Born Under Punchcards: When Govt Met Web 2.0
Tech reformer Jennifer Pahlka's new book Recoding America explores how government got so broken and how we might fix it.
Between March and April 2020, unemployment in American surged from 4.4% to 14.7% due to the pandemic. Tens of millions remained unemployed by the summer. In California alone, 4.4 million were receiving unemployment benefits. But the state’s Employment Development Department (EDD) estimated that more than 200,000 claimants were stuck in limbo, waiting to be contacted to address some supposed flaw in their application. Many were desperately phoning the EDD, with just one call in a thousand even getting through. So they called their state legislators, who swiftly translated their frustrations into a demand for action. At the direction of Governor Gavin Newsom, the EDD went on a hiring spree, adding five thousand new staffers in just a few months. And guess what: the problem got worse. Every new person the agency hired made it take longer to get a benefit check to an unemployed Californian.
Why? As Jennifer Pahlka explains in excruciating and necessary detail in her timely and valuable new book, Recoding America: Why Government is Failing in the Digital Age and How We Can Do Better, the systems that the state’s employment agency uses to review and process unemployment claims are decades old. Like “layers of sediment you might excavate in an archeological dig,” the agency’s technology goes back to the at least the 1980s, with its primary database, the Single Client Database, running on an IBM mainframe programmed in COBOL, a language invented in 1959. To automate complicated queries into its data, EDD employees started writing macros, simple keyboard shortcuts, to let them avoid having to write more code in COBOL. As newer languages like Visual Basic, C# and the .Net framework came into fashion, they used those too. “Today, the green-screen emulator used by an EDD claims processor has several rows of brightly colored buttons that provide access to dozens of these macros,” Jen writes, noting that these tools have grown chaotically over decades, with new ones written whenever the agency needed to implement a new workflow or respond to some change in the rules or regulations. A more recent push to modernize the EDD’s systems created two new and different ways of accessing the Single Client Database, with the result that now an applicant for unemployment might have a record in both—making it even harder to resolve questions about claims.
Now, here’s the kicker. As a result of this accretion of systems, the training manual for an EDD claims processor is 800 pages long. And reading it alone “does not give you the skills you need to do the job, since much of the knowledge involves workarounds that are passed along from employee to employee but not consistently known across teams or even written down,” Jen writes. One group of experienced claims processors had someone who referred to himself as the “new guy” because he had only been on the team for 17 years and he was still getting up to speed on how the system worked.
So, why didn’t hiring more staff help the EDD work down its backlog of unresolved claims during the height of the pandemic? Because only highly experienced claims processors could do the work. And due to Newsom’s hiring spree, many of them were instead being diverted into training the newbies. Not only that, each experienced processor now handling these trainees got deluged with hundreds of requests for help from them, generating untold hours of extra work. As a result, by the time Jen—who was appointed to a special state taskforce to investigate the claims backlog by Newsom—got, with several talented colleagues, into the heart of what was going on, the EDD’s actual backlog of claims had ballooned to 1.2 million. Even worse, they discovered that one system that the department used to try to catch fraudsters—requiring that a claimant’s name, date of birth and Social Security number exactly match the record in other databases, like the Social Security database—had the perverse effect of enabling thousands of fraudulent applications generated by mobsters using hacked personal records to sail through, while snarling up the claims from honest people who didn’t know how their name was exactly recorded by Social Security. (This is something that happened across all fifty states, she grimly notes.)
Recoding America is riddled with similar examples: an effort by the Air Force to update the technology used by Global Positioning System satellites that got bogged down by an obscure reference in the Pentagon’s own technology architecture framework that appeared to require using an outdated data processing tool; an online application for veterans benefits that only worked with an outdated version of Internet Explorer and Acrobat Reader; well-intentioned ballot initiatives by reformers aiming to free ex-cons from the burdens of minor drug arrests leading to almost no expungements because of the paperwork obstacles; and of course, the most infamous case of the technical dysfunction of government, the 2013 meltdown of HealthCare.gov on the day Obamacare was supposed to open the door to health care insurance for millions.
