← Priorities Podcast
Priorities Podcast
States are trying to reduce SNAP error rates to keep the program alive
00:24:57
Changes to the Supplemental Nutrition Assistance …
Loading summary
Transcript19 lines
- [00:02]
A
Hello and welcome to State Scoop's Priorities podcast. I'm Sophia Foxowell, a reporter for statescoop. This week we're talking about what states can do to reduce their payment error rates in order to comply with upcoming federal changes to snap. But first, here are the biggest state IT stories of the week. Texas governor Greg Abbott last week appointed Amanda Crawford, the the state's chief information officer and executive director of the Texas Department of Information Resources, as the state's commissioner of insurance. Officials say they'll soon announce a replacement to lead the state's technology department. New York State's technology office this week announced Eleanor Hornier Toombs, a United Nations University researcher, has been appointed as chief artificial intelligence officer, replacing Sharia Amin, who spent less than a year in the role. The state also named a new chief digital officer. Florida is strengthening animal welfare protections with the passage of Dexter's Law, a new statute designed to prevent repeat animal abuse and improve accountability across the state. Named after a shelter dog whose inhumane death in 2023 sparked public outrage, the law responds to growing concerns that individuals convicted of animal cruelty and were still able to adopt or obtain pets with little oversight. Reducing error rates in food assistance programs like the Supplemental Nutrition Assistance Program has become a top priority for states, especially as more of the financial and administrative burdens shift from the federal government to state and local agencies. HR1, the federal budget bill that passed last year, hill increases work requirements for SNAP recipients, shifting 50 to 75% of administrative cost to states and potentially requiring states to contribute to benefit amounts based on their error rates. Rebecca Piazza, executive director of SafetyNet strategy at Code for America, stressed the urgency for states to find effective solutions before 2027 in order to comply with federal changes such as adopting automation and low risk AI tools to improve accuracy. So at a high level, what does HR1 kind of change about how states are expected to manage SNAP moving forward and how does technology kind of come into play when managing those types of expectations and changes?
- [02:31]
B
Thanks. So SNAP is a really key program in the safety net. It's really I just want to ground us in this first. It's one of the most impactful anti hunger programs that we have in the country. It's a program that millions of people depend on. You know, it's meant to be a supplementary program, but really a lot of people depend on it for all of their grocery budget. In HR1, there are a number of changes to how it's administered that make it a very different program than it has been to date. It impacts both the number of people who are receiving benefits by expanding work requirements and the number of people who are subject to them. And it changes the administration of the program in ways that are going to create a number of more administrative and financial burdens to states. So looking at work requirements, there have been a number of places throughout the country that have not been subject to them. There have been a number of groups where people have been exempted. Thinking of veterans, people who are homeless, those people are now going to be subject to work requirements. And that's going to put a lot more strain on programs and program administrative staff. There's going to be a lot more work that they need to do with recipients in order to validate work, whether people are meeting those requirements. And that's going to mean that a number of thousands of people are going to lose benefits because they can't show compliance with those requirements. And I say that intentionally. It's not necessarily because they're not working, it's because showing compliance with those requirements is so difficult. What you're also seeing is a big shift in the cost of administering the program going to states. So the HR1 bill changed right off the bat the percentage of administrative costs that states are responsible for from 50% to 75%. And so that is in many states, you know, tens or hundreds of millions of dollars. And then it also introduced a new potential cost share for the SNAP benefit amount for the first time. I think something that's a big distinction is there's the cost to run the program, which is administrative costs, and then there's the benefit cost, which is the actual money that is loaded onto EBT cards that people use to pay for food. And states have never been responsible for a portion of that benefit amount to date. But now HR1 is introducing new rules where states may be responsible for a portion of those benefit amounts. And that would mean hundreds of millions of dollars to, in some states, even billions of dollars a year that states would now be responsible for for the SNAP program.
- [05:11]
A
Those are some pretty expensive changes. We're talking about more administrative burdens, we're talking about more financial bur that's on the state. And you're talking about also the recipients of SNAP who might not be able to comply with the new work requirement changes. Other because they can't show documentation, they are unable to fill out applications immediately. And then we're also talking about processing because I do also know that a lot of these state agencies are operating on kind of decades old like aged infrastructure. How are they expected to manage all of these Changes, especially if on the back end their infrastructure and like Digital Skeleton is not able to support these changes.
