Stress testing procedures on the Australian banking sector : the impact of announcement & disclosure events

(1)

Stress Testing Procedures on the Australian Banking Sector:

The Impact of Announcement & Disclosure Events

Amsterdam Business School

Name Norman Ross Caird

Student number 11084731

Program MSc Business Economics

Specialization Finance Number of ECTS 15 ECTS Points

Supervisor Dr Tanju Yorulmazer

Target completion 7th July 2016

Abstract

This study examines the impact of stress testing events on the Australian banking sector. That is, it analyses the impact to share price returns due to announcement and disclosure of stress testing programs carried out by regulatory supervisors, rating agencies and international banking authorities on the domestic Australian banking sector. In particular, 11 stress testing programs performed over the 2006 to 2014 timeframe are examined using OLS, GARCH and EGARCH methodologies. The analysis examines the impact to Australian bank’s share price returns in an attempt to ascertain if a relationship exists between the announcement and disclosure of stress testing regimes and whether there are identifiable impacts on the returns of banking stocks. The findings imply that investors reacted more profoundly to certain episodes, irrespective of the party carrying out the stress testing procedures.

(2)

2

Table of Contents: Continued

5. Data Collection and Stress Testing Events………...24

5.1. Australian Banking Data Collection……….24

5.2. Stress Testing Events & Procedures in the Aust. Banking Sector…...27

6. Empirical Results………...37

6.1. Results: Ordinary Least Squares (OLS) Regression………37

6.2. Results: GARCH Regression………...37

7. Discussion and Analysis……….42

7.1. Ordinary Least Squares (OLS) Regression………..42

7.2. GARCH Regression………...44

8. Robustness Checks……….45

9. Conclusion………...47

10. References………...49

(4)

4

Statement of Originality

I hereby declare that the submission of this thesis is my own work and that all material (to the best of my knowledge) contained within this document encloses no previously published material by another person (except where referenced in text).

Signature:

Norman Ross Caird 11084731

Acknowledgements

I would like to thank my thesis supervisor Dr. Tanju Yorulmazer for his support, comments and feedback over the three month thesis period.

(5)

5

1. Introduction

This section includes an overview of the thesis, describes the purpose of the research, lists the research hypotheses and details the general outline of the thesis document.

1.1 Overview

Since the 2008 financial crisis, many banking sector regulatory authorities have instigated regular stress testing regimes with the aim to restore confidence, safeguard financial stability and highlight any potential concerns within their financial sectors. The introduction of stress testing as part of the Supervisory Capital Assessment Program (SCAP) in the United States in February 2010, paved the way for macroeconomic stress testing to become a regular feature utilized by many regulatory authorities around the world. In some situations, regulatory authorities exploit stress testing procedures to provide market participants with valuable information that is otherwise unlikely to be disclosed given the opaque nature of their banking systems.

This study aims to measure the effects on Australian banking institutions due to the announcement and disclosure of stress testing practices. Past studies examining the effects on market confidence in US and European settings have revealed conflicting results. However, the majority of past publications seem to suggest that stress testing exercises led to an improvement in market confidence. Stress test design, the degree of disclosure surrounding testing results, as well as macroeconomic and political factors have all been linked to display significant impacts on the actual outcome of a stress test (Ong and Pazarbasioglu, 2013).

The Efficient Market Hypothesis (Fama, 1969) states that asset prices should fully reflect all information that is immediately available to the market participant. If this statement holds, than markets are efficient and only the arrival of new information can cause changes in current price levels. Hence, the disclosure and announcement of stress testing regimes could convey new information that may affect stock pricing and returns for those institutions subjected to stress testing procedures. My proposed study looks to investigate this proposition further, by examining the impact on banking stock returns given announcement and disclosure of stress testing practices; and if impacts are detected, a comparison between various stress testing regimes carried out by differing regulatory bodies and industry insiders.

(6)

6 1.2 Definitions

In the appendices, a list of abbreviations and their corresponding definitions have been provided to allow for greater insight and clarity surrounding the terminology included in this document.

1.3 Purpose of Research

The purpose of this research is to ascertain whether any detectable influences associated with stress testing regimes performed on the Australian banking sector are observable amongst the returns of corresponding institutions. That is, to investigate whether stress testing exercises performed on the Australian banking sector produced new information about the financial adequacy of each institution, and as a result the impact market participants impose on institutional share pricing and returns. To provide both robust and accurate results, abnormal returns will be calculated using the following methodologies- Ordinary Least Squares (OLS), Generalised Autoregressive Conditional Heteroskedasticity (GARCH) and Exponential Generalised Autoregressive Conditional Heteroskedasticity (EGARCH) models. The latter methods allowing for conditional variances to be modelled instead of constant variance as assumed under basic OLS methodology.

This study utilizes methodologies (GARCH and EGARCH) that have not been employed in prior studies examining impacts on financial stocks associated with stress testing regimes. For this reason, this particular study is unique. Prior studies (see Related Literature section) have mostly incorporated varying OLS approaches. The secondary approaches integrated in this study help to control for issues associated with heteroskedasticity and volatility clustering. Further explanation is provided in the Methodology section.

Lastly, no prior studies have been performed examining the impact on Australian banking institutions due to the announcement and disclosure of stress testing regimes. Hence, this study will ‘fill a gap’ relating to this topic. Numerous prior studies have examined stress testing programs undertaken on US and European banking and financial stocks, but little literature surrounds testing procedures on smaller banking sectors. Thus, although this study focuses on the Australian banking sector, the general findings and implications may be applicable throughout other stress testing operations performed on smaller banking sectors.

(7)

7 1.4 Research Hypotheses

The following hypotheses will be explored in greater detail in this thesis. This research takes a quantitative approach, testing both hypotheses via statistical inference.

Hypothesis 1:Markets anticipate the announcement of a stress testing procedure and react accordingly – hence, expect detectable changes in the cumulative abnormal returns of banks due to the announcement of a testing regime to be performed on a particular institution or entire banking sector. Hence, the proposition of performing stress testing procedures may cause market participants to formulate a view on which particular institutions will be affected by the testing, or likely capital shortages associated with particular institutions.

