Project Snapshot: PolicyEngine
PolicyEngine is a nonprofit that provides a free, open-source web app enabling users in the US and UK to estimate taxes and benefits at the household level, while also simulating the effects of policy changes. By combining tax and benefits data, PolicyEngine helps individuals and policymakers better understand the impacts of existing policies and proposed reforms, using microsimulation models built from legislation and enhanced survey data.
Organization Description + Mission
PolicyEngine is a non-profit that seeks to compute the impact of public policy for the world. Through their free, open-source application, they want to make policy information accessible, and help users understand policy effects at the individual and population level. By giving policymakers and residents greater access to policy rules and their effects, PolicyEngine hopes to promote more democratic policymaking processes.
Approach to Eligibility
PolicyEngine’s web app allows users in the US and the UK to estimate their benefits as well as their taxes at the household level. On an individual level, PolicyEngine can answer questions like, “Am I eligible for free school meals?” The platform can also evaluate complex policy proposals and reforms to simulate the impacts of policies for a household and society.
PolicyEngine also helps users in the U.S. and U.K. understand how changes to a policy might affect someone’s tax liability or benefits eligibility, and allows users to simulate the impacts of customizable policy reforms. Because many households pay taxes and also get benefits, combining tax and benefits estimates on one site allows PolicyEngine users to get a fuller picture of how different policies and policy changes may impact their household and society more generally. Combining tax and benefits information also makes the tool more relevant to users across the income spectrum.
Rules Structure + Technology
PolicyEngine started by building the OpenFisca UK and US microsimulation models. These models bring multiple sources of information into a unified data set, including rules from legislation, survey microdata, and administrative totals. PolicyEngine UK uses machine learning to enhance survey microdata (including reweighting data, for example, to account for underreporting of high or low incomes) which they then use to build their microsimulation models. They plan to extend this approach to the US in the coming months. These models feed into PolicyEngine’s free, open source web-app user interface.
PolicyEngine makes the information used on their site available to other developers via Python packages and a Web API. PolicyEngine also has an API explorer, which displays metadata for any variable or policy parameter in their system, including links to the legal code section that defines it, and allows users to search for these.
In the near future, PolicyEngine hopes to continue improving user experience, by incorporating features like a sign-in experience so that users can save policy simulations. They also hope to improve the tool’s mobile optimization. Currently, PolicyEngine’s simulation model does not incorporate macro-economic impacts or account for potential behavior changes in response to policy reforms; to address that they also plan to add extrapolation factors. Importantly, because PolicyEngine’s unified data set and rules engine are open source, they can be used for other projects.
How They Write Rules
PolicyEngine starts with legislation to understand existing laws on benefits eligibility and taxes. PolicyEngine is built on top of OpenFisca which is a free open source platform used to write rules as code. OpenFisca was developed by the French Government in 2011. Developers can work with policy experts to translate rules into code on the OpenFisca platform, or build applications that call coded rules in OpenFisca through a web API. (For more on OpenFisca for rules as code, see Cracking the Code: Rulemaking for humans and machines.) They can also update these rules as policies and regulations change. OpenFisca was initially built to create tools for predictive policy analysis such as LexImpact, a tool that allows members of France’s parliament to simulate possible changes in tax law to better understand their potential impact on residents. PolicyEngine’s model also enables them to include policy proposals for simulations on their web app.
PolicyEngine is now incorporating models for tax and benefits programs in all 50 states and DC, and enhancing their survey data for microsimulation. They are also working to add new benefits programs to their engine (including TANF, general relief, childcare subsidies, and housing subsidies). The project demonstrates the potential for rules as code approaches to enable evidence based policymaking and improve policy literacy.
