Research tooling · Personal
otter
An econometrics toolkit for Python: regressions and their diagnostics, forecasting, and result tables ready to go into a paper. Simulation is one part of it, not the point. Built on top of my data loaders and used for work on wealth concentration and monetary policy transmission.
Why it exists
The questions I work on are how wealth concentration affects consumption and investment, and how it changes the transmission of monetary policy. These are distributional questions, where aggregate series obscure the variation of interest.
Investigating them requires repeated simulation over real data with explicit uncertainty. In practice each attempt began with the same setup work: fetching series, aligning frequencies, and reconciling vintages.
What it does
It is the third layer of a three-part stack. fred-loader and census-loader handle access, turning federal APIs into readable requests. otter sits above them and handles the modeling: Monte Carlo simulation and statistical machinery that takes the loaded data as its input.
Keeping access separate from modeling keeps the loaders generally useful and keeps the framework focused on the statistics rather than accumulating fetching code for each new source.
Where it stands
Ongoing personal research, currently mid-refactor: it grew alongside the loaders and is being tightened as their conventions settle.
The published work it supports is The 100% threshold, which vendors this framework in for its simulation machinery.