Hypothesis 2:The outcome of stress testing procedures and the disclosure of testing results impact bank abnormal returns (markets may or may not react favourably as a consequence of the disclosure of the testing results – dependent on the market’s perception of the bank ). Hence, expect detectable changes to CARs for various institutions as a result of stress testing. 1.5 Disposition

This thesis is an empirical study investigating the relationship between stress testing procedures and their impact on the market reaction of Australian banking stocks. In order to achieve its intended purpose, this paper is organised in the following structure. Section 2 outlines background information concerning stress testing procedures in the financial sector, as well as provides a brief overview of the Australian banking sector. Section 3 encompasses a literature review examining past research findings related to stress testing exercises in the finance sector (mostly US and European studies), as well as further information surrounding stress testing in an Australian setting specifically. Section 4 outlines the methodological approaches undertaken in this particular paper related to event study procedures. That is, the incorporation of OLS, GARCH and EGARCH techniques to examine cumulative abnormal returns for each of the Australian banking institutions. Section 5 encapsulates the data collection processes and further elaborates on the specific stress testing events carried out in the Australian banking sector. Sections 6 and 7 detail the empirical results and summarise the discussion & analysis of the findings. Section 8 presents further robustness checks around the methodological approaches. Lastly, sections 9 and 10 provide conclusions and an extensive list of the relevant reference sources consulted and utilized in formulating this thesis.

(8)

8

2. Background Information

2.1 Overview of Stress Testing in the Financial Sector

From an individual financial institutional perspective, “stress testing” describes a range of techniques that attempt to measure the sensitivity of a portfolio to a set of extreme but plausible shocks (Jones et al., 2004). The objective is to gain a greater understanding of the sensitivity of the portfolio to changes in various risk factors imposing stress on the portfolio. As described by the Committee on the Global Financial System (2000), stress testing estimates the exposure to a specific event, but not the probability of the event occurring. Thus, stress testing can provide information on how much of a given quantity could be lost under a given scenario, but not how much is likely to be lost. Stress tests are simply analytical techniques used to produce a numerical estimate of a particular sensitivity (Jones et al., 2004).

Stress tests were originally developed for use at the portfolio level (micro perspective), to examine and understand the risks posed to a trading book by extreme movements in market prices. They have now become widely-utilised risk management tools by financial institutions since the early 1990s (Cardinali and Nordmark, 2011). In the late 1990s, the International Monetary Fund (IMF) became one of the first institutions to introduce macroprudential stress testing as part of their assessment of the Asian financial crisis. The IMF’s application of macro stress testing led to the realization that instability in a financial sector can spread quickly, prompting a new broader approach of assessment to focus on systemic risks and stability (Borra, 2015). Following the onset of the financial crisis in 2008, stress testing became an essential tool for regulatory authorities around the world to evaluate the risks associated with capital shortfalls, assess the overall stability of financial systems and as an effort to try to provide reassurance to nervous market participants.

The first major stress testing event to be performed following the 2008 financial crisis was the testing regime undertaken as part of the US Supervisory Capital Assessment Program (SCAP) in 2009. Ben Bernanke, the US Federal Reserve chairman at the time stated that “SCAP provided investors with credible information about banks’ financials and that the public disclosure of the test results helped to restore confidence in the banking system” (Bernanke, 2013). This claim has been backed by past empirical studies such as Morgan et al. (2013) and Ong & Pazarbasioglu (2013), which confirm the perception that SCAP provided improved investor sentiment.

(9)

9 Following the success of the US SCAP program, the Committee of European Banking Supervision (CEBS) carried out their own stress testing regime on the European banking sector in order to restore confidence in the soundness and stability of their financial markets. The CEBS European stress tests of 2009 and 2010 differed from the US SCAP in that they tried to incorporate exposure to sovereign debt into their testing procedures (Schuermann, 2012). Furthermore, implementation of the stress testing procedures proved more challenging due to the large number of different countries partaking and varying banking systems involved.

In 2011, the European Banking Authority (EBA) assumed responsibility of stress testing in Europe with the key aim of monitoring and providing crisis management. A greater emphasis was placed on disclosure, as it was identified as a crucial element that limited the effectiveness of previous European stress testing programs (Schuermann, 2012). Since the 2011 test, the EBA in cooperation with the European Systemic Risk Board (ESRB) and the European Central Bank (ECB) have conducted further rounds of stress testing exercises, with the latest cycle of testing announced in February 2016. Back in the US, the Fed Reserve undertakes annual supervisory stress testing regimes incorporated as part of the Dodd-Frank Act Stress Testing (DFAST) and the complementary Comprehensive Capital Analysis and Review (CCAR) programs. Each year, the Federal Reserve undertakes stress testing to assess whether the largest bank holding companies (BHCs) operating in the US have sufficient capital to absorb losses and continue operations during adverse economic conditions.

2.2 An Overview of the Australian Banking Sector: 2006-2016

There is substantial evidence in published papers and financial regulatory authority publications that the Australian banking sector displayed considerable resilience during the recent financial crisis. The Australian banking sector is largely controlled by four dominant institutions (WBC, CBA, NAB and ANZ) that reported combined net profits near the $8 billion mark over the last six months of the 2008 calendar year (Allen and Powell, 2010). In contrast to this figure, the results of other banking sectors in equivalent OECD peer countries (such as the US and UK) suffered crippling losses with approximately $50B wiped out from the combined balance sheets of the 5 largest US banks and reported losses totaling £20B from the 5 largest UK banks over the same period (Reserve Bank of Australia, 2009).

The four dominant Australian banks all reported ratings upgrades in 2007 (from AA- to AA status) due to the “progressive structural strengthening of their financial profiles as well as

(10)

10 continued development of their risk-management capabilities”, with further confirmation of this high rating status delivered in July 2009 (Allen and Powell, 2010). Ratings agency Standard and Poor’s (S&P) reaffirmed the Australian banks high rating status by stating that the strengths of the Australian banking system lie with “an expectation of continued satisfactory earnings, well-controlled credit losses, adequate capitalization, well-managed funding and liquidity as well as a conservative risk appetite” (Standard and Poor’s, 2007). However, in reality, given the interconnectedness of the global banking system the Australian banking sector was not isolated and faced similar exposure to the threats facing other worldwide banking sectors (uncertainty leading to pressures on costs and availability of global funding). During the financial crisis, impaired assets and loss provisions of Australian banks rose sharply (to a lesser degree than global counterparts) putting downward pressure on bank share prices and asset values. The leading banking index (S&P/ASX200 Bank Index) comprising the largest six banks plunged nearly 60% in a 15-month period from Oct 2007 to January 2009 (Allen and Powell, 2010). Furthermore, the Australian banking sector experienced greater consolidation with the merger of Adelaide Bank and Bendigo Bank in 2007, and the acquisition of St George Bank by Westpac Banking Corporation in October 2008.

