The debate over what to do about the largest U.S. banks is reaching a new peak as government auditors and regulators prepare to address two issues: Whether some banks enjoy special advantages because they are deemed too big to fail, and whether tougher capital requirements are needed to make failures less likely.

One of the big banks' aims is to debunk the idea that they enjoy a taxpayer subsidy -- in the form of lower borrowing costs -- because creditors assume the government will always bail them out. In February, Bloomberg View, using a 2012 paper by economists Kenichi Ueda of the International Monetary Fund and Beatrice Weder di Mauro of the University of Mainz, estimated the annual taxpayer subsidy to be $83 billion for the 10 largest U.S. banks.

The majority of studies, with the exception of some done by the banks, have also found the subsidy to be very large. The Government Accountability Office is planning to offer its own take on taxpayer support to banks in a pair of reports, one coming in mid-November and the other in spring 2014.

In a recent article published in American Banker, Abby McCloskey, a program director at the American Enterprise Institute, argues the banks' position. She says Bloomberg View's $83 billion estimate unfairly cherry-picked data from the crisis year of 2009. The true number should be a negative $14 billion, she says, if one takes into account the cost of new regulations imposed by the Dodd-Frank Act.

McCloskey's analysis repeats the flaws of many who have gone before (see here and here), including citing an IMF subsidy estimate that predates the Ueda and Weder di Mauro work. Perhaps most important, she confuses the question at hand: Should we be asking whether big banks borrow at lower rates than small banks, or whether government support allows big banks to borrow at lower rates than they otherwise would? The latter is the right question, and the one that Ueda and Weder di Mauro were trying to answer in their paper. It should also be the focus of researchers at the GAO.

True, Ueda and Weder di Mauro looked at the credit-rating uplift big banks received thanks to government support in 2009. If you think about it, 2009 was the rare occasion when the credit-rating companies, which have every incentive to side with the big banks, had little choice but to recognize fully how dependent the banks were on taxpayers. The researchers then looked at borrowing-cost data from 1920 to 1999 to put a percentage-point value on the rating uplift. They used the historical borrowing-cost data because their aim was to calculate the typical annual value of the subsidy, which they found to be 0.8 percentage point. If they had used borrowing-cost data from 2009 alone, the result would have been much larger.

McCloskey's confusion about the nature of the too-big-to-fail subsidy leads directly to another misconception: that regulatory compliance costs could somehow offset the subsidy. Compliance may or may not put a greater burden on larger banks (I suspect executives at small banks would disagree). In any case, the costs do nothing to reduce the taxpayer subsidy or the incentive to preserve it by becoming as big and systemically threatening as possible.

In other words, U.S. taxpayers are paying big banks to put the economy in danger, which is crazy. The best solution is to make banks less likely to fail by requiring that they finance themselves with more equity capital, which absorbs losses in bad times.

Regulators have taken a small step in that direction by proposing that bank holding companies have at least $5 in capital for every $100 in assets -- a 5 percent leverage ratio that is a bit more than the global minimum of 3 percent. The proposal, to which banks are adamantly opposed, falls far short of the 20 percent that economists have argued would be best for the economy, but at least it's a start. Let's hope they stick to it.

(Mark Whitehouse is a member of the Bloomberg View editorial board. Follow him on Twitter.)