Did the Market Signal Impending Problems at Northern Rock? An Analysis of Four Financial Instruments

Abstract

The academic literature has regularly argued that market discipline can support regulatory authority mechanisms in ensuring banking sector stability. This includes, amongst other things, using forward-looking market prices to identify those credit institutions that are most at risk of failure. The paper's key aim is to analyse whether market investors signalled potential problems at Northern Rock in advance of the bank announcing that it had negotiated emergency lending facilities at the Bank of England in September 2007. A further aim of the paper is to examine the signalling qualities of four financial market instruments (credit default swap spreads, subordinated debt spreads, implied volatility from options prices and equity measures of bank risk) so as to explore both the relative and individual qualities of each. The paper's findings, therefore, contribute to the market discipline literature on using market data to identify bank risk-taking and enhancing supervisory monitoring. Our analysis suggests that private market participants did signal impending financial problems at Northern Rock. These findings lend some empirical support to proposals for the supervisory authorities to use market information more extensively to improve the identification of troubled banks. The paper identifies equities as providing the timeliest and clearest signals of bank condition, whilst structural factors appear to hamper the signalling qualities of subordinated debt spreads and credit default swap spreads. The paper also introduces idiosyncratic implied volatility as a potentially useful early warning metric for supervisory authorities to observe.

1. Introduction

The academic literature has regularly argued that market discipline can support regulatory authority mechanisms in monitoring banking sector stability (Evanoff and Wall, 2001aFlannery, 1998). This includes, amongst other things, using forward-looking market prices to identify those credit institutions that are most at risk of failure. An extensive literature has empirically analysed the risk sensitivity of subordinated debt (SND) yields and equity prices (Flannery and Sorescu, 1996; Flannery, 1998; Sironi, 2003) and, to a lesser, degree the competing ability of different instruments to identify problem banks (Gropp et al., 2006; Evanoff and Wall, 2001b). In addition, some banking supervisory authorities are showing an increasing interest in the use and quality of market signals on bank condition to support more traditional monitoring mechanisms (European Central Bank (ECB), 2004, 2005; Schmidt, 2004; Persson and Blåvarg, 2003).

On 14 September 2007 the UK bank, Northern Rock (NR) announced that it had negotiated access to emergency lending facilities at the Bank of England. This news precipitated the first run on a UK bank since 1866. The run only subsided when the government announced on the evening of 17 September 2007 that all deposits would be guaranteed. This crisis provides an opportunity to investigate the risk-signalling qualities of investors in a modern financial market context. The paper's primary research aim is, therefore, to analyse whether market investors did signal potential problems at NR in advance of the bank announcing that it had negotiated emergency lending facilities. Financial innovations in the form of new market instruments, such as credit default swaps (CDS), and the rapid rise in UK credit institution SND issuance, facilitate an analysis of multiple financial instruments in identifying bank failure. In so doing, the paper examines the signalling qualities of a number of instruments and, therefore, furthers the debate on using market data in banking supervision.

The structure of the paper is as follows. Section 2 comprises a literature review of the use of market information in bank supervision and whether it could enhance market and regulatory authority discipline. This section also examines the limited literature on the ability of different market instruments to identify problem banks. Section 3 summarises the key events that culminated in the run on NR and provides a context as to whether the market was able to signal impending difficulties at NR. Section 4 sets out the research methodology used in this paper and explains the contributions made to the existing literature. Section 5 presents the empirical results and the final section provides a summary of the main academic and regulatory implications of the findings.

2. Supervisory Use of Market Information

2.1. Market discipline and market information

Supervisory authority interest in the use of market information as a market discipline mechanism is becoming increasingly evident. For instance, bank supervisors in the US have used market information ‘informally’ for supervisory purposes since the end of the 1990s. The market indicators frequently used are: debt ratings; stock prices; SND spreads; expected default frequencies (EDFs) extracted from equity values; market capitalisation; asset volatility; and analysts’ opinions (see Furlong and Williams, 2006). Recently, the Swedish Central Bank has begun to harness market information to complement its conventional analysis of banking sector soundness, which is mainly based on financial statements. They favour equity-based indicators, such as EDFs and Distance-to-Default (DD),¹ because of the stock market's liquidity (Persson and Blåvarg, 2003). Finally, the ECB (ECB, 2004, 2005) embraces a macro-prudential approach and constructs aggregate measures of the DD that reflect the risk of the banking system as a whole. However, although banking authorities in several countries are currently using market information to complement traditional supervisory tools, they seem reluctant to trigger specified corrective action based exclusively on financial market signals.

