Interbank markets have been brought to the fore by the global crisis. From a macroeconomic viewpoint, interbank rates matter because they are a key part of the monetary policy transmission mechanism. Under normal circumstances they are determined almost one-to-one by policy rates, and their variations are rapidly transmitted to the entire term structure, eventually affecting borrowing conditions for households and firms. Figure 1 documents how this state of affairs came to an end in the summer of 2007, and helps explain why central banks worldwide have become so concerned about interbank spreads (the differential between interest rates on unsecured and secured deposits). A further, strong increase in the spreads took place in mid-September 2008, when the default of Lehman Brothers hit the markets. Despite having since declined to pre-crisis levels, spreads have started to inch upwards again in recent months. The root causes of the rise in the spreads, as well as those of the fall, have not yet been fully understood (Taylor and Williams 2008).
Figure 1. Interbank rate spreads (basis points: daily data)
Source: Reuters, Datastream. Notes: The dollar spread is computed as the differential between the LIBOR USD 3-month rate and a composite index of contributors on the bid side for repo 3-month with Treasuries as collateral. The euro spread is computed as the difference between the Euribor and the Eurepo rate of the same maturity.
In a recent study (Angelini et al. 2009) we analyse the micro- and macro-economic determinants of the spreads using bank-specific data from the e-MID, a screen-based trading facility used by Italian and other European banks to exchange uncollateralised funds. This unique database contains information on individual interbank trades at maturities ranging from overnight to one year. Matching them with key characteristics of the banks selling and buying funds (balance sheet and profit and loss account data, ratings, etc…) we complement and qualify the results of previous analyses, and answer several unaddressed questions. Our main results are as follows.
Taylor and Williams (2008) argue that the rise in the spread was largely driven by an increase in the risk premium. Our findings confirm this result and allow us to break down the risk premium into its components:
- The riskiness of banks (the quantity of risk), and;
- The compensation investors require per unit of risk (risk aversion).
About two thirds of the average increase in the spread observed since August 2007 can be traced to higher risk aversion. By contrast, borrowers’ riskiness, at least as captured by our observable measures of bank-specific default risk (rating, capitalisation, profitability), played a minor role.
These proxies remain relatively constant across the pre-crisis and crisis periods, or improve slightly. A decline in the share of cross-border transactions is observed, especially after the collapse of Lehman brothers, but no significant differences emerge between rates charged to national counterparties and those involving a foreign borrower, both prior and after the onset of the crisis.
Overall, this evidence does not lend support to theories focusing on adverse selection effects and limited participation as an important determinant of the increase of interbank spreads.
Accounting procedures also contributed to the increase in the interbank spread. Quarter-end and year-end data on outstanding interbank loans appear in banks’ internal reports and in public accounting releases. Both have a bearing on the amount of regulatory capital. In principle, banks must respect capital requirements at any time, but end-of-period data are particularly relevant as they allow interested parties – the top management, regulators and external investors – to formally check the requirements. Therefore, banks have an incentive to curtail lending on those days, causing rates to increase.
While this pattern is well-identified throughout our entire sample, after August 2007 it accounts for large jumps in the spread (up to 60 basis points for maturities between two weeks and one month). This suggests that during the crisis banks have become more reluctant not only to incur interbank exposures, but also to make them public.
Likewise, the substantial increase in the volatility of the spread observed since August 2007 largely reflects aggregate volatility. The standard deviation of the spread over the crisis period is 15 basis points if computed using raw data on individual 3 month maturity transactions compared with only 4 points if computed on the same data in deviation from daily means. Bank characteristics explain only a modest portion of this residual cross-sectional variance of the data.
The role of liquidity
There is an ample body of theoretical literature emphasising liquidity shocks as a key driver of interbank crises. It has been suggested that that the rise in the spread observed since August 2007 was due to an aggregate (funding) liquidity shock, exacerbated by the steady decline in the ratio between banks’ liquid assets and total assets observed worldwide over the last two decades. If this were the case, banks with more liquid assets should have paid lower-than-average spreads. However, measures of borrowers’ liquidity turn out to be unrelated to the spread, both before and during the crisis.
