Washington ? Stress-testing top banks has turned out to be a terrific stress reducer. Like a medical patient who takes off on a euphoric binge after the biopsy comes back negative, bank stocks have staged a heady rally, driving a broad recovery in the markets and talk that the end of the recession may be nigh. But the real significance of the stress tests goes deeper. They answer the perplexing long-range question: When the financial system emerges from this crisis, how can it be prevented from blowing up again?

Without an answer to that question, we are in for a rough time. After their recent thrashing, financiers won’t take excessive risks again for a while. Yet bankers’ memories are short, and the crisis may ultimately induce more risk. Successive bailouts have revealed that the debts of large financial institutions are backstopped by the government: If those institutions go down, bondholders won’t lose money, so they might as well finance risk with no end. In essence, all of Wall Street has been transformed into Fannie Mae and Freddie Mac, the giant housing lenders whose implicit government guarantee led them down the road to ruin.

So we need some serious risk controls, and the question is: What sort? Once upon a time, governments kept the lid on bank risks by regulating structure: Banks that took deposits were not allowed to underwrite securities or branch across state lines. Good luck reviving that approach. Unscrambling banks that have become not merely national but global is beyond the competence of government. In an age when companies are global, the banks that service them are naturally global, too.

Starting in the 1980s, when the old structural controls began to crumble, government came up with the approach of requiring banks to hold capital. It seemed an elegant idea: Capital was the catchall security blanket for all manner of risk-taking, ranging from foreign-exchange trading to plain commercial loans. But how much capital? Mistaken answers to that question explain most financial blowups, in the current crisis and before.

The original approach taken by regulators in the 1980s was to require banks to hold a fixed proportion of capital to loans. But loan volumes are a preposterously crude measure of bank risk-taking. A bank that lends $1 million to a foreign company may be taking a greater risk than a bank that lends $2 million but hedges out the currency risk. Or a bank that lends $3 million and also splits its money among multiple borrowers. Or a bank that lends $4 million while hedging the currency, diversifying among customers and using other tricks of modern finance to insure against default.

The shortcomings of regulators’ capital-to-loan ratios gave rise to a new measure: capital-to-risk. Investment banks and hedge funds pioneered this method; shortly before the current crisis, bank regulators blessed it. Lenders were supposed to plot the ups and downs of their holdings, factoring in currency hedges and so forth, and figure out the most that they could lose on them; then they would hold that much plus a cushion in their capital reserve. But this method also has shortcomings: Past ups and downs fail to capture what might happen in the future. For example, models based on historical patterns in real estate caused a lot of lenders to hold too little capital against the risk of mortgage default.

If loan volumes and backward-looking risk measures are not the right metrics, what might be better? The answer has been understood but not properly implemented by Wall Street for at least a decade, and it is the same one that the Obama administration has come up with: forward-looking stress tests. Rather than looking at how their portfolios would have behaved in recent history, banks must force themselves to imagine how their portfolios would respond to future shocks. What if the U.S. economy shrinks for the next two years? What if the dollar goes into free-fall? What about a war with North Korea or an earthquake in Tokyo?

Of course, the art of the stress test is deciding which scenarios to test for, and that is why government must have a role. A rational bank will want to manage risk sensibly, but the more capital it holds, the less it stands to profit, so it will always be tempted to ignore potential storms. But what is rational for an individual bank may not be rational for society. If a bank blows itself up, it can take others down with it, damaging the economy and handing taxpayers the bill. So government has a duty to force banks to plan for bad scenarios.

The administration’s stress tests are the template for this new approach. The measure of their success is not whether they cause the rally in bank stocks to continue, pleasant though that rally is. The real question is whether the administration forces the banks to raise the capital they are lacking, even at the risk of ending the market rally — and then whether it makes stress-testing a permanent feature of bank regulation.

— Sebastian Mallaby is a fellow at the Council on Foreign Relations.