Sunday, 3 March 2013

Do fiscal consolidations make banks safer?

A research team at the ECB thinks they do.  They have published a working paper to that effect. I admit that my immediate reaction was sceptical: "they would, wouldn't they", given how wedded EU institutions are to the cause of fiscal consolidation. But they have produced a lovely elegant mathematical model that appears to demonstrate this. And they use heaps of real data to test the model (though see my note at the end of this post about sample bias). So in the interests of keeping an open mind, I read the paper.

The researchers take their terms of reference from, among others, a rather good BIS paper from 2011 that showed that banks  are adversely affected by higher sovereign risk due to losses on their holdings of sovereign debt. However, the BIS paper is mainly focused on funding, whereas the ECB researchers aim to show that banks' capital strength improves as a direct consequence of fiscal consolidation. They use the Tier 1 capital ratio as the key measure of capital strength.

The Tier 1 capital ratio is the ratio of shareholders' funds (equity) to risk weighted assets. Assets (loans and investments) held by banks are of varying degrees of riskiness depending on who the counterparty is, what the market is like and whether there is collateral. Risk weighting reduces the value of less risky assets by a percentage related to their total risk: so for example, if a £100k mortgage is weighted at 50% it will be shown on the risk weighted balance sheet as £50k. Broad risk weightings for different classes of asset and different types of risk are set by the Basel committee, but are actually calculated by banks themselves, in some cases using their own models. In general, the more risky an asset, the higher the risk weighting. It might surprise some people to learn that the sort of loans we want banks to make - unsecured loans to small businesses - are about the riskiest loans possible and carry a risk weighting of 100%, whereas loans against property are weighted at 50% or less. The target ratio of Tier 1 capital to risk weighted assets set by the regulator governs the amount of lending that a bank can do. Clearly, the lower the risk weight, the higher the total value of loans that can be made without need for more Tier 1 capital. As banks' principal source of income is interest on lending, is it any surprise that banks prefer lower-risk loans and charge high rates of interest on the riskiest loans, such as those to SMEs?

For most of the period from which the researchers draw their data, the risk weighting of sovereign debt was zero, even for debt that markets consider risky. This means that sovereign debt holdings don't appear at all on the risk weighted balance sheet - they require no capital. But they pay interest, though at a lower rate than riskier assets. Therefore it is in banks' interests to hold sovereign debt. And the more sovereign debt a bank holds relative to assets with higher risk weightings, the higher their Tier 1 capital ratio for the same unweighted balance sheet size - which reduces market perception of their own riskiness and therefore reduces their funding costs.

The BIS paper argues that if they are able, banks respond to increasing sovereign risk by adjusting their investment portfolios to reduce their holdings of more risky sovereign debt and increase holdings of less risky assets. Note that this is not a change in the overall holding of zero-weighted assets - it is a portfolio adjustment within the zero-weighted asset category. Eurozone banks may substitute less risky sovereign debt from another Eurozone country: other banks may substitute cash or US Treasuries for their own country's debt (dollarisation is a common response to increasing sovereign risk). This is of course a rational response to an increasing sovereign risk premium, although it is unlikely that distressed banks would sell their entire holdings of their own governments' debt, because they can use it as collateral for central bank funding when they are shut out from interbank markets. The toxic relationship between banks and governments remains a matter of considerable concern in the Eurozone.

The researchers correctly describe the BIS paper's suggestion that banks adjust their zero-weighted asset portfolios when sovereign risk increases. But they then go on to suggest, as their basic hypothesis, that when sovereign risk decreases banks will increase their holdings of sovereign debt (my emphasis):
"Banks may therefore react to changes in sovereign risk through adjustments on the asset side, i.e. changes in the portfolio composition. By the same token, a fi scal adjustment should trigger more appetite for government securities by banks, as treasuries are perceived to be safer after a fi scal consolidation.
This is a complete non sequitur. Country diversification within the sovereign risk category, as BIS suggests, doesn't change overall holdings of sovereign debt - it just changes their composition. But an increase in total holdings of sovereign debt implies reduced holdings of riskier assets such as corporate and household loans. That is a very different matter and not necessarily a good thing, as I shall explain.