The appeal of market-based information over accounting and supervisory information in the banking supervisory context is threefold. First, market data represents the aggregated opinions of a large number of market participants. Second, the data is forward-looking in contrast to the retrospective nature of accounting data. Finally, it has high frequency and is publicly available in a timely manner. Conversely, for some market variables there is no simple relationship between changes in those variables and the probabilities of bank default leading to potentially misleading signals. For example, an observed increase in stock prices could be caused by higher profits or by an increase in the volatility of asset returns. Accordingly, prices based on debt instruments, whose holders’ incentives are more aligned with those of the supervisory authorities, could be more informative to the supervisory authorities than equity prices. Appropriate market signalling is also dependent on deep and liquid markets and standardised securities. Moreover, market prices may contain factors other than credit risk, which can cloud ‘signal-to-noise’ qualities.

A review of the empirical literature on market discipline and market signals highlights two important research questions:

• 1
  ‘Does market information accurately reflect contemporaneous information about a credit institution's condition?’
• 2
  ‘Does the market incorporate this information in a timely manner so as to add information to supervisory assessments?’ And, attached to this, ‘Can market participants predict individual bank or system-wide fragility?’

Hamalainen (2007) conducts an extensive literature review of the existing market discipline studies on the first research question. Generally, the findings are that equity prices/returns and SND yield spreads do accurately reflect banks’ risks as measured through balance sheet or other market indicators.

The second question, one of timeliness, is more pertinent to this paper. Flannery (1998) neatly summarises the empirical findings: ‘market assessments have at least a plausible chance of providing timely, accurate information that supplements the supervisory agencies’ traditional ways of gathering and assessing bank quality’. Subsequent research has supported this conclusion (Berger et al., 1998; Evanoff and Wall, 2001b; Jagtiani and Lemieux, 2001).

A small literature has examined which market instruments may provide better predictive qualities of bank failure. Three studies have compared the predictive qualities of SND yield spreads and equity-based market indicators, such as DD (Persson and Blåvarg, 2003; Krainer and Lopez, 2004; Gropp et al., 2006). All found that equity market indicators provide more value far from default, whereas yield spreads have a tendency to react close to default. Another study, by Swidler and Wilcox (2002), compares Implied Volatilities (IVs) from exchange-traded options with share prices and sub-debt yield spreads. Their findings suggest that there exist important co-movements between the three market indicators. However, the IV estimates diverge at different times from the paths followed by stock prices and credit spreads. Evanoff and Wall (2001b) identify that, in establishing an early-warning system, the desire is to minimise the misclassification of problem banks as non-problem banks (type-I error). Given that the supervisory mission is to identify a relatively few ‘bad’ banks amongst a large sample of mostly ‘good’ banks, the classification problem regulators might face in practice is highly relevant. Gropp et al. (2006) show that the different properties of bond and equity-based indicators help to reduce the risk of type-I errors. Equally, the findings of Swidler and Wilcox (2002), and the statistical analysis of the spread-risk relationship conducted by Bliss (2001), lend support to the idea of combining various risk indicators to improve the accuracy of bank distress forecasts.

2.2. The FSA's use of market information

Having established the potential importance of market signals, it is also important to recognise the way in which supervisory authorities utilise this information. The Financial Services Authority (FSA) (2008) has specifically addressed the question of market information and how this was used within the FSA. The report found significant weaknesses in the flow of information between different parts of the FSA and also concluded that none of the teams reviewed realised that their role was to include identifying outlier firms from peer analysis. This is not to say that the teams did not receive any information that would enable them to conduct such analyses. The banking sector team and relevant individual supervisory teams did receive information on banks’ results and share price movements. However, the report goes on to state that the information: ‘does not appear to have been well used by supervisors and …  was considered a low priority’. Beyond simple share price movement analysis the only other external market data that appears to have been used were credit ratings and analysts’ reports. Once again, however, the report identifies confusion about responsibilities for obtaining and analysing external information. In particular, there was: ‘no clear mechanism for ensuring that a supervisor received all relevant credit ratings and analysts’ reports for their firms’. The report concludes that market data does provide considerable opportunities to enhance supervisory analysis. However, there is also the question as to whether the supervisory authorities should be widening their information scope to incorporate simple market data from other financial instruments, such as SND and CDS and, where possible, performing more complicated analyses on existing data sources. This paper will accordingly provide some analysis based on the NR context and offer some suggestions.

3. Northern Rock: the Events Leading up to the Run on the Bank's Deposits

NR was granted emergency funding facilities from the Bank of England on Friday 14 September 2007. The salient timescale events preceding, and during, this crisis are detailed in Hall (2007). The key dates for the purposes of this study are now discussed in order to provide a context for the subsequent analysis.

