Journal Issue
Share
Article

Singapore

Author(s):
International Monetary Fund
Published Date:
August 2008
Share
  • ShareShare
Show Summary Details

I. Assessing the Stability of Singapore’s Banking System in a Regional Context1

This chapter proposes a methodology for assessing the stability of the banking system in Singapore in a regional Asian context. The methodology allows for a quantification of the evolution of default interdependence of Singaporean and selected regional banks. The main results of the analysis indicate that the largest Singaporean banks have remained resilient to the global financial turmoil and have generally been less affected than regional banks operating in Singapore.

A. Introduction

1. The direct impact of the global credit turmoil on Singaporean banks has been limited so far. Credit default swap (CDS) spreads for Singaporean and Asian banks increased significantly since the second half of 2007 on the back of the global credit turmoil and have remained elevated even after a decline since March 2008. However, the reported subprime related exposures and estimated losses of Singaporean banks are lower than elsewhere in Asia, and in the United States and Europe. Moreover, Singaporean banks have made prudent provisions against losses on exposures to the U.S. subprime related assets.

Banks: Credit Default Swap Spreads

Stock Market Indices

(2000M1=100)

2. This chapter assesses the stability of the banking sector in Singapore using a novel methodology. The methodology is still work in progress and the results should be interpreted with care. The central insight is to derive the probabilities of default (PoDs) of a sample of Singaporean and regional banks and estimate the joint probability of default (JPoD) and other conditional measures of banking sector stability. The chapter proceeds as follows: Section B summarizes the analytical framework, Section C presents the main finding, and section D concludes.

B. Analytical Framework

3. The analysis provides novel measures of banking stability and contributes to the modeling of default risk. It extracts market information to assess potential contagion effects and the resilience among Singaporean and selected regional banks operating in Singapore.2 The exercise provides a new methodology and complements the study presented in the special feature in the 2007 Monetary Authority of Singapore’s Financial Stability Review.3

4. The central idea behind the methodology is to treat the banking system as a “portfolio of banks.” The estimates of banking system stability capture risks of the individual banks and interdependencies of the banks in the portfolio.4 The model is applied to a portfolio of Singaporean banks only and one of Singaporean and regional banks.5

5. The analytical framework uses two variables to estimate different measures of banking stability: daily equity prices and daily CDS spreads.6 Equity prices are used to calibrate the initial conditions for model simulations.7 Based on the PoDs for individual banks extracted from CDS spreads, the framework then computes a multivariate density function (PMD) to capture the implied distribution of asset values of the banks included in the portfolio.89 The PMD embeds the default dependence among the banks in the portfolio and allows for the estimation of the JPoD of the bank portfolio under consideration.10

6. The JPoD represents the unconditional probability of default of all the banks in the sample, i.e. the tail risk of the system. The JPoD accounts for the nonlinear dependencies among banks in this portfolio of banks.11 In periods of financial distress, the JPoD of the banking system may experience larger fluctuations compared with those of the PoDs of individual banks because of stronger interdependencies during times of stress.

7. The JPoD provides the base for calculating conditional measures of banking stability:

  • The Banking Stability Index(BSI), which shows the expected number of bank defaults, conditional on at least one bank defaulting.12
  • The Default Dependence Matrix(DDM), which is a matrix of pairwise conditional probabilities of default, indicating the probability of default of a bank in the row, given that a bank in the column defaults;
  • The Conditional Systemic Relevance Factor(SRF), which reflects the probability of default of all the banks in the system conditional on the default of a specific bank; and
  • The Conditional Resilience Factor(RF), which indicates the opposite of the SRF, namely the probability that a bank defaults conditional on the default of all the other banks.

8. The methodology is subject to some data limitations when applied to emerging market countries. First, inputs (equity prices and CDS spreads) used may not be the best proxy for estimating banks probabilities of default, especially in times of global market turmoil.13 Second, limited data availability on CDS spreads constrains the choice of banks, which may result in a sample that is not fully representative of the banking sector under consideration.