Regulatory oversight of the Australian banking sector is mainly supervised by both the Reserve Bank of Australia (RBA) and the Australian Prudential Regulatory Authority (APRA). However, in broad terms, APRA has responsibility for the setting of prudential standards and for supervision of institutions. APRA utilizes industry-wide stress testing procedures, horizontal reviews as well as more defined analysis techniques to identify and monitor potential risks requiring prudential action. The use of stress testing extends beyond the prudential requirements for regulated institutions to undertake their own stress testing, with more advanced testing procedures being performed in the recent past to gain greater understanding of banking sector vulnerabilities. As is the case with other financial regulatory bodies around the world, APRA has stepped up and improved its stress testing procedures since the 2008 financial crisis. The majority of the work under these testing regimes is performed away from the public eye. This confidentiality allows APRA to investigate extreme scenarios and “push the bounds” of institutional and systemic viability without potentially raising public concern (RBA, 2012).

(11)

11

3. Literature Review

The International Monetary Fund (IMF) and the World Bank have been using top-down and bottom-up macroeconomic stress testing as an integral component of the Financial Sector Assessment Program (FSAP) since 1999 (Chan-Lau, 2013). Since the recent financial crisis and its inclusion in requirements stipulated under Basel III, stress testing programs have been implemented by regulatory authorities around the world to assess the soundness of their domestic banking systems. As proposed by Chan-Lau (2013), in some instances the outcomes of such testing regimes guide policy decisions concerning banking sector recapitalization decisions, with testing results being reported in financial stability reports published by central banks. In the Australian setting, APRA acts as the appropriate supervisory body and discloses only rudimentary findings of testing regimes in biannual reports (called Financial Stability Reviews) issued by the RBA.

Past studies have presented differing outcomes concerning the subsequent impacts of the disclosure of stress testing results on bank holding companies (BHCs). Candelon & Sy (2015) utilize an event study approach comparing market reactions following both the announcement of stress testing procedures and disclosure of stress testing results in the US and EU from 2009 to 2013. Following the disclosure of results, they report statistically significant cumulative abnormal returns (CAR) for EU banks exposed to stress testing conditions in 2010, 2011 and 2012. However, they find only one statistically significant result for the disclosure concerning the group of EU banks not exposed to the stress-testing program in 2011. Hence, reactions to the release of stress testing results may depend on perceived conditions in the market at the time of the publication; as well as the procedures, scope and opaqueness of the stress testing regime followed.

Earlier studies have also investigated the effects an announcement of an upcoming stress testing procedure has on a particular financial system. In the case of the Candelon & Sy (2015) study, the announcement of stress testing procedures reported statistically significant findings for both EU stressed and non-stressed banks in 2011 & 2012. In a similar fashion, the conclusions drawn from US stress testing scenarios revealed significant findings only for the disclosure of results for stressed banks in 2009 and 2012.

(12)

12 In a paper concentrated solely on the stress testing procedures instigated following the introduction of the 2009 SCAP in the US, Morgan et al. (2014) investigated the effect on abnormal returns for banks displaying less favourable balance sheet circumstances following the onset of the financial crisis. Their study examined 19 bank holding companies (BHCs) and postulated that the market had developed its own inferences regarding which particular banks would reveal capital gaps in their balance sheets prior to the disclosure of stress testing results, and that the magnitude of gaps became apparent with the testing regime. This consequently led to those particular BHCs displaying larger capital gaps experiencing more negative abnormal returns. Hence, stress testing procedures produced valuable market information reducing the opacity and uncertainty surrounding banking institutions at the time of testing.

Morgan et al. (2014) and Candelon & Sy (2015) both display similar methodological approaches. That is, they utilize 2-stage event study approaches (see MacKinlay, 1997) based on the market model. Both studies use daily stock return data for each individual institution under examination and regress the corresponding returns against a representative financial index. This approach was followed by other European-based studies. Petrella G. and Resti A. (2013) examined the 2011 European Banking Authority (EBA) stress test exercise and found discernable evidence that stress tested banks had significantly larger announcement date cumulative abnormal returns than non-stress tested banks. Their study also found that the market was not able to anticipate the test results effectively, an outcome that is consistent with the proposition of greater bank opaqueness prior to the disclosure of stress testing results. Their findings provide further evidence that stress tests reveal information related to the strength of individual banks, as well as produce valuable intelligence for market participants leading to a reduction in bank opacity.

Studies by both Ellahie (2013) and Alves et al. (2014) on the EBA stress testing regimes in 2010 and 2011 further reinforced the findings of Petrella G. and Resti A. (2013). Ellahie (2013) revealed that the 2011 EBA stress test reduced informational asymmetry while Alves et al. (2014) conclude that stress tests conveyed new information and that the outcomes were not anticipated by the stock market. Moreover, Alves et al. (2014) postulate that the publication of the outcomes of the stress tests had a stronger impact on the stock prices of riskier financial institutions.

(13)

13 Flannery et al. (2015) examine implications regarding cumulative abnormal returns and trading volumes for US bank holding companies (BHCs) resulting from recent US Federal Reserve stress testing - under the Comprehensive Capital Analysis and Review (CCAR) and Dodd-Frank Act Stress Tests (DFAST) scenarios. They extend the event study approach by incorporating the three-factor model specified by Fama & French (1992, 1993) to examine both absolute cumulative abnormal returns and cumulative abnormal trading volumes for each of the BHCs under review. Their findings reveal that stress-tested BHCs experience both significant price and volume effects, with the impacts more pronounced for more levered and riskier firms. Similarly, Bird et al. (2015) examine potential disclosure effects in capital markets following the CCAR scenarios, with links to significant absolute abnormal returns. Relatedly, Glasserman and Tangirala (2015) uncover a significant correlation between projected supervisory losses for BHCs in the 2013 and 2014 DFAST testing regimes, and no statistically significant findings between projected loss rates and abnormal returns at the disclosure of other US stress testing results. Hence, they conclude that Federal Reserve stress testing procedures have become more predictable and less informative over time.

Goldstein and Sapra (2012) show that greater disclosure of bank stress testing results leads to better market efficiency & resource allocation via improved market discipline as well as greater overall financial stability (however, disclosure can exacerbate bank-specific inefficiencies). They also highlight that disclosure of results can achieve the macroprudential role of helping stabilize the financial system as a whole, but not necessarily the microprudential role of providing market discipline for specific individual banks.

In a recent working paper released by Fernandes et al. (2015), the capital market implications of stress tests in the US since the onset of the global financial crisis are examined. In particular, the authors endeavour to determine the capital market reactions to information released by stress tests, as well as if and how disclosure of these tests affects information generation and processing in capital markets. Their results indicate that there is new information in stress tests and public disclosure helps reduce informational asymmetries and uncertainties, especially during times or market distress (Fernandes, 2015). Furthermore, their findings reveal that public disclosure of stress testing methodology and results does not seem to have reduced private incentives to generate information. The authors believe that these findings have important policy implications and draw upon the workings of Borio et al.

(14)

14 (2013) to conclude that macro stress tests are ill-suited as early warning devices, but can be effective crisis management and resolution tools (Fernandes, 2015).

