Analyzing how crime rates may have been affected by the Legalization of Cannabis
This project was inspired by an article I read on Forbes:
In summary, the article alleges that criminals are crossing state borders to snatch up all the Cannabis they can and transport it back to their home state. In particular the author mentions a rise in "home invasions, violent crimes and robberies" perpetrated by people coming in from other states. The goal of my project is to test these claims of "Weed Pirates" by looking at crime data from Seattle and Portland.
To observe the immediate effects of legalization I began by finding crime data from cities that had fully legalized as test groups and data from cities that had medical programs as controls. After digging through several cities I decided to use Seattle as my test and Portland as a control.
This focus of this project is weed piracy, so I decided to look at crime types I associated with piracy: Homicides, Thefts, Robberies, and Burglaries.
I used a One-Tail Z-Test with the following hypotheses to test the articles claims for each crime category:
Critical Value = 0.05, Bonferonni Corrected to 0.00625
H(null): There is no significant rise in violent crime rates coinciding with legalization of Cannabis
H(alt): There is significant rise in violent crime rates coinciding with legalization of Cannabis
(note: crime rates for Portland in 2015 are imputed averages from 2010 to 2014)
| City | Homicide | Theft | Burglary | Robbery |
|---|---|---|---|---|
| Seattle | 0.472 | 0.536 | 0.999 | 0.5 |
| Portland | 0.999 | 0.974 | 0.00008 | 0.425 |
From our graphs and p-values above I do not find enough evidence of increasing rates of violent crimes to accept our alternate hypothesis. Instead, I find more evidence that violent crimes trend downward once legalization occurs.
The most challenging part of this project was sifting through crime datasets and finding a city that could serve as a viable control against Seattle. In addition, each city had a slightly different method of data entry, so cleaning followed a similar but different process each set. Furthermore, my data only covers reported crimes - if a crime gets unreported or miscategorized it will affect our analysis.
While I found some significant results, there are a few other directions I'd like to explore:
- Identify more control cities to compare against Seattle.
- Apply the same process to cities that participate in a different part of the product cycle.
- Dig deeper into the data I have to determine if proximity to dispensaries has an effect on neighborhood crime rates.
- Apply clustering to crime data to see if perpetrators were high during arrest. (Only viable for certain datasets.)
- Python
- Numpy
- Pandas
- MatPlotLib
- GitHub