C. Main Findings

9. An assessment of banking sector vulnerabilities involves a careful evaluation of a broad spectrum of indicators, in changes and levels. Changes bespeak the impact from global financial reverberations, while levels provide evidence of resilience. In the case of Singapore, changes in the indicators confirm international spillovers, but the absolute levels of the measures suggest that overall resilience of the Singaporean banking system remains high. As regards linkages with regional banks, evidence is more ambiguous. In particular:

  • The PoDs and the JPoDs of both bank groups analyzed increased as the global credit crisis unfolded. Between the end of July 2007 and the end of March 2008, the average PoD of the institutions in the “Singaporean bank” portfolio increased by about 7 times, while their JPoD increased by a larger factor.14 The average PoD of the group of “Singaporean and regional” banks increased by 3 ½ times, while its JPoD rose by even more (Figures I.1 and I.2). The significant difference between the magnitude of increases in the average PoDs and the JPoDs reveals large increases in default interdependence among the banks during this period of financial distress. The JPoD of the Singaporean-regional bank portfolio is smaller than that of the Singaporean bank portfolio, which could reflect diversification gains and/or a nonrepresentative sample. Because of the data limitations mentioned above, this evidence needs to be interpreted carefully.
  • The BSI shows signs of an adverse impact from the global turmoil on Singapore’s banking sector, but largely through regional banks (Figure I.3). For the group of Singaporean banks, the BSI increased by only 0.5 reaching 1.6 between mid-2007 and the end of March 2008. Thus, before the onset of the market turmoil, only a partial default at one bank was expected, if another bank in the sample defaulted. For the group of Singaporean and regional banks, the BSI increased by about 1.3 reaching 2.7, implying that the presence of regional banks in the sample add some vulnerability or channels of contagion.
  • The DDM for the Singaporean banks suggests that the conditional pairwise probabilities of default increased somewhat since the second half of 2007 (Table I.1). In particular, the default interdependence between two Singaporean banks was higher than it was between each of these and the third bank.
  • The DDM for the Singaporean and regional group of banks suggests, that the regional banks in the sample could depend more on the Singaporean banks than vice versa. Although the conditional PoDs of the Singaporean banks rose since mid-2007, they stayed below the conditional PoDs of the regional banks throughout the whole period (Table I.2). These results are consistent with the view that worsening liquidity or solvency conditions at regional banks would have a modest effect, if any, on the Singaporean banks. However, this could also reflect weaker balance sheets of regional banks compared to Singaporean banks, which could imply that the former would probably face financial strains if the latter do (i.e. in response to a common adverse shock).
  • The conditional resilience factor (RF) suggests that Singaporean banks are resilient to an adverse systemic event. Among the Singaporean banks, one stands out as more resilient (Figure I.4) and the other two banks demonstrate almost identical resilience, confirming the finding from the DDMs that these two banks are becoming increasingly interlinked. Among the regional banks in the sample, one bank stands out as less resilient (Figure I.4).
  • Finally, the conditional systemic relevance factor (SRF) has been very similar for all the banks. However, its magnitude goes down when the group of regional banks is also taken into account (Figure I.5), most likely reflecting diversification gains.

Figure I.1.Marginal Probabilities of Default

Figure I.2.Joint Probability of Default and Average Marginal Probabilityx

Figure I.3.Banking Stability Index

Table I.1.Singaporean Banks: Default Dependence Matrix (DDM).15
DateSing.bank 1Sing.bank 2Sing.bank 3Average1
6/29/07
Sing.bank 11.000.340.020.18
Sing.bank 20.251.000.030.51
Sing.bank 30.020.031.000.52
9/28/07
Sing.bank 11.000.570.100.33
Sing.bank 20.501.000.100.55
Sing.bank 30.090.111.000.55
12/31/07
Sing.bank 11.000.690.140.41
Sing.bank 20.471.000.120.56
Sing.bank 30.100.121.000.56
3/31/08
Sing.bank 11.000.810.310.56
Sing.bank 20.631.000.280.64
Sing.bank 30.240.281.000.64

Row average.

Row average.

Table I.2.Singaporean and Regional Banks: Default Dependence Matrix (DDM).16
DateSing.bank 1Sing.bank 2Sing.bank 3Reg.bank 1Reg.bank 2Reg.bank 3Average1
6/29/07
Sing.bank 11.000.130.110.110.130.100.12
Sing.bank 20.101.000.150.060.120.060.28
Sing.bank 30.090.181.000.070.130.050.29
Reg.bank 10.210.170.171.000.150.150.33
Reg.bank 20.150.190.180.091.000.090.31
Reg.bank 30.480.360.310.380.401.000.49
9/28/07
Sing.bank 11.000.320.290.300.320.280.30
Sing.bank 20.291.000.380.230.350.210.43
Sing.bank 30.270.401.000.250.340.200.44
Reg.bank 10.270.240.241.000.230.230.39
Reg.bank 20.290.360.340.231.000.230.43
Reg.bank 30.540.460.410.490.501.000.57
12/31/07
Sing.bank 11.000.410.380.370.410.320.38
Sing.bank 20.281.000.400.240.360.200.44
Sing.bank 30.260.401.000.240.350.190.43
Reg.bank 10.320.310.311.000.290.260.43
Reg.bank 20.280.360.340.231.000.210.43
Reg.bank 30.650.590.540.600.631.000.67
3/31/08
Sing.bank 11.000.590.570.550.600.540.57
Sing.bank 20.461.000.580.430.550.410.59
Sing.bank 30.440.581.000.430.540.390.59
Reg.bank 10.510.500.511.000.500.480.60
Reg.bank 20.440.520.510.401.000.400.56
Reg.bank 30.700.670.640.670.701.000.73

Row average.