In a similar vein, the exceptional refinancing operations launched by the ECB since August 2007 had some direct dampening effect on the spread only in the post-Lehman period (although we believe that central bank interventions prevented the crisis from spiralling out of control, and therefore had an important but a hard-to-measure indirect effect on the spreads). Overall, this evidence suggests that low funding liquidity was not a major driver of the rise in the spread, and that theories of interbank contagion triggered by aggregate liquidity shocks may not be capturing the full essence of the current crisis.
We also fail to support the hypothesis that a shortage of capital among lenders contributed to the increase of the spread. Lender liquidity appears to be unrelated – or even positively related – to the spread (a positive relationship could reflect predatory behaviour in the interbank market, or merely signal high risk aversion). This evidence reinforces the view that a shortage of funding liquidity was not an important driver of the spread. Finally, large lenders charged lower rates.
During the pre-crisis period borrower characteristics, in particular, observable measures of creditworthiness (rating, capitalisation) were not an important determinant of the cost of funds. Virtually the only significant characteristic was size; large banks could borrow at a discount. Since August 2007, better rated, better capitalised borrowers obtain a discount. However, the difference between rates paid by good borrowers compared with bad borrowers remains altogether modest. Overall, these results support the theories maintaining that banks are unwilling to engage in peer monitoring, for example because they believe that the central bank will intervene in case of a crisis, or because monitoring incentives are reduced by the availability of public noisy signals about borrowers’ creditworthiness.
Improving borrowing conditions
On the surface, the effect of the borrower size on the spread during the crisis is ambiguous. Considering that key features of large banks (complexity, international diversification) make them likely to be more exposed to the crisis, one could have expected their borrowing conditions to deteriorate in relative terms. On the other hand, the classical too-big-to-fail argument, in the light of the wave of bailouts at the end of 2007 and in 2008, would point to an opposite effect. Our estimates show that the discount enjoyed by large banks has increased during the crisis, suggesting that latter considerations have offset the former, and that moral hazard risks may have increased.
The results summarised thus far remain broadly unchanged if the analysis is extended to include the period following the Lehman Brothers failure. Over this extended period, it is particularly interesting to look at the effect of borrowers’ size on the spread. Again, how this effect might have changed is ambiguous. On the one hand, the Lehman default could have shattered the faith in the too-big-to-fail implicit warranty, causing a decline in the discount to large borrowers. On the other hand, the sudden acceleration of the financial crisis after this event makes it equally plausible that lenders may have been reinforced in their belief that no (other) large bank would be allowed to fail. The estimates predominantly support the latter view: the discount enjoyed by large banks survives the Lehman failure virtually unchanged.
Addressing the too-big-to-fail problem
Kane (2010) argues that the size discount enjoyed by large institutions may be viewed as a proxy for the value of safety-net support captured by financial firms, which represents “a cogent way to measure what authorities ought to mean by ‘systemic risk’”. He proposes that individual financial firms should produce reliable estimates of their subsidy, which should be validated by regulators and made public, as a means to improve transparency and general awareness of the too-big-to-fail problem. While Kane falls short of proposing that a levy or a capital surcharge be tied to the estimated subsidy, this could be the obvious next step, as proposals in this class are indeed being considered in the current debate. Of course, the estimates of the too-big-to-fail subsidy should be sufficiently reliable, if they were to be used as a key ingredient of such a regulatory scheme.
Our results represent a step forward relative to existing figures, as they control for borrowers’ creditworthiness, disentangling a discount purely based on size, from one based on soundness. To conclude, our main findings are that:
- Aggregate factors (risk aversion and accounting practices), rather than bank-specific ones, seem to have driven the increase in the spread
- Funding liquidity, capital shortage and central bank interventions were not important drivers of the spread
- Lenders did not strongly discriminate based on borrowers’ creditworthiness
- During the crisis, borrowing conditions improved for large banks suggesting that the too-big-to-fail moral hazard remained
Angelini, P, A Nobili and C Picillo (2009), “The interbank market after august 2007: what has changed, and why?”, Banca d’Italia, Temi di discussione n. 731.
Kane, E (2010), “Redefining and containing systemic risk. Paper presented at the 2010 financial markets conference “Up from the ashes. The financial system after the crisis”, Federal Reserve Bank of Atlanta, May 11-12.
Taylor, JB and JC Williams (2009), “A Black Swan in the Money Market”, American Economic Journal: Macroeconomics, 1:58-83.