Language Shift Modeling

I worked on calculating rank correlation coefficient values for generational changes and local census tract changes. I also reconfigured the project structure to filter by metro, tract, year, generation, and then language. I then calculated aggregated statistics like migration and speaker populations for each generation and tract. Hopefully, I’ll be able to apply these to some functional interactive app in the coming weeks.

I shifted the project focus since last time. Namely, I’m focusing on language shift, and upon realizing that there is no PUMA-level data for the decennial census, reduced the data to only use the American Community Survey. Today, I mainly worked on filtering the ACS data through some Rust code (example JSON output is shown below).

The code splits up the raw data (which includes all surveyed participants) into chunks split up by location and generation for each language being studied (currently the 5 most spoken languages in California). I then am able to calculate aggregated statistics like the immigration rate for each year within a generation, total language speakers per year, and the linguistic diversity of a given region along with the linguistic diversity of each constituent PUMA (as shown below).

I downloaded some Census data to understand the format, and started drafting some Rust structs and types to figure out how to store and manipulate data efficiently.

I’m working on my first project! This is so exciting. I can’t wait to share more updates as I build.