The Hidden Price of a Free Choice

A program can be free and still be expensive. A health plan can offer abundant choice and still deprive people of control. A subsidy can improve one outcome while quietly creating another burden that never appears in the spreadsheet.

This points to a deeper problem in policy design: we often measure the price of an intervention without measuring the difficulty of using it.

That omission matters because people do not encounter programs as abstract bundles of benefits and costs. They encounter them through waiting rooms, forms, travel, uncertainty, confusing menus, delayed rewards, and decisions made while sick, tired, or financially stressed. The real effectiveness of a policy depends not only on what it provides, but also on what it asks human beings to understand, predict, and do.

The most useful way to connect cost effectiveness analysis with behavioral economics is to treat them as two halves of the same question:

What does an intervention cost, and what kind of human decision does it require in order to work?

A serious evaluation must answer both.

The spreadsheet is not the experience

Imagine a program that provides school materials to low income families. Its budget includes staff salaries, transportation, procurement, and administrative overhead. But the program also requires parents to attend three meetings, each involving a long journey and several hours away from work. Those hours may not appear as a line item, yet they are real costs. If the program were expanded, the burden on families would affect participation, retention, and ultimately the measured impact.

Now imagine that the materials were donated during the pilot. The financial accounts might suggest an unusually cheap intervention. But if comparable goods would need to be purchased in a larger rollout, the pilot has understated its true resource requirement. A free input is not necessarily a costless input. It may simply be a cost paid by someone else, or postponed until scale.

Cash and in kind transfers create a related puzzle. From the perspective of the implementing organization, a transfer may look like an expenditure. From the perspective of a social evaluation, it may be better understood as a transfer of purchasing power rather than a resource consumed. Yet it still matters enormously for the program’s design and for the lives of recipients. The accounting category and the human meaning are not identical.

These distinctions reveal why cost effectiveness is more than a ratio. A ratio such as dollars per additional student attending school appears precise, but it can conceal three different questions:

1. What resources did the intervention use?
2. What burdens did participants bear to receive it?
3. What benefits, beyond the headline metric, did it produce?

A conditional payment might increase attendance, for example, while also helping a household smooth consumption, reduce financial stress, or improve bargaining power within the family. Judging it only by attendance may make the program look inefficient, even if attendance is merely one of its benefits.

This is not an argument for adding every imaginable benefit until every intervention looks successful. It is an argument for making the unit of analysis honest. A policy should be evaluated as a lived system, not just as a financial transaction.

The hidden variable is decision difficulty

Behavioral economics adds a crucial insight: many policies fail not because people lack motivation, but because the policy asks them to make decisions that human beings handle poorly.

Health care is an especially vivid example. Choosing whether to buy insurance, selecting among plans, deciding when to seek treatment, and estimating future medical needs all involve uncertainty, time, and complexity. People must assess probabilities they do not understand, compare prices that are difficult to calculate, and predict how they will feel under health conditions they have never experienced.

These are not minor imperfections in an otherwise rational process. They are systematic sources of error. People often misjudge probabilities, place too much weight on immediate costs, and struggle to forecast their future preferences. Someone who is healthy today may underestimate the value of insurance because the premium is concrete while illness is abstract. Someone who becomes sick may discover that the emotional impact is less devastating than anticipated, meaning that people can also overestimate how much a future disability will reduce their happiness.

The result is a peculiar mismatch between policy design and human capacity. Reformers may say that individuals should take greater responsibility for choosing insurance or health care. But responsibility is not the same as competence, and a choice architecture can assign responsibility without providing the conditions needed to exercise it well.

Consider a menu of twenty insurance plans. In theory, more options allow each person to find a better fit. In practice, the extra options may increase comparison costs, make important differences harder to detect, and encourage people to focus on a visible premium rather than on deductibles, exclusions, or future risk. The apparent freedom of choice can become a tax on attention.

That tax belongs in the analysis of the policy. If a subsidy produces a technically affordable plan but requires recipients to navigate an opaque system, then its effective benefit is lower than its nominal value. A person who fails to enroll because the process is confusing has not received the value promised on paper.

We can express this with a simple mental model:

Effective impact equals stated benefit multiplied by usability.

Usability includes time, clarity, predictability, trust, and the number of decisions required. A program with a large theoretical benefit but low usability may produce less real world value than a smaller program that people can reliably access.

This also changes how we interpret nonparticipation. It is tempting to conclude that people do not value a service if they fail to use it. But nonuse may reflect transaction costs, present bias, uncertainty, or an inability to predict the future. A person may value insurance deeply in principle and still fail to enroll before a deadline. A parent may value an education program and still miss meetings because the immediate cost of attendance is more salient than the distant benefit.

The policy question is therefore not simply, “How much is this benefit worth?” It is also, “How much mental and practical work must a person perform to obtain it?”

From cost accounting to friction accounting

The intersection of these ideas suggests a broader framework: friction adjusted evaluation.

Traditional cost analysis tracks money and physical resources. A more complete evaluation tracks at least four layers of cost:

1. Resource cost

These are the obvious inputs: staff, facilities, transport, procurement, technology, and funds transferred to beneficiaries. They matter because a program that works at small scale may require very different resources at national scale.

2. Participation cost

This includes travel, waiting, paperwork, missed work, childcare, and other burdens imposed directly on participants. Participation costs can be especially important when the program serves people with the fewest spare resources. One hour of waiting is not socially neutral when the person waiting loses wages or must arrange care for a child.

3. Cognitive cost

This is the work of understanding the program, comparing alternatives, remembering deadlines, estimating risk, and predicting future needs. Cognitive costs are often invisible because they are not paid to anyone. Yet they can determine whether a person makes a good choice or makes any choice at all.