- [05:56]
B
Absolutely. The technology that runs these programs is another barrier to meeting these new requirements. And the fact is that with the change in administrative cost sharing, that means that if states want to improve that infrastructure to rely more on technology in order to, you know, better serve the people in the program, they're going to have to pay more in order to do that. So it really creates a bind for states where they're being held to higher standards. They're being held to higher standards while the program rules are becoming more complex, and then they're being given less money to be able to spend on either personnel or improving technology in order to meet those standards. So it really creates sort of a perfect storm of challenges for states in meeting these new expectations. Now, I want to say, you know, program integrity is really important. You know, it is important that, you know, states are able to run the programs as expected, but they need the resources, they need the technology, and, you know, they need the support in order to be able to really meet those expectations.
- [07:06]
A
I mean, on top of all of this, we even talked about error rates in terms of, like, that is just to fulfill the natural requirements of providing these benefits to often the most marginalized groups of society that really need them. But if states cannot prove that they have reduced their error rates or they. I'm drawing a blank on what the actual threshold is now. But they need to show that they have reduced their error rates and if they don't, then they will receive even less funding from the federal government to provide already a strained system. Can you talk about what states can do there to comply and reduce their error rates?
- [07:48]
B
Sure. Let me first walk a little bit through the wonkiness of the payment error rates and how that intersects with what we were talking about. So I mentioned that states could be responsible for some of the benefit amounts for the first time ever, and that is dependent on the payment error rate in the state. So HR1 changes the rules so that if states only continue with the way things are today, with not having to pay towards the payment, sorry, the benefit amount if their payment error rate is under zero percent. Now, I want to say really clearly upfront, you know, your payment error rate is not fraud. This is. Yes, the person who was administering the case, you know, calculated the benefit amount incorrectly. Whether that's giving someone too much in benefits or too little. And so this is not a measure of somebody did something, you know, fraudulent or tried to cheat the system. This is really you know, was that calculation correct in the amount that someone should receive in SNAP benefits? Because that's one thing that makes SNAP kind of complex. You know, unlike Medicaid, where you only have to determine whether somebody gets the benefit or not, with snap, you need to determine exactly how many dollars a person or a family should receive. And the calculations to perform that are really complicated. So if the error rate is over 6%, then a state will need to contribute towards that benefit amount that gets loaded onto EBT cards. They'll have to contribute 5%, if it's between 6 and 8%, 10% if it's between 8 and 10%. And if it's over 10%, the state will need to pay 15% of benefit amounts. And that's really a significant amount of money. When you're talking about the impact on state budgets, as I mentioned, hundreds to potentially hundreds of millions to potentially billions of dollars depending on the state. And this is something that states are very much grappling with right now because, you know, while the penalties don't kick in, they don't have to start paying them until 2027. The cases that determine that error rate, that determine what they need to pay in 2027 are being worked today. So this is something that states are very much trying to focus on and very much trying to reduce that error rate. And frankly, it was not a large amount of lead time in order to start updating systems and processes to start getting to the place where those error rates were going down before the clock started ticking. Another. Oh, go ahead.
- [10:20]
A
I was just thinking back to. They didn't have a lot of time, they didn't have a lot of leeway. I'm thinking back to the government shutdown of late year when states were also being threatened with not receiving benefits for SNAP recipients. And they had to really scramble and prepare to scramble any like state or local resources reaching out to food banks, like generating any type of extra money in the budget or allocating that extra money in order to make sure recipients in their state did not go without. And it feels like that same type of urgency is being applied even though there is no government shutdown currently, but they are preparing to just receive one. An onslaught of probably new applications given inflation and the economy, but also knowing that they're not going to have as many resources. What are you seeing at Code for America in terms of states and finding solutions to troubleshoot this?