NR opened 2007 by reporting record profitability in the previous financial year (NR, 2007a). The markets reacted positively to the news with the share price rising 64p to close at 1212p; only 12p off the record high achieved 3 weeks earlier.² In response, NR increased its strategic target for return on equity to 20–25% for 2007. The strong performance continued into 2007 with NR issuing a positive quarterly trading statement on 2 April 2007 (NR, 2007b). NR was on course for an 18% increase in full-year profits, driven by sustained lending growth and benign economic conditions. The bank's trading statement had identified, however, that due to an unexpected rise in the central bank's policy rate (Bank Rate) in January 2007 the resultant Bank Rate-Libor gap remained noticeably higher than in 2006, but the bank did not believe that its profit growth forecast required adjustment.³ Nevertheless, NRs reliance on securitisation to fund its rapid growth, and how this contrasted with the other two mortgage banks’ strategies, was not lost on the market (Hughes, 2007).

NR issued a Pre-close period statement to the London Stock Exchange (NR, 2007c) on 27 June 2007, which stated that it expected full-year profits to fall short of analysts’ forecasts. This profits warning appeared to act as the impetus for the market to identify and report the weaknesses in NR's business model; in particular, the bank's significant reliance on wholesale funding⁴ to support its rapid growth and the mismatch between its funding sources, based on Libor rates, and its revenue sources, based on the Bank Rate. Accordingly, these factors became the focus in subsequent reporting on NRs performance (Croft and Tett, 2007).

On 25 July 2007 NR released its Interim results for the six months to 30 June 2007, and the Chief Executive reiterated that profits would be adversely affected by interest rate movements (NR, 2007d). Investors’ concerns over NR intensified as a result of the US sub-prime turmoil in late July and early August 2007.⁵ The announcement by Bear Stearns that it needed to refinance two large hedge funds exposed to the sub-prime market led to the seizure of credit markets and central banks pumping unprecedented amounts of liquidity into the financial system from 9 August onwards. The sharp rise in Libor rates in early August raised concerns that higher funding costs would squeeze NR's margins and limit growth (Thal Larsen, 2007a), forcing the bank to make a further profits warning (Hill, 2007). On top of this, the closure of credit markets prevented NR from distributing securitised assets and accessing its all-important wholesale funding sources (Thal Larsen, 2007b).⁶ Concerns over NRs access to reasonably-priced funds continued in the market (Hume, 2007) until the Tripartite announcement on the morning of Friday 14 September 2007. This stated that the Bank of England was to provide NR with access to emergency funding at penal rates of interest. Simultaneously, NR issued a Stock Exchange statement which emphasised the impact of the funding mismatch between Sterling Libor and Bank Rates on its future profitability (NR, 2007e). This news initiated a run on the bank's deposits, which only subsided when HM Treasury announced on Monday 17 September 2007 an explicit full guarantee of NRs existing deposits. Following the Tripartite announcement, NR quickly amassed large borrowings from the Bank of England.⁷ Ultimately, on 17 February 2008, given the scale of NRs reliance on central bank funds, the Government nationalised NR.

4. Research Methodology and Data

The paper contributes to the literature on using market information as a predictor of bank risk by answering the following research questions:

• • 
  Did market investors signal increased risk at NR in advance of the bank requiring emergency central bank assistance?; and
• • 
  What can we learn from the NR context concerning the use of market signals in banking supervision and the signalling qualities of different financial instruments?

The study furthers the literature on market signalling in six ways. First, most existing studies have applied proxies for bank failure (Gropp et al. 2006; Krainer and Lopez, 2004). In contrast, NR represents a clear case of individual bank failure. Second, all existing studies have covered substantial sample time periods where there was more than one instance of ‘bank failure’. As a consequence, their analyses are partly clouded by the ‘failure’ of a number of institutions in their sample. Third, financial innovations in the form of the creation of new market instruments, and the rapid rise in UK credit institution SND issuance in the past 15 years (Hamalainen et al., 2010) and the consequential deepening of the market, provide an opportunity to examine the market signalling qualities of a number of financial instruments. Fourth, this paper introduces a further potential metric in signalling individual bank risk-taking through extracting implied idiosyncratic volatility from banks’ traded option contract prices. The fifth contribution is that, the single country study of this paper provides a cleaner analytical context in which to examine the issues at stake. This is because studies that cover numerous European countries typically have to accommodate the influence of different government responses to bank fragility. The UK context of this research paper is particularly pertinent in this respect, because there has been an implicit belief that UK banks do not benefit from ‘too-big-to-fail’ (TBTF) type conjectural guarantees (Sironi, 2003).⁸ Finally, almost all papers in this area use data from ‘benign’ or tranquil sample periods to discriminate between ‘bad’ and ‘good’ banks. One of the contributions of this paper, therefore, is the analysis of the informational content of bank security prices in an unprecedented macroeconomic environment.

