To recap: HSAs were introduced in the 2003 Medicare bill, and now Bush wants to expand them. Basically, you purchase a high-deductible policy—one with lower premiums that forces you to pay, say, the first $2,000 of your medical costs out of your own pocket—and then you can get a savings account into which you can deposit tax-deductible money each year, up to $2,000, to pay for those out-of-pocket costs. The money rolls over each year. As Ezra Klein points out, it's a good deal if you're healthy (or have a spare $2,000 stuffed under your mattress), not a great deal otherwise.

The idea behind thinking that HSA plans will help the uninsured is that high deductible policies have lower premiums than traditional insurance, and that market competition among companies offering high-deductible plans will reduce costs further. Plus, since patients with these policies will have to pay a lot out of their own pockets, they'll be more careful about spending money on frivolous care, which will constrain costs. It's like magic!

Now true, this didn't work in South Africa, where HSAs became massively popular yet health care costs skyrocketed—as it turns out, individual consumers have less ability to bargain down the costs of medical services than large insurers do. And there's no conceivable way HSAs can reduce the vast bulk of health care costs in America: The rule of thumb, after all, is that 20 percent of patients account for 80 percent of costs, and since these are generally catastrophic costs over $2,000, HSAs won't change the amount consumers are spending. They just can't.

On the other hand, early "trials" with HRAs (a similar system) have shown that the accounts might prove moderately effective at constraining costs for the rest of us—see this hardly-unbiased Aetna study or John Bertko's testimony about Humana, Inc.'s experience with a "consumer directed health plan." Whether this is because workers were being smarter about their care or because they were foregoing much-needed care is an open question. Meanwhile, since the healthy people are all fleeing traditional insurance plans and signing up for HSAs, the premiums for traditional policies could rise unduly, although there are, in theory, a few ways that employers can reduce this "adverse selection."

Back to the main point. Even if "consumer-directed health plans" do control costs and reduce premiums, they probably won't make a significant dent on the numbers of those without insurance. As this Kaiser study shows, people generally don't have insurance because either their employer doesn't offer it, they can't afford the premiums, or they can afford the premiums but can't afford the surcharges for their age or pre-existing conditions. The reduced cost of high-deductible insurance won't make a difference for many of these people, especially if they would have to pay more out of pocket. (Many of the uninsured also don't have $2,000 to spare on out-of-pocket costs.)

Moreover, most of the uninsured are in the 10 or 15 percent tax bracket, and would not see enough of a tax benefit to take advantage of the HSAs. A Commonwealth Fund study pointed out that HSAs will mostly make a difference to higher-income uninsured individuals (those making over $50,000) who currently can afford insurance but just don't value it enough.

Now in his FY2005 budget, the president proposed to allow premiums for high-deductible policies to be claimed as a deduction. This would substantially reduce the cost of insurance—and cost the Treasury $25 billion over 10 years. On the other hand, Edwin Park and Robert Greenstein of the Center on Budget and Policy Priorities estimated that this deduction would increase the number of uninsured by 350,000—inducing some people to buy insurance on the one hand, but inducing many employers to drop health coverage on the other. On the other hand, the HSA Coalition, an advocacy group, argued that Bush's proposal would reduce the number of uninsured by 1.2 to 4.3 million, although this is based on data from an HSA company whose customers may not be representative of the general population. (See CBPP's response here.)

So at absolute best, the U.S. government could spend about $35 billion and cover 4 million new people. At worst, this would have no effect. Ultimately, there's no free fix for achieving universal coverage. Uwe Reinhardt has often said that barring comprehensive reform, a good rule of thumb is to assume that covering the 45 million uninsured—or the 60 million Americans who go uninsured at some point during the year—will cost at least $80-100 billion a year. (Jonathan Gruber has shown that the cheapest way to do this—i.e., covering the uninsured without inducing a lot of the currently-insured to drop their coverage and go for what the government's offering—is simply to expand Medicaid.) Any proposal that falls short of this rough dollar amount, as Bush's is sure to, is not likely to do much for the uninsured.