- [11:13]
B
I would say I'm excited to get the solutions. One other thing I'll say is that the focus is pretty singularly on payment error rate in states. And as I mentioned, I think that's very important. I think that program integrity is critical. However, payment error rate needs to be balanced with other really important measures of the program. What we're seeing right now is that because states, you need to get that payment error rate down, they're asking people to provide more paperwork, more documentation in order to justify entries on applications. And, you know, it. It's logical, it's preservation in order to help preserve budgets. But what that means is that people are losing benefits because they can't show that paperwork, because the process of applying is getting so. Or recertifying is getting so complex that they give up or drop out of the system. And so because of that, you need to balance the payment error rate with other measures of how many people who should receive benefits are making it through the process. How long is it taking to make it through the process? Is it taking a week for people to get benefits, or are those numbers going up to 60 days? You know, because that makes a big difference when somebody is relying on these benefits. So it's creating some very skewed incentives with this singular focus on payment error rate. Now, I will say rational behavior, but. But skewed behavior. And so, you know, we at Country America are really trying to help states with this. You know, how can they help reduce that payment error rate while not having those secondary impact impacts of people losing benefits because the process gets so complex. So, you know, we're taking a lot of approaches. We're really encouraging states to use data to prioritize where they should be focusing. You know, what are the areas of highest impact of bringing down that error rate? You know, how can they learn quickly? You know, look at measures that might sort of show, you know, the impact of system changes or process changes. They're making, you know, more quickly than those error rates that really are calculated on longer timelines. And then, you know, also really looking at helping caseworkers work efficiently but still accurately. You know, are there ways that, you know, repetitive tasks can be automated? Are there ways that data can be improved? Frankly, sometimes even looking at, you know, what are the workarounds that caseworkers have put in place, because that sometimes signals something that needs to be changed. It's not necessarily failure of the system. It's a signal that the system needs to be improved. And then also making sure that people are providing timely, accurate information by making those notices that go out to participants easier to read, easier to understand, helping clarify language around program rules and requirements because participants want to do the right thing. It's just that the rules are so complicated, the language is so complicated, it's sometimes hard to comply. So how do we make it easy to do the right thing? How do we make it easy to understand how to do the right thing?
- [14:31]
A
I really appreciate you saying that, that everyone who is a recipient of these benefits really does want to ensure that they're getting them. And in order to get them, they want to make sure that they are compliant. They want to make sure they're filling out forms correctly. They want to make sure they're getting their applications in on time or any other documentation to comply with these new require. And government for their part in administering these benefits. You're absolutely right. It's their responsibility to also make it easier so these recipients can follow through and make sure they're getting these. These benefits. I'm curious, you mentioned updating, like finding these little signals within systems, whether there's workarounds for that caseworkers are doing or something that continues to show, like mistakes. What are you seeing in terms of those signals? Like, are there any common patterns that you're finding when you're, when you're looking through states and helping them work through solutions, Are there any common errors that you're finding?
- [15:31]
B
I think that we're seeing a lot of instances where people, maybe participants aren't sure that they need to report new information. And so how can we help people understand when they need to provide that information, how they need to, and what actions they need need to take? I think also helping caseworkers really get the full picture of an application as well. And one thing that has been really successful and productive is sort of flagging potential things that maybe a caseworker should take a second look at. Not necessarily saying it's an error, but areas in a case where you frequently see errors. So confirming if expenses exceed income, is that accurate? I mean, unfortunately for many people who receive snap, that is accurate, but it's something to take a second look at to ensure that that isn't something that was an error and an unintentional data entry mistake. But I think, you know, helping caseworkers really, you know, take a close look at the case, you know, proactively identify and mitigate, you know, common errors and then also helping participants better do their part as well.
- [16:45]
A
Do new tools like any automation or artificial intelligence, do those play a role in terms of making these systems work more efficiently?
- [16:54]
B
Yep. I know everybody loves to hear about how AI is helping things these days, and I do want to Say, up front, I know that, you know, there are sometimes, you know, concerns about AI both, you know, in state agencies because it might be new, and also from participants. And I do want to differentiate that there are different types of AI and there are different materials maturity levels of AI. So I think that one thing that we're really encouraging to start are the sort of de risked and proven types of AI that are out there that can provide a lot of efficiency and accuracy gains. You know, one thing that comes to mind with that are, you know, using AI to look at a document that somebody has uploaded and first of all, being able to give immediate feedback of, you know, is the image of this document blurry? You know, should the participant scan it again because the system can't read it. And then also, you know, it can help characterize, did they upload a pay stub or did they upload their lease? You know, how can AI categorize documents in the case file so that it's much faster and easier for the caseworker to find the right document that they're looking for? And then even getting to the point where it might extract data from that document to determine, you know, take information from that pay stub and fill it in in the right parts of the file. So those are really proven uses of AI technology. Those aren't really the cutting edge things that maybe people think of with AI right now, but they're all really effective and save a lot of time and improve accuracy. Other areas where AI is very proven and can be used is around what we call entity resolution, which is matching cases that are actually referring to the same person, but the data needs a bit of a fuzzy match in order to get there. So, you know, we see this in case files where maybe somebody really needs benefits. And so they've applied three times, and so they have three different cases in the system. You know, how can we really combine those files so that we have all of the information in one case and the system can be smart enough to know that those are really three applications from the same person? Or how can we use that fuzzy, you know, AI entity resolution logic to match, you know, somebody's snap case with a file in another system, like saying that they're getting a disability benefit so that they shouldn't have to comply with work requirements. You know, how can we do that in order to streamline and make things easier, both for the caseworker and for the participant?