4. Error cost

This is the damage caused by predictable mistakes. A confusing insurance menu can lead someone to choose inadequate coverage. A delayed benefit can be discounted too heavily. A complicated enrollment process can exclude exactly those people least able to absorb uncertainty.

A program can be inexpensive in the first category and costly in the other three. That is why low administrative spending does not automatically mean high efficiency. Sometimes an institution reduces its own costs by shifting work onto beneficiaries.

The distinction resembles a restaurant that advertises a cheap meal but requires customers to cook half of it themselves, bring their own utensils, and wait an hour in line. The price is low because part of the production process has been transferred to the customer. In public policy, that transfer is often treated as efficiency even when the people doing the extra work are the least equipped to do it.

Friction accounting does not require placing a dubious monetary value on every minute of confusion. It requires identifying the friction, measuring it where possible, and testing whether it changes who participates and what outcomes follow. A short survey about time burdens, administrative experiments comparing simple and complex forms, or enrollment data by demographic group can reveal costs that a budget cannot.

The danger of optimizing one visible outcome

A second connection concerns the temptation to optimize a single metric. Cost effectiveness analysis needs a defined outcome, but the chosen outcome can become a tunnel through which every other consequence disappears.

Suppose a conditional transfer increases school attendance at a certain cost per additional day attended. It may look less attractive than a cheaper intervention that produces the same attendance effect. But perhaps the transfer also reduces hunger, stabilizes household income, and gives parents greater freedom to make decisions. The attendance statistic is real, but incomplete.

The same problem appears in health care. QALYs and similar measures are useful tools for comparing health improvements, but they rely on judgments about how much different health states affect well being. People are not always good at predicting those judgments in advance. They may imagine illness as more unbearable than it proves to be, or fail to appreciate the value of adaptation, social support, and changed priorities.

This creates a measurement paradox: a policy can be evaluated using preferences that are themselves shaped by poor prediction. If people systematically misforecast their future well being, then asking what they would prefer in advance may not fully reveal what improves their lives afterward.

The answer is not to abandon measurement. It is to use multiple perspectives and remain modest about any single metric. A robust evaluation might distinguish:

• The outcome directly targeted by the program.
• Other outcomes that plausibly matter to recipients.
• The burdens required to obtain the benefit.
• The distribution of benefits and burdens across groups.
• Whether the program improves people’s choices or merely asks more of their attention.

This last distinction is particularly important. A policy may produce a favorable average outcome while worsening inequality in decision quality. Sophisticated participants may navigate a complex subsidy successfully, while people with less time, education, or confidence receive less value. The average can improve even as the system becomes more dependent on individual administrative skill.

A useful test is to ask: Does the policy reward resources that are unrelated to the policy’s stated goal? If obtaining health coverage depends heavily on literacy, spare time, or comfort with probability, then the system is partly selecting for administrative competence rather than health need.

Designing programs that respect human limits

The practical implication is not that every program should eliminate choice or that beneficiaries should be protected from all complexity. It is that complexity should be treated as a design decision with measurable consequences.

Start by collecting cost data during implementation, not months later when forgotten burdens must be reconstructed. Ask participants how much time they spend traveling, waiting, learning, and complying. Record donated goods and services, and distinguish between money that is consumed and money transferred to households. These steps make the cost side more truthful.

Then test the decision environment itself. Compare a long form with a short form. Compare a large menu of plans with a curated set. Compare a deadline based process with automatic enrollment and an option to opt out. Compare a subsidy that requires active annual choices with one that adjusts predictably over time.

The goal is not to make every decision for people. It is to reserve human attention for decisions where personal judgment genuinely adds value. In many cases, good policy will combine a simple default with transparent opportunities to choose differently. That structure acknowledges behavioral limits without treating citizens as incapable.

Finally, evaluate more than the immediate target. If a program has several legitimate objectives, state them before judging its efficiency. A transfer may be valuable partly because it improves education and partly because it provides security. An insurance reform may matter partly because it reduces financial catastrophe and partly because it makes medical choices easier to navigate.

The best intervention is not necessarily the one with the lowest cost per unit of measured impact. It is the one that produces valuable outcomes without exporting hidden work and avoidable error onto the people it is meant to help.

Key Takeaways

1. Count the costs people bear directly. Include travel, waiting, missed work, childcare, and required meetings when they affect participation or scale.

2. Treat complexity as a real policy cost. Measure the time and effort required to understand options, complete forms, compare plans, and meet deadlines.

3. Separate the target outcome from the full value of the program. Track secondary benefits such as income stability, reduced stress, autonomy, and protection against risk.

4. Design for predictable human errors. Use clear defaults, fewer unnecessary choices, reminders, automatic renewal where appropriate, and simple explanations of uncertainty.

5. Ask who benefits from complexity. If success depends on spare time, education, or administrative confidence, the policy may be distributing value according to navigation skill rather than need.

The deepest lesson is that efficiency is not merely about producing more benefit with fewer dollars. It is about building systems in which people can actually convert resources into well being.

A free service that demands a day of travel is not free. A generous subsidy that requires a maze of decisions is not fully generous. And a policy that gives people responsibility without accounting for the difficulty of responsible choice may be shifting costs rather than reducing them.

The future of serious policy evaluation lies in joining two disciplines that are often kept apart: the discipline of counting resources and the discipline of understanding minds. Once we do that, the central question changes. We stop asking only how cheap an intervention looks from above, and start asking what it feels like, and what it requires, from within the life of the person who must use it.