There Goes the Neighborhood: Localized Economic Effects of Marijuana Retail Sales in Colorado

You can see the research paper resulting from this work here.

Recreating the Results:

Here is how to generate the result files:

1. Download the zip file from current repository.
2. Open Average Growth Rate Comparison.ipynb, click Cell -> Run All on the menu bar. Then, a group of histograms showing the general trend of difference between control and treatment group will be generated.
3. Open Linear Regression .ipynb, click Cell -> Run All on the menu bar. Then, separate linear models for each outcome variable will be created and fitted. Their corresponding coefficients and p values will show up and be exported to Linear Regression Results.csv. In the same notebook, a group of scatter plots showing how outcome variable changes vs. different cumulative store months (outliers removed) can also be seen.

Recreating the Data:

If you'd additionally like to recreate our data gathering processes, the original CO state license information can be found here. The get_data.py file includes basic instructions and all of our code used to manipulate these original manually-collected records, make API calls for Google Maps address matching and geocoding, and transform and save the resulting data. The get_census_data.ipynb notebook can be run in full to collect and save relevant tract-level Census variables.

A Guide to the Files:

Core Files

tract_similarity_match.ipynb

• Divide control and treatment groups, and apply propensity score matching.

Data_Prep_DID.ipynb

• Prepare data needed for later Difference-in-Difference analysis.

Average Growth Rate Comparison.ipynb

• Compare control and treatment groups.

Linear Regression.ipynb

• Linear regression model to explore the relationships and p-values, also check the validity of the results got from 'Average Growth Rate Comparison'.

Linear Regression Results.csv

• P values and coefficients from the linear regression model.

get_census_data.ipynb

• Use [Cenpy] (https://github.com/cenpy-devs/cenpy) to get relevant Census economic variables for relevant years

get_data.py

• Wrangle existing data from co_cannabis_stores.csv, add addresses from Google Maps API, add census tract and county numbers, output both tract-level and store-level data

Visualizations

Growth Rate Comparison.png

• Histogram (result image) generated from 'Average Growth Rate Comparison'.

Scatter Plot.png

• Scatter plots of how outcome variable changes vs. different cumulative store months, outliers removed.

visualizations/Tableau_images_screenshots folder

• Screenshots of all visualizations that created using

Tableau files folder

• Proof dispensary datasets and the imported spatial data and graphs generated by Tableau