VIKRAM PANDIT’S reign as chief executive of Citigroup ended Tuesday after what may have been the roughest five years for any Citi boss. Mr. Pandit took over in 2007 on the eve of an epochal financial crisis, which was triggered in part by over-investment in mortgages by Citi and others. Citi survived only with the help of federal aid, including $45 billion in capital and a U.S. backstop for $301 billion worth of toxic assets.

Taxpayers eventually recouped the aid, plus a profit. But perhaps more than any other institution, Citi — the original financial “supermarket” — epitomized the dilemma of “too big to fail.” The crowning irony was the recent suggestion by Sanford Weill — who first expanded Citi’s business to encompass investment banking and other risky activities — that banks “be broken up so that the taxpayer will never be at risk, the depositors won’t be at risk, the leverage of the banks will be something reasonable.”

Mr. Weill was jumping aboard a bandwagon already occupied by former Federal Reserve chairman Paul Volcker and members of Congress from both parties. But the idea has not fully taken hold in U.S. policy, despite the Dodd-Frank regulatory reform bill’s provisions bolstering bank capital and pushing federally insured institutions out of proprietary securities trading.

As matters now stand, the assets of the six largest U.S. banks equal 62 percent of gross domestic product, compared to 18 percent in 1995. These banks service 56 percent of all mortgages and hold 35 percent of deposits. Given that their deposits are federally insured, it’s small wonder that no one quite believes the specter of “too big to fail” has been banished — and that the giants enjoy cheaper access to market funding as a result. At the second presidential debate, GOP candidate Mitt Romney made that very point.

The latest call to shrink the banks came from a particularly influential source. On Oct. 10, Federal Reserve Board Governor Daniel K. Tarullo floated the idea of capping banks’ non-deposit liabilities — borrowings they use to finance themselves — at a certain percentage of GDP. Mr. Tarullo’s suggestion was quickly seconded by James Bullard, the president of the Fed’s St. Louis branch. Sen. Sherrod Brown of Ohio, a Democrat, has proposed a bill that would do something similar, and he has teamed up with Sen. David Vitter (R-La.) to seek higher capital requirements for banks.

In short, there is broad support for some form of restriction on the size and complexity of federally insured commercial banks. The problem, of course, is how to translate that seemingly straightforward idea into a workable law. Depending on where you set the cap, Mr. Tarullo’s idea could break up all six of the biggest banks, some of them or none. Critics of the proposal argue that the U.S. economy benefits from having at least a few big players in global finance, which can take advantage of economies of scale and thus provide services efficiently across borders.

There is merit to arguments on both sides, but it should be possible to manage the trade-off between whatever benefits large-scale banking produces in the short-run and the risk of future losses to taxpayers if a large, complex institution fails. Recent experience counsels resolving that trade-off in favor of safety.