Systemic Risks in the Danish Market for Uncollateralised Overnight Deposits
This chapter analyses systemic risks in the market for uncollateralised overnight deposits. For this purpose, systemic risk is defined as the risk that difficulties experienced by one institution may trigger a chain reaction resulting in other institutions also experiencing problems. Such contagion may threaten financial stability. Specifically, the consequences if a major market participant unexpectedly has difficulties meeting its obligations in the money market are simulated. The analyses operate with two distinct scenarios. The first one analyses the effect on the other institutions' capital buffers; the second one the effect on their liquidity buffers.
The simulations show that the effect on the capital buffers is the greater of the two. On three of the ten days analysed, other institutions are subsequently affected. In the liquidity buffer scenario, only one institution is affected on one day.
Overall the systemic risks are deemed to be limited. However, it cannot be ruled out that unexpected difficulties experienced by individual participants may have systemic effects.
The uncollateralised overnight money market
The Danish money market is the market for loan agreements and interest-rate derivatives in kroner with a maturity of up to one year between credit institutions. The money market, also known as the interbank market, is mainly used to exchange krone liquidity and to manage short-term interest-rate positions. Consequently, exposures are established between the institutions.
Overnight loans account for around two thirds of the liquidity exchanges in the money market. The turnover comprises uncollateralised deposits, repos and foreign-exchange swaps.
At end-2003, Danmarks Nationalbank's monetary-policy counterparties comprised 117 credit institutions. In the period 22 September 2003 to 18 March 2004 approximately half of them had at least one uncollateralised overnight transaction with a counterparty. On average, the total market for uncollateralised overnight deposits among the monetary-policy counterparties is approximately kr. 9 billion per banking day. The market is relatively concentrated with the six largest institutions accounting for around two thirds of the turnover.
Method and limitations
The basic purpose of the analysis below is to examine whether unexpected events in one institution trigger a chain reaction leading to problems for other institutions and thus threatening financial stability. Only uncollateralised overnight deposits among Danish institutions with a current account at Danmarks Nationalbank are considered (i.e. Danmarks Nationalbank's monetary-policy counterparties). Box 17 describes the data.
|
Box 17
|
|
|
The data for the main part of the analysis is a calculation of the turnover in uncollateralised overnight deposits in the period 22 September 2003 to 18 March 2004. The starting point for the calculations are payments made via Danmarks Nationalbank's payment system, Kronos. The underlying intuition is relatively simple. For instance, if on Monday bank A grants an overnight loan of kr. 100 million to bank B, the transaction will entail a payment of kr. 100 million from bank A's to bank B's current account at Danmarks Nationalbank on the same day. When the loan is repaid with interest the following day (Tuesday), a payment of kr. 100 million plus interest from bank B's current account to bank A's current account occurs. A search algorithm thus runs through payments between current accounts and – on the basis of a number of assumptions regarding interest and amounts – identifies uncollateralised overnight deposits. In this way a matrix can be set up, by means of which the counterparties' mutual exposures can be calculated. This enables calculation of any systemic effects for this market segment. To check the positions calculated, they have been compared with reports from a questionnaire survey for selected banks. This random sampling confirms that it is possible to identify uncollateralised overnight deposits via the method described. |
|
| Note: This method is inspired by an analysis of the systemic risks in the US interbank market, see Craig Furfine: Interbank Exposures: Quantifying the Risk of Contagion, Journal of money, credit and banking, Vol. 35, No. 1, February 2003. An equivalent analysis method is used by Stephen Millard and Marco Polenghi from the Bank of England in The relationship between the overnight interbank unsecured loan market and the CHAPS Sterling system, Bank of England Quarterly Bulletin: Spring 2004 | |
The analysis is based on an extreme scenario assuming that the institution with the largest gross deposits of uncollateralised overnight deposits unexpectedly ceases to meet its obligations on a given day. This is assumed to lead to direct losses for the institutions that have provided uncollateralised loans to the institution in question. Some of the other institutions may thus experience problems if the amounts are sufficiently large. This can trigger a chain reaction whereby several other institutions may have problems meeting their obligations.
