IMF wrong to label Deutsche world’s riskiest bank, says economist

Co-developer of risk methodology used by IMF says it misapplied it when labelling bank riskiest G-Sib

"Most of [Deutsche Bank's] travails are not generating connectedness to other G-Sibs," writes Yilmaz

The International Monetary Fund erred in labelling Deutsche Bank the biggest contributor to systemic risk among large banks, says the economist who co-developed the methodology the fund employed. Kamil Yilmaz, of Koç University in Turkey, now claims it was incorrectly applied.

In its financial system stability assessment for Germany in June 2016, the IMF stated Allianz was the largest contributor to systemic risk in Germany, and that “among global systemically important banks (G-Sibs), Deutsche Bank appears to be the most important net contributor to systemic risks”. The IMF’s analysis was widely reported, and blamed for prompting falls in the value of Deutsche’s share price.

The assessment relied on a methodology for measuring the systemic risk posed by an institution developed by Yilmaz and economist Francis Diebold of the University of Pennsylvania. Risk.net subsequently reported on criticism of the methodology, and Yilmaz now claims the IMF incorrectly applied it. He wrote a blistering critique of what he saw as the IMF’s flawed application, which he decided against publishing, but has now shared it with Risk.net.

“If the connectedness methodology is a set of pieces of a jigsaw puzzle, the authors put only a couple of these pieces together. It turns out that with just a few pieces, they figured out the whole picture and declared that Deutsche Bank was the most important net contributor to systemic risk in the global financial system,” Yilmaz writes.

The Diebold-Yilmaz methodology tracks the connectedness of volatilities in the stock returns of financial firms through the use of variance decomposition, which indicates the amount of information each variable contributes to the other variables in a vector autoregression. A vector autoregression captures the linear interdependencies among multiple time series.

Yilmaz argues the IMF erred in focusing on return connectedness in its study, rather than volatility connectedness. The former metric is variable, but can increase in both good times and bad; volatility connectedness tends to be high in bad times and lower in good times.

In a 2011 paper describing the methodology, Diebold and Yilmaz say connectedness measures derived from variance decomposition provide “natural and insightful measures of connectedness among financial asset returns and volatilities”.

“In our work, we always use volatility, because volatility connectedness is asymmetric,” Yilmaz tells Risk.net. “In bad times, volatility connectedness is very high. In good times, it is very low.”

Deutsche Bank is not a Lehman or AIG at the moment. Most of its travails are idiosyncratic and are not generating connectedness to other G-Sibs

Kamil Yilmaz, Koc University

When Yilmaz performed an analysis of volatility connectedness among G-Sibs covering the period from November 2007 to February 2016, he found Deutsche Bank only ranked fifth, behind Citi, JP Morgan, BNP Paribas and Wells Fargo.

“Deutsche Bank is not a Lehman or AIG at the moment,” he writes. “Most of its travails are idiosyncratic and are not generating connectedness to other G-Sibs.”

Udaibir Das

In response, the IMF says the organisation’s own research shows a stronger correlation between returns and volatility connectedness than assumed by Yilmaz. “Based on our work thus far, both system-wide return and volatility connectedness do move closely with each other over time. They tend to be more elevated during financial stress and economic recessions,” says Udaibir Das, head of the IMF’s financial stability assessment division.

The IMF uses a range of academic models in its financial sector assessment program, which analyses the financial sector resilience of participating countries.

“We use peer-reviewed academic models in our applied work,” says Das. “We adjust them given the idiosyncrasies of structures and systems. As we use and learn, we are looping that back into our internal work, and we will be happy to provide authors of the models feedback too.” 

Deutsche Bank declined to comment for this article.

An in-depth look at different methodologies for assessing institutions’ systemic riskiness can be found here.