Considerable research has been undertaken on both the US and EU stress testing procedures in recent times. However, outside these two jurisdictions, a more limited quantity of studies have been performed. Although many regulatory institutions around the world have instigated stress testing procedures during the last decade or so, relatively few have proceeded to the extent of both the US Federal Reserve and the EBA to disclose their stress testing results, methodologies and scope. This has particularly been the case since the 2009 SCAP stress testing operations in the US and the stress testing procedures undertaken by the EBA in 2010. It is a widely-held belief that the extensive disclosure of the SCAP testing results was critical to stabilizing the US banking sector, helping to reduce informational uncertainties and bank opaqueness. Thus the motive for the EBA to follow the lead of the US Federal Reserve in 2010. By disclosing stress testing outcomes, the EBA sought to increase transparency in the European banking sector, with a particular emphasis on exposing the uncertainties posed by sovereign risk and to increase investors’, depositors’ and other creditors’ confidence in the financial sector (Alves et al., 2014).

An interesting paper by Allen and Powell (2010) further highlights the risks posed to the Australian banking sector during the recent financial crisis. A comparison to their global counterparts reveals that Australian banks continued to display strong earnings & credit ratings and adequate capitalization during the financial crisis. However, it was noted that Australian banks, like the majority of their global counterparts, experienced substantial deterioration in the market value of their assets. Armed with this knowledge, Allen and Powell (2010) utilized KMV/Merton structural methodology and investigated default probabilities to discover that Australian banks’ default probabilities were only slightly superior to their global peers.

The resilience of the Australian banking sector during and in the intermediate years following the onset of the 2008 financial crisis was in part due to sound fundamentals and thorough prudential and supervisory frameworks in place (Bologna, 2010). This resilience and the strength of the whole Australian economy in those years is largely postulated to be owing to the support contributed by the huge upsurge in resource demand the country experienced. The so-called “mining boom” came on the back of unprecedented Chinese demand for core metals and energy commodities (iron ore, coal, natural gas) in the wake of double-digit

(15)

15 Chinese growth figures. Since the downturn in the extraordinary commodity price peaks of 2011, the Australian banking sector has continued to be fueled by strong domestic housing demand. Recent risk assessments (including stress tests) of the Australian banking sector place considerable focus on the large household debt levels and behaviour of the housing market. The non-uniform growth in house prices across Australia (with the major capital cities of Sydney & Melbourne experiencing much greater price growth than other cities and rural areas) complicates attempts to control or regulate housing market lending practices (RBA, 2012; RBA, 2014). Hence, any reversal of the price rise trend could potentially leave banks with a host of loans backed by reduced collateral prices (possibly less than the original value of the loans). This proposition, along with reduced access to funding on international markets, and a slowdown in the economies of vital trading partners (particularly China) are key considerations or factors used in creation of recent stress testing regimes in Australia (APRA stress testing regimes in 2010, 2012 and 2014).

(16)

16

4. Methodology

In this section, background information concerning the methodologies for this study are outlined to provide the reader with greater insight of the approaches taken and a better understanding of the outcomes and findings.

4.1 Theoretical Background - Efficient Market Hypothesis

The undertaking of an event study concerning the impact to share prices and returns due to the particular features of an event (such as the impact associated with the disclosure of stress testing results on banking returns) requires a thorough understanding of the Efficient Market Hypothesis (Fama, 1969). The fundamental essence of the Efficient Market Hypothesis (EMH) is the proposition that current stock prices fully reflect available information about the value of the firm, and there is no additional circumstance (above what is already known in the marketplace) to earn excess profits by using this information. Three key principles underlie the theory; weak-form, semi-strong and strong-form efficiencies. The weak-form efficiency principle basically states that the current market price of the institution reflects all past publically available information and that no form of technical analysis can be effectively utilized to benefit investors’ evaluations. Semi-strong principle proposes that the current market price of a stock reflects all publically available information but is susceptible to instantaneous change if new information is uncovered. Those who subscribe to this version of the theory believe that if new information is uncovered, than investors can boost their returns to a higher degree than other market participants. The last principle, strong-form efficiency states that even concealed or hidden information known only to institutional insiders is captured in the market price of the firm (Fama, 1969; Sheppard, 2014).

For the purposes of this study, an understanding of the key principles underlying EMH theory plays a critical role in interpreting return changes associated with disclosure of stress testing results. The degree of disclosure by the appropriate regulatory authorities, as well as the extent of bank opaqueness surrounding disclosure dates, may impact the associated magnitude of abnormal returns. Given the readily available information in the public domain, as well as the highly analysed nature of banking stock returns, we can assume that current pricing levels in the marketplace reflect all past information and that instantaneous changes to market pricing will occur upon the release or announcement of influential information. Hence, we can presume that our study reflects semi-strong form efficiency.

(17)

17 4.2 Event Study Approach

The principle purpose of an event study is to use financial information to measure the effect of an economic event on the value of an institution (Sheppard, 2014). Suitable events could include situations related to an earnings announcements, the issuance of debt or equity by a firm, merger and acquisition agreements and of course circumstances related to government or central bank announcements concerning macroeconomic variables. For event study methodology to be effective, it is assumed that the appropriate event is linked to some degree to the underlying price of the assets or institutions being investigated in the study. To capture the impact or economic effect of the event on the institution, the behaviour of the firm’s prices or returns need to be examined around the event date.

Performing analysis via an event study approach requires several stages. Firstly, it is essential to define the event (or multiple events) of interest. It is necessary to determine the timeframe over which analysis will take place. This timeframe is known as the event window. Once the firm or collection of firms for analysis have been selected, it is necessary to determine the normal return that would be expected if the event under investigation did not occur. This can be defined as E[R*it │Ωit], where R*it is defined as the return for firm i (i = 1,…,N) at time t

and Ωit is the conditioning information for the normal performance model (Sheppard, 2014).

It is necessary to determine the parameters of the normal performance model over a timeframe called the estimation window. This period falls before the event window in such a way as to minimize any leakage of information prior to the event period.

To determine the abnormal return for the firm, it is necessary to calculate the difference between the actual return and the expected normal return. This can be defined as:

ε*it = R*it - E[R*it │Ωit]

The diagram below distinguishes the three timeframes or windows surrounding an event. That is, the event date is assumed to occur at time t=0.

(Estimation Window) (Event Window) (Post-event Window) │………..│………...…..│………..……..│

(18)

18 To determine the impact of an event on the institution, it is essential to calculate the abnormal return over the event window while defining the expected normal return ex ante or assuming the event is exogenous to impacts on the price of the institution.

