Data and Models

This is an evergreen post. It’s linked from the header, and will be updated periodically.

Lambda School Incentives

Google Sheet
Post: Lambda School’s Incredibly Naive Incentive Alignment
Description: Modeling Lambda School’s ISA incentive structure with consideration of taxation. A fully functional California income tax calculator.
Findings: There is only a brief window where student and LS returns grow in tandem. LS is mostly not aligned with students, or actually incentivized against their interests. Based on student outcome data, only around one third of LS students fall within the incentive alignment window.

Independence Referendums

Google Sheet
Post: Base Rates on Secession
Description: Historical data on Independence Referendums from Wikipedia subjected to sensitivity analysis.
Findings: The rate of referendums per country per 4 year period varies by 4x depending on when you start counting. It’s at least 0.01, and at most 0.04 if you start counting right as the Soviet Union collapses.

TSMD and Intel R&D Spending

Google Sheet
Post: Isolated Demands for Rigour in New Optimism
Description: I extend the dataset from Bloom et al.’s Are Ideas Getting Harder to Find? to include 2016-2019 data from 10-K filings, adjusted for exchange rates
Findings: Intel nominal R&D peaked in 2018. TSMC nominal R&D continues to increase exponentially.

Herd Immunity Guesstimate

Google Sheet
Post: Contra StatNews: How Long to Herd Immunity?
Status: Out of date, useful as a template for future modeling.
Description: Toy model of time to herd immunity based on vaccine acceleration, cases per day and different immunity thresholds.

Giving What We Can Growth

Google Sheet
Post: Why Hasn’t Effective Altruism Grown Since 2015?
Description: Scarped data for GWWC member count, dollars pledged and donations actualized.
Findings: Growth has continued, growth rate peaked in 2015.

Open Philanthropy and Good Ventures Allocation

Google Sheet
Post: Why Hasn’t Effective Altruism Grown Since 2015?
Description: Grant databases from Open Philanthropy and Good Ventures aggregated by focus area and year.
Findings: Both increased rapidly from 2015 to 2017, and have sort of stagnated since. Global Health & Development is around 34% of all Open Philanthropy giving. Scientific Research is another 16%.

Substack Author Experience

Google Sheet
Post: How to Become Famous on Substack Overnight (in Ten Years)
Description: Top 25 free Substack authors from the old leaderboard and the time they’ve spent on Substack. Years of previous experience and notable work for the top 10.
Findings: Top Substack authors have around 10 years of prior experience in research or writing.

Byrne Hobart’s Portfolio

Google Sheet
Post: The Byrne Hobart Portfolio
Description: Scaped stock ticker disclosures from Byrne Hobart’s The Diff. Historical prices at the time of disclosure and performance since.
Findings: Subject to some assumptions, this portfolio dramatically outperformed hedge funds.

Constitutional Amendments

Google Sheet
Post: Wake Up, You’ve Been Asleep for 50 Years
Description: Proposal date of constitutional amendments that were later successfully ratified.
Findings: There were long gaps with no proposals leading to ratification from 1804 – 1865, 1870 – 1909 and 1972 – present.

Founders of Top 10 YC Startups

Google Sheet
Post: Replying to Robert Wiblin on Young Rationalists, Empirical Estimates of Golden Handcuffs
Description: Investigation into the backgrounds of top YC founders as ranked here.
Findings: Median age at founding was 27, median age now is 37. Median age of startup is 9 years. Founders attend elite universities, but don’t typically work at large or prestigious companies.

Google Alumni Estimate

Google Sheet
Post: Empirical Estimates of Golden Handcuffs
Description: Based on historical headcount data and simulated churn, I estimate how many people have ever worked at Google up through 2010.
Findings: Although Google’s employee count in 2010 was 24,400, I estimate there were an additional 25,400 ex-Google employees.

Model of Correlated Returns

Google Sheet
Post: Correlated Returns are Insufficient for True Alignment
Description: Modeling correlated returns in domains without proper incentive alignment, followed by seemingly correlated returns adjusted for opportunity cost.

Miike vs Tarantino

Google Sheet
Post: No One is Even Trying
Description: Rotten Tomatoes critic and audience score for the top 10 movies of Takashi Miike and Quentin Tarantino.
Findings: Miike’s top 10 movies are as well reviewed as Tarantino’s. Meanwhile, Miike has made 10 times as many movies overall.