Share
Embed
Transcript
29 slides
-
Death, Taxes, and GUIs Anderson Frailey and Hank Doupe Open
Source Policy Center – American Enterprise Institute -
About Us • Hank Doupe ◦ Core maintainer: COMP, ParamTools
• Anderson Frailey ◦ Core maintainer: Tax-Brain 2.0, TaxData 2 -
State of Open Source Policy Analysis • Open Source Policy
Center, est. 2013 ◦ Open Models == Better Policy Analysis ◦ Better Policy Analysis == Better Policy • Incubates a wide range of open source policy models • Many other groups, like the CBO and NY Fed, have released code and data in recent years • Policy Simulation Library, est. 2018 ◦ Home to a number of open source models for policy analysis 3 -
Tax-Brain 2.0 • Integrates multiple tax models into one interface
• Built with COMP in mind to make the models available to everyone • Also works as a simple Python package for those who want to code 5 -
Tax-Calculator • Individual income and payroll tax microsimulation model •
Used by presidential campaigns, journalists, policy analysts, students, etc. • Backend of Tax-Brain 1.0 • Core Maintainers: Martin Holmer (Policy Simulation Group), Matt Jensen (AEI) 7 -
Behavioral-Response • Python package for partial-equilibrium behavioral responses to changes
in federal income and payroll tax system • Core Maintainers: Martin Holmer, Matt Jensen 8 -
Problems 1. You need to know all of the models
well to use them 2. There’s no way for non-technical users to take advantage of the models 9 -
15 COMP • An open-source utility for running and sharing
computational models • Empowers analysts who need to use and share results from these models, but aren’t able or willing to use the programming APIs • Provides access to models too resource intensive to run locally or that use sensitive data • Modelers share for free, and compute costs are paid for by users or is sponsored by a benefactor -
16 Why we did it • Tax-Brain 1.0 is sweet!
Every model should have its own webapp! -
17 Problems 1. No standard publishing process 2. Custom inputs
and outputs technology 3. Too much of the modeling process was in the web developer’s hands 4. Dependency hell -
18 Making of COMP • Standard inputs and outputs •
Three python functions: - Get the inputs - Parse and validate the user inputs - Run the model with the user inputs • Custom environments that are decoupled from the webapp -
19 COMP Tax-Brain Page loads Get Default Values views.py User
modifications Parse & validate inputs If errors: Display errors Else: Call Run Model Create tables, graphs, etc. Display Output {Default Values} {User Inputs} {Parameters, errors} {Tables, Graphs, etc.} -
COMP Road Map • Automatic publishing process • Richer model
inputs and outputs pages • Scalable infrastructure • REST API • Your models! 22 -
Tax-Brain 2.0 Road Map • General refactoring • Performance improvements
• Automated tax analysis reports • More models! 23 -
OG-USA • Overlapping generations model used for fully dynamic analysis
of tax policy reforms • Outputs macroeconomic effects of tax policy • Core Maintainers: Jason DeBacker (University of South Carolina), Rick Evans (University of Chicago) 24 -
Business-Taxation • US corporate and pass-through business tax model •
Core maintainer: Cody Kallen (University of Wisconsin- Madison) 25 -
How to Get Involved • Low hanging fruit: COMP website
design • Use the projects! We need feedback • Help with code architecture 28 -
Thank you. Contact Us Anderson Frailey Twitter/GitHub: andersonfrailey [email protected] Hank
Doupe GitHub: hdoupe [email protected] https://github.com/PSLmodels/Tax-Brain https://github.com/comp-org/comp