Initially it is assumed that the institutions lose their entire exposure. Subsequently the sensitivity is analysed if a given proportion of the exposure is recovered[1].
|
Box 18
|
|
|
For two reasons, branches of foreign institutions are not included in the analyses at present. Firstly, in a Danish context they are too small in relation to the parent group. Secondly, it is difficult to distinguish between liquidity and capital in the branch and group, respectively. As described in the chapter on branches of foreign credit institutions, the development indicates that foreign branches will gain increasing importance in the individual countries. |
|
It is important to realise that uncollateralised overnight deposits only account for a part of the transactions among these institutions. The analysis in this chapter therefore gives only a partial picture of the systemic relations.
Analyses of this type are typically based on problems in the institution with the largest gross deposits. However, it should be noted that the institution with the largest gross deposits is not necessarily the one that could generate the greatest systemic effects.
There are almost 190 banking institutions and mortgage-credit institutions in Denmark, of which approximately 70 do not hold current accounts at Danmarks Nationalbank. The latter are typically small institutions that effect payments, etc. via correspondent agreements with other institutions. The analysis in this chapter only comprises institutions with a current account at Danmarks Nationalbank, which suggests that any systemic effects are underestimated if a bank acting as a correspondent bank ceases to meet its obligations.
The simulations cover two different scenarios. The first one takes into account the institutions' capital in excess of the statutory 8-per-cent solvency ratio or 4-per-cent core capital[2]. The second scenario considers the institutions' access to liquidity in a situation where their loans are not repaid. The detailed assumptions underlying the two scenarios are described below. In general the scenarios are simplified. The analyses do not take into account the fact that the situation described may occur over some time and thus offer an opportunity to react. The results should therefore be interpreted with caution.
|
Box 19
|
|
|
Analyses of systemic risks based on estimation of the banks' bilateral exposures have also been performed in other countries. On the basis of transactions in the US payment system, Fedwire, Craig Furfine estimates the systemic risks in the US inter-bank market attributable to uncollateralised overnight deposits. The method applied in the analysis resembles the one used by Danmarks Nationalbank in this analysis. For instance, an attempt is made to quantify the effect if the institution with the largest deposits fails. Furfine's analysis shows that if 60 per cent of the exposures are ultimately lost (i.e. a recovery rate of 40 per cent), two to six banks are subsequently affected to a degree that entails a risk that they will also fail. The institutions experiencing difficulties are typically relatively small, but in a few cases banks with balance-sheet totals of approximately USD 30 billion will also suffer serious problems. Via a questionnaire survey, Sveriges Riksbank has looked into the systemic effects between the four largest banks. Unlike Furfine's analysis and the analysis in this chapter, Sveriges Riksbank has data from only few reporting dates at its disposal. The result of the simulations shows that even on the assumption that the full exposure is lost, no banks lose their entire core capital on the days under review, and thus they do not subsequently experience difficulties. If, however, the analysis parameter is the access to liquidity, the survey shows a risk on the days in question that several of the other participants will incur liquidity problems. |
|
| Source: Craig Furfine: Interbank Exposures: Quantifying the Risk of Contagion, Journal of money, credit and banking, Vol. 35, No. 1, February 2003. Sveriges Riksbank, Financial Stability Report 1, 2003 | |
Scenario 1 – capital buffer
This scenario considers the exposure in relation to an estimate of the institutions' capital in excess of either the statutory 8-per-cent solvency ratio or the 4-per-cent core capital requirement relative to risk-weighted assets. The amount in question comprises the institution's excess capital reserves and revenues in the current year.
The simulations thus assume that an institution will subsequently experience problems if the total capital buffer is less than the institution's exposure vis-à-vis the institution not meeting its obligations. It is also assumed that the institutions are unable to change the composition of their risk-weighted assets in the short term.
Scenario 2 – liquidity buffer
The second scenario analyses the institutions' access to liquidity in relation to their bilateral exposures. If a given institution cannot meet its obligations, institutions with claims on the institution in question will not receive the liquidity they expect. If these institutions cannot raise sufficient alternative liquidity, a chain reaction may be triggered. At present no analyses have been made to disclose the extent to which this may disrupt the general settlement of payments.
