This model was a first go at forecasting cocaine trafficking seizures by the DEA. The data is stride and the model was one of the most accurate I ever put together. The data parameters were set by myself in conjunction with agents at the DEA. It forecast cocaine trafficking seizures for 50% purity or greater for 5 kilos or more., the absolute count of them. The predictors are important and have more to do with the way DEA agents manages their informants than market forces, as the max on small seizures used in the model was capped at 4 ounces as the charges are different for individuals caught with over 4 ounces. Everything is in log differences. The Macroeconomic variables in the model are the M1 currency component, the unemployment rate and GDP. All data was aggregated to quarterly data and the M1 and unemployment are averages of the monthly time series. For the small predictors both the count and total netgrams from 0.1 grams to 112 grams are used. The out of sample forecast error MAPE is under 6 for the model. Enjoy!

​

The Predictors Do File:

https://drive.google.com/file/d/1V7d3kMV8D_KM4QK-MD9lJk7nvooYyKS0/view?usp=drive_link

​

The VAR:

https://drive.google.com/file/d/1xqaOmE8THm11hP9r1nHXjRgDdpq_S4Sr/view?usp=drive_link

​

The Data:

https://drive.google.com/file/d/1oM4denf4HvrQNzPjI6T3fyyZ1sZh4fGn/view?usp=drive_link

https://drive.google.com/file/d/1WGd-n4MdoICJ_kQz07ca9J0rgCkhUnrw/view?usp=drive_link

https://drive.google.com/file/d/1Mw1jT76oMfJWzMkvowBgHGJzZ4jzc9xd/view?usp=drive_link

https://drive.google.com/file/d/1D-k6KUmRnjsGKozvI24JU9hnLWXRJX4D/view?usp=drive_link

https://drive.google.com/file/d/1unyfp_gzRQ9-u6gY14OQ1Suuceagb7iZ/view?usp=drive_link

https://drive.google.com/file/d/10i0zJHgjLbHFDh1voZKAZVk5mDjWCtXa/view?usp=drive_link

https://drive.google.com/file/d/1mQiSb_fns17zSqz-o7bi7x47ORkuHSXj/view?usp=drive_link

​

​

​

​

​

​

​