Last updated October 25, 2026
For tax year 2022, about 4 in 5 eligible US taxpayers received the Earned Income Tax Credit, according to participation estimates from the US Internal Revenue Service (IRS) and Census Bureau, leaving about 1 in 5 eligible taxpayers without it. In Canada, just over 2 in 5 eligible children and young people had received the Canada Learning Bond by the end of 2024, according to Employment and Social Development Canada's 2024 statistical review.
Eligible people complete public benefits applications when agencies remove the steps they cannot finish alone, say plainly what they would receive, follow up after the first notice, and offer hands-on help where applications stall. At The Decision Lab, we are an applied behavioral science firm, and our public-sector work with Health Canada, the World Bank, the City of Rome and the Government of Lesotho starts from the same position: when eligible people do not act, the process is usually what stops them, and any fix has to be tested with the people it is meant to reach.
What works best to increase public benefits uptake
In the order we would prioritize them, based on the research in this guide and our own public-sector work:
- Remove or simplify steps, such as repeated document requests and accounts people must create before they can start, and use data the agency already holds to pre-fill or verify information where program rules allow.
- Give hands-on help to people who stall partway through an application.
- State the likely benefit amount in every notice, and send a follow-up when the first one gets no response.
- Test each change against the existing process before scaling it.
These four priorities draw on the seven fixes covered later in this guide. Outreach on its own sits lowest on the list because, in the largest outreach study in this guide, it did not reach the people furthest from applying.
Why eligible people do not apply for government benefits
Researchers call the time and stress of dealing with government "administrative burden." In their study of administrative burden in citizen-state interactions, Donald Moynihan, Pamela Herd and Hope Harvey separate the cost of finding out about a program from the cost of complying with its rules, and add the psychological cost of stigma and stress on top. In benefits programs, we see those costs show up as five barriers to public benefits enrollment.
- People do not know they qualify, or what they would get.
- Notices and forms ask more than people can manage in one sitting.
- Logins and identity checks stop people before the application begins.
- Office access and deadlines screen out the people with the least slack.
- Messages meant to reduce stigma rarely move behavior, and sometimes backfire.
1. People do not know they qualify, or what they would get
A program that people have not heard of, or think does not apply to them, gets no applications however simple the form is. Saurabh Bhargava and Dayanand Manoli tested this in Bhargava and Manoli's IRS field experiment, mailing new notices to 35,050 tax filers who had not claimed the credit after a first notice. Adding the size of the benefit to the headline of a simplified notice raised the share who claimed from about 1 in 4 to about 1 in 3, compared with the same simplified notice without the amount.
2. Notices and forms ask more than people can manage in one sitting
In the same IRS experiment, the share of filers who claimed rose from 14% with a complex notice to 23% with a simpler version, a gain of 9 percentage points, or about two-thirds. The gain from simplification was roughly twice as large for the lowest-income filers as for those above the median. Application forms carry the same risk: each extra document request or ambiguous question is a point where a person under strain can put the form down and not come back.
3. Logins and identity checks stop people before the application begins
The Digital Benefits Network at Georgetown's Beeck Center reviewed 158 online applications across six US programs, from food assistance to unemployment insurance, in Georgetown's Digital Benefits Network dataset on digital identity in public benefits, published in 2023. About 3 in 4 applications required people to log in or create an account before they could start, and about 1 in 3 required or prompted some form of identity proofing. Just under half of unemployment insurance applications used biometric checks. Each of these steps protects against fraud, and each is also a place where an eligible person can drop out before the agency knows they exist.
4. Office access and deadlines screen out the people with the least slack
When the US Social Security Administration closed field offices, disability applications in the surrounding areas fell by 10% and new recipients by 16%, compared with areas where offices stayed open, according to Deshpande and Li's study of field office closings and disability applications. In the areas studied, applications from people without a high school diploma fell by 14%, against 5% for college graduates.
Deadlines work the same way, as Tatiana Homonoff and Jason Somerville found when they studied renewals for the Supplemental Nutrition Assistance Program (SNAP), the US federal food assistance program, in San Francisco. There, caseworkers assigned interview dates at random within the last month of a person's eligibility. In the SNAP recertification study, people given a late-month interview, and so only a few days to finish their paperwork, were 10 percentage points less likely to renew than people given an early-month interview. The people who lost benefits this way were no less needy than the average recipient.
5. Messages meant to reduce stigma rarely move behavior, and sometimes backfire
Stigma is real, but in the evidence so far, rewording a letter to address it is a weak lever. In the Bhargava and Manoli experiment, describing the credit as an earned reward had no effect on claims compared with the plain notice. Telling people that many of their peers claimed the credit lowered the share who claimed by about a fifth. We treat stigma as something to test for in each population, since a single sentence in a notice is unlikely to shift it.
How government agencies can improve public benefits uptake: 7 evidence-based fixes
1. Measure where applicants drop out before redesigning anything
Most agencies know how many applications they approve and far less about how many were started and abandoned. We start public-sector projects by mapping each stage of the process, and for benefits those stages run from awareness through starting an application, passing login and identity checks, submitting, approval, first payment and renewal. In our behavioral diagnostic for the World Bank in Uganda, we used a standard behavior change model to sort field research into 26 specific barriers to adopting clean cookstoves, including the fact that the cheapest model was too heavy for most women to carry home on their own. No message would have fixed that one, and benefits processes have their equivalents, such as a document people cannot get or an interview time they cannot make.
2. Redesign the process before adding outreach
For the people furthest from applying, the evidence favors changing the process over adding outreach. Elizabeth Linos, Allen Prohofsky, Aparna Ramesh, Jesse Rothstein and Matthew Unrath ran six preregistered field experiments with a combined sample of about a million, reported in Linos and colleagues' study of nudges and Earned Income Tax Credit take-up. They varied the content, design, messenger and mode of the outreach, and found no evidence that any version raised filing or claiming among households that had not filed a tax return, compared with households that received no outreach.