The paper analyses the market signals for NR compared to a set of eight peer UK banks in order to assess whether market investors signalled increased risk at NR in advance of requiring emergency assistance.⁹ The selected banks, including NR, are the nine largest in the UK and represent over 80% of the UK banking system's assets. The signalling qualities of different financial instruments are examined in the process. The data comprises CDS spreads, SND spreads, implied volatilities from options prices and equity measures of bank risk from the beginning of 2006 (or 2007) to the announcement of NRs access to emergency liquidity facilities on 14 September 2007. The CDS data is sourced from Credit Market Analysis, the options prices from Reuters and Datastream Thomson Financial, and the equity and SND data from Datastream Thomson Financial.

In addressing the first research question, the small sample size prevented the use of a standard ‘early-warning’ model in the same vein as Gropp et al. (2006). The research question was, therefore, answered in a graphical and descriptive statistical manner, similar to Persson and Blåvarg (2003). Graphical representations of the different market signals are presented to identify trends, with key market dates mapped onto the graphs to support the analysis. Adopting descriptive statistics prevents statistical significance testing of market instruments’ predictive qualities. However, in light of the study's unique context and the clear under-utilisation of market information by the supervisory authorities, the study contributes to the two posited research questions. The study's second research question, the signalling qualities of different financial instruments is answered in the process through comparing and contrasting the outcomes.

5. Empirical Findings

This section presents the research findings and is structured by market instrument type. The section also brings together the key themes to compare the signalling qualities of each financial instrument and their potential supervisory application.

5.1. Equities

Each of the bank's equities is listed in the FTSE100 and, therefore, has a liquid market. In line with the existing literature the equity indicators examined are: stock prices/returns, trading volumes and distance to default.

Equity prices. The appeal of using equity prices as a signalling mechanism is that the data is readily available, but there is no obvious link between equity prices and default risk (Persson and Blåvarg, 2003). In the NR context, the share price had been consistently falling since its record high in February 2007 and it was identified in the House of Commons Treasury Committee's (2008) report as a potential warning signal. The deterioration in NR's share price becomes apparent after the profits warning of June 2007, especially when it is compared with peer mortgage banks – see Figure 1. This may well be because the profits warning highlighted, for the first time, the fundamental shortcomings of NRs business model to a wider audience.

Figure 1. UK mortgage banks’ equity prices, rebased value (% of 01/01/07 value).

The deterioration in NR's share price is all the more startling when UK banks’ share prices are rebased back to January 2006 and compared with all sample banks. To enable comparisons with all of the other sample banks, we created a value-weighted portfolio of equity returns from the other eight sample banks and calculated the cumulative weekly return difference between January 2006 and 13 September 2007 for each mortgage bank and its respective portfolio of other banks. NR was the best-performing UK bank share during 2006 (see Figure 2).¹⁰ As with the equity price signal graph, NR's equity returns began to fall from February 2007 onwards with the profits warning signalling a clear deterioration compared to the sample banks.

Figure 2. UK mortgage banks’ cumulative weekly return difference (%) with portfolio of other UK banks.

Equity trading volumes. Like equity prices, equity trading volumes have no direct link with default risk. At best, analysis of trading volume data can simply indicate to the supervisory authorities that potential issues may have been identified by equity market participants that require further investigation.¹¹ In the NR case, simple peer analysis with other mortgage banks¹² illustrates consistently higher trading volumes in NR's shares after the profits warning in June 2007 (see Figure 3). The higher volumes may simply represent investors searching for potential value stocks, but the jump to average trading volumes above 10,000 does warrant further investigation. The analysis of trading volume, taken separately, does not allow one to draw meaningful statistical inference with respect to the direction of the pressure exerted on stock prices by the abnormal trading activity observed after the profits warning release. However, the corroborating evidence on the time evolution of stock prices and returns enables us to refine our interpretation. Specifically, the abnormal trading activity observed after June 2007 was clearly induced by an excess of sell orders resulting in downward pressure on prices. The high spike in trading in early August 2007 suggests that the market had by then understood the implications of the seizure in wholesale credit markets on NRs business model. In a similar vein, the rise in Bradford & Bingley's (B&B) trading volumes from June 2007 might have alerted the supervisory authorities to potential issues that again merited further investigation.

Figure 3. UK mortgage banks’ equity trading volume (10 day MA).

Distance to default (DD) An alternative approach to extract information from equities is based on option-pricing theory and involves treating equities as call options on the company. This enables investors’ implicit views of default risk to be determined in the form of the DD measure. This calculates the number of standard deviations a banking firm is away from its default point (i.e. the asset value at which the firm will default or have zero market net worth). First, we estimate the asset value and asset volatility for each bank from the market value and volatility of equity and the book value of liabilities by solving numerically (using the Newton-Raphson method) a non-linear system of two equations: (1) the value of equity, modelled as a call on the bank's assets using a standard options pricing-based framework; and (2) the instantaneous relationship between assets and equity volatility derived from Ito's lemma. Then, we compute DD using the KMV framework described in Crosbie and Bohn (2003). The final step in our derivation of a sophisticated equity-based indicator would be to convert the DD for each bank in our sample to a probability of default or a term structure of default probabilities. Unfortunately, as we do not have access to a historical default database, we scaled the DD to a probability using Merton's model. As a consequence, the default probabilities are undervalued, essentially zero for all banks, except NR, after August 2007. This is because the empirical distribution of default rates exhibits wider tails than the normal distribution under the original Merton model.

