How can Washington have spent so much time debating health care and still have failed to come up with a reasonable reform? By starting from the wrong premises. There are three basic questions to be asked about the design of any health-care system. If you answer these questions the right way, the path to reform is rather easy and straightforward. But if you answer even one question the wrong way, you will get caught up in a cycle of complexity — with fix upon fix, each trying to patch up the problems created by the previous one.

Hint: Right answers are ones that give people incentives to meet their own needs without imposing costs on others. Wrong answers are ones that give people perverse incentives to impose costs on others while pursuing their own interests.

The first question is: How is health insurance to be subsidized? Is the government subsidy to be uniform across all ways of obtaining insurance (as it is in most countries)? Or will the subsidies differ according to how the insurance is acquired? Currently, we have the latter condition: The subsidies are heavily tilted toward employer-provided group insurance and away from individually acquired insurance. Every dollar of premium paid by an employer avoids a 15.3 percent FICA payroll tax and — in the case of a middle-income family — a 25 percent federal income tax and a 6 percent (on average) state income tax. Avoiding these taxes amounts to receiving a subsidy equal to almost half the cost of the insurance. People who must purchase their own insurance, by contrast, get virtually no tax relief. At the workplace, insurance is purchased with before-tax dollars. Away from the workplace, it must be purchased with after-tax dollars. The poor individual purchaser must earn twice as much money to be able to buy the same insurance.

The health-reform bills in Congress would create even more lopsided subsidies — this time tilted toward individual purchase in a “health insurance exchange,” and away from purchase through one’s employer. Under Sen. Edward Kennedy’s bill, the highest premium anyone would have to pay in the exchange would be 12.5 percent of his income. The additional cost would be paid by the taxpayers. In the House bill, the highest premium would be as low as 3 percent of income for someone earning up to 150 percent of the poverty level ($33,000). That means a low-income worker would pay only $992 for a family plan whose actual cost is $13,000 (about average for employer-provided coverage). The net subsidy would be $12,008. By contrast, if an employer purchased this same insurance, the implicit subsidy would be only the amount of payroll tax he avoided, or $5,060.

You don’t have to be a math genius to figure out that the exchange is a better deal. Indeed, if a worker had been getting insurance from his employer, he and the employer could drop the plan at work, agree on higher wages instead of premiums, purchase the same plan in the exchange, and end up with almost $7,000 in extra take-home pay. Great deal for the worker — but that gain will be at the expense of the taxpayers. And after millions of people make this choice, the taxpayers will end up paying billions of dollars in subsidies, without any real change in health-insurance coverage.

Nonuniform subsidies create havoc. Under the current system, they make portability — the ability to keep your health-insurance plan when you change jobs — virtually impossible. Under the bills in Congress, more than half the population will try to shift from group to individual coverage — creating an enormous taxpayer burden without any real benefit.

The second question is: How is health insurance to be priced? Are the premiums going to reflect expected costs, as they do in other insurance markets? Or are they going to be independent of expected costs? Under “community rating” (the pricing scheme preferred by congressional Democrats), insurers must charge the same premium for a given plan to everyone, regardless of health condition. This creates perverse incentives for both buyers and sellers.

Buyers will tend to choose lean plans when they are healthy (and don’t expect to use many services) and rich plans when they are sick (and plan to use a lot). They will tend to choose HMO coverage when healthy (and choice of doctors does not seem very important) and fee-for-service insurance when they have a serious problem (and choice of doctors is critical). More generally, if the premium is the same for everyone, people will tend to underinsure themselves when they are healthy and overinsure themselves when they are sick.

Sellers of health plans will seek to attract the healthy (on whom they make a profit) and avoid the sick (from whom they incur losses). After enrollment, they will tend to overprovide to the healthy (in order to keep them and attract new enrollees) and underprovide to the sick (in order to encourage their exodus and discourage new enrollees like them).

Much regulation would be needed to try to keep insurers from acting on these incentives. But there is a better way: Let premiums reflect real risks. One out of five seniors is enrolled in a (private) Medicare Advantage plan that receives a risk-rated premium, calculated on the basis of up to 75 different variables. In this system, health insurers will no longer want to drive the sickest enrollees away, because these patients will be paying higher premiums. Employers could do the same for their employees.

The third question is: How much health insurance is to be paid for by a third party (one other than the recipient or the provider)? Will a third party pay every bill (as in Britain and Canada)? Or will individuals be able to self-insure, managing some of their health-care dollars in an account they own and control? If the former, patients will have incentives to overconsume, using up health-care resources until their value at the margin (the value of the last unit purchased) is almost zero. And health plans will have to employ rationing devices to control demand. If patients are using their own money, however, they won’t spend a dollar on care unless the care is worth a dollar.

There is already a good model for this patient-centered alternative. In the Medicaid Cash and Counseling pilot program, homebound disabled patients manage their own budgets. They can pay market prices and hire and fire those who provide them with health services. Satisfaction rates approach 100 percent (something unheard of in health care anywhere else on the planet). But other Medicaid patients have no such freedom. They cannot add to Medicaid reimbursement rates and pay market prices at walk-in clinics, surgicenters, and urgent-care centers. They cannot choose a doctor they like and pay his usual fee. I have no idea what the satisfaction rate is in Medicaid generally. But since being in Medicaid is only marginally better than being uninsured, I suspect it is not very high.

On these three questions, just one wrong answer makes a sensible health-care policy highly unlikely. But with three out of three answers wrong (the situation of Congress), workable health reform is a metaphysical impossibility.