Identifying the Risk Transmission Mechanisms within the Jamaican Financial System: The Conditional Value-at-Risk Approach

Transcription

1 Identifying the Risk Transmission Mechanisms within the Jamaican Financial System: The Conditional Value-at-Risk Approach Jide Lewis 1 Financial Stability Department Research and Economic Programming Division Bank of Jamaica Abstract During times of financial crises, losses tend to spread contagiously across financial institutions, threatening the system as a whole. This paper, which follows closely the work of Adrian & Brunnermeier (2009), will explore systemic risk vulnerabilities within Jamaica's financial sector by estimating the conditional value at risk (CoVaR) between various publicly listed deposit taking institutions. That is, the value at risk (VaR) estimates of the entities within the financial system conditional on other institutions being in distress. Further, the study will quantify the extent to which characteristics, such as leverage and size, interbank activity and other balance sheet factors, can predict systemic risk contribution using panel data regression techniques. Finally, to the extent that the framework can quantify the incremental contribution of each institution to overall system risk of the system then the framework should also prove useful in the identification and on going monitoring of systemically important institutions within the financial sector. This approach will not only assist in the enhanced high frequency forward looking macro prudential monitoring of the sector but also assist in the identification of the risk transmission mechanisms within the Jamaica financial system. Keywords: Risk Transmission, CoVaR, Contagion JEL classification: G10, G18, G20 1 The views expressed in this paper are those of the author and in no way represent an official position of the Bank of Jamaica.

2 1.0 Introduction During times of financial crisis, losses tend to spread contagiously across financial institutions, threatening the financial system as a whole. For example, after the intensification of the global financial crisis in 2008, the Jamaican financial system came under severe stress. This stress was triggered by margin calls which came after Government of Jamaica (GOJ) sovereign bonds spreads widened in the aftermath of the collapse of Lehman Brothers. As the market value of the collateral backing the margin arrangements of these specific institutions decreased, their counterparty credit risk exposure increased for entities exposed to them via the interbank market during the latter part of 2008 and into 2009 (see Figure 1). Contemporaneously, there were significant outflows from the economy due to the significant increase in calls to repay margin arrangements on GOJ global bonds and the termination of some repurchase arrangements and lines of credit with overseas brokers and distributors. These events promulgated increased levels of volatility across the inter bank, foreign exchange and stock markets. 2 As the correlation between the market values of assets and liabilities increased, so did the systemic risk of the financial system. The Bank of Jamaica (BOJ) responded to these challenges by establishing a special loan facility for security dealers and deposit taking institutions (DTIs) with US dollar liquidity needs to repay margin arrangements on GOJ global bonds. An intermediation facility was also implemented within the domestic interbank market in both foreign and local currency to enhance the flow of credit in the system. Finally, the Central Bank sold foreign currency to authorized dealers in the foreign exchange market to augment the supply of hard currency and reduce the volatility in that market. 3 Experiences such as the financial stress period in the later part of 2008 underscore the importance of regulators to develop forward looking measures of systemic risk that capture the increase in tail comovement during times of financial stress. 4 Value at Risk (VaR), a measure commonly used by portfolio managers and regulators, summarizes the worst loss over a target horizon that will not be exceeded with a given level of confidence (Jorion, 2007). However, Adrian and Brunnermeier (2008) point out this stand alone measure of risk does not capture too big to fail or too connected to fail risks which are the significant drivers for the systemic risk exposure of the financial system at any given time. Further, Brunnermeier, Croket, Goodhart, Perssaud, & Shin (2009) note that a systemic risk measure should identify the risk to the system posed by institutions which are so interconnected and 2 See the Bank of Jamaica (2008) for more details. 3 This amounted to approximately US$917.8 million. 4 By the end 2009 the Government of Jamaica (GOJ) was in advanced stages of negotiating a Stand by Arrangement with the International Monetary Fund (IMF) and other Multi lateral Financial Institutions (MFIs) for balance of payments support of USD$2.4 billion. The GOJ also undertook a significant debt re profiling exercise of its domestic debt of $700.0 billion or 65.0 per cent of GDP in January The transaction targeted participation of domestic bond holders, with the aim of doubling the average age of the domestic debt profile while lowering the interest costs of GOJ by an average of 850 basis points. The transaction was dubbed the Jamaica Debt Exchange (JDX) The GOJ also took many steps to address the market uncertainty, focusing on core policies that would entrench fiscal discipline and restore confidence in the domestic economy (Bank of Jamaica, 2011). 2

