The perceived risk of a big U.S. bank failure has receded in recent years, but more study is needed to understand whether the improvement is due to government policies or simply an improved economic outlook, a top Federal Reserve official said on Friday.

Studies using measures of market risk including credit default swaps show "that the size of the too-big-to-fail problem has fallen over the past couple of years but remains large," Narayana Kocherlakota, president of the Minneapolis Federal Reserve Bank, said in remarks prepared for a conference sponsored by the Boston Fed and Boston University.

For any given financial institution, he said, "it could be that creditors believe that there is little likelihood of that financial institution becoming distressed," perhaps because new rules require banks to put up more capital.

It could also be that creditors believe that a government bailout is unlikely, suggesting that other policies -- such as the requirement banks devise blueprints for a wind-down should they become insolvent -- are working, he said.

But metrics could be improving "simply because creditors' assessments of future macroeconomic conditions improve," he said.

Teasing apart the reasons for the improvement in the too-big-to-fail problem is key to understanding whether approaches like those enshrined in the 2010 Dodd-Frank financial reform act are having the intended effect, Kocherlakota said.

The wide-ranging law, written in response to the 2007-2009 financial crisis, aims to reduce the likelihood of banks failing and to lessen the cost to society if they do.

Kocherlakota has urged the Fed to adopt guideposts for policy in terms of unemployment and inflation, and on Friday reiterated his view that without such metrics "it is challenging to know whether monetary policy is overly accommodative or not."

The same point can be made for the too-big-to-fail bank problem, which Congress has set out to resolve, he said.

"The public can only hold Congress and its delegees responsible for achieving this mandate if there are quantitative measures of the size of the too-big-to-fail problem," he said.