- [19:30]
A
I love that you brought that up, because my next question is going to be, are you seeing a lot of These agencies or states kind of change their data sharing practices. So better integrate their data between these often really silo systems that probably should talk to each other because I'm sure there's a bunch of overlap between participants who are receiving one type of state benefit versus another.
- [19:53]
B
That's something that I think is really powerful. I know that a lot of the focus on, you know, in civic tech where Code for America is a lot of times on the front end of systems, you know, making forms easier to understand, helping people, people understand program rules. But I think that back end data is just so powerful in an area that you know, provides a lot of opportunity. You know, how can we be using information that the state already knows about you to make it simple, simpler for you to keep your benefits, to show compliance with those work requirements in order to simplify recertification and allow you to really keep those benefits? You know, that has really impact for everyone. It makes it cheaper for the case to. Sorry, cheaper for the state to administer the case if they already have a lot of information that they trust and can rely on. It makes it simpler for participants and you know, it provides that level of accuracy that I know is a big focus right now with that emphasis on payment error rate.
- [20:55]
A
I'm curious what your, what your thoughts are given how much of the burden is going to be shifted to states versus being with the federal government. I'm curious what you think this is going to do to the relationship between states and the federal government in terms of these really penny pinching, high taxing measures in terms of like the mistakes in the error, like the payment error rates. But if states can't depend on the federal government to provide a bulk of these benefits, which is going to result in possible harms and loss of, you know, funding for their recipients. I'm curious what you think that's going to do in terms of the symbiotic relationship, relationship between states and the federal government. Will states not be able to rely on the federal government as much? Will they try and take more ownership since they know it's such a big burden? I'm curious what you think.
- [21:45]
B
I think that's a great question. I don't know how it's going to impact the relationship between states and the federal government, but what I can tell you is that it really puts states ability to offer SNAP at risk. I think SNAP has been such a critical program in our country for so many decades that people just assume it will always be there. But the fact is it's actually optional for states to participate in SNAP and to offer that. And if states have to pay hundreds of millions of dollars to billions of dollars to continue to participate in snap, they might not be able to afford to do that. And that's something that is, I think, going to be something to keep an eye on in 2027 and is as states are putting together their look at in 2026, the states are putting together their 2027 budgets, you know, our state saying that they may not be able to offer SNAP in the future. I think that's something that's a real possibility. And even if they are able to offer it, you know, what are the other changes in the budgets that are going to need to happen in order for states to continue to offer snap?
- [22:53]
A
Absolutely. This could be a compounding factor situation where in order to preserve this entity that is is so important, so many people rely on, they might have to take money from other also equally essential programs in order to continue to fund it or greatly reduce the amount of funding that they're able to offer to these, these recipients. My last question for you is, are you hopeful that states will be able to figure out solutions, whether it's administratively on the data end, using more innovative tools or just being able to, to better comply? Are you hopeful that in the allotted time between now and when the changes take effect in 2027, are you hopeful that they'll be able to make it?
- [23:37]
B
I have to be hopeful. I, I believe in this program so much and it's so critical. I feel like, you know, I don't know if states are all going to be able to get under 6%, for example, with their error rates that I don't think is going to be possible. But there are huge benefits even from going from a 10% error rate to an 8% ER rate. You know, what that really means is that means tens of millions of dollars that a state no longer needs to find in order to continue to offer the program. So, you know, I think every inch of ground that a state is able to take is critical. And especially as those improvements to SNAP happen in ways that really sort of ensure that people can continue to receive benefits that it doesn't increase burden to the place where it creates a barrier to receiving that, that.
- [24:25]
A
Thank you to Rebecca Piazza for participating in that conversation. You can subscribe to the Priorities podcast@monities podcast.com and wherever you get your podcast while you're there, be sure to leave a review or rating on the podcast page. That small extra step helps more people like you find the show. This podcast is a production of Scoop News Group in Washington, D.C. adam Butler and Carlin Fisher help put it together. Until next week, I'm Sophia Foxoell. Thanks for listening.