The DD results are presented in Figure 4. In line with our previous equity findings, equity investors’ perceptions of risk at NR deteriorated very rapidly after the profits warning in June 2007 compared to the sample banks. Subsequently, although all sample banks were reporting declining DD indicators, NR was perceived as the most likely to default, with the smaller mortgage banks as the next most likely.

Figure 4. Time evolution of the Distance-to-Default (DD) indicator for UK banks.

5.2. Time-varying estimates of Idiosyncratic Implied Volatilities from banks’ exchange-traded option contracts

An alternative to historical volatility (HV) measures of bank share prices as a basis for supervisory action is provided by implied volatilities (IVs) from exchange-traded options on bank share prices. The appeal of implied rather than historical volatility estimates is threefold. First, IV is inherently a forward-looking forecast of the future volatility of the underlying asset as opposed to a backward-looking measure. This implies that IVs may contain reliable information about stress events not captured in time series of past returns. Second, recent findings from the options literature indicate that IV levels and changes contain useful incremental information about future return volatility (Dennis et al., 2006; Diavatopoulos et al. 2008). In the banking context, Swidler and Wilcox (2002) show that IV estimates have lower error forecasts of future volatility than HVs and significantly improve forecasts. Third, the market for option contracts on bank equity is deep and sufficiently liquid to provide reliable signals to supervisors. All of the banks included in our sample (except B&B) have traded option contracts.

Detailed information on bank equity options contracts showed, as expected, that the option market is deeper and more liquid for the largest UK banks. Nevertheless, the contracts on the two mortgage lenders’ were actively traded during the analysed period with option open interest reasonably high.¹³ Not surprisingly, the highest Put/Call open interest ratios were observed for NR (7.28:1 on average), indicating that investors were starting to conjecture that the stock price would go down.

The paper utilises a ‘standardised’ measure of IV using a methodology that is similar to Swidler and Wilcox (2002). From a banking supervisor's perspective, it is useful to distinguish between fluctuations in IV time-varying estimates due to changes in bank risk profile and fluctuations induced by the changes in the volatility of the market. To decompose the total implied volatility into systematic and idiosyncratic components it was necessary to construct a barometer of market volatility in the same manner as individual banks’ IVs, i.e. by taking IVs from Calls and Puts traded on the FTSE100 index that are near the money. Following previous studies on IV dynamics (Diavatopoulos et al. 2008), we estimate the individual bank stock's implied idiosyncratic volatility (IIV)¹⁴ using a variance decomposition based on the discrete-time version of the Ornstein-Uhlenbeck stochastic process.

Table 1 describes the statistical distribution of the measures of volatility implied by the market prices of the options contracts on UK listed banks’ stocks (IV and IIV) after the release of the profits warning by NR and before the central bank granted financial support. The highest values for both measures are observed for the two mortgage lenders, Alliance & Leicester (A&L) and NR. It is worth noting that the IIV represents a substantial portion (almost 70%, on average) of the total implied variance. This is an important result because, from a micro-prudential perspective, bank supervisors should be more concerned about changes in bank-specific risks than changes in volatility of the whole market.

Table 1.   Implied Volatility (IV) & Implied Idiosyncratic Volatility (IIV) Estimates This table reports the statistical distribution (mean, standard deviation, minimum, and maximum over the period 26 June – 14 September) of our two measures of volatility implied by the market prices of the options contracts on UK listed banks’ stocks: Total Implied Volatility (IV) and Implied Idiosyncratic Volatility (IIV). We infer a ‘standardised’ measure of IV based on the nearest two ‘at-the-money’ options series – one above and one below the underlying price – using values from the nearest expiry month options (the options series switches to the next available month on the first day of the expiry month). We next interpolate between the two IVs to calculate an estimate of the IV for a hypothetically ‘at-the-money’ Call/Put option. Finally, we average the IVs of the two (Call and Put) option contracts to obtain the IV for options with the strike price nearest to the underlying bank stock price. As all option contracts on UK listed banks’ stocks traded on LIFFE are American-type options, the IVs are computed using the Cox-Ross-Rubinstein binomial tree model to take into account the possibility of early exercise. To compute the idiosyncratic component of implied volatility (IIV), we use a decomposition of the total implied variance based on the discrete time version of the Ornstein-Uhlenbeck stochastic process, which takes into account the mean-reversion behavior of IV. As a benchmark for the option market volatility and sentiment, we use the implied volatility on the FTSE 100 index options, calculated in the same manner as the individual banks’ IVs, from both Calls and Puts that are near the money. Source: authors’ computations based on data extracted from Reuters 3000 Xtra and Datastream Thomson Financial.

