The Air Force was stunned last month when the House voted not to buy six F-22 stealth fighter jets next year. After all, Congress already had approved the purchase of several of the airplanes to be used in testing. The Pentagon's acquisition chief had pledged that the $70 billion program was ready to cross the threshold from development to production. For an appropriation bill to turn off the fiscal faucet at that point seemed unthinkable.

The truth is that only a gambler or a spendthrift would commit his own money to building this fighter in quantity based on the limited amount of testing the Air Force has completed. If the Department of Defense had exercised more responsible stewardship of the public's money, Congress would not have been left to wield the budget ax.

The F-22 is the Air Force's premier program, ardently promoted by its advocates as the key to continuing U.S. air superiority and a linchpin of our future national security. But it is also one of the most complicated and expensive fighters ever attempted. To prove that it works as promised, the Air Force plans to conduct an extensive series of flight tests as part of the plane's development phase.

The development phase overlaps markedly with the airplane's production timetable, however. The overlap, termed concurrency, means that tests may not expose serious problems until a large number of the fighters already are built or in various stages of assembly. It's as though the American people were being asked to buy a fleet of newly designed racing cars scarcely road-tested by the manufacturer.

The costs of rework on a system rushed into production with insufficient testing can be enormous -- a lesson the Air Force still is learning with the B-1. Although that bomber entered service nearly 15 years ago, its complicated electronic systems do not work right despite billions of dollars in modifications.

Congress has expressed concern about F-22 program concurrency for several years. The Defense Department has countered that concurrency posed only a moderate level of risk -- one outweighed by the monetary savings that would result from manufacturing the test airplanes and the production versions in quick succession. But until the contractor, Lockheed Martin, built the first test fighter and delivered it to the Air Force in 1997, the arguments on both sides were largely theoretical.

By late last year, evidence had mounted that Lockheed was encountering serious problems. The titanium castings that attach wings to bodies were more difficult to produce than expected. Welding difficulties slowed deliveries of key parts. Assembled planes experienced nagging fuel leaks and problems with the brakes and air conditioning. Hardware and software associated with the airplane's avionics -- the on-board electronic systems that allow it to identify itself to friendly forces, communicate, navigate and jam or spoof enemy radars -- were behind schedule.

These slowdowns delayed the test program and thus significantly raised the risks posed by the rush to production. By December 1998, the Air Force had accomplished only a fraction of the flight tests it had hoped to finish by that point. And although those tests demonstrated that the F-22, once assembled, can take off, fly around and land, two of the fighter's most acclaimed features -- stealth (reduced visibility by sensors viewing the airplane from any angle) and advanced avionics -- will not be ready for testing for some time. Yet these are precisely the features that the Air Force says will distinguish this jet from all its competitors and allow it to "own the skies."

Faced with these problems, Pentagon acquisition chief Jacques Gansler ought to have directed the Air Force not to commit money to production until it resolved the worst of the manufacturing problems and tested key features. Instead, he postponed the internal review that the Pentagon's own rules require and reported to Congress that the risk involved in committing to production was acceptable.

Now the Air Force threatens that any pause in building the planes will increase costs by more than $6 billion. It is half right. The cost of this plane is going to rise -- but not because of this delay. The program's cost is going to rise in any case.

In 1996 and 1997 the Air Force's cost-estimating team, the Defense secretary's analysts and the Congressional Budget Office projected that the F-22 would cost about 20 percent more to build than the Air Force claimed. In response, Lockheed Martin and the Air Force devised dubious cost-control measures -- from selling the fighter abroad to double-counting claimed savings. Secretary of Defense William Cohen, ignoring the warnings of his own staff, accepted the Air Force's view and asked Congress to do the same. The difficulties that Lockheed Martin has encountered in building test airplanes and the added labor hours that resulted from them reinforce the conclusion that the major price increase predicted three years ago is all but inevitable.

Right now, however, cost growth may be the least of our worries. The question is whether the F-22 will work the way it's supposed to work. With only a fraction of testing finished and key capabilities of the airplane too immature to test, there is no way to know. The procurement pause the House has endorsed will give the Air Force time to test the planes it already has bought, rather than rushing to produce an unproven system that would ultimately cost even more to fix. If the airplane is as important to national security as the Air Force says, how can it not be worth the time it takes to get its most important features right before building it in quantity?

The writer, a senior research fellow in the Security Studies Program at the Massachusetts Institute of Technology, was assistant director for national security in the Congressional Budget office from 1994 to 1997.