to link people into families.
• Our approaches to defining households need to be expanded to incorporate more information that would provide a diversity of households. We began with address and last name information, but we can get more specific in our definitions by adding other variables, including for more granular Census comparisons or probability calculations.
• Orienting to the WMLAD data and determining relevant datasets and variables utilized much of our summer, shortening our time to work on approaches. We leave behind a database with a clear schema to hopefully shorten the onboarding period for future - researchers. Our approaches to household creation should be evaluated with more different methods. Some of those implemented already include comparison to each other and to the Census, and mapping.