Bank	Implied Volatility (IV) 26/06–14/09					Implied Idiosyncratic Volatility (IIV) 26/06–14/09
	Mean	Median	Std. dev.	Min	Max	Mean	Median	Std. dev.	Min	Max
Alliance & Leicester	34.47%	34.57%	3.58%	23.07%	41.77%	27.02%	26.67%	2.84%	16.51%	34.86%
Barclays	32.71%	32.43%	5.09%	24.28%	42.79%	19.80%	20.53%	3.85%	12.19%	28.09%
HBOS	30.12%	27.39%	8.25%	21.74%	47.45%	21.19%	16.77%	7.85%	10.11%	39.47%
HSBC	19.93%	20.36%	3.54%	13.98%	29.30%	14.72%	14.43%	2.78%	9.08%	22.55%
Lloyds	25.86%	25.31%	2.57%	20.74%	32.46%	18.89%	19.07%	2.86%	10.93%	23.84%
Northern Rock	39.33%	37.49%	11.80%	26.69%	70.77%	28.09%	25.06%	10.41%	8.79%	63.64%
Royal Bank of Scotland	29.77%	29.86%	5.04%	22.14%	38.60%	21.47%	20.62%	2.65%	16.99%	29.54%
Standard Chartered	28.96%	29.12%	3.40%	24.93%	36.95%	13.72%	15.06%	4.30%	4.00%	19.48%

In order to visualise the way in which the options market reacted to the deterioration of bank financial conditions prior to the NR crisis, Figure 5 plots the time evolution of our IIV indicator for our sample of banks between 1 January and 31 December 2007. As was the case with the equity indicators, the two mortgage lenders included in our option sample were perceived as more risky compared to the larger UK banks. Equally, the options market flagged bank-specific signs of vulnerability only after NR issued its profits warning in June 2007. Moreover, despite the emergency loan granted in September 2007, concerns about NRs insolvency persisted after that date.

Figure 5. Time evolution of Implied Idiosyncratic Volatilities (IIV) of UK banks.

5.3. Subordinated debt

The market discipline literature has frequently espoused the potential benefits to the regulatory authorities of SND because, in contrast to more senior debt holders, the incentive of SND holders to monitor and limit bank risk-taking is more aligned with that of the supervisory authorities (and hence taxpayers) (Benink and Wihlborg, 2002). This implies that they would exert a greater restraint on bank management and, as such, are a suitable instrument of market discipline (Hamalainen et al. 2003). UK banks have been issuing SND in record levels since 1999 and, in value terms, are the second largest issuers of publicly-issued debt after US banks (Hamalainen et al., 2010). Subsequently, the market has become extremely liquid and, therefore, offers risk-signalling potential. In this study, we focus on sterling-issued SND because this is the predominant currency in which mortgage banks issue, and because existing studies have shown that the currency of denomination does influence SND spreads (Hamalainen et al., 2007).

Figure 6 presents the findings for the mortgage banks for bonds with ten years and less to maturity. This clearly shows that NR spreads had been declining until the profits warning in June 2007. Subsequently, the spreads began to rise, but it is only with the start of the credit crisis in early August 2007 that NRs spreads widened significantly to potentially signal investor concerns. By then, the regulatory authorities were well aware of NRs potential funding difficulties.

Figure 6. UK mortgage banks’ SND spreads (bps), 10y and less.

For comparative purposes, Figure 7 shows the SND spreads for a selection of larger UK banks for bonds with less than 10 years to maturity. Prior to the turbulence in credit markets the spreads were generally lower than for mortgage banks. However, the onset of the credit crisis brought mortgage banks’ and non-mortgage banks’ SND spreads closer together. Therefore, it is difficult to deduce any bank-specific risk signals.¹⁵

Figure 7. Selected large UK banks’ SND spreads (bps), 10y and less.

5.4. Credit default swaps

Growth in CDS trading in the past ten years provides an opportunity to explore the risk-signalling qualities of this financial instrument. The characteristics of this credit derivative are such that the spread should provide a clear measure of default risk in the underlying company's debt. CDS prices are available for Senior and SND debt; however, in this study, only senior debt spreads are used. This is because one of NRs peer mortgage banks, B&B, does not have tradable CDS based on SND. The previous SND analysis showed that B&Bs SND spreads widened in late August 2007 and, therefore, for comparative purposes it is important to include this bank.

