The Cost of Proving It Again

The appointment letter gives a time. It does not provide the time.

A man needs to complete an interview for a public service. The appointment falls during a shift he cannot easily change. He calls to ask about another slot, waits, leaves a message and returns to work. By the time someone calls back, he is unable to answer.

From the office, the record may look simple: an interview has not been completed. From his side, it is a small coordination problem with consequences for the household budget.

Both descriptions concern the same event. A good system should be able to see both.

Public programs need rules. They may have limited funds, eligibility conditions and duties to protect against mistakes and misuse. Asking for information can serve those purposes. The question is whether the route for gathering it helps the institution reach a correct decision, or quietly substitutes persistence and free time for the facts that are supposed to matter.

A right still needs a route

Imagine that a service is available to everyone who meets a clear set of conditions. The policy sounds complete. Yet a person must discover the service, understand the conditions, gather the evidence and deliver it through an accepted channel.

Each step creates a possible gap between being eligible and receiving a decision.

A person may never hear about the program. They may misunderstand a question. Their employer may take time to provide a record. An uploaded file may fail without a clear error message. A letter may arrive after the person has moved. None of these possibilities tells us whether the person meets the policy’s actual conditions.

This is why process design belongs inside the policy debate. The public promise is carried through the process. If the process fails, the promise may not reach the person it was meant to serve.

The same concern applies outside government. A worker may qualify for reimbursement but struggle to produce a receipt in the required format. A customer may be entitled under an agreement to a repair but unable to find the correct request route. An institution can preserve the appearance of a benefit while making access depend on practical resources it never names.

We should measure those resources. Time away from work, travel, repeated document requests and time spent waiting for an answer are real costs, even when no fee appears on the screen.

Test the step before blaming the person

The most useful research asks whether changing a specific part of the route changes what happens.

In a Los Angeles County SNAP experiment, applicants were randomly offered an added, applicant-initiated unscheduled interview route. Interviews and verification remained. Approval rose, especially early. Later estimates include reapplications. Scheduled interviews were not replaced.¹

Flexible interviews changed approvals

Additional approvals per 1,000

Measure	Value
Scope	Los Angeles, 2020–21; randomized offer; 64,126 applications
By day 5	+139.0 [132.5, 145.5]
By day 30	+62.3 [53.9, 70.7]
By day 150	+21.9 [13.7, 30.1]
Axis	0, 75, 150
Limits	Adjusted effects; approximate 95% confidence intervals. Day 150 includes reapplications, not continuous receipt.

Source: Giannella et al., 2024, Tables 5–6

Figure 3. A flexible interview changed access without changing eligibility rules.

The purpose of an interview should be to obtain reliable information. A schedule is one way to organize that work. If another safe route can gather the same evidence, it deserves consideration. The institution needs to test the route and pay for enough staff to run it.

There may be other explanations for a missed appointment. A person may have changed their mind, misunderstood eligibility or no longer need the service. The responsible response is to investigate, with data capable of distinguishing possibilities where that is feasible. A record of noncompletion by itself is a weak basis for a judgment about character.

I would like institutions to be much slower to use words such as uncooperative when their evidence consists of a failed connection. Perhaps the person did refuse to cooperate. Perhaps the institution offered one narrow window during a complicated week. Those situations call for different remedies.

This is not an instruction to keep every case open forever. Deadlines can be necessary. It is a request to make the route workable before treating failure to complete it as proof that the person did not care.

Help changes who comes through

We should also test an appealing assumption: if a process becomes easier, the people with the greatest need will necessarily benefit most.

Finkelstein and Notowidigdo randomized outreach to older Pennsylvania Medicaid enrollees likely eligible for SNAP but not enrolled. At nine months, enrollment was 5.8 percent in the comparison group, 10.5 percent with information and 17.6 percent with information plus application help. New participants reached by the interventions were, on measured characteristics, less sick and had higher net resources than comparison-group enrollees. Easier access increased take-up without automatically concentrating gains among the most disadvantaged.²

This is a reason to improve the question, rather than retreat to the belief that difficulty sorts people fairly.

A person with an urgent need may face barriers that a letter or a short call does not remove. Another person may be close to applying already and respond quickly to a reminder. Different kinds of help can reach different groups. An agency should decide which gap it is trying to close and check whether the intervention reaches those people.

The choice is not between making everything difficult and declaring every simplification a success. There is room for a more serious design: reduce unnecessary work, provide help suited to the people still excluded, and track accurate decisions as well as total participation.

This requires resources. It is easier to announce a streamlined service than to fund the staff who answer questions when the streamlined path fails. A reform that removes one office visit but creates three unresolved phone calls may have moved the burden again.

The paper is a means

Consider a request for proof of address. The institution has a reason to know where someone lives. It accepts a particular kind of bill as evidence. A person without that bill is told to return when they have it.

