What this covers
Jennifer Pahlka, founder of Code for America, speaks with Russ Roberts about why government digital services fail and what lightweight, design-driven interventions can repair them. The conversation centers on a counterintuitive diagnosis: the problem is rarely that government lacks money, talent, or even intention to serve citizens well. Instead, Pahlka identifies three structural culprits—slow and expensive procurement habits treated as iron law rather than convention, regulatory complexity born from taxpayer distrust, and a broken feedback loop between civil servants and the people they serve. The remedy involves applying user-centric design principles, publishing data through open APIs rather than building monolithic systems, and giving government workers visibility into where their processes fail real users. Examples thread through: a food stamp application redesigned from 45 minutes to 7 minutes; a school-assignment tool built in 10 weeks instead of 2 years; a transit-planning system that cut modeling time from weeks to seconds.
The conversation widens into harder territory: whether Empathy in Service Design actually improves outcomes or simply adds unused features, whether government improvements ever shrink bureaucracy or merely redeploy its capacity, and how much responsibility citizens bear for the regulatory straightjackets they've imposed on their own government. Pahlka argues that the rules constraining innovation are legal fictions—what officials claim is impossible is usually just unconventional—and that reframing the debate away from big government versus small toward "does it work as intended" could shift incentives. Roberts presses back on whether private-sector momentum and Market Feedback Loops have fundamentally different power to drive change than voluntary external help like Code for America's. The exchange also touches on SNAP participation rates, the honor or stigma of public service, and whether government should concentrate on what citizens cannot accomplish alone.
Pahlka argues that government's failures in delivering digital services stem not from a lack of caring or talent but from procurement habits, regulatory constraints born of distrust, and missing user-empathy feedback loops; lightweight, design-driven technology can make government work as intended.
- Government's slow, costly procurement is mostly habit and conventional wisdom, not legal requirement
- Civil servants want to serve the public but lack tools and visibility into what fails users
- Taxpayer distrust produced complex rules that ironically make spending less effective
Faster, user-centric procurement and delivery are legally permissible but require challenging conventional government wisdom.
- When government workers say they cannot fix problems the way Code for America can, the limitation is true in practice but not in law; doing things the way that is right for users—including faster, user-centric technology procurement—is actually legal, and it is the conventional wisdom that needs challenging.
“'Yeah, that has been true. But it's true in practice. It's not actually true in law.'... doing it the way that's right for users is actually totally legal... It's the conventional wisdom that needs to be challenged”
- A school-assignment mapping app that Code for America Fellows built in about 10 weeks would have taken roughly 2 years and about $2 million if it had gone through the standard government procurement process.
“we found out after we spent about 10 weeks getting this up and running...that if it had gone through a procurement process, it would have taken about 2 years and about $2 million to build.”
- By placing startup-skilled teams into government using user-centric design and platform thinking, Code for America's fellowships do not just build products but demonstrate to government workers that a fundamentally different, faster way of solving problems is possible, giving them a replicable model.
“what they are really doing is demonstrating how we could do this fundamentally differently. And then government people's feedback--they generally are smart enough to know that there is a better way and that other ways are possible.”
- Demonstration projects have huge value in getting people to change practices, but after six years an organization cannot have everything be a demonstration project; sustainable solutions and using the community as capacity are needed to scale.
“Demonstration projects have a huge value in getting people to change practices. But after 6 years of it, you can't have everything be a demonstration project.”
Government's poor service stems from technology built for internal needs rather than user experience and design.
- Government's poor citizen experience is fundamentally a problem of design and process redesign, not of having too little technology; government has tons of technology, but it is built for government's internal needs rather than for user needs.
“it isn't really technology. It's design. [?] process redesign... There's tons of technology in government. It's just that the language we use isn't built for government needs, not for user needs.”
- California's online food stamp application took 45-50 minutes, didn't work on mobile, and couldn't save progress, while many low-income applicants only had access at libraries whose computers time out after 30 minutes; Code for America's redesigned app takes about 7 minutes on mobile because it accounts for how people actually use technology.
“it takes about 45 minutes online...and it doesn't work on mobiles...the library computers time out after 30 minutes. And the application takes about 45-50 minutes. And it won't let you save your work... we have an application that takes about 7 minutes on mobile phone”
- Empathy alone leads to adding every feature users request, producing software with a thousand unused features; pairing empathy with design skills translates user needs into software that works for everybody, which generally requires fewer features and less complexity, not more.
“you do that a thousand times, and you get software with a thousand features that no one uses... our process is very much grounded in empathy, but with the skills of design... it actually needs to have fewer features, less complexity not more.”
Specific government processes fail users at scale, as shown by SNAP's complexity, low participation, and wasteful administration despite massive spending.
- Despite California's reputation as a Blue state, it has the second-lowest rate of SNAP participation among eligible people in the country, with only Wyoming lower, in part because many eligible people cannot navigate getting on the program.
“we have the second-lowest rate of participation in SNAP. The only state with lower participation by eligible people is Wyoming.”