The calculation of the expected normal return can follow a number of predetermined theoretical models (constant-mean-return model, market model, multifactor model etc.). All these models assume that the returns for the firms are jointly multivariate normal, and identically and independently distributed over time (MacKinlay, 1997). This particular study will follow the market model approach, which allows for the possibility of isolating the impact of the event from other general market movements. It is assumed that the market return (Rmt) and the institutional returns (Rit) display a stable relationship, hence, the

following relationship applies:

Rit = αi + βi Rmt + εit where E[εit] = 0 and var[εit] = σ2 εi

The market model for the expected normal return is defined as: E[R*it │Ωit] = 𝛼̃i + 𝛽̃i R*mt

Ordinary Least Squares (OLS) techniques will be utilized on the estimation window observations to gauge an estimate of 𝛼̃i and 𝛽̃i.

For a linear regression performed using OLS techniques to be considered the Best Linear Unbiased Estimator (BLUE), the assumption that the error terms have a constant variance (Var[εi│X] = σ2 ∀ i) and are uncorrelated with one another (Cov[εi,εj│X] = 0) must be made

(Sheppard, 2014). However, as past empirical studies have exposed, the return variance is not constant over time and may exhibit periods of higher volatility and episodes of little variation. This phenomenon is called volatility clustering and can lead to heteroskedastic estimation issues. To account for this effect, past empirical studies have utilized the Autoregressive Conditional Heteroskedasticity (ARCH) model first proposed by Engle (1982).

To describe the ARCH model, consider the following:

Let the return for an institution between times t-1 and t be equal to a conditional expectation and an additional term (ηt) describing the stochastic error. Hence, we arrive at the following

equation:

rt = ∑𝑘𝑖=1 i 𝑏 xt,i + ηt where ηt │Ωt-1 ~ N(0, σt2)

From this equation, it is assumed that the error terms follow normal (N) distribution conditional on the information (Ω) provided at time t-1. However, due to the fluctuating

(19)

19 nature of variances over time, there is an expectation that the unconditional distribution of ηt

will display fatter tails than that corresponding to normal distribution (Engle, 1982): ηt = σt εt where εt ~ is I.I.D. and N(0,1)

With εt displaying unit variance, the conditional variance (σt) will depend upon the past

values of squared errors and can be summarized by the following equation, known as the ARCH(q) model:

σt2 = ω + α1 η2t-1 + α2 η2t-2 +…+ αq η2t-q

where ω and α must be non-negative values to produce positive values for the variances at any time (Sheppard, 2014).

An alternative approach to the ARCH (q) model could be to implement Bollerslev’s (1986) Generalised Autoregressive Conditional Heteroskedasticity (GARCH) model. This method is similar to that of the ARCH(q) model in that it relies on past values of squared errors, however it also incorporates the past conditional variances such that the GARCH (p,q) is:

σt2= ω + α1 η2t-1 + β1 σ2t-1 + …+ αp η2t-p + βq σ2t-q

When applying the GARCH (1,1) model, to effectively calculate the expected normal return requires the following conditional variance to be applied to the mean model in order to capture any heteroskedastic effects existing in the data set (Sheppard, 2014):

εit │Ωt-1 ~ N(0, σt2) = ω + α1 η2t-1 + β1 σ2t-1

This approach has been followed in a number of past event studies including Wang et al. (2002), Batchelor & Orakcioglu (2003), and McKenzie et al. (2004). A study by Hansen et. al. (2009) shows that the GARCH(1,1) model provides superior forecasts for the volatility of daily stock returns. The GARCH (1,1) is the base GARCH approach followed in this particular study.

One potentially unfavourable feature of the GARCH (1,1) model is that it fails to account for any asymmetry associated with positive and negative shocks arising in the market (otherwise known as leverage effects). Past researchers have examined the asymmetric consequences that arise from the information affecting an institution’s revenues. Bandi and Reno (2012) find that the asymmetry negatively increases risk, leading to a decline in asset values as a consequence of a fall in revenues with an associated increasing debt ratio. This increased leverage results in more volatile share pricing (Sheppard, 2014). Nelson (1991) proposed the introduction of the EGARCH model (Exponential GARCH model) as a novel solution to deal with this problem. The EGARCH model applies a logarithmic conditional variance such that

(20)

20 the variance remains positive even if parameters display negative values. Furthermore, Nelson (1991) introduced a leverage term into the base GARCH model equation to account for the asymmetric effect by increasing or decreasing the effect imposed by the error term (Sheppard, 2014). Hence, the Nelson (1991) model in its EGARCH (1,1) form:

ln (σt2) = ω + β ln(σ2t-1) + α│ εt−1

σt−1 │+ γ │ εt−1 σt−1 │

Note, β refers to the coefficient on the new logarithmic GARCH term while α relates to the ARCH term and γ the so-called leverage term. Hence, if γ is significant and different from zero there will exist asymmetry in the estimation period (Sheppard, 2014). The advantage of the EGARCH model is that it allows volatility clustering and leverage effects to be taken into account (Nelson, 1991).

Hence, to account for estimation inaccuracies associated with financial data as well as provide robust checks and more accurate results, abnormal returns will be calculated using Ordinary Least Squares (OLS), Generalised Autoregressive Conditional Heteroskedasticity (GARCH) model and Exponential GARCH (EGARCH) models.

4.3 Estimation of Abnormal Returns

For any given banking institution i at time t, the abnormal return (ARit) will be calculated by

the following:

ARit = Rit – [𝛼̃i + 𝛽̃i R*mt] where Rit is the actual return for institution i at time t

and Rmt is the return on the market index over time t

The term in parenthesis (the market model for the expected normal return) was calculated earlier. Using these corresponding values for each institution (i) at each time period (t), it is now possible to determine the average excess returns for time period (ARt) and the

cumulative abnormal return (CAR) for each institution (CARi). That is:

ARt = ( 1

N) ∑ AR 𝑁

𝑖=1 it where N is the no. of firms with excess returns during time t

CARi,k,l = ∑𝑁𝑡=𝑘ARit where CARi,k,l is for the period t=k until t=l.

Once the abnormal returns have been determined, it is possible (via a standard t-test) to ascertain the significance of the deviation from the normal return. That is, a simple hypothesis test (defined below) with the decision to reject H0 if │tcalc│ > tcrit :

(21)

21 H0: (CARi)k,l = 0 (abnormal returns cannot be distinguished from zero)

H1: (CARi)k,l ≠ 0 (abnormal returns can be distinguished from zero)

To determine tcalc: tcalc = CARk,l

SE(CARk,l)

Any announcement or disclosure of stress testing results are considered one-day events for which the actual date is defined as t. However, the event window will be extended to include one day prior and one day post the event date. Hence, a three-day event window [t-1, t+1] will firstly be analysed to calculate abnormal returns. By extending the timeframe of the event into a window period, the risk of information being leaked a day prior to the event date as well as capturing the effect of a slower response by investors can be integrated into the model (Petrella and Resti, 2012). As highlighted earlier, it is necessary to determine the parameters of the normal performance model over the estimation window timeframe. This window will cover a period extending from 210 days to 30 days (180 days in total) prior to the event window. Further robustness checks have incorporated varying the length of event and estimation window periods. Numerous past event study papers have investigated varying timeframes in both window periods - see MacKinlay, 1997 and Morgan et al. (2013).

