Banking fragility and money

Another day, another LSE lecture.

This one features David Miles from the Bank of England and is chaired by Charles Goodhart. It's on monetary policy and banking fragility. There are two aspects of this problem that are particularly important.

The first is transmission mechanisms – an important aspect of monetary macroeconomics. Simple monetary models highlight that there is a relationship between the amount of money in the economy and the amount of economic output (consumption, investment or both). From this we can infer that printing money may increase GDP.

Slightly more sophisticated theories give a causal explanation of how this can happen. The central bank prints money and uses it to buy bonds. More money (and fewer bonds) mean that the price of bonds rises – this is simple supply and demand. The price of bonds is the inverse of the interest rate (that is, if a bond pays £50 per year interest and it costs £1000, interest is 5%. If the bond price goes up to £2000, it still pays £50, so interest rates are now 2.5%). Thus, when bond prices go up, interest rates fall. And when interest rates fall, it is easier for companies to borrow money to invest, and for people to borrow money to buy flat-screen TVs. Thus, GDP increases.

The banking question is this: if the financial system is broken, can monetary macroeconomics still work? If companies can’t borrow money anyway because the banks won’t lend it, or if people can’t spend money because the ATMs have stopped working, it won’t make any difference if the central bank buys all the bonds in the world. The situation in 2008 wasn’t as extreme as that, but in some ways the banking system did stop working. The interbank markets, where banks lend money to each other, seized up because every bank was worried about every other. This meant that even if the average interest rate had fallen, money could not find its way to where it was needed. Governments were able to borrow money very cheaply, but lots of companies could not; and nor could consumers, because the biggest channel of consumer finance (remortgages) were also on hold. Thus, the GDP-boosting effect of monetary policy did not work as it usually would.

One argument is that when government can borrow and spend but companies cannot, Keynesian government spending policies are the only way to sustain economic activity. Government spending has its own problems, which need to be weighed up against how serious and how long-lived the banking problems are likely to be. But it’s an argument that can be had.

So I imagine this will be one of the topics this evening: how does the stability of the banking system affect the efficacy of monetary policy, and what’s the appropriate course for macroeconomic policy?

The other key question is in the opposite direction: how do macroeconomic changes lead to banking fragility? If they do, then we could have a vicious cycle where macroeconomics hurts the banks; and the banks in turn prevent monetary policy from being effective. Unfortunately, there is a tendency, built into the financial system, for banks to be overexposed to macroeconomic fluctuations. The cause of this: asset correlation.

One of the main problems in 2008 was that the banks were all doing the same things. Of course, some banks were too big, and they were all highly interconnected. But even had they been smaller and less confusingly connected – like the savings and loans (building societies) in the US in the 80s, the crisis could still have happened. This is because they were all overinvested in property at the same time. They had lent – directly, or through mortgage-backed securities, or via the US government agencies Fannie Mae and Freddie Mac – trillions of dollars secured on people’s homes. When the housing market slowed and went into reverse – a macroeconomic effect which could have been limited to just the property sector – all of these loans and securities were at risk at the same time. This meant that (nearly) every bank was in trouble, and it was impossible to let just a couple of them go bust in the hope that the rest would get by OK.

My contention is that this excessive correlation of assets was a main cause of the financial crisis. But before looking at regulatory solutions, we must ask why the market didn’t correct for this automatically? There are three answers to this.

The first is information-driven herding. Banks are always seeking better returns at lower risks (as they should do). But when a new type of financial product is introduced, it is unlikely to provide that right away. Any new instrument – an insurance policy, or a new type of bond, or anything else – suffers from uncertainty about how it should be valued. Some brave people might give it a try, but while the market is small, and there is a shortage of data about past performance, it will be discounted for uncertainty. It will also be discounted for low liquidity, because if there aren’t many people trading the new product, it might be hard to offload if you need to switch strategies. These effects make the product less profitable than it would otherwise be.

However, once a particular asset class becomes more popular, these problems go away. Once there is a market to trade it, the liquidity discount disappears. And once there is a longer history of data in the asset, the uncertainty discount goes away – though there is still (as there should be) a discount for the measurable risk of the asset going wrong.

This is effectively a learning process. The community of traders as a whole gradually learns about new types of asset; and the ones they have learned most about are the ones that are considered safe to trade. This in turn helps people to learn more about them, and the effect reinforces itself. So, once an asset class – mortgage-backed securities, say – becomes reasonably popular, a cascade will make it more and more broadly accepted across the financial sector.

This results in a herding effect – where banks and traders focus excessively on a small number of asset types, and bank balance sheets become less diversified.

The second reason the market can’t fix this on its own is because of positive externalities. As a society, we want people to experiment with many new ideas – across the worlds of finance, business, science and social policy. Part of the reason we want this is because the people who do the experiments don’t capture all the value. If a trader in a bank tests out a new investment in a green technology incubator, everyone can see whether it is successful or not. So the bank bears the whole cost if it fails, but if it succeeds, everyone else gets some of the value – because they can copy the idea without having taken the initial risk.

So for this reason too, banks will underinvest in new ideas and overinvest in old ones that have already been tested.

Finally, there’s a negative externality – the risk of systemic failure. Correlation of bank assets is a danger to the banks, but it’s a bigger danger to the rest of us – as we saw in 2008. Therefore, some of the costs of conformity are passed onto the rest of us; and banks don’t have enough incentives to avoid excessive correlation.

So the market has three strong tendencies towards conformism. This exposes the rest of us to high risks – because macroeconomic problems often start in a single sector, and if that sector (like property) is one in which banks are overinvested, the effects will quickly spread to the rest of the economy. We should therefore seek a way to correct this.

My proposal is to require banks to hold capital reserves which depend on how closely their assets are correlated with those of other banks. If a bank’s investments are very similar to the investments of other financial institutions, they should have to keep larger buffers of spare capital.

A valid question is: how do we measure this correlation? We can’t easily predict which assets will move in line with each other. There are a number of ways to come close, however.

One is to require banks to trade future-dated asset swaps with one another. Imagine a contract which gives me the option to buy $1 million of JP Morgan’s balance sheet and sell $1 million of Citigroup’s balance sheet, on 31 July 2016. If those balance sheets are closely correlated, the contract will have a very low value. If the balance sheets are very different, the contract will have a high value because it will offer me the opportunity to make a good profit.

We need only look at the market value of this instrument (and the corresponding swaps between a series of other banks) to find out how much capital each bank should have to hold. The lower the value of these contracts, the higher the capital reserves.

The existence of these instruments would create a group of traders with a powerful incentive to examine the underlying assets of each bank and look at the relative risks of each. And the capital reserves would either themselves improve the safety of the banking system, or drive the banks to diversify their assets and thereby protect the system.

Tickets for tonight’s event are sold out so I can’t invite you along; but let me know if you are interested in hearing what went on. There should be a podcast on the LSE website in a few days.


Popular posts from this blog

What is the difference between cognitive economics and behavioural finance?

Is bad news for the Treasury good for the private sector?

Dead rats and dopamine - a new publication