Energy paper (Calomiris, Çakir Melek, Mamaysky 2021) code

Step 1: Construct variables

• TextProcessing/ -- Reuters archive analytics
• VariableConstruction/ -- code to construct dependent and explanatory variables
• Analysis/ -- SDF construction
  • other parts of this (OOS runs tests and and outliers plots) largely deprecated (as of 7/2/2024)
• Reconciliation/ -- code to compare new topic model (circa 2024) to the prior topic model (from the KC Fed working paper version)
  • the code in the Reconciliation/ directory compares some files from 2023/ to files in 2020-11-16/
  • see description in the data section below

Step 2: Run analysis, first in-sample, then OOS

• InSample/ -- all the in-sample forward selection and bootstrapping code
• OutOfSample/ -- various out-of-sample tests

Partial output of analysis:

• data/ -- data generated after all the processing steps, which is then used in the IS and OOS analysis (after Step 1 is done)
• Intermediate data generated during Step 1 are stored in /shared/share_mamaysky-glasserman/energy_drivers/
  • README file on the grid contains additional information about directory structure
  • 2023/ -- contains the intermediate data for round 2 of MS submission
    • contains the rolling topic model intermediate data
    • contains the final OOS analysis results
    • also contains a new version (circa 2023-2024) of the full-sample topic model from 2020-11-16/ (KC Fed WP version)
  • 2020-11-16/DataProcessing -- contains intermediate data for the initial MS submission
    • info/ contains the monthly lists of articles with text stats, like sentiment, entropy, topic allocations
    • Louvain/ contains the topic model, i.e., words in topics
    • topic_allocation/ shows headlines of articles and their topic allocations in monthly files
  • Any other directory in the shared grid folder is deprecated and no longer used (as of 7/2/2024)