Figure 8 presents the results for mortgage banks on one-year maturity CDS contracts. On an absolute basis, both NR and B&B are viewed as more risky than A&L. In the summer of 2007, the market for mortgage bank CDSs dried up, until mid-August, when NR registered significant widening spreads. This is undoubtedly reflecting the funding fears in the wholesale markets at the time. Notably, the same concerns are not applied to B&B and A&L. This, in turn, suggests that investors were specifically becoming concerned with NRs creditworthiness. Analysing the same data on a relative basis, however, clouds the picture. All of the non-mortgage banks’ spreads were also widening significantly in late July 2007, albeit from very small bases, and, therefore, the considerable relative rise in NRs spreads was matched by similar relative rises in the larger UK banks’ spreads. In this case it would be hard to argue that bank-specific risk factors are being signalled by the CDS market.¹⁶

Figure 8. UK mortgage banks’ CDS spreads (bps), 1y Senior.

5.5. Comparative signalling analysis of the four financial instruments and its implications for the supervisory authorities

Of the four financial instruments analysed in this study, equities appear to provide the earliest and clearest signs of potential concerns about NR prior to the emergency funding announcement in September 2007. There were bank-specific falls in prices and returns, trading volume registered a markedly upward shift, and the DD indicator rapidly deteriorated. However, the bank-specific signs only became apparent after NR issued a profits warning in late June 2007. It is at this stage that the bank explained how funding mismatches would prevent analysts’ profit forecasts from being met. This disclosure appears to have highlighted to the market the weakness in NRs business model and, therefore, subsequent increases in the Libor-Bank Rate relationship cemented further falls in NRs share price.

The second instrument of bank financial distress analysed in this study is the idiosyncratic component of IV from banks’ exchange-traded option contracts. This forward-looking market indicator has generally been neglected in the market discipline literature. As with equity indicators, the mortgage lenders appear to be more risky compared to the other UK banks a couple of months before the Bank of England decided to take action.¹⁷

SND spreads and CDS spreads were considerably slower in indicating concerns with NR. With hindsight, both instruments reported significant jumps in NRs spreads following the seizure in wholesale credit markets in August 2007 and further marked rises in early September 2007. In that sense these markets did react prior to the bank requesting funding assistance from the Bank of England. However, these signals were extremely late and also very difficult to decipher for bank-specific risk elements.

In the case of SND spreads, it was difficult to extract clean, bank-specific signals, because finding bonds of comparable maturity issued by more than one mortgage bank was not possible and so term structure components were also reflected in spreads. Therefore, despite the growth of the UK bank-issued SND market over the past ten years, finding comparable SND instruments is difficult for risk-signalling purposes. This problem explains why proponents of SND market discipline propose standardised debt structures.

In the case of CDS spreads, the findings illustrate that the CDS market is still not sufficiently deep to extract clean signals. There was no trading in mortgage banks’ CDS for some time during the summer of 2007, despite volatility in the credit markets. Therefore, liquidity risk is an issue for CDS spread signals. In addition, it is difficult to decipher bank-specific risk-signalling. For example, in absolute terms, NR was consistently perceived by the CDS market as one of the riskier UK banks. However, the onset of the liquidity crunch in August led the larger UK banks’ CDS spreads to rise dramatically and, therefore, their relative spreads increased considerably compared to NR. Under these circumstances, should NR, or the larger banks, have received attention from the supervisory authorities?

Apart from comparing the signalling quality of the four financial instruments, the findings also contribute to the academic debate on the timeliness of market information compared to supervisory authority actions. The supervisory authorities became concerned with NR when the wholesale credit markets dried up on 9 August 2007, prompting regular communication with the bank from then on. Incorporating these developments with the market signalling results illustrates that the equity market did provide some evidence of investor concern in advance of the regulatory authorities’ action. In contrast, however, the SND and CDS markets did not signal timely concerns, due to the structural limitations discussed above, which hamper their signalling capabilities. This paper has found evidence to suggest that the regulatory authorities could make more use of market information to improve their monitoring of bank risk, but that efforts should initially concentrate on equity market signals. The results also indicate that the regulatory authorities could adopt a more rigorous and sophisticated equity analysis than is currently undertaken in order to monitor market sentiment. There were a number of relatively straightforward equity indicators that should have alerted the supervisory authorities to, at least, undertake further investigation into NR. However, this does not necessarily mean that earlier intervention could have prevented NRs rapid deterioration in liquidity. The profits warning was not released until late in June 2007 and the unprecedented collapse in wholesale credit markets started in early August 2007. Within a month, NR was approaching the Bank of England for emergency funds.