Row average.

Figure I.4.Resilience Factor

Figure I.5.Systemic Relevance Factor

D. Concluding Remarks

10. This chapter provides an indicative assessment of the vulnerability of Singapore’s banking system. The methodology builds on market indicators to derive measures of banking system stability and can be extended to perform stress testing of the banking system. In addition, the methodology provides technical improvements over other methods to assess financial stability. For example, the estimated measures of vulnerability account for time-varying dependencies among various banks. Thus, they go some way toward capturing dynamic interdependencies among banks during times of financial distress.

11. Although the results need to be interpreted with care, overall they point to two main findings:

  • Ripple effects from the global credit crisis have been felt, but overall resilience of Singapore’s banking system remains strong; and
  • Regional bank integration appears to have an ambiguous impact, as diversification gains may counter the opening up of addition channels for financial contagion. This said, sample selection may also play a part in shaping this result.
References

    Diebold, F., J. Hahn, and A. Taylor,1999, “Multivariate Density Forecast Evaluation and Calibration in Financial Risk Management: High-Frequency Returns on Foreign Exchange,”The Review of Economics and Statistics, vol. 81, n. 4, pp.661-73.

    • Search Google Scholar
    • Export Citation

    Goodhart, C. and M.Segoviano,2008, “Banking Stability Index,”IMF Working Paper, forthcoming.

    Monetary Authority of Singapore, 2007, “Financial Stability Review,”Singapore.

    Segoviano, M.,2008, “The CIMDO-Copula: Robust Estimation of Default Dependence under Data Restrictions,”IMF Working Paper, forthcoming.

    • Search Google Scholar
    • Export Citation

    Segoviano, M.,2006a, “The Conditional Probability of Default Methodology,”Financial Market Group, London School of Economics, Discussion Paper 558.

    • Search Google Scholar
    • Export Citation
1

Prepared by Elena Loukoianova and Miguel Segoviano (both MCM).

2

The methodology proposed here can be applied to alternative inputs for calculating PoDs of individual banks and JPoDs of different bank groups.

3

The special feature used panel data econometric estimates and analyzed data for the period from January 2002 to December 2006—prior to the onset of the credit turmoil in mid–2007. This analysis found that the East Asian banking systems became more resilient after the Asian financial crisis, the default risk of banks declined, and contagion among banks also declined. The MAS attributed lower default risk to income diversification.

4

Goodhart and Segoviano, 2008. Box 1.5 in the April 2008 GFSR presents an application of this methodology for a group of large financial institutions.

5

A number of multinational banks have branches in Singapore and thus have a bearing on domestic financial stability. Lack of branch-level data precludes, however, a quantitative analysis of the issue.

6

The data for both variables are from Bloomberg.

7

See Segoviano (2008) for details on calculation of PoDs.

9

Under the probability integral transformation (PIT) criterion, the PMD produced by nonparametric techniques is an improvement over standard parametric PMDs used for modeling portfolio credit risks (See Diebold et al. (1999) for details).

10

The methodology proposes a novel nonparametric copula approach—which assumes neither a particular distribution nor parameters, thus making possible a better fit to the data. The structure of linear and nonlinear dependencies among banks in a system can be represented by copula functions. This approach infers copulas from the joint movements of PoDs of individual banks, thus avoiding the difficulties involved in explicitly choosing and calibrating individual measurements of banks’ defaults. This is the main contrast between and traditional copula modeling approaches, as explicit calibration in most cases is difficult because of data constraints.

11

Accounting for nonlinear dependencies changing over time is a relevant technical improvement over most risk models, which typically account only for dependencies that are assumed constant over the cycle.

12

The BSI is conditional of a default of any bank in the system, but not a specific bank. Moreover, such a default might never materialize.

13

In the current episode, the risk landscape and consequently hedging strategies have shifted significantly since June 2007. The discrepancy between CDS spreads and bond prices, coupled with the increasing illiquidity of the credit markets (especially for CDS spreads of local banks), has clouded the information embedded in the CDS spreads.

14

The average PoD is defined here as a simple average of the PoDs of individual banks in a group.

15

Probability of Default over one year of a bank in a row, conditional on the default of a bank in a column.

16

Probability of Default over one year of a bank in a row, conditional on the default of a bank in a column.

Other Resources Citing This Publication