Exploiting information asymmetry in Specialized Disclosures

This is a project to examine whether it is possible to use differences in information reported by companies in Form SD disclosures to do something cool. Not sure what exactly.

Current idea:

1. Create a dataset of every supplier or refiner (SOR) in 2023 using EX-1.01 of Form SD.
2. Identify SORs shared between companies (e.g. Tesla and Apple)
3. Exploit differences in information reported to do something cool.

Implementation

1. Download all FORM SD using datamule
2. Create a csv with columns 'filing_date', 'cik' (company unique identifier), and 'text' (text of the document)
3. Convert text to a structured dataset using txt2dataset
4. Explore data
   1. Force directed network graph of SORs and companies, coloring shared SORs differently
   2. Cluster Graphs to identify patterns in shared SORs (Turns out, almost every refinery is connected, so not needed)
   3. Treemap of SORS by country

Share exploration results

1. Linkedin
2. Twitter

Ask for help connecting with people doing research in this space. Ideally an expert who knows what to look for / what questions would be interesting to answer.