WASHINGTON — Here’s a phrase you can expect to hear a lot in the next few days: “According to the CBO.” The CBO is the Congressional Budget Office, the official scorekeeper of the costs of proposed legislation. Rarely has a CBO report been more anxiously awaited than the analysis released Thursday of the proposed changes to the Senate health care reform bill. Democrats are delighted with the bottom-line analysis that the measure would save $138 billion over the next 10 years, and as much as $1.2 trillion in the second 10 years — all this while expanding coverage to 32 million people who would otherwise be uninsured.

So Democrats will be pointing to this preliminary CBO score as if it is engraved on stone tablets. Republicans will profess their respect for the CBO and proceed to argue that its estimates should not be taken too seriously in this instance. What I’m about to say may come as a surprise, but I think the Republican argument is closer to the correct one. To crow, as did House Speaker Nancy Pelosi, that the package is “a triumph for the American people in terms of deficit reduction” is premature at best, delusional at worst.

Like the Supreme Court, it can be said of the CBO that it is not final because it is infallible; it is infallible because it is final. Its score is the one that counts. Unlike the Supreme Court, however, the accuracy of the CBO’s judgments is subject to testing over time. It’s in the nature of the CBO’s work that its assessments are projections — informed projections made by incredibly smart people working with sophisticated economic models, but projections nonetheless.

On a measure of this complexity and this magnitude, these projections are, even more than usual, a matter of art rather than science. The longer the time horizon, the greater the fuzziness. The CBO is more than happy to acknowledge these limitations. In analyzing the likely impact of the Senate health care bill in the second 10 years, the CBO projected “a broad range between one-quarter percent and one-half percent of gross domestic product.” This is wiggle room amounting to more than half a trillion dollars over the decade. As the CBO put it, “The imprecision of that calculation reflects the even greater degree of uncertainty that attends to it, compared with CBO’s 10-year budget estimates.”

That’s one reason to take the cost estimates with a grain of salt. Here’s another: Of the $138 billion saved in the first 10 years, $70 billion represents premiums collected for a new long-term care program, money the government will have to pay in benefits later.

And here is the accompanying tablespoon of salt: The CBO is required to assume that Congress will do what it promises. So, for example, Congress promises in the measure to cut several hundred billion in Medicare spending. Sometimes such promises have come to pass. Other times, as in the current difficulty with scheduled cuts in Medicare reimbursements for doctors, they are put off because of a public — or politically connected — outcry.

One big reason for the CBO’s long-term assessment of major cost savings involves the excise tax on high-cost insurance plans. This is the tax that, in the face of opposition from labor unions and others, has been diluted to almost nothing — a measly $32 billion, compared to the $149 billion in the original Senate bill — during the first 10 years. Will the tax really be collected in 2018 — long after many of those voting for it will have left office, long after the benefits it is helping finance have kicked in? In one important way, the revised proposal improves the tax: Beginning in 2020, it will be indexed at the regular rate of inflation, not the Senate-passed inflation rate plus one percentage point. But that change may make it even less likely to take full effect, because it will hit more people sooner, meaning that they’ll start complaining sooner.

Likewise, the revised measure would change the way that government subsidies to help people purchase insurance will be calculated, beginning in 2019. These would end up growing more slowly over time. Ask yourself: If subsidies end up covering a smaller share of people’s premiums, what do you think a future Congress will do?

I hope that the CBO’s projections prove correct. I wouldn’t bank — or base my vote — on it.