The proportion of a bank's balance sheet that is made up of zero-weighted assets is determined by its degree of risk aversion. A bank that wishes to take no risk at all with depositors' money - say a full reserve savings bank - will only hold assets with zero risk weighting. At the opposite extreme would be, say, a peer-to-peer lender which has no capital of its own but acts a a pure intermediary: lending losses rebound to depositors, who have to make their own arrangements for protection. The vast majority of banks sit somewhere in between those two extremes. Assuming that management is rational (which is perhaps a rather heroic assumption), it is reasonable to suppose that the zero-weighted proportion is always optimal for any bank in relation to its business model, the economic circumstances and the regulatory stance. Without any change in business model, economic circumstances or regulatory stance, therefore, banks have no reason either to increase or decrease total holdings of zero-weighted assets due to changed perception of risk in one, or even several, sovereigns. So although banks may increase holdings of a particular sovereign's debt in response to perception of reduced risk, under normal circumstances they would do so by reducing holdings of other risk-free assets, not by reducing holdings of riskier assets. An increase in TOTAL holdings of zero-weighted assets indicates higher risk aversion by banks and/or regulators. The researchers not only make no mention of this, they design their entire model to test the thesis that total holdings of sovereign debt will change as a direct response to changes in sovereign risk without any change in bank or regulatory risk aversion. This is in my view a serious error.

The risk of sovereign debt is determined by current and anticipated economic realities in the country concerned. The authors assume that fiscal consolidation always reduces the risk of government securities. This implies that economic fundamentals should always improve as a consequence of fiscal consolidation. But in the light of recent history in the Eurozone, this assumption must be challenged. Front-loaded fiscal adjustment causing severe economic contraction increases the risk of default on government debt, since collapsing economic activity leads to lower tax take, making it more difficult for countries to meet their debt obligations. Therefore severe and poorly-constructed fiscal consolidation can actually make government debt more risky, not less.

The researchers exclude indirect economic effects of fiscal consolidations on bank balance sheets. They consider only direct effects - by which they mean deliberate adjustment of asset portfolios solely in response to changed perception of sovereign risk arising from fiscal consolidation. The authors describe expected indirect effects of fiscal consolidation thus:
A second, and indirect, channel would be related to the macroeconomic e ffects of fi scal consolidations. If fi scal adjustment leads to an economic downturn, it would increase the likelihood of non-performing loans and write-o ffs. If those eff ects are strong, one should observe more investment in government securities when a country enters a period of fiscal consolidation.
So the prospect of increased losses encourages banks to look for safer assets. But economic contraction does not just cause bankruptcies and loan writeoffs. It also reduces demand for loans. The combined effect of loan defaults, reduced demand for loans, reduced willingness of banks to offer loans except to very good risks, and shortages of safe assets, is that bank balance sheets tend to shrink in recessions, both in absolute and - particularly - in risk weighted terms. The 2009 data in the authors' Table 2 show this effect dramatically. Risk lending simply collapsed in the 2009 recession.

Conversely, in boom times banks increase their risk lending - again, the data in Table 2 show this clearly (see figures for 2005-7). So the "safety" of bank balance sheets is a cyclical phenomenon. This is partly due to the fact that loan defaults are cyclical - more loans default in recessions than in booms - and partly due to the fact that banks are more likely to be risk-averse in recessions, even if there is no change in the regulatory stance. Any bank facing increased loan losses is bound to be nervous about taking on more high-risk loans.

But fiscal consolidations tend to be associated with recessions, partly because they are an inevitable political response to higher deficit spending in a recession, and partly because extracting money from the private sector in order to improve the fiscal position inevitably reduces economic activity, at least in the short term.  It seems likely therefore that banks increasing holdings of government debt and reducing risk lending is not a direct response to fiscal consolidation as the authors suggest, but a response to increased fragility of bank balance sheets in an economic downturn. In other words, the direct effect that the researchers identify - deliberate rebalancing of the total asset portfolio towards less risky assets - may in fact be due to increased risk aversion rather than perception of reduced sovereign risk. After all, if a sovereign is perceived as less risky, its economy should be in good shape and we would therefore expect banks to be increasing risk lending in that country, not reducing it. If banks really do increase their holdings of sovereign debt during fiscal consolidations, that suggests banks, at any rate, don't regard the economic effects of fiscal consolidations as positive.

To be (slightly) fair to the authors, they do try to control for the issues I raise here.  Section 4.1 of their analysis describes the approach they take to controlling for bank crises. Dummying out bank crises is sensible. Though is a three-year lag between crisis and consolidation a sufficient time delay? Banks have not yet recovered from the 2007/8 crisis, yet the researchers' criteria would include the current fiscal consolidations in the UK and much of the Eurozone - only the Greek and Irish consolidations started too close to the crisis for them to be included.  But the real issue is not bank crises, but correlation between fiscal consolidations and recessions. The researchers explicitly include the business cycle in their model by means of an output gap variable, but they don't control for correlation between fiscal consolidations and economic downturns. I suspect they wouldn't have much data left if they did.