It is assumed that the immediate liquidity buffer is the institutions' maximum intraday access to current-account overdrafts against pledging collateral to Danmarks Nationalbank[3]. Thus other options to raise liquidity, including any drawing facilities at other institutions, are disregarded. It is assumed that the problems experienced by the largest participant occur in the morning, before the payment systems open. Consequently, access to liquidity once the payment systems open constitutes the maximum liquidity buffer[4].
Results
For analysis purposes the 10 days with the highest turnover in uncollateralised overnight deposits, in the period 22 September 2003 to 18 March 2004, have been selected. On these 10 days, an average of 20 institutions had at least one uncollateralised overnight money-market loan. The average turnover per day was approximately kr. 16.5 billion. The highest turnover was kr. 18.7 billion.
Chart 39 shows the number of institutions that lose more than their capital buffer or liquidity buffer on these days if the institution with the largest deposits unexpectedly experiences problems meeting its obligations. In the current capital situation, this would on average entail that less than one institution would subsequently experience problems. On the day with the largest effects, four institutions are affected, while no effect is registered on seven days. There is no systematic picture of the institutions which are typically hit. As regards liquidity, the effect is less pronounced. Only one institution has insufficient liquidity during these 10 days.
The number of links in the chain may vary. If an institution's problems in meeting its obligations entail direct problems for other participants, and the effects subsequently cease, this is an effect of the "first order". If the problems experienced by the institutions affected in the first instance lead to difficulties for other institutions, this is an effect of the "second order", etc. On the day when four institutions are affected, a capital-buffer effect of the second order is seen. On the other two days when effects are registered they are only of the first order. Since only one institution is affected in terms of liquidity, these effects are by definition of the first order.
As mentioned, the simulations are based on the period's 10 days with the largest turnover in uncollateralised overnight deposits. To get an idea of whether the results are representative, additional simulations have been performed for 10 days with turnover close to the average turnover in the period. On these days the number of institutions with insufficient capital buffers is reduced to one institution on one day. At the same time, the liquidity buffer proves to be sufficient to prevent any systemic effect.
To assess the sensitivity of the results, it is assumed that the institutions' current capital and liquidity buffers are reduced. This naturally entails that other institutions subsequently experience difficulties. Chart 40 shows that if the capital buffer is reduced to 70 per cent of its current value, an average of just over one institution per day will experience difficulties. In terms of liquidity, the current liquidity must be reduced to 20 per cent before more than one institution a day finds itself in difficulties.
If it is assumed that the institutions do not lose their full exposure, but only part of it, the effects are correspondingly lower[5]. Simulations show that if institutions recover just 25 per cent of their exposures, the current capital buffers will not entail systemic effects on the days analysed.
| Sensitivity to buffer size |
Chart 40
|
 |
|
| Note: Average per day for the 10 banking days in the period which saw the largest turnover. | |
| Source: Danmarks Nationalbank. | |
[1] In reality the exact loss is not known until some time later. Studies involving US banks show that the percentage recovered varies between 40 and 95 per cent. See Craig Furfine, Interbank Exposures: Quantifying the Risk of Contagion. Journal of Money, Credit and Banking, Vol. 35, No. 1, February 2003.
[2] I.e. either the part of the core capital exceeding the statutory 4 per cent or the part of the liable capital exceeding the statutory 8 per cent (both in relation to risk-weighted assets).
[3] The maximum intraday overdraft access is determined by the current-account balance, the pledgeable value of certificates of deposit, approved bonds pledged to Danmarks Nationalbank, and utilisation of the automatic collateralisation agreement.
[4] In addition, repayment of uncollateralised overnight money-market loans typically takes place from the morning, when the payment systems open, which makes it plausible to operate with liquidity access at this time.
[5] This is only relevant for the simulations of the capital buffer, since the recovery percentage is calculated with a time lag. For liquidity, the problem occurs immediately, and consequently there is no point in looking at recovery.