What makes the Healthcare.gov debacle so important is that for the first time in memory, the success of a top White House initiative actually depended on a service visibly working when everyone was watching, and its failure taught many key leaders, starting with President Obama, a critical lesson: government is only as good as the interactions it has with actual people. Passing legislation or issuing an executive order may get politicians the headlines they crave, but unless the implementation of those decisions is done well, many Americans will not benefit and their distrust of government will only grow worse.
So Why Don’t We Have Better Government?
Here, I should pause to say that I’ve known and been friends with Jen since I met her in 2007 at eTech, a conference run by internet publishing guru Tim O’Reilly, who was not then yet her spouse (they married in 2015). In 2009, she helped put on the first Government 2.0 conference, back when O’Reilly, who had coined the term “Web 2.0” was touting the notion of “government-as-a-platform.” Soon after, she had a brainstorm. Like “Teach for America,” which sought to make public service cool by evangelizing teaching as a career for students from elite universities, she envisioned that “Code for America” would give designers and developers who had offered their skills to political campaigns a way “to contribute to the business of fixing what’s broken in America.” I’m quoting from her original pitch document, which she shared with the Sunlight Foundation (which I was then consulting for as a senior advisor), and which led to Sunlight giving CFA its very first grant. It remains one of our best decisions.
Back then, Jen imagined that Code for America would “transform city governments across America by enlisting the nation’s most promising developers to apply the principles of Web 2.0 to civic problems” and hoped to make local government more collaborative, participatory and transparent, to get more IT talent into government, and at the most abstract “to change the culture of government.” While CFA’s model has changed dramatically since it launched, deemphasizing placing teams of tech fellows in various cities (which required convincing them to pay CFA for the service) and instead working directly as a contractor to state and county government agencies in need of big fixes, that larger goal of changing government culture hasn’t wavered.
If you are someone who works in government, or cares about how government works, Recoding America is must-reading. It fills an absolutely critical hole in what passes for mainstream media reporting on government. That is, instead of telling us who is ahead in the horserace or parsing some policy debate, it gets down into the real details of how government agencies actually implement laws and regulations. (This isn’t just a media gap, by the way—I could offer dozens of examples of advocacy groups and campaigns that tell would-be supporters all they need to do to get X to happen is “elect the right leaders and hold them accountable” as if that translates magically into tangible outcomes.) The fact that it pulls you into that story via tech is arguably incidental, or rather, it’s because enough decision-makers in government realized that they needed to open up their processes to outsiders (“techies”) who they imagined could cure what ailed their systems. As Mikey Dickerson, a top Googler who left Silicon Valley to help save Healthcare.gov and then was the first administrator of the US Digital Service (an agency Jen helped set up during a sabbatical from Code for America) said to Tara Dawson McGuinness and Hana Schank in their book Power to the Public, “Try saying you’re going to the VA to be a management consultant. No. Don’t bother. The technology label is a hack that works right now.”
I learned a ton from Recoding America about many of the reasons our government agencies have gotten into such unproductive patterns of behavior. Jen offers several: an over-reliance on judicial-like systems of review that were imposed starting in 1946 by Congressional Republicans who had just wrested control from New Dealers and wanted to constrain the power of all the new government programs created by FDR; a longstanding commitment to the “waterfall” approach to building government services that frontloads all the conceivable requirements into a massive “Request for Proposals” that advantages giant contractors good at navigating the procurement maze but then guarantees the delivery of products that have been built to suit the needs of government rather than users; a disdain for the mechanics of implementation by politicians and technocrats far more attracted to the higher status of policy-making; the constant tweaking of programs and services by legislators mainly interested in earning headlines and careless of how those changes just add layers of paperwork and bureaucracy to agency leads already struggling to avoid being punished for failing to follow the letter of every law.
That said, I finished Recoding America feeling that despite all her archeological digging, Jen avoided a key reason for our collective failure to fix government: because our democracy is also broken and captured. She dances near this idea a couple of times throughout the book, not just in noting how it was anti-New Deal Republicans who first threw legal sand into the gears in the late 1940s, but also recognizing that in many cases of social policy that was ostensibly aimed at helping poor people or minorities who experienced discrimination, policymakers either expressly chose to disadvantage some constituencies (think of redlining of loans by the Federal Housing Authority) or, because they believed more in punishing the poor than uplifting them, they made it incredibly hard for people to access benefits that they were entitled to.