As per the Data Collection section, appropriate share price information for each of the relevant Australian BHCs was obtained via the Wharton (WRDS) – Computstat Capital IQ Database (any additional information relating to gaps in the WRDS data was sort via Yahoo Finance). Prior to importing the Australian BHC data through into Stata, it was necessary to first manipulate and organize the data in Excel to allow for latter ease of estimation techniques and statistical measures regarding the calculation of abnormal returns in Stata. Hence, the determination of return calculations and natural logarithmic (ln) return calculations for each firm was performed in Excel. Hence, ln return calculations were determined via the following:

rt = ln Pt

Pt−1 = ln (Pt) – ln (Pt-1) where Pt is the observed closing price of a

particular firm on day t, while Pt-1 is the

previous day’s (t-1) closing price for the same firm.

Upon importing this data into Stata, it was necessary to incorporate S&P/ASX 200 Index and All Ordinaries Index return data into the same file. Return data for both the S&P/ASX 200

(22)

22 Index and All Ordinaries Index was extracted from Yahoo Finance over the same corresponding period as the BHC share price data (January 2005 – April 2016). Using the merge command in Stata, it was possible to incorporate the data for both indices along with the corresponding BHC share price and return data for each respective day over the 10-year period.

4.4 Calculating Abnormal Returns under OLS

Under the market model approach mentioned earlier, the natural logarithm (ln) of return for each of the various individual BHCs were regressed against the S&P/ASX 200 Market Index over the predefined estimation period of 180 days (210 days to 30 days prior to the event date). An estimation of the expected normal return was preformed via undertaking separate regressions for each respective BHC, using the corresponding data within the estimation windows for each individual BHC (hence, α and β values were determined for each respective BHC).

Within the respective event windows (11 for each institution, except both Adelaide Bank and St George Bank as these particular BHCs were subjected to merger and acquisition activity during the period under investigation) for each of the 11 BHCs under review, it was necessary to subtract the expected normal returns from the actual returns in order to obtain the abnormal returns (as per equation outlined in the Methodology section). Average abnormal returns and cumulative abnormal returns were calculated as outlined in the methodology section, and standard t-tests determining the significance of each cumulative abnormal return per individual institution and event episode were performed (see Empirical Results and Appendix sections – tables of results are outlined for all three methodological approaches).

Due to the presence of heteroskedastic effects (data with non-constant variance), estimation via the standard OLS approach will return standard errors and confidence intervals that are too narrow (Skog, 2010). However, GARCH and EGARCH models treat heteroscedasticity as a variance to be modelled inside the regression (Engle, 2001; Nelson, 1991). Hence, for each methodological approach (OLS, GARCH or EGARCH), there were 101 t-test values corresponding to each calculated cumulative abnormal return (11 event episodes, 9 institutions, plus one t-test value for both Adelaide Bank and St George Bank following the first event episode in 2006).

(23)

23 4.5 Calculating Abnormal Returns under GARCH and EGARCH

Following the determination of abnormal returns under the OLS approach, robustness checks were run utilizing both the GARCH and EGARCH methodologies. These more advanced techniques were applied to account for heteroskedastic properties existing in the dataset. The calculation of the abnormal returns under the GARCH and EGARCH methods share similarities with the OLS model. That is, they both incorporate the market model in the process for determining the expected normal returns for each institution. However, both further include their own methodological terms (see earlier commentary in Methodology). Although previous event study investigations have utilized GARCH and EGARCH methodologies, no prior studies have been carried out examining the abnormal returns related to stress testing procedures for banking institutions. It is appropriate to apply such methodologies in this study, as several of the stress testing dates fall around extended periods of pronounced volatility (such as the global financial crash of 2008, and the sovereign debt crisis of 2012). Figure 1. below displays return volatility charts for each of the “Big 4” institutions.

Figure 1: Return Volatility Charts for the “Big 4” Australian Banking Institutions

Figure 1: Graphs displaying the positive and negative return variance (in percentages) for each of the “Big 4” Australian banking institutions. The four charts exhibit the existence of volatility clustering, where periods display high and low volatility clusters. Each of the four charts span from January 2005 to April 2016. Source: Wharton WRDS Compustat Capital IQ Database, Yahoo Finance and own calculations.

(24)

24

5. Data Collection and Stress Testing Events

The following section briefly describes the procedures undertaken to collate the various data sources used in this study, as well as provide an overview on each of the stress testing events performed on the Australian banking sector since 2003.

5.1 Australian Banking Data Collection

Australian banking data was retrieved via both the APRA and RBA websites. APRA collects and publicly provides monthly banking statistics for all banking institutions operating in Australia. This data is useful to provide an insight into the size and share of each respective institution’s loan book, asset and deposit holdings. Table 1a outlines the total residential assets, total gross loans and total deposits for each of the nine domestic Australian banking institutions under review in this study. From the table below, it is possible to gauge the degree of market dominance associated with the four largest (the “Big 4”) Australian banking institutions (WBC, CBA, NAB and ANZ).

Table 1a: Descriptive Statistics - Australian Banking Sector Composition

Institution Name (ASX Code)

(amounts are specified in $millions AUD) Total Resident Assets % Total Gross Loans % Total Deposits % Westpac Banking Corporation (WBC) 778420 22.9 509145 22.1 397316 20.5

Commonwealth Bank of Australia (CBA) 723061 21.3 552146 24.0 478626 24.8

National Australia Bank Ltd (NAB) 627468 18.5 421055 18.3 334165 17.3

Australia & New Zealand Banking Group

(ANZ) 566405 16.7 362872 15.8 303080 15.7

Macquarie Bank Ltd (MQG) 84469 2.5 43162 1.9 43033 2.2

Bendigo and Adelaide Bank Ltd (BEN) 59283 1.7 48216 2.1 48353 2.5

Suncorp-Metway Bank Ltd (SUN) 58977 1.7 49486 2.1 40232 2.1

Bank of Queensland Ltd (BOQ) 47554 1.4 36819 1.6 40168 2.1

AMP Bank Ltd (AMP) 14674 0.4 12559 0.5 6293 0.3

Other 432756 12.8 266187 11.6 242507 12.5

Sum Total (excl. Other) 2960310 87.2 2035499 88.4 1691267 87.5

Total 3393066 100.0 2301686 100.0 1933774 100.0

Notes: Only domestic Australian banking institutions are listed in the table above (that is, all foreign subsidiary and branch banks are included as

part of ‘Other’ category). The data presented corresponds to Australian banking data (as of the 31st

March 2016) reported monthly by each individual institution to the Australian Prudential Regulatory Authority (APRA).

