Tim Harford thinks the government has a problem with failing. But not the problem you probably think it has with failing. Harford’s worry isn’t that the government fails too much. It’s that it fails too little, and too unenthusiastically.

Harford, a British economist, is the author of the compelling new book, “Adapt: Why Success Always Starts With Failure.” In it, he warns that “we face a difficult challenge. The more complex and elusive our problems are, the more effective trial and error” — which is to say, failing and learning from those failures — “becomes, relative to the alternatives. Yet it is an approach that runs counter to our instincts, and to the way in which traditional organizations work.”

Health-care costs prove his point perfectly. Few policy problems are more confounding than the inexorable rise in health-care spending. But the only way to fix it runs counter to both our instincts and our political system: We need to allow ourselves to fail — often, enthusiastically and, above all, constructively.

When you hear the words “health-care costs,” it’s good to apply a quick mental auto-correct. What people really mean are “sick-person costs.” When 5 percent of patients account for 50 percent of spending, you’re not talking about the costs that most Americans with health-care insurance rack up over the course of a year. You’re talking about the costs racked up by a tiny fraction who suffer from serious health conditions.

If you travel internationally, you’ll see that other countries have come up with fairly simple ways to keep those costs down. One option, which every other developed nation employs, is to simply use the blunt force of government purchasing power to set low prices. That’s how you get American seniors heading to Canada to purchase American-made pharmaceuticals at a steep discount. Wal-Mart has nothing on those guys. But it’s not just the Canadians. And it’s not just drug costs. A 2007 McKinsey Co. study surveyed a range of health-care delivery systems and concluded, after making adjustments for purchasing power, that we were overpaying for health-care services and products to the tune of $500 billion a year. That’s about 20 percent of our total annual health-care spending.

Our system doesn’t allow Medicare to strike those sorts of bargains. On the health side, if Medicare doesn’t pay enough, doctors can opt out of the program. On the political side, legislators are reluctant to impose deep cuts on their good friends and campaign contributors in the medical industry. When Congress passed the Medicare Prescription Drug Benefit in 2003, for instance, it explicitly prohibited Medicare from bargaining for better prices on drugs.

Another option is to price people out of care. This is essentially the solution proposed by Rep. Paul Ryan of Wisconsin. The Congressional Budget Office estimates that if Ryan’s plan is enacted, in 20 years seniors will pay an average of 70 percent of their insurance costs, compared with about 30 percent under the current system. But Ryan’s plan appears dead.

So what’s left is finding savings in the health-care system itself. And there are opportunities there. The best estimates suggest that between one-fifth and one-third of all care is wasted. But figuring out which treatments are helping and which are hurting is easier said than done. Reforming health-care provider pay so that care gets better, instead of simply stingier, is tricky. Upgrading our health-care information infrastructure has been a decades-long battle. In effect, we’ve been failing at these tasks.

But we have not been failing in the right way. Encouragingly, the Affordable Care Act might change that. There’s the Center for Medicare and Medicaid Innovation, whose job is to generate pilot and demonstration projects that might help us learn to do things differently, and better. There’s the clunkily named Patient-Centered Outcomes Research Institute, which is tasked with compiling evidence on the effectiveness of various treatments, and when and how they’re best used. And there’s the Independent Payment Advisory Board, a group of experts empowered to take the best of these experiments and replicate them through the Medicare system.Taken together, those agencies could do a lot of constructive failing and a bit of real succeeding, too.

The only question is whether we’ll let them. In a two-party political system, one party is often better off if the customers are very, very unhappy. In fact, the minority party’s route to majority status is to keep the customers unhappy, which means contributing to failures and then obstructing the governing party’s efforts to fix them.

Identifying, preserving and highlighting policy failures is a great way to win an election. But from a policy perspective, it’s a bad way for the government to fail. It makes the governing party overly averse to policy risks and leaves the political system incapable of learning from mistakes.

“Any politician knows he can have 50 policies going well and one failure,” Harford tells me from his hotel room in Seattle, “and that failure will dominate the next campaign. So the politician is just desperate to avoid provable failure.” That means politicians won’t discover the kinds of successes that emerge from constructive failures.

So when it comes to health-care costs, politicians who fear failure -- or exploit it for political gain -- should ask themselves what, exactly, they think the alternative is. If we can’t learn to fail well, then we’re going to, well, fail.