New research suggests – perhaps unintentionally – that leverage ratios are a poor indicator of impending severe bank stress or failure, casting doubt on the heavy reliance placed upon them in stress tests and efforts to make the banking system safe.
As part of a broader study, a working group looked at data on 117 banks, and tried to find levels of key leverage ratios that could be used to separate the ultimate fortunes of the bank. For example, did banks with a Tier One ratio below 4 always fail? Did banks with Tier One ratios above 10 per cent always succeed?
The short answer is that these levels were not very instructive. One level the paper suggests is 4 per cent for the Tier One ratio. But the actual finding is that in 50 per cent of cases, banks with more than 4 per cent did not go on to suffer severe stress, and banks with less than 4 per cent, did. Fifty per cent is rather low, however: could one similarly ‘identify’ the fate of banks by tossing a coin?
The paper says that there is no “correct” method to identifying these levels, though it seems that this should have called for a classic regression analysis. Had the data been regressed, we would also be able to gauge the significance of the findings, and possibly reject these leverage ratios altogether as determining the ultimate fate of their banks.
There are two points here. One is the troubling thought that the political need to brandish numbers and targets – the need to be seen to be doing something, and the (perhaps arbitrary) choice to be doing something with leverage ratios – has forced some perfectly competent people into producing some apparently incompetent numbers. The research tells us that the levels identified are a better bet than alternative levels; but it says nothing about the sense of using a level at all.
The other point is that limits on leverage are a key plank of efforts to stabilise the financial system, and this research suggests that they are not relevant. It is always tempting to form goals around what is measurable, but temptation is no indication of suitability. It might well be that no such indicator exists.
From the FT by Emma Saunders