(25)

25 Further data (i.e. individual daily share price data) related to each of the individual institutions was sourced via the Wharton (WRDS) Compustat Capital IQ Database. This database contains daily share price, trading volume and other secondary information related to all Australian financial institutions. It is necessary to sort and filter through the data listings keeping only the banking institutions (authorized deposit-taking institutions). Note too, it is important that the data entries related to the trades associated with the primary shares are incorporated into the master data file (secondary share trades are listed for some of the institutions displaying small trading quantities, drop these entries from the final dataset. There exclusion should not influence or bias the overall findings of the investigation). Any missing share price and trading volume data related to any particular institution was sourced via Yahoo Finance.

The full daily share price dataset for the 11 institutions (including Adelaide Bank and St. George Bank – both of which were merged and acquired by separate institutions in 2007 and 2008 respectively) contains 27747 individual entries, over a timeframe spanning from the start of 2005 until April 2016. It is necessary to merge a proxy for a representative Australian stock market index (ie S&P/ASX 200 Index, All Ordinaries Index) spanning over the same timeframe. This proxy will be used (along with the institutional share pricing data) during the regression analysis as a core component in the market model. It is also necessary to replicate the 27747 individual entries into 11 identical subsets spanning the same timeframe, each subset will be utilized to examine the effects corresponding to different event episodes. Hence, the final dataset contains 288117 separate entries to examine the 11 event episodes across the 11 individual institutions (101 unique ‘institutional-event’ entries – 11 event episodes for each of the 9 institutions and single event episodes for both Adelaide Bank and St. George Bank respectively, corresponding to the 2006 stress testing event). Although the events are explicitly listed under 8 event headings – events 4, 5 and 6 all contain both announcement and disclosure episodes (hence, the 11 event episodes explicitly tested).

Table 1b (on the next page) outlines further descriptive statistics related to the Australian banking sector. It highlights the distinct domination the four largest institutions hold over the Australian banking sector. A clear portrayal of the increasing market dominance with regards to the core banking attributes (asset and loan holdings) since 2002 is provided in the Figures 2a and 2b. Further charts (Figure 2c and 2d) included in the Appendix depict the composition of the Australian banking sector.

(26)

26

Figure 2a: Australian Banking Sector- Total Resident Assets Figure 2b: Australian Banking Sector- Total Gross Loans

Table 1b: Descriptive Statistics: Australian Banking Sector

This table outlines the summary statistics for all Australian banks (hence, it excludes foreign bank branches and subsidiaries operating in Australia) over the period July 2002 to April 2016. The “All Institutions” category includes the 9 independent institutions operating post-2008 (hence, after the acquisition of St. George Bank by Westpac). The “Big 4 Institutions” include Westpac Banking Corporation (WBC), Commonwealth Bank of Australia (CBA), National Australia Bank (NAB) and Australia New Zealand Banking Corporation (ANZ).

All Institutions Big 4 Institutions

Mean Standard Deviation Median 25th Percentile 75th Percentile Mean Standard Deviation Median 25th Percentile 75th Percentile Total Resident Assets ($mil) 200857 200942 57749 34084 379229 408105 62562 412143 368535 451713

Total Gross Loans ($mil) 132236 136181 40531 20527 258659 278947 42617 278170 252111 305006

Total Deposits ($mil) 111143 108897 34225 24206 201137 222539 40193 216934 196692 242781

100,000 200,000 300,000 400,000 500,000 600,000 700,000 800,000

900,000 Australian Banking Sector: Total Resident Assets ($ million AUD)

WBC CBA NAB ANZ MQG BEN SUN BOQ AMP ADE STG

Source: APRA Monthly Banking Statistics

100,000 200,000 300,000 400,000 500,000

600,000 Australian Banking Sector: Total Gross Loans ($ million AUD)

WBC CBA NAB ANZ MQG BEN SUN BOQ AMP ADE STG

Source: APRA Monthly Banking Statistics

Notes: Figures 2a and 2b above display the growth in Total Resident Assets and Total Gross Loans of the domestic institutions constituting the Australian banking sector since mid-2002 (excludes foreign bank branches and subsidiaries operating in Australia).The acquisition of St George Bank (STG) by Westpac(WBC)is clearly depicted in Figure 2B. As a consequence of this action, Westpac rose ahead of Commonwealth Bank (CBA) to become the leading Australian Bank on the basis of Total Resident Assets and Total Gross Loans.

(27)

27 5.2 Stress Testing Events and Procedures in the Australian Banking Sector

This section outlines the stress testing procedures performed on the Australian banking sector since 2003, with varying degrees of public disclosure for the scope, methodologies and results dependent on the respective testing authority. Each distinct testing regime is designated by a separate event heading and plotted on a timeline chart at the end of this section. Note though, events 4, 5 and 6 contain multiple event episodes that will be tested – dependent on whether each episode is an announcement, release of initial results or full disclosure incident.

Event 1 – Late 2003, “Panama Project” by APRA

In late 2003, APRA performed its first stress testing procedure as part of its so-called “Project Panama”. This investigation was aimed at highlighting the potential risks posed to the Australian banking sector by a considerable downturn in the housing market (an approximate 30% drop in prices in the residential housing sector and default rates climbing to as high as 3.5%). The testing results revealed that the banking sector could easily withstand the adverse conditions, with modelled loss rates well within minimum capital requirements. The results were disclosed by the APRA chairman at the time, who stated that the substantial housing market correction “would not of itself be a cause of undue alarm” for the banking sector (Hughes, 2003). It is largely considered that this testing event was not a “true” macroeconomic stress testing event, as this particular program lacked key stress testing considerations that later regimes incorporated (it failed to focus on the potential causes for the downturn nor on the secondary effects imposed on other ADI lending or profit sources). However, it did provide valuable insight for both the regulatory authority and lending institutions surrounding the shortcomings in monitoring and the availability of vital information concerning lending risks, loan-to-valuation ratios (LVRs) and insurance status of loans (Hughes, 2003). This awareness would aid the development of future testing procedures.