6. Summary and Conclusions

The findings support the existing literature on the predictive qualities of market information and indicate that equity market indicators provide more value far from default, whereas SND and CDS have a tendency to react close to default. The time-varying estimates of IIVs extracted from option prices clearly deserve further attention because the informational content of this forward-looking indicator appears to be potentially useful to bank supervisors. From a micro-prudential perspective, the most encouraging result is that the idiosyncratic component represents a substantial portion (more than 70%) of total IV for a banking firm.

The paper's findings also support the existing market discipline literature in that it can prove difficult to extract clean risk signals from SND spreads and, thus, careful judgement must be exercised when interpreting them (Hancock and Kwast, 2001). In the case of CDS spreads, the paper has shown that the market is not sufficiently-deep in small bank CDS for supervisors to be able to currently rely on these market signals.

From a supervisory perspective, the paper has shown that equity market information may have provided the authorities with some predictive signals of impending default at NR, whilst SND and CDS spreads only possibly signalled concerns once the FSA was already communicating with the bank. Furthermore, the paper highlights the potential benefits to the regulatory authorities in observing a number of early warning metrics concurrently. In so doing, the paper introduces a further metric, IIV.

Footnotes

• 1

  DD combines stock price information with stock volatility and leverage, and measures the number of standard deviations away from default, where default is defined as the point at which assets are just equal to liabilities. This property makes it a useful indicator from a supervisory perspective.

• 2

  Within a month, HSBC was reporting spectacular losses in its US mortgage business due to sub-prime lending. This announcement is generally viewed as the first major sub-prime loss to be reported.

• 3

  Bank Rate is the interest rate that the Bank of England charges banks for secured overnight lending and is the UK central bank's key interest rate for enacting monetary policy.

• 4

  Llewellyn (2008) reports that only 30% of NR's funding came from stable retail deposits compared to 44% for A&L, 48% for B&B and 50% for HBOS. Equally, the Bankscope ratio Net Loans/Customer funding was 299% for NR, compared to figures of between 126% and 155% for the other three banks.

• 5

  See Gorton (2009) for an in-depth analysis of the sub-prime related debt financial structures (such as CDOs and CDO²s) that are ascribed as being the trigger for the liquidity crisis in the summer of 2007.

• 6

  Subsequently, the House of Commons Treasury Select Committee's investigation into the NR crisis has shown that the seizure in funding markets on 9 August 2007 prompted the FSA to contact NR, because it perceived the bank to be at risk from the freezing of financial markets. Thereafter, the FSA and NR were in daily telephone contact (House of Commons Treasury Committee, 2008).

• 7

  By the end of 2007, this amounted to £27 billion in emergency lending to replace £28 billion in retail and wholesale funding that had deserted the bank.

• 8

  The UK government's ‘rescue’ of NR depositors on 17 September 2007 does suggest a possible policy shift, but the paper focuses on the period leading up to the failure of NR. Furthermore, the government's NR guarantee arrangements explicitly state that SND instruments are excluded.

• 9

  The eight banks comprise: HSBC; Barclays; RBS; HBOS; Lloyds; Standard Chartered; Bradford & Bingley; and Alliance & Leicester. The last two banks, along with NR, are informally distinguished from the other six through the term ‘mortgage banks’ because they were originally mutual building societies and most of their business still revolves around housing loans. These three banks are also the smallest ones in the sample.

• 10

  The figure illustrates that an investor buying a NR share in January 2006 would have realised by the end of 2006 a superior cumulative return of 7% compared to buying a portfolio of equities of the peer banks.

• 11

  For example, we thank a banking analyst for pointing out that increasing trading volumes in a firm's shares can be associated with a sign of increasing uncertainty in the firm's direction.

• 12

  Comparisons with the trading volumes of the non-mortgage banks were not made because the latters’ trading volumes are considerably higher.

• 13

  This is important as IV estimates tend to be noisy for thinly-traded option contracts (Dennis et al., 2006).

• 14

  To our knowledge, this is the first paper that proposes the use of idiosyncratic implied volatilities as a bank risk metric for supervisory purposes.

• 15

  We also examined SND with longer maturities for both mortgage banks and larger UK banks. In both cases, yield spread patterns follow similar paths to their shorter-dated counterparts.

• 16

  We also analysed senior tier, five-year maturity CDS contracts for mortgage and non-mortgage banks and the same points made in the one-year CDS discussion, were confirmed.

• 17

  Ultimately, the nationalisation of NR in February 2008 did not represent the final problem to be experienced by the UK's mortgage banks. In July 2008, following a crash in profits due to excessive exposure to the UK housing market, and heavy reliance on wholesale funding markets, Alliance & Leicester was wholly taken over by Banco Santander. Equally, during the midst of the banking crisis in September 2008, Bradford & Bingley lost its independence due to a crash in profits resulting from excessive exposure to riskier elements in the UK property market. The bank's loan book was nationalised and its deposits and branch network were sold to Banco Santander.