The purpose of this research project was to determine how much of an economic impact (if any) Amazon has on communities by analyzing economic and demographic data before and after the entrances of Amazon facilities. Work-to-residence origin-destination data provided estimations on the number of employees for various facilities and where these employees lived. Based on this data, there were two groups of Amazon facilities in California that stood out as having the best available data for further analysis. For each of these groups, employee residence was aggregated by census tract. Then, the census tracts with the most employees for each facility were identified. These tracts were compared to a group of control tracts derived from a synthetic control model. The treatment and control groups were compared using difference-in-difference models with various dependent variables such as unemployment rate, number of people employed, percent of families with income below the poverty level, and percent of households with food stamps or other SNAP benefits. Based on the difference-in-difference models, the unemployment rates for both groups were significantly smaller after Amazon opened its facilities.

One of the biggest challenges of the analysis was finding geographically and temporally granular economic data. Any analysis on company or local policy impact is very difficult without this data. Future analysis on this topic should consider several elements. First, Amazon facilities are endogenous in terms of both place and time. Second, a facility’s opening date is very difficult to define because a facility may take years after construction completes to hire all its employees. Third, other methods of finding employee residence may better indicate which local areas are most affected by Amazon facilities. Lastly, future analysis should consider other states in addition to California.