How to Fix the Government
Things did not go smoothly when, as the coronavirus pandemic took hold in spring 2020, Congress turned to the unemployment system to help people who found themselves out of work. The under-resourced state agencies that carry out the day-to-day administration of the unemployment system would now have to get tens of billions of dollars into the pockets of the newly unemployed—which, by and large, they did. But they also had to avoid making payments to grifters who didn’t qualify. That proved more difficult.
Consider Michigan, the state where I live. In pre-COVID times, Michigan’s unemployment agency would wait at least 10 days before paying a claim to allow the agency to verify an applicant’s work history. During the pandemic, with claims skyrocketing, Michigan officials came under intense bipartisan pressure to “get money out the door faster.” As part of that effort, the agency decided the 10-day hold wasn’t strictly essential. Even without it, an automated system called Fraud Manager would still review every claim and flag suspicious ones.
That was the plan, at least. But, as is typical with technical matters, the agency had contracted out responsibility for running Fraud Manager. And the contractor, unbeknownst to the agency, had set Fraud Manager to run only in the evenings. That approach worked fine when the 10-day hold was in place, because Fraud Manager would inevitably run before any claim was paid. When the 10-day hold was eliminated, however, the agency started to pay claims the same day they were received—before Fraud Manager could review them.
It took the state unemployment agency more than two months to notice the problem, according to a subsequent audit by Deloitte. In the meantime, the agency paid out an estimated $1.5 billion—that’s billion with a b—in claims that should have been flagged for possible fraud.
What’s wild about this story isn’t the absurdity of the mistake or its outsize financial consequences. What’s wild is that it’s completely typical. Every state unemployment agency struggled with fraud claims. Notoriously, a rapper known as Nuke Bizzle bragged about scamming California for more than $1.2 million. (He is now serving a six-year prison sentence.) The inspector general at the U.S. Labor Department estimates that as much as $163 billion in pandemic-related unemployment benefits shouldn’t have been paid.
And the issue isn’t just unemployment insurance. As Jennifer Pahlka explains in her indispensable new book, Recoding America, similar tech problems plague government across nearly every program and at all levels—federal, state, and local alike. Addressing those tech problems would not only help us avoid billion-dollar mistakes in the future; it is also crucial, Pahlka argues, to the broader project—one that is starting to catch the attention of both pundits and policy makers—of improving the government’s capacity to do what we have collectively asked of it.
Pahlka is a former deputy in the White House Office of Science and Technology Policy and the founder of Code for America, a nonprofit that aims to help government agencies with their tech issues. She excels at describing in cogent, accessible prose why government is so bad at tech.
That sounds like it should be easy. It’s not. You need someone with technical expertise and an insider’s understanding of how a complex system operates at the line level. But you also need someone who can understand the policy and political forces that have shaped the system—and can see how the lessons from one program might generalize to other programs.
Pahlka has all that, and what makes her account so forceful—and what will make it ring true for anyone who has worked in government—is that it’s not a story about bumbling civil servants or venal politicians. It’s not even that the government can’t afford to hire people with the needed tech skills. Instead, it’s about the structures and incentives that make it more attractive for agencies to check bureaucratic boxes than to design usable tech.
When you apply for food stamps, for example, you have to demonstrate that you’re eligible. But some of the applications have swollen to absurd lengths—Michigan’s used to be 40 pages long—and ask detailed, confusing, ridiculous questions about applicants’ financial situation. A lot of people give up before they finish, even if they’re eligible. In the old days, that was a literal paperwork problem. Now that applications have moved online, it’s a problem with code—and the user experience is not top of mind for many government coders.
At the tail end of her time in the White House, Pahlka was trying to convince a career bureaucrat over at the Office of Management and Budget that simplicity had to be the watchword when it came to digital services. To drive her argument home, she pointed to some of these food-stamp applications, which among many other things asked if applicants owned any burial plots. “Why on earth,” she asked, “does government want to know about burial plots?”
The bureaucrat was unmoved. Yes, he knew all about the burial plots. He used to work at the agency that oversaw the food-stamp program—and had written the regulations about burial plots. Why would he call for including such an inane question on forms that were already far too long? “Congress said to assess their assets,” he said. “A burial plot is an asset.” He was going to follow his instructions to the letter, the consequences be damned.
[Deborah Pearlstein: How the government lost its mind]
For Pahlka, this check-box mentality is at the root of much government dysfunction. We have a penchant in the United States for holding civil servants accountable not for the quality of the public services they provide, but for strict compliance with programmatic requirements. When the inspector general comes knocking, that’s what’s evaluated. So, too, with courts. It’s not their job to take a holistic view of whether the agency is doing its job effectively. They ask whether the agency has jumped through the prescribed procedural hoops.
In prior work, I’ve railed about the way this “procedure fetish” in American law has hampered effective governance. Pahlka brings to vivid life how a cover-your-butt culture that prizes legalistic compliance above all else is especially pernicious for government tech. Policy makers layer requirement upon requirement without considering whether the benefits of complexity outweigh the costs. Even when policy makers give agencies some flexibility, the bureaucracy often transforms suggestions into rigid requirements, which are then slavishly followed. The public interest gets forgotten along the way.
In other words, Pahlka’s book isn’t just about tech. It’s about the American administrative state, and it’s a call for paring back the rigid rules that make it so hard to govern, and for rebuilding government’s ability to do its job effectively. In this, Pahlka joins ranks with the likes of Brink Lindsey, Misha Chellam, Alec Stapp, and Ezra Klein, who are all beating a similar drum about the need to improve the government’s ability to meet our collective aspirations.
This nascent “state capacity” movement—also known as supply-side or abundance progressivism—has arisen in response to the urgency of the housing crisis, the massive land-use changes needed to pivot to renewable energy, and the exorbitant costs of new public transit. Pahlka would add to that list the deplorable performance of government tech. We need a government that can build, whether it’s wind farms or websites.
How we get there from here is a hard question. For now, the movement’s origin in housing, renewables, and transit means that it’s associated with the political left. Its emphasis on construction and deregulation, however, means it holds some bipartisan appeal. Maybe there’s room for progress out of the limelight. For her part, Pahlka doesn’t see government capacity as an especially partisan issue. Big problems with government tech, she argues, typically arise because of implementation failures that both Republicans and Democrats would like to avoid.
That’s a departure from the conventional account on the left. Return to the story of unemployment insurance that I opened with. Part of the reason the tech worked so badly is because of decades of underinvestment in a safety-net program that Republicans dislike. In other words, the problem was not just implementation. It was also a kind of sabotage. As Jerusalem Demsas here at The Atlantic has written, “State capacity is downstream of ideological commitments: When we have political consensus, we have state capacity, and when we don’t, we don’t.”
Demsas is certainly onto something. Better tech is harder for programs that face political headwinds. Pahlka doesn’t disagree, but she suggests that we can sidestep some of the deeper challenges—for now—by focusing on the relatively low-salience domain of policy implementation. Food-stamp applications didn’t ask those in need about burial plots to harass them or to surreptitiously shrink the food-stamps program. It really was just bad implementation.
Which is another way of saying that there are some low-hanging fruits to pluck, if we care enough to try. Implementation, Pahlka writes, “can’t be beneath the attention of our most powerful institutions, and it can’t be beneath our attention as a public.” When Americans no longer see government websites as laughingstocks—when applications ask sensible questions, when submitting your taxes online is easy, when signing up for health insurance doesn’t require a Ph.D.—maybe the politics will follow. You’ve got to start somewhere.
Leave a Reply