How big are your banks? Because the ones in the U.S. are pretty big!
summary
Analyzes a 5+ GB FDIC balance sheet panel for Q1 1992–Q4 2015 using Python, Pandas, Scikit-learn, Bokeh with a GitHub Python scraper to bulk the raw reports. Finds Zipf’s law (α=2) is usually rejected; a power law with α≈1.9 fits the upper tail better, implying heavier tails, undefined theoretical mean, and much greater concentration. Reports number of FDIC banks halving (~14,000 → <7,000), largest bank share rising to ~12–13%, and the top 10 controlling >50% of sector assets.
AI-written summary · about
chapters (5)
- 0:00Banking sector simulation models
- 6:05US banking consolidation trends
- 13:55Power law models for bank size
- 28:33Summary of findings
- 29:45Q&A on European banks
related talks
45mInteractive Analysis of (Large) Financial Data Sets
57mPulling Back the Green Curtain: Data Forensics, Power BI, and Dataviz
40mSearching for Shady Patterns: Shining a light on UK corporate ownership
37mPanel: The Hidden Value behind Cryptoanalytics
20mA Deep Dive into Analytics Delivery Behind one of the Largest North American Banking Acquisition..
35mHow many dataframe libraries do you need to change a lightbulb?
19mStock data analysis with
42mUsing Python pandas for scientific Research