- California spends billions administering food assistance through expensive IT systems, phone calls, and paperwork that are far more complicated than needed; redirecting that administrative spending into actually giving people food would be a better use of taxpayer dollars regardless of one's view on the program.
“If you take the couple of billion dollars that we are spending probably in California alone on expensive IT systems, many, many different phone calls, tons and tons of paperwork, and just put that into giving people food, that would be a better use of taxpayer dollars”
- By documenting applicants' experiences, Code for America found a county was sending a form not required by any law, policy, or regulation that caused people to abandon the process; once flagged, the county director discovered she didn't know it was happening, stopped it, and more applicants completed the process.
“our staff called the county that was sending that form out, and said, 'I don't think the form that you're sending is required by any law, policy, or regulation...' And the County Welfare Director said, 'We don't send that out.'”
- SNAP, a roughly $70 billion program, is the program most highly correlated with health and education outcomes for children, and may save money because well-fed kids are less likely to act out, be placed in Special Ed, and more likely to perform well in school.
“SNAP (Supplemental Nutrition Assistance Program), which is a $70 billion dollar program across our country, is a program most highly correlated with health and education outcomes for kids.”
- Roughly $42 billion a year is spent through charity on essentially welfare-type needs while $470 billion or more is spent in taxpayer dollars; if public spending disappeared, private spending would grow but likely not to $470 billion.
“about $42 billion a year is spent through charity on sort of essentially [?] issues... $470 billion or more is spent on taxpayer dollars.”
Demonstration projects and data-driven redesign convince government workers that fundamentally different problem-solving is possible and replicable.
- The most common response Code for America gets from government officials when shown data about failing processes is gratitude, because government lacks IT systems instrumented to reveal what works and what doesn't, contradicting the assumption that bureaucrats resist change.
“that response of 'Thank you for letting me know,' is actually the most common response that we get from government officials, by far... we don't have the ability in government to create IT systems that are instrumented to know”
- Public servants' reputation suffers partly because they lack the tools to serve the public; giving them effective tools can rekindle belief in their ability to do their jobs, as shown by a Detroit official with 28 years of service who had stopped believing she could serve the public until a Code for America project restored that belief.
“'I went into public service to serve the public, and I've been doing it for 28 years; and I have just realized that many years ago I stopped believing I could serve the public. And what your project did for me was make me believe that again.'”
- The Adopt-a-Hydrant app showed that simple civic technology can be built in a single day and that ordinary people are willing to take responsibility for a piece of physical public infrastructure on their street, a pattern that generalized to tsunami sirens and other assets across many cities.
“a lot of people were really surprised that a). this technology is so easy to do--this was like a day-long project, the first instance of it. And that people want to help.”
Public distrust of government created procurement rules that waste money; citizens must accept responsibility for trusting and enabling effective governance.
- The constraints that make government bad at digital and innovation exist primarily because taxpayers don't trust government and have created complex procurement regulations to control spending, which ironically results in money being spent poorly; the public must own responsibility because distrust plus heavy rules won't produce the desired outcomes.
“it's primarily because taxpayers don't trust government, and have created...very complex procurement regulations... Which results, ironically, in it being spent often not that well. But the taxpayer has to own some...responsibility”
- Government is an important institution but needs to be updated to fit the 21st century, where services could work better, outcomes could improve, and costs could be reduced if people take collective responsibility for changing how government works.
“we are a national non-profit, a network of people who believe that government is an important institution but needs to be updated to fit the 21st century. We think government services could work better, outcomes could be better, costs could be reduced”
Publishing data with APIs and reducing government's build scope enables distributed innovation by many parties.
- Applying the 'do less' principle, government can publish data with good APIs rather than building entire interfaces itself, enabling many parties to build offerings for different users; the College Scorecard exemplifies this, with its federal data now powering Google's responses to college queries.
“Do you need to build a whole website? Or, can you get the data out there...with great APIs on them in such way that other people can build the interfaces to that data... it's now the data that Google accesses when it tries to respond to a query”
- A tool that began as a game for citizens to play with transit routes became Remix, a startup whose software lets transit planners model the cost and Title VI compliance effects of route changes in seconds rather than the days or weeks the process previously took.
“Something that was a several-day process to sort of model out what would happen if you changed a transit route now takes 10 seconds.”
- Code for America's real scale comes less from the fellowship than from a network of volunteers in about 75 cities who, as community members, partner with city hall staff on local projects.
“the scale really comes from the amazing network of volunteers who do this, in about 75 cities, who aren't--they are just part of the community.”
- Contrary to the assumption that government can never do technology and innovation as well as the private sector, some of the best people in the consumer tech world are currently working for government and producing high-quality work.
“some of the best people in the consumer tech world right now are working for government. And doing this. And we just need to--I would love more people to look at those folks and look at the quality of the work that they are putting out.”
- The framing of big government versus small government is less powerful than asking whether government does what we intend it to do; regardless of one's views, no taxpayer wants their dollars failing to achieve the intended effect.
“this notion of sort of big government or small government is less powerful than government that does what we intend it to do.”