The Proof of Harm

The surveillance that welfare systems apply to their recipients is not primarily designed to detect fraud. This is worth stating at the outset, because the fraud narrative is the one most often used to justify the surveillance’s intensity, and the fraud narrative is not supported by the evidence in any jurisdiction where the evidence has been examined carefully.

Benefit fraud rates in most wealthy countries are consistently estimated at between one and two percent of total welfare expenditure—lower than tax fraud, lower than corporate subsidy misuse, and sufficiently small that the administrative cost of the surveillance designed to prevent it frequently approaches or exceeds the cost of the fraud itself.

The surveillance persists not because it is effective at preventing fraud but because it is effective at something else: the visible demonstration that the system is not being exploited. The demonstration is the product. The product is not produced for the recipients. It is produced for the political audience that funds the system and whose continued support requires evidence that the support is not being abused. The recipient is the site of the demonstration. The demonstration is performed on their time, at their expense, using their energy, in conditions that the recipient did not choose and cannot avoid without losing the support they depend on.

 

The architecture that produces this is worth understanding as architecture rather than as policy failure, because understanding it as architecture clarifies why it persists across political changes that claim to reform it. The welfare system in most wealthy democracies was designed within a set of incentive structures that reward the demonstration of oversight and penalise the appearance of laxness. The politician who is seen to be rigorous about welfare eligibility has demonstrated responsibility to the taxpayer. The politician who is seen to be generous with welfare eligibility has exposed themselves to the charge of enabling dependency. The incentive is not neutral between rigour and generosity. It rewards the appearance of rigour regardless of whether the rigour is producing outcomes that justify its cost.

The bureaucracy that administers the system inherits these incentives and adds its own. The administrator who processes claims quickly and without extensive documentation has fewer records to defend their decisions against appeal or audit. The administrator who requires extensive documentation, repeated verification, and periodic re-assessment has produced a record that justifies each decision against any subsequent challenge. The incentive is not neutral between efficiency and documentation. It rewards documentation regardless of whether the documentation requirement is serving the recipient or the administrator’s liability protection.

The result is a system whose design makes extensive surveillance and repeated proof requirements the path of least institutional resistance. The surveillance is not a conspiracy. It is the predictable output of an incentive architecture whose primary pressures point consistently toward the visible demonstration of oversight rather than toward the welfare of the person the oversight is supposed to protect.

 

The person in the system encounters the output of this architecture as a set of requirements whose relationship to their actual situation is often difficult to discern. They are required to demonstrate, at regular intervals, that they remain eligible for the support they are receiving. The demonstration requires documentation: medical certificates, work capacity assessments, job search logs, income declarations, housing confirmations, statements from healthcare providers, reports from social workers, evidence of the compliance activities the system has specified as conditions of the payment. Each document has a cost: the time to obtain it, the transport to reach the office or the clinician who provides it, the financial cost if the provision is not free, and the cognitive and emotional cost of re-engaging with the evidence of the condition that produced the need for support.

For a person with a mental health condition, the requirement to repeatedly document the condition—to produce clinical evidence that the condition continues to exist, to describe its symptoms in the terms the assessment instrument requires, to submit to evaluation by clinicians who may have different frameworks from the treating clinicians who know the actual history—is a requirement to re-inhabit the condition in its most legible administrative form. The legible administrative form is not the form in which recovery occurs. Recovery tends to be incremental, variable, and not well captured by the standardised instruments the assessment uses. The assessment instrument was designed to produce consistent, comparable, defensible outputs across large numbers of cases. It was not designed to capture the specific trajectory of a specific person’s recovery through a specific condition in a specific context.

The energy spent producing the documentation is energy not spent on recovery. This is not a metaphor. It is a description of a resource constraint operating on a person whose resources are already limited by the condition the system is supposed to be supporting them through. The system requires the proof of harm at the expense of the capacity to reduce the harm.

 

The epistemic problem is the one that receives the least attention and is in some respects the most consequential. The welfare system has, over time, developed extensive capacity to know whether its processes are being followed and very limited capacity to know whether its processes are producing the outcomes the processes are designed to support. It knows how many assessments it has conducted. It knows how many people have been found eligible and how many ineligible. It knows its processing times and its appeal rates and its compliance rates. It does not know, in any systematic way, whether the people who received support are faring better as a result of having received it, whether the people who were found ineligible were correctly assessed, or whether the assessment instruments are producing accurate determinations or producing defensible ones, which are not the same thing.

This is the Goodhart dynamic operating within welfare administration. The system’s metrics are process metrics: the assessment completed, the document received, the compliance activity recorded, the payment issued or withheld. The outcomes the process is supposed to produce—the person’s increased capacity for self-sufficiency, the reduction of suffering, the support of recovery, the maintenance of dignity—are harder to measure, require longer time horizons, and are not within the process’s direct control in the way that the process metrics are. The system therefore knows its process well and its outcomes poorly, and the system’s incentives reward good performance on the process metrics without requiring good performance on the outcome metrics.

The progressive degradation this produces is not the system becoming cruel, though it can produce cruelty. It is the system becoming less capable of knowing whether it is succeeding. The system that has optimised its processes to the point where the processes run smoothly and produce consistent defensible outputs has optimised away the feedback mechanisms that would tell it whether the outputs are serving their purpose. The appeal rate is a proxy for this feedback. The appeal success rate is a more specific proxy. Both are gamed by the system’s incentives in the same way that all proxies are gamed when they become targets.

 

The proof requirement shifts the relationship between the recipient and the system from one of support to one of demonstration. The recipient who is asked to prove their trauma has been repositioned. They are no longer primarily the person the system exists to support. They are the evidentiary source for the system’s legitimacy demonstration. They are proving, to the system and through the system to its political audience, that the support being provided is support that a legitimately suffering person deserves. The suffering must be documented. The documentation must be current. The currency must be maintained through repeated assessment. The assessment must be survived, in the clinical sense, before it can be passed, in the administrative sense.

The person who is successfully supported by the system is eventually, in most welfare frameworks, a person who no longer requires the system’s support. The exit from the system is the measure of the system’s success. The exit requires the person to demonstrate that they have recovered sufficiently to no longer meet the eligibility criteria that admitted them. The demonstration is often the same assessment instrument, now applied to a different purpose: instead of proving that the condition exists, the person must prove that it no longer exists to the degree that justified the support.

The condition may not cooperate. Conditions that produce welfare dependency are often conditions that do not resolve cleanly within the assessment instrument’s categories. The person with a chronic condition does not recover from it. They manage it. The management is ongoing. The assessment instrument that was designed for acute conditions requiring temporary support is not well designed for the chronic condition requiring sustained accommodation. The person whose condition is chronic is required, at regular intervals, to prove to a system designed for acute conditions that their chronic condition continues to exist at a level that the acute-condition instrument can recognise as sufficient to justify support.

 

The surveillance that accompanies this is less a watching of the recipient than a watching of the system’s own process. The compliance requirements, the activity logs, the appointment records, the documentation chains—these are the system watching itself operate, producing the record that demonstrates the operation has occurred. The recipient is required to participate in the watching. Their participation produces the data that the system uses to know that it has done what it was designed to do.

What it was designed to do and whether it is doing it are questions that this data does not answer. The data answers: was the process followed? It does not answer: is the person faring well?

The two questions have different answers.

The system is very good at answering the first.

The system has limited means of answering the second.

The second question is the one the system was established to answer.

The system answers the first question, consistently and at great administrative expense, and calls this welfare.