Event 2 – Sept. 2006, IMF FSSA Program

The first macroeconomic stress testing procedure of the Australian banking sector was performed as part of the IMF FSSA Program in September 2006. Under the FSSA procedure, three stress test settings were performed. The first, involved a macroeconomic stress scenario of the five largest domestic banks over a three-year time period. Conditions mirroring a short but sizeable recession in the economy pertaining to an initial slump in housing prices (approx. 30% drop in housing prices), a substantial increase in the unemployment rate (from

(28)

28 5% to a peak near the 9% mark), exchange rate shocks to the Australian dollar and increased associated funding costs for banks. The second scenario involved the same banks running a series of single factor stress tests to examine the sensitivity of their banking and trading portfolios due to interest rates shocks. While the last testing routine involved analyzing mortgage portfolio stress thresholds for two regional banks with heavy exposure to the housing sector (IMF Country Report No. 06/372, October 2006).

In a similar fashion to the 2003 testing regime, the large Australian banks under the IMF 2006 FSSA program exhibited considerable resilience, with no near-term stability issues. While their financial performance deteriorated in response to shocks to their mortgage portfolios and funding costs, the banks withstood the adverse macroeconomic scenario relatively well (IMF Country Report No. 06/372, October 2006). The IMF 2006 FSSA report further elaborated on the specific factors that may have explained the favourable testing results at the time. This included the advantageous economic conditions1 the Australian economy had experienced in the 15 years prior to 2006, the portfolio composition of banks and their significant mortgage exposure to the housing market (which is generally resilient to a short-lived recession) and the reliance of the Australian banking sector on the wholesale funding market (both the domestic and offshore wholesale funding markets).

The IMF 2006 FSSA program was generally considered to be a valuable learning exercise and an effective mechanism to address the financial stability and systemic risks posed to the Australian banking sector. It was timely given the preparations being put in place for the necessary compliance stipulations surrounding the Basel II risk-management framework. The only criticism the RBA expressed concerning the IMF 2006 FSSA program related to the design of the macroeconomic scenario. That is, it involved a domestic Australian recession alongside an ongoing expansion in the global economy. All previous recessions in Australia had corresponded with global downturns. Hence, had a weaker global economy been incorporated in the FSSA testing procedure, it is likely that a significantly more challenging situation would have arisen for the Australian banking sector. This is the situation that would

1_{The IMF cite 15 consecutive years of economic expansion, Real GDP growth averaging 4% annually during} 1992 to 2005 and the unemployment rate falling to a 30-year low of approx. 5% in late 2005 as key advantageous economic conditions. The resilience of the Australian economy has been noteworthy; growth has not fallen below 2% in any calendar year since 1991, despite shocks surrounding the Asian crisis in 1997-1998, the global IT slump in 2000-2001, and a severe drought in 2002-2003 (IMF, 2006).

(29)

29 arise and confront the banking sector only two years later with the onset of the financial crisis in October 2008.

Event 3 – June 2010, APRA Chairman Address (#1)

On June 9th 2010, the chairman of APRA publicly addressed a meeting of the Australian Business Economists in Sydney to discuss the disclosure of results from a recently performed stress testing regime. Although APRA had been conducting stress testing procedures behind “closed doors” for a number of years prior to this date, this address was the first time the chairman of the main regulatory body fronted the public to elaborate on procedures and results surrounding the performance of a macroeconomic stress testing program on the Australian banking sector. In his 10-page address, Laker outlined the role stress testing provides as a risk-management tool and the part APRA plays in coordinating stress testing regimes in the Australian banking system. Although the Australian regulatory bodies (APRA and RBA) provide recurring updates via the bi-annual release of the RBA Financial Stability Review documents, these publications can only be considered as updates of regular supervisory monitoring outcomes and control procedures.

The 2010 stress testing program involved a three-year macroeconomic scenario developed in conjunction with both the RBA and the Reserve Bank of New Zealand (RBNZ). The scenario investigated the impacts of further deterioration in global economic conditions, this time with a particular focus on China (in the decade prior to 2010, China had risen to become the leading significant trading partner for Australia – ahead of the US, Japan, South Korea and Europe) and gave consideration to the ongoing fragility of North Atlantic banks dragging recovery efforts and contributing to wider spreads in global funding markets. These conditions generated a hypothetical economic downturn in the Australian economy significantly worse than those experienced during the recession in the early 1990s (and conditions far worse than those modelled four years earlier under the IMF 2006 FSSA program). In his speech, Laker disclosed a number of specific macroeconomic parameters used in the testing regime including a sharp 3% contraction in real GDP in the first year, a rise in the unemployment rate to 11%, and peak-to-trough falls in both residential house pricing of 25% and commercial property of 45% (Laker, 2010). He further elaborated on the stress testing procedures followed during the course of the program, as well as detailed the main findings in relation to the participating ADIs as a collective. The results suggested that none of the banks would have breached the 4% minimum Tier 1 capital requirement set out under the Basel II framework (Bologna, 2010). No explicit comment was specified in the

(30)

30 chairman’s address related to the 8% (Tier 1 and Tier 2) capital requirement. Laker’s concluding remarks surrounding the 2010 APRA stress test program were that “it provided important evidence that the Australian banking system has the capital resources to weather an economic contraction much worse than that expected during the depths of the global financial crisis”.

Event 4 – Sept. 2010, Fitch Ratings Stress Testing

On September 30th 2010, credit ratings agency Fitch announced that it intended to stress test the four largest Australian banks (WBC, CBA, NAB and ANZ) to investigate the potential impacts surrounding a slump in Australian housing prices and the consequential spike in mortgage defaults. Their aim was to determine the bearing such effects would have on the portfolio of Australian residential-backed mortgage securities that they rated. At the time, Fitch’s Australian managing director announced that “over the last few months, Fitch has received numerous inquiries as to the sustainability of Australian residential property prices and the possible impacts of a correction” (White, 2010). He followed this statement with an expectation of the proposed findings by declaring that “over the short-to-medium term, a downturn is not Fitch’s central expectation. The agency is performing its stress test exercise on ratings impacts under the hypothesis of an imminent housing market correction”.

Fitch displayed clear motives for pursuing its investigation, with many industry insiders describing the ‘housing bubble’ as the one overhang in the Australian economy posing a clear risk to the stability of the system. According to RBA estimates, home prices in Australia’s capital cities had risen 41% on average over the previous four year period whilst those in the US and UK had experienced falls of 28.4% and 18% respectively over the same timeframe (Zappone, 2010). At the time, some proponents estimated that Australian banks had 60% of their loans secured by residential property leading to a large direct exposure of the Australian economy to this domestic asset bubble.

The preliminary testing results of the Fitch stress testing regime were released on the 13th October 2010. The findings delivered ‘manageable’ results under all three stress scenarios (mild, medium and severe). The final results were reported on the 25th January 2011, leading the ratings agency to declare that “Australian banks and mortgage insurers could absorb the losses should home prices plunge and related securities would maintain their ratings”. These findings further reinforced the conservative lending standards and the strength of the