Michael Greenstone is the Milton Friedman distinguished service professor of economics and director of the Becker Friedman Institute for Economics, both at the University of Chicago.

We should start paying people to get tested for covid-19.

Nearly everyone agrees that more testing is critical, but months into the crisis there is no plan for how to do it. The United States does not have the infrastructure to conduct millions of tests per day and aggressive contact tracing of the sort South Korea and Taiwan have done. The programs those countries employed are also not an easy fit with U.S. civil liberties and privacy protections.

An alternative is to learn about covid-19’s prevalence among all Americans. It’s unclear what share of the population is currently infected. Nor do we know how many have been exposed and developed antibodies, which would indicate how far we are from the approximately 70 percent necessary to attain herd immunity. This basic question has enormous influence on the pace at which society can confidently reopen — yet we are not even on a path to answers.

In at least a dozen states, testing capacity outstrips the supply of patients. Another approach is urgently needed.

No existing or planned covid-19 surveys meet the two fundamental conditions for testing results to reveal the true rate of spread.

First, the people invited to be tested must be representative of the general population. They should be drawn randomly. So far, much of what is known about U.S. cases comes from specialized populations: the sick and symptomatic, and essential workers. In Taiwan, 158 tests have been conducted for every confirmed case. In the United States, the rate is just eight and drawn largely from the groups mentioned above

Second, among those invited to be tested, those who show up must also be representative of the general population. Current approaches fall apart here because only a fraction of those able or invited to be tested follow through. The major covid-19 prevalence studies to date — Santa Clara, Iceland and even ongoing National Institutes of Health testing— have relied on the self-motivated or volunteers.

The problem is that volunteers differ from people who choose not to volunteer in ways that cannot be readily observed. This is why reliable polling is not based on specialized populations, such as people who have a particular interest in a topic, but on random sampling from the general population. In terms of the pandemic, people who are sick or suspect that they have been exposed might be more likely to seek testing, such that volunteers’ rate of infection could exceed that of the general population. In the other direction, the very health-conscious might jump at the opportunity to be tested, and their rate of infection might be below average.

Unrepresentative testing cannot produce a reliable compass to guide policy. If testing is not mandatory, individuals decide whether the benefits outweigh the costs: if the information and contribution to the public good are worth physical discomfort, general hassle and the invasiveness of sharing health information. These are not small problems. In Iceland, only 33 percent of those invited to be tested for covid-19 took up the offer; two-thirds of those invited decided the costs outweighed the benefits and went untested.

There is, however, another option. If economics has taught us anything, it is that financial incentives can change behavior. Offering payments for completed tests could change the calculus of those invited to be tested.

Some people might agree for $25. Some might follow through at $50 or $100. With each higher incentive, more people are likely to conclude that testing is worthwhile.

Absent the reliability of everyone being tested, plausible assumptions suggest it is possible to learn the infection rate in the general population by comparing how people’s infection rates vary with their willingness to be tested.

Governments could implement this approach inexpensively to learn about covid-19’s prevalence in their jurisdictions. Some colleagues and I are working to launch a program in Chicago, where we expect to determine a reliable measure of citywide prevalence with only 2,000 tests. We plan to draw a random sample of the population, invite the full sample to be tested and randomly offer incentives for completed tests.

This approach doesn’t require millions of tests or several hundred thousand contact tracers, averting privacy concerns. It is unhindered by shortages in testing supplies and protective equipment for those administering the tests.

Once launched, the tests could be repeated regularly to track how the disease evolves and even the protection offered by developing antibodies. The results could reliably be a guide to safely unlocking our economy. At lower rates of spread, partial reopenings could be based on data, not hope. And if we reach herd immunity, decision-making becomes easier.

Until we implement covid-19 testing that produces representative results, it is impossible to know the current prevalence of the disease or of past infection in the population. Policymakers need this information to navigate between the goals of protecting health and reopening the economy; without it, we all pay the cost in lives and livelihoods.

Better to pay smaller amounts, upfront, to encourage people to get tested.

Watch Opinions videos:

Read more: