13 September 2017

Are the Bank of England’s stress tests fit for purpose?


The 2007-8 financial crisis had lots of causes. But one crucial problem was that banks’ errors were correlated: nearly all of them made the same bets, and they bet big. So you’d think that regulatory policy, now in the hands of the Bank of England, would steer clear of anything that made that problem worse, by encouraging banks to put all their eggs in one basket. Quite the opposite is true.

Stress tests are simulations, or models, where regulators imagine scenarios — shocks — that could hit a country’s financial sector, and see if banks could stand the stress. It’s a good question to ask: do banks have enough capital to save them if hit by a housing crisis, banking crunch and economic downturn at the same time. But we should not be reassured by the Bank’s latest tests — as a new Adam Smith Institute paper, out today and written by Professor Kevin Dowd of Durham University shows us.

One problem is the way the Bank measures the value of a banks assets. How much is a house worth? Is it worth the amount you paid for it, plus a few years of projected house price rises, less projected costs and expected property taxes, or is it worth the price it can fetch on the market if you sold it? The Bank of England’s measures assume the former. But if a bank is in trouble, it can’t sell its assets for book value — it can only offload them for what someone is willing to pay. Now this isn’t usually a problem; often market values simply fluctuate around book values, which prove to be a better long-term estimate. But it is precisely in crises when the differences matter most.

Around the world, regulators have moved towards trying to measure financial firms’ assets and liabilities in more detail. Instead of looking at a simple leverage ratio — total liabilities compared to shareholder capital — they try and weight some of those liabilities by their risk, and allow some other safe and liquid assets to count as capital. This allows them to create risk-weighted leverage ratios. So far, so good, in principle.

But it turns out that financiers are as good at regulatory arbitrage as they are when it involves buying things cheap in one place and selling them expensive in another. For every degree of risk there are more and less risky securities. For example, Eurozone government debt (including Greece’s) was all rated at zero from the perspective of the Basel regulations.

This sort of risk weighting carried through to the stress tests. Given how inherently difficult it is, in practice it gives banks a major incentive to take the riskiest assets for each level of regulatory risk-weighting to make their balance sheets look better. Worse than that, these particular assets are the same for all of the banks, herding them towards similar business models. At best, if the Bank of England is right about risk, then this trades more regular, milder crises, for rarer, bigger crises. At worst, if it is wrong, it gives us both worse crises and bigger ones.

This is why the literature is catching up with Professor Dowd, who has been giving this warning for three years in a sequence of ASI papers. For example, an American Enterprise Institute paper by Paul Kupiec found that “complex multi-equation models produce the least accurate forecasts. Simple models have less explanatory power within the estimation sample but produce more accurate out-of-sample forecasts.”

Notwithstanding those who claim they “predicted” the financial crisis despite neither mapping out nothing but its vague contours, nor calling the top of the boom, nor making any money off it, we all missed the last shock. Stress tests are more likely to prepare us for the last crisis than create a diverse banking system with the sorts of capital buffers that would absorb the next, if and when it does come. It’s hard to give up on something you’ve invested in, but we implore Mark Carney’s Bank of England to try.

Ben Southwood is Head of Research at the Adam Smith Institute