Chapter 3: The Open Source Revolution

In May 2011, a small team at France Stratégie—the French government's policy analysis agency—released something unusual: the source code for a tax and benefit calculator .^[1]

They called it OpenFisca. The premise was simple but radical: tax and benefit rules should exist not just as legal text but as executable code. Give the system a person's circumstances—income, family structure, location—and it would calculate their taxes and benefits. Change a parameter—a tax rate, an eligibility threshold—and see the effects immediately.

The code was released under an open-source license. Anyone could use it, modify it, extend it. The French government had decided that the logic of its tax-benefit system should be a public good.

This was the beginning of the open source revolution in policy modeling. It would take a decade to spread, and it remains incomplete. But it represents something fundamental: the idea that the rules governing citizens' lives should be not just publicly available but publicly computable.

Rules as Code

The concept went by various names: "rules as code," "legislation as code," "machine-consumable rules." The idea was always the same: laws and regulations should be expressed in a form that computers can execute, not just humans can read.

This wasn't new in principle. Tax agencies had been encoding rules in software for decades—that was what George Sadowsky had done at Treasury in the 1960s, what every government tax system did by the 2000s. But those implementations were proprietary, hidden inside agency systems. Citizens experienced the outputs of coded rules (a tax bill, a benefit payment) without access to the logic.

The open source revolution proposed transparency: publish the code, let anyone run it, enable verification and experimentation.

In 2018, New Zealand's Service Innovation Lab launched "Better Rules"—a collaboration between Inland Revenue, the Ministry of Business, Innovation and Employment, and the Parliamentary Counsel Office .^[2] The team spent three weeks translating legislation into Python code, demonstrating that rules could be drafted in both human-readable and machine-readable form simultaneously.

Estonia's Chief Information Officer called it "the most transformative idea" he'd seen .^[2] The OECD took notice, eventually publishing "Cracking the Code: Rulemaking for Humans and Machines" in 2020—a primer for governments on what rules as code could mean .^[3]

By 2022, OpenFisca had been deployed on four continents .^[1] France used it for Mes Aides, a citizen-facing benefits calculator. New Zealand built a rates rebate application. Tunisia, Senegal, Australia, Canada, and others adapted the framework for their own systems.

The OECD named OpenFisca's approach an "Innovation of the Year" at the World Government Summit. The European Commission recognized it as the most innovative open-source software in their Joinup program .^[1] A small French project had become a global movement.

The Promise and the Gap

The promise was intoxicating. If tax and benefit rules were published as code:

Citizens could check their own calculations. Rather than trusting that an agency computed their benefits correctly, anyone could run the same logic themselves.

Reformers could model alternatives. Policy proposals wouldn't require access to government systems. Anyone with a laptop could simulate how a new benefit structure would work.

Errors could be found and fixed. Open code meant open scrutiny. Bugs in benefit calculations—which affected real people's lives—could be identified by outside researchers.

Innovation could flourish. Nonprofits, journalists, and entrepreneurs could build applications on top of official rule logic, creating tools the government never imagined.

But between the promise and reality lay significant gaps.

Technical barriers. OpenFisca required Python programming skills. Most citizens—most policy researchers, even—couldn't write code. The rules were technically public but practically inaccessible.

Data problems. Microsimulation requires not just rules but data: a representative population to simulate. OpenFisca encoded the rules; it didn't solve the data challenge. Without microdata, you could calculate individual scenarios but not aggregate impacts.

Trust gaps. Even with open code, who would trust it? Governments were wary of unofficial calculations contradicting official ones. Citizens didn't know how to evaluate competing estimates.

Maintenance burdens. Tax codes change constantly. Someone had to update the models, track legislative changes, fix bugs. Open source meant anyone could contribute; it didn't mean anyone would.

The gap between OpenFisca's elegant framework and actually usable policy analysis remained wide.

A Researcher's Frustration

In 2019, a researcher named Max Ghenis founded the UBI Center, an open-source think tank focused on universal basic income policy .^[4] The mission was to produce rigorous research that could inform UBI debates—research that anyone could verify because all code and data would be public.

The challenge was immediate: UBI proposals interacted with the entire tax and benefit system. To model a $1,000-per-month UBI, you needed to account for how it affected income taxes, benefit phase-outs, work incentives. You needed to trace effects across the income distribution. You needed data on real households.

Ghenis discovered the tools that existed: Tax-Calculator for US federal taxes, OpenFisca for the framework. But putting them together into a usable research platform was frustrating. Tax-Calculator focused narrowly on income taxes. OpenFisca-US was nascent. Neither had the web interface that would let non-programmers explore policy options.

And for state-level analysis—crucial for UBI proposals that often targeted states—the tools barely existed at all.

"I kept hitting walls. I'd want to model a policy and discover that nobody had encoded the relevant benefit program. Or the model existed but hadn't been updated in years. Or it worked but required expertise I didn't have to run it."

The frustration wasn't unique to UBI research. Anyone trying to analyze cross-cutting policy reforms faced the same barriers. The open-source revolution had produced components, but no one had assembled them into something ordinary researchers—let alone citizens—could use.

What Was Missing

Looking back, the gaps were clear:

Integration. The tools were fragmented. Tax models didn't talk to benefit models. Federal systems didn't connect to state systems. No one had built the comprehensive picture.

Accessibility. Running a microsimulation required installing software, preparing data, writing code. The barrier to entry was too high for most potential users.

Data infrastructure. Open-source rules were necessary but not sufficient. Without open (or at least accessible) data, you couldn't do population-level analysis.

Sustainability. Open-source projects depended on volunteer maintainers who could lose interest, change jobs, or simply burn out. Long-term maintenance was nobody's job.

Trust and validation. How did you know if a model was accurate? There were no systematic comparisons to authoritative sources, no uncertainty quantification, no institutional credibility.

The open-source revolution had proved the concept. OpenFisca showed that legislation could be code. Tax-Calculator showed that rigorous tax modeling could be open. UKMOD showed that major governments could use open-source tools.

But the revolution was incomplete. The tools were promising components, not finished products. Using them required expertise that most researchers lacked and all citizens lacked.

For the promise to be fulfilled—for ordinary people to actually be able to model policy and understand how it affected them—someone would need to assemble the pieces.

Toward Part II

The researchers who had built these tools understood the gaps. Holly Sutherland at Essex knew that EUROMOD's accessibility was limited. Matt Jensen at AEI knew that Tax-Calculator served a niche audience. The OpenFisca team knew that encoding rules was only half the battle.

What none of them had done was build the full stack: rules plus data plus interface plus institutional credibility. A platform that could take a researcher's question—or a citizen's question—and return an answer they could trust.

That challenge would require not just technical work but organizational building. Someone would need to fund ongoing maintenance, hire engineers, establish relationships with official data sources, build trust with policymakers.

In 2021, the UBI Center researcher who had been frustrated by the tool gaps decided to address them directly. PolicyEngine was born—first for the UK, then for the US—as an attempt to complete what the open-source revolution had started .^[5]

That story is the subject of Part II. But it only makes sense in the context of what came before: Orcutt's vision of simulating society from the bottom up, the tax model wars that concentrated analytical power in government institutions, and the open-source revolution that began to democratize access without yet completing the job.