These attitudes are in flux now, and as Jen notes with her discussion of the movement to undo the harms of the War on Drugs, some people in positions of government authority are even beginning to move proactively to liberate people from the burden of, say, an old marijuana arrest, rather than force them to do all the paperwork. But the reason for this change isn’t wonks or tech nerds who got invited into government agencies and, in the guise of being glorified management consultants, convinced the powers-that-be that the drug war was a disaster. The reason for this culture change is good old-fashioned political organizing that developed the power to make less punitive drug laws the new norm.
The problem is that we live in an oligarchy. As Martin Gilens and Benjamin Page put it back in 2014, “Multivariate analysis indicates that economic elites and organized groups representing business interests have substantial independent impacts on US government policy, while average citizens and mass-based interest groups have little or no independent influence.” If few, if any, politicians are spending time trying to clear the records of former felons or trying to make it easier for the needy to access food benefits, a big part of the reason is that those people have very little power in the existing system. There is no lobbying organization for the interests of unemployed people and very few that speak on behalf of the formerly incarcerated or residents of public housing. Most elected officials spend an inordinate amount of time talking to wealthy people, since they need to raise millions to win and stay in office, and over time they are far better versed in why they shouldn’t touch the carried-interest loophole than in why they should care or have interest in streamlining a SNAP application.
Jen isn’t wrong to say that “in our system as it stands today, there is little incentive for elective leaders to build twenty-first-century state capacity.” But it’s naïve to write that “there is still someone above them, and in fact above all three branches of government: us.” Yes, in the very next sentence she acknowledges that “there is a staggering amount wrong with our electoral system and our campaign finance laws,” but in an age of minoritarian capture of many state legislatures as well as the Supreme Court, plus the heavy tilt of the Electoral College and the ongoing attacks on voting rights, the word “staggering” is doing a lot of heavy lifting in that sentence.
What’s still valuable about Recoding America is that we aren’t faced with an either/or choice here. We don’t have to either figure out how to fix government or to fix politics. We have to do both. Being able to explain so clearly how government is broken and how it can be fixed is vital to that other project. For why should be bother trying to fix politics if the results coming out of government won’t change? And people working on fixing politics should take heed: allowing our broken government to continue unchanged is also making are democracy weaker. As Jen notes, participating in means-tested programs significantly reduces the chance that someone will vote. Bad experiences with welfare agencies, the criminal legal system, immigration or even the process of getting a construction permit all erode faith in our system of government. So politicos have to recognize that the work doesn’t end when an election is won, the work is only beginning. At least thanks to Recoding America, we have more of a map and a way through the territory than we had before.
—Bonus link: Go to RecodingAmerica.us to learn more.
Odds and Ends
—Sarah Jones has a great piece in New York about how Moms for Liberty is starting to meet its match in progressive networks like Defense of Democracy and Red, Wine & Blue.
—Russell Berman has a silly piece in The Atlantic arguing that “AI could save politics” by dramatically reducing the cost of running for election. He writes, “With AI’s ability to handle a campaign’s most mundane and time-consuming tasks—think churning out press releases or identifying and targeting supporters—candidates would have less need to hire high-priced consultants. The result could be a more open and accessible democracy, in which small, bare-bones campaigns can compete with well-funded juggernauts.” And I can flap my wings and fly to the moon.
—Don’t miss Simoine Stolzoff’s story for The Verge about the inside fight to unionize Kickstarter, the first major tech company to get a union.
End Times
Just a taste of what’s ahead for 2024.
It’s not just gummint. ANY sufficiently mature organization runs on word-of-mouth institutional knowledge of whose monitor has the sticky notes with the magic incantation to unlock the next obstacle. What’s more, the substance of what an institution does (as opposed to what it professes) is encoded in the interstitial spaces of those procedures. Staff loyalty is to that system and those procedures because that is the basis on which they are hired, fired, promoted, demoted and paid. It is adherence to procedures that is measured and manage. The system resists change and replacement in equal measure because the principles on which is operates are only partially understood. Means come to have only a tenuous connection to ends, which inevitably become an exogenous variable. So, not only the system is a black box but so is the external world that it is supposed to affect.