Their attempt to control for economic distress using the exchange rate to the US dollar is flawed, too. It is  pointless for those countries in the sample that are Euro members, since the Euro's value relates to the performance of the Eurozone as a whole. And for the same reason it is pointless for countries whose currency is pegged to the Euro, such as Denmark. It is also pointless for the "safe haven" countries Japan and Switzerland, since the value of their currencies is driven more by investor perception of risk in the US and Euroland than by fundamentals in those countries. And of course it is utterly pointless for the largest part of the data set, namely US banks. That doesn't leave many countries in the data set for whom the exchange rate could be used as a predictor of distress.

But there is an even bigger issue underlying all of this. The researchers show that banks do in fact increase their holdings of zero-weighted assets and reduce risk lending in response to fiscal consolidations. I dispute their analysis of the reasons for this. But banks reducing risk lending when fiscal consolidation is associated with economic contraction (which it usually is) is bad news. A lending bank that won't take risks is not much use. And banks that substitute government debt for risk lending to corporates and households are worse than useless - they are toxic for the economy for two reasons, firstly because they withhold essential investment, and secondly because in competing with investors such as pension funds for government securities they raise prices and push down yields, wrecking returns on savings and hurting people on fixed incomes.*

In the looking-glass world of finance, safety is an illusion, and the search for it is dangerous. Fiscal consolidation may indeed force banks to increase total holdings of risk-free assets, but in my view that would be a response to balance sheet fragility in a recession, not improved sovereign risk. And this would not necessarily be a good thing. All proposals that I have seen for portfolio changes to make banks safer founder on the problem that a safe bank is also a risk-averse one, and risk-averse banks are bad for the economy. We want banks to accept and manage risk - not avoid it.

Related links:

Fiscal consolidations and bank balance sheets - Cimadomo, Hauptmeier & Zimmermann (ECB)
The impact of sovereign credit risk on bank funding conditions - BIS
The illusion of safety - Coppola Comment
The deadly quest for safety - Coppola Comment
(and the rest of my posts on fear and safety too)

* Yes, I know that's what QE does. But the central bank isn't substituting safe assets for risk lending. It is trying to remove safe assets from circulation to encourage banks and investors to take more risks. A laudable ambition, though it doesn't seem to work very well.

Note on the data: The data sets in the ECB team's analysis suffer from sample bias. German banks predominate in their bank data sample, exceeded only by US banks - which they later exclude because of concerns that US banks aren't like European ones. There are two problems with this. The first is that US banks are most of the data set - not surprisingly, because the US is an enormous country with a fragmented banking system:  by excluding US banks, the researchers effectively make this a European rather than global analysis, despite leaving in Japanese banks. Secondly, the fact that European banks and Japanese banks aren't like German ones passes unnoticed. Indeed, in the commentary the German bias is evident. For example, on p.10:
"We think of savings banks and credit cooperatives as mostly providing loans to small businesses...."
The point they are making is valid, namely that small banks don't generally invest in corporate bonds. But in the UK, for example, small banks and credit unions don't mostly provide loans to small businesses. They mostly provide secured loans to households.

Germany also predominates in the fiscal consolidation episodes identified by the IMF, although Italy is a close second. The key samples used for this analysis are therefore skewed towards German banking structures and German fiscal policy. It almost amounts to an analysis of whether German fiscal consolidations made German banks safer. That might indeed be a useful piece of research, though perhaps the Bundesbank would have been a more appropriate sponsor. But for this to be a reasonable analysis of whether IN GENERAL fiscal consolidations make banks safer, I would have expected the researchers to have retained representative US data and controlled for German dominance.


  1. First of all, you cannot ever be an academic because you can explain things too well and too simply (in contrast to terrible and too complicated!).

    Just two comments:
    1. So the "safety" of bank balance sheets is a cyclical phenomenon - Minsky again I see. It actually makes much more sense than the article you comment.

    2. Quote from Blanchard&Cottarelli's Ten Commandments for Fiscal Adjustment in Advanced Economies: "You shall not front-load your fiscal adjustment, unless financing needs require it." Amazing how they fail to live up to their own suggestions...

    Sample bias is a major issue in most academic papers. You can practically show that everything is correlated to everything if you try enough. Yet, we forget that it's merely people interacting with each other

  2. You explain things so simply Frances....
    Clay :)