At this point, two questions have become tangled. Does the person live at the address? Do they possess the preferred document?

The document may be a reliable and convenient way to answer the first question. It should not become the purpose of the exercise. If another safe and lawful form of evidence can answer the question, the process should explain whether and how it can be used.

Some requirements come from law. Some come from policy. Some come from software choices or local habits. Frontline staff should know which kind they are applying. Otherwise a change that needs a legislative decision and a change that needs a form edited can both be presented as impossible.

This is a modest but powerful place to begin an institutional review: list what each requested item is meant to establish. If nobody can explain the purpose, investigate before continuing to collect it. If two items establish the same thing, ask why both are needed. If a record is already available through an authorized and reliable route, ask whether the person needs to supply it again.

Privacy must remain part of that discussion. Reusing data can save effort and still create risks if information moves too widely or is wrong. The answer should not be unrestricted sharing across every institution. It should be a limited, lawful arrangement with a clear purpose, appropriate permission and a way to correct an error.

The person should not have to choose between repeating every fact forever and losing control of where the facts go.

A small error can become a long delay

A date is typed in the wrong format. A scan cuts off one corner. A name appears differently in two records. These are ordinary possibilities in any process handled by people and machines.

The response to them matters.

A clear request identifies the problem and the acceptable way to fix it. A poor request says only that the information is incomplete. The person must guess which field failed, resubmit material and wait again. The institution may then count the new submission as another piece of work, while the applicant counts it as another lost evening.

For a serious decision, it can be appropriate to pause until the record is correct. But the pause should be connected to a clear action. A person needs to know what is missing, why it matters, when it must arrive and how to confirm that it was received.

An error message is part of the service. It should be written with as much care as the invitation to apply.

There is also a question of proportion. An institution may need to reject a document that cannot be trusted. It may be able to correct a harmless formatting issue without requiring a fresh application. The rules should distinguish these situations instead of forcing all errors through the most severe response.

Staff need authority for reasonable correction, and the boundaries of that authority should be recorded. Otherwise one applicant receives help while another receives a refusal for the same problem. Kindness alone is an unreliable administrative policy when access depends on meeting the right person.

What the budget leaves out

An office can report that its costs have fallen after moving a service online. That may represent a real improvement. It may also leave out work now done by applicants, family members, libraries or community organizations.

To judge the change, we need to ask what happened across the whole route.

Did people complete the task more quickly? Did accurate decisions arrive sooner? Did the new system create failures for people without reliable devices or private places to upload documents? Did staff spend less time on routine entry but more time untangling errors? Were people who abandoned the process included in the assessment?

These are proposed questions for evaluation, not claims about every digital service. Online access can be excellent. A person may welcome the chance to complete a task after work instead of traveling to an office. The benefit should be preserved while gaps are fixed.

The mistake is to assume that removing a staffed channel automatically removes the need that channel met.

A mixed service may cost more to run than a single digital route. That additional cost should be compared with the value of correct access and the burdens avoided. We should be honest about the public choice. If a society wants a service to reach people with varied circumstances, it must pay for some variety in the route.

It is easy to praise inclusion until the budget for answering the phone appears.

Integrity deserves a better measure

A program needs protection against incorrect awards, deliberate abuse and careless administration. Those are real concerns. They do not justify assuming that every difficult step is a useful protection.

A well-designed check should target the fact that matters and have evidence that it improves decisions. An unnecessary obstacle can deter people who meet the rules while doing little to stop a determined false claim. Whether that happens in a specific program is an empirical question. It should be investigated rather than settled by the reassuring appearance of strictness.

There are costs on both sides of an error. Giving a service to someone who does not qualify can waste resources and undermine trust. Failing to provide it to someone who does qualify can defeat the purpose of the program. A responsible institution should examine both, including how long an error lasts and how hard it is to correct.

The measures should be chosen carefully. A high rejection rate is not automatically evidence of strong integrity. A high approval rate is not automatically evidence of fair access. We need information about the correctness of the decisions and the population that tried, or could reasonably have tried, to use the service.

This is harder work than counting completed files. It is closer to the work the institution was created to do.

The applicant’s time belongs in the room

Imagine the office reviewing the man’s missed interview. Instead of closing the discussion with the record of a missed call, it traces the attempts on both sides. It discovers that he called during the available hours and the return call came while he was working. It offers a route he can use and completes the interview.

That resolves one case. The larger improvement comes if the office notices that the same pattern recurs and changes the process for later applicants.

The rule does not need to lose its purpose. The evidence can still be checked. The decision can still be made under the proper conditions. What changes is the assumption that all coordination costs belong to the person seeking help.

This is where fairness becomes concrete. A service can respect public money and a person’s afternoon. It can ask for proof without making uncertainty humiliating. It can record a missed step without inventing a moral story about the person who missed it.

The appointment letter should still give a time. The institution should also help make that time possible.

Sources and limits