Canada is taking the process route with the Canada Learning Bond. From 2028, eligible children born in 2024 or later will be enrolled automatically at age four if they have a Social Insurance Number and no education savings plan of their own, and Employment and Social Development Canada expects automatic enrolment to bring about 130,000 more children a year into the program. Where full automatic enrollment is not possible, pre-filling forms with data the agency already holds and asking for documents only at the point they are needed both cut the number of decisions a person has to make before anyone can help them.
3. Rewrite notices so they are short and state the amount
When agencies do send notices, the Bhargava and Manoli results give a clear starting template: put the likely benefit amount in the headline and make the next step a single action. In our work, we test redesigned notices against the version already in use, because a notice that reads better to staff does not always perform better with applicants.
4. Send a second notice
In the IRS experiment, about 1 in 5 people who had ignored a first notice claimed the credit after receiving a second one, months later. A follow-up costs little compared with the first outreach, and agencies often skip it.
5. Offer hands-on help with the application itself
In the studies below, information alone helped some people and assistance helped far more. Amy Finkelstein and Matthew Notowidigdo ran a randomized SNAP outreach trial with about 30,000 older Pennsylvanians who were enrolled in Medicaid and likely eligible for food assistance. Over nine months, 6% of the comparison group enrolled, against 11% of those who received a letter and a reminder postcard and 18% of those who also got phone help with the application. The H&R Block experiment on the Free Application for Federal Student Aid (FAFSA), run by Eric Bettinger, Bridget Terry Long, Philip Oreopoulos and Lisa Sanbonmatsu, found a similar split: information alone had no measurable effect, while completing the form with families raised submission among dependent students from about 2 in 5 to more than half.
6. Let people start before asking them to prove who they are
Where fraud rules allow, we recommend letting applicants check their likely eligibility before they create an account, and tracking drop-off at the identity step as its own figure. An agency that only measures completed applications cannot tell whether its identity check is stopping fraud or stopping eligible people. Every identity process weighs fraud prevention and program integrity against privacy and equal access, and the right balance differs by program. Offering a route through verification that does not depend on a smartphone or a credit history keeps the check from screening out the people the program is for.
7. Schedule renewals and deadlines around people's capacity
In San Francisco, the SNAP recertification study showed that a few days of difference in time to complete paperwork changed who kept their benefits. Agencies can give everyone the longest workable window before a deadline, and reuse information from the original application at renewal instead of asking for it again.
How to test benefits enrollment improvements
A change to a benefits process should be judged against the process it replaces, measured stage by stage and broken down by group. Staged or randomized rollouts make this easy, because some applicants keep the old process for a few weeks and become the comparison group. Controlled experiments give the strongest evidence, but they are not always practical for a government program, for example when a rule change must reach everyone at once. In those cases, we compare against similar regions or earlier periods and say plainly how much weaker that evidence is.
Testing by group matters because the same change lands differently across communities. For the City of Rome's participatory budgeting campaign, we tested message framings with nearly 10,000 residents, measuring their intent to take part and comparing how different communities responded, so that the strategy would reach groups whose voices are often underrepresented.
Behavioral science for benefits enrollment at The Decision Lab
The research earlier in this guide is the direct evidence on benefits enrollment, and none of it is ours. Our published case studies do not yet include a benefits enrollment project. What we bring is adjacent public-service work and the testing method described above. These projects differ in subject, but in each one people had to complete a demanding, high-stakes task with limited time, attention, documents, digital access and trust in the institution asking.
We helped redesign Wellness Together Canada, a mental health platform funded by the Government of Canada that was used by more than 1.2 million people in its first year. People in distress are more prone to information overload and give up quickly if they cannot find what they need, a risk any benefits applicant faces halfway through a long form. For the Government of Lesotho, we built a behavioral framework to increase voluntary tax compliance, where the task was getting citizens to complete a required filing. With Health Canada, we worked on organ donation governance, where the behavior that needed to change belonged to the officials running the system. Our guide to behavioral science in the public sector covers this wider work.
Frequently asked questions about public benefits enrollment
Why do eligible people not apply for government benefits?
Many eligible people never learn they qualify or what they would receive. Others start and stall at paperwork, logins, office visits or deadlines. Research on US disability programs found that closing local offices cut applications most among people with less education and income, so these costs fall hardest on the people a program is usually meant to reach.
How can government agencies improve public benefits uptake?
Agencies usually get the largest gains by changing the process, for example by enrolling people automatically from data they already hold, as Canada plans for the Canada Learning Bond from 2028. Hands-on help with the form comes next. Clearer notices that state the benefit amount, plus a second notice, add smaller gains on top.
Do reminder letters increase benefits take-up?
Reminder letters help mostly people who are already close to applying. A second IRS notice led about 1 in 5 people who had ignored the first one to claim their credit. But six large experiments with a combined sample of about a million found that well-designed outreach did not move households that had not filed a tax return at all.
Does identity verification reduce benefits enrollment?
It can, and the agency running the check is best placed to find out by how much. Counting how many applicants reach the identity step and how many get past it, split by age and by device, shows whether the check is stopping eligible people. Alternatives such as in-person or mailed verification can then be tested against that baseline.
What does a behavioral science firm do for benefits enrollment?
A behavioral science firm finds where and why eligible people drop out, then redesigns the process around those points and tests each change against the existing version. At The Decision Lab, we bring public-service work for clients including Health Canada and the World Bank, and we report results by group so gains among easy cases cannot hide losses elsewhere.
