Speaker Paul D. Ryan (R-Wis.) holds up a Department of Health and Human Services report on insurance premiums. (Alex Wong/Getty Images)

Peter Orszag was director of the Congressional Budget Office from 2007 to 2008 and director of the Office of Management and Budget from 2009 to 2010.

The new Congressional Budget Office score of the American Health Care Act is devastating. In 2026, 23 million fewer Americans would be insured with it than without it. The response of the secretary of health and human services and other defenders of the Republican bill? Attack the Congressional Budget Office.

I had the honor of running the CBO a decade ago and will be the first to admit it’s not perfect. But it’s also far better than the alternatives, and most of the critiques are off base.

As an example, Health and Human Services Secretary Tom Price claims that “the CBO was wrong when they analyzed Obamacare’s effect on cost and coverage, and they are wrong again.” Yet it is Price himself who is more wrong than right.

It is true that the CBO substantially overestimated the number of people who would obtain coverage through the Obamacare exchanges. But one important reason was that it was too pessimistic about how many people would lose employer coverage and therefore have to obtain coverage on the exchanges in the first place.

The CBO was almost spot-on in estimating the key figure: the overall number of people without insurance. In July 2012, after the Supreme Court ruling that the Medicaid expansion under Obamacare would have to be a state option rather than a requirement, the CBO estimated that 30 million people would be uninsured in 2016. The actual number? 27 million.

Not too bad. If I underestimate the percentage of free throws made in a basketball game but correctly predict the final score, Price may claim that I was far off on the free throws. I’d be quite pleased with the end result.

The CBO’s analysis of the premiums under Obamacare have also been quite accurate. Average marketplace premiums this year are almost exactly where the CBO initially predicted they would be.

Given this rather impressive track record, is there any reason to doubt the CBO’s ability to accurately assess the current bill? The modified version admittedly poses a particularly difficult challenge: estimating how states would respond to the various waivers the bill allows. The waivers would, among other things, allow states to lighten up on what benefits insurance plans have to cover and also permit plans to set prices based on a person’s health. The CBO’s view was that the combination would make markets unstable in areas covering roughly 15 percent of the non-elderly population — more that 40 million Americans.

Once again, proponents, not appreciating the results, attacked the referee. Rep. Mark Meadows (R-N.C.), for example, asked, “What planet did they come up with this?” The author of the key amendment providing for the waivers, Rep. Tom MacArthur (R-N.J.), scoffed that the CBO was “trying to answer questions that I think it would be better where they say, ‘I don’t know.’ ”

How did the CBO analyze the effects of the waivers? Its analytical approach was similar to what it did after the Supreme Court’s Medicaid ruling. Then, the CBO had to estimate how many states would expand Medicaid voluntarily. Instead of evaluating what would occur in each individual state, it made a simplifying assumption about the share of the national population that would live in expansion states. The result? It projected that 10 million more people in 2016 would be covered by Medicaid because of Obamacare. The actual number was  14 million. Given the complexities involved, that’s a remarkable performance.

In estimating how many states would take up the waivers under the American Health Care Act, the CBO tried to gauge the overall odds across the nation as a whole rather than examining the potential response within each particular state. As the basis for its estimates, the CBO looked at past behavior across states (for example, how many states had previously mandated coverage of maternity benefits and mental health benefits, how many had previously required community rating, and so on) and current market conditions. One wonders how Meadows would instead suggest approaching the question.

The furious reaction to the CBO score of the American Health Care Act seems a bit like Keynes’s famous quip about Euclidean geometers in a non-Euclidean world who, frustrated that “in experience straight lines apparently parallel often meet, rebuke the lines for not keeping straight — as the only remedy for the unfortunate collisions which are occurring.” Proponents of repeal-and-replace keep colliding with the CBO and then rebuking it — rather than trying to figure out why the plans keep falling short.