Averages do not predict individual experiences, and individual experiences do not negate the average. They can contrast without presenting a conflict. All too often, policymakers lose sight of this simple truth – with repercussions for their credibility, and for programs that could otherwise make millions of Americans better off.

Consider the case of the expanded Child Tax Credit. In 2021, it delivered as much as $3600 to every child in the U.S., distributed monthly, whether the parents were working or not. Real-time surveys showed what research later confirmed: If the government gives money to low-income families with young children, they spend it primarily on food, shelter and paying down debt. While the policy was in effect, it lifted an estimated 2.9 million children out of poverty.

Yet Congress ultimately let the credit expire, in part because the political discussion went awry. Opponents pointed to parents using the money to buy drugs or stop working. Proponents focused exclusively on the average effect, as if everyone used the money wisely – taking on a burden of proof that no program could possibly meet.

By recognizing the flaws and addressing them, advocates could have made a more persuasive case. Yes, some parents bought drugs, but depriving them of the tax credit would in no way counter the undeniable problem of addiction. Yes, some parents stopped working, but that says less about the specific policy than about the nature of work in the U.S.: no paid sick days, no requirements for predictable scheduling or time off between shifts, no protection from being fired for missing work to care for children.

Unemployment insurance is another example. On average, it works fine, providing people who have been laid off with a temporary cost-of-living subsidy while they search for new jobs. But it also fails individuals who, through no fault of their own, can’t find another job quickly enough – because their town’s biggest factory shut down, because they were replaced by robots, because the labor market was particularly weak. A system that considered such cases – by, say, providing extended benefits combined with robust retraining – could ensure that fewer people fell through the cracks and make the whole country more productive.

Recognizing both the rules and the exceptions, the aggregate effect and the individual experience, is a much more inclusive way to make policy. People’s concerns are driven by the lazy workers and drug addicts they’ve seen, by what has happened to them or to their friends. Saying their observations don’t matter is akin to denying their existence. They matter, even if they don’t dictate policy.

Economists like me forget this at our own risk. When we argue about the chances of a recession, or about what the Federal Reserve should do, we say things that directly contradict people’s lived experience. Broad statements such as “price growth is slowing” and “the labor market remains strong” inevitably ring false to families struggling to afford groceries or suffering from layoffs. The economy is large, with about 165 million workers. Anything that can be true will be true for someone. As a friend of mine recently put it: “I can’t afford rent anymore; I’m in a recession.”

No doubt, drawing far-reaching conclusions from anecdotal observations would be a horrible way to design policy. My point is that experts make a similar mistake when they belittle people’s observations, as if averages predicted individual experience. They don’t, and policymakers will be more effective if they keep this in mind.