3 Figure 1. Rates in the Inter Bank Market Figure 1 The disruption in inter bank activity post September 2008 reflected itself f in increased periods of non trading in the interbank market as well as sharp and persistent increases in inter bank rates. large thatt they can cause negative risk spillover effects on other institutions. That is, having computed the VaR for a given financial institution it is natural to ask to what extent it contributes to the VaR of other financial institutions. For example, how much of the risk of Bank 1 is due to the activities of Bank 2, and from the perspective of a regulator, how much of the system s risk is due to the activities of Bank 2? 5 In this vein, Adrian and Brunnermeier (2009) put forward a reduced form measure of systemic risk that is relatively easy to interpret. CoVaR measures the degree of risk externalities that a single institution can place on the financial system. The marginal risk decomposition captured by the CoVaR measure can then identify unintended concentrations of riskss within the financial system. The regulator, once armed with marginal risk contributions of each participant in the financial system, can then monitor whether these risk contributions are within tolerablee limits andd if necessary conduct additional on site monitoring of a particular institution, or if necessary, impose additional capital adequacy related sanctions. There are three broad advantages to using the CoVaR measure when undertaking the evaluation of systemic risks in the financial system. Firstly, the measuree focuses regulatory efforts on the welfare deteriorating aspects of financial activity which is the primaa facie argument for regulation. Specifically, the CoVaR measure focuses on the contribution of each institution to overall systemic risk which, in aggregate, captures excessive risk taking which affects the stability of the entire financial system. 5 Pearson (20 002), for example, discusses these kinds of issues from the perspective of a portfolio manager. 3

4 Secondly, the same measure, is specific enough to capture the increase in the risk of individual institution j when institution i falls into distress. In this context, systemic risk is the probability that, if one institution is in distress, it can possibly trigger other institutions to also be in distress, which can consequently lead to bank run and the collapse of the financial system when a certain number of institutions are affected. Thirdly, this concept can be easily applied to other measures of tail risk such as expected shortfall, which is ideal since the downside aspects of the VaR measure are well documented in the literature. This study intends to estimate the contribution to systemic risk of selected financial institutions, estimate the incremental contribution to the downside risk of an institution contingent on other entities undergoing stress in order to understand the potential for contagion during time of market stress, and use panel regressions to understand the risk transmission mechanisms in the Jamaican financial system. The paper outline is as follows. Section 2 discusses the data used in the estimation of the systemic risk indicator, CoVaR, while Section 3 discusses some of the stylized facts about the evolution of risk and returns of listed financial institutions on the local stock market. Section 4 presents the methodology behind the measure of systemic risk. Section 5 outlines the empirical results emerging from the framework as well as the econometric assessment of the risk transmission mechanisms within the Jamaican financial sector. The paper concludes in Section 6 with a summary of the findings and some policy implications Dataset The dataset included weekly equity prices of five listed financial institutions in Jamaica, covering the period January 2002 to December Equity market capitalizations (MKTCAP) for each institution are then computed by taking the product of the shares outstanding, which was available on a quarterly basis for the Jamaica Stock Exchange (JSE), and the weekly equity prices. Leverage ratios (LR) are obtained from the monthly balance sheets of these institutions and are converted into a weekly series using cubic spline interpolation. 7 Market valuations of total assets are then derived in weekly frequency and can be computed accordingly using equation (1). (1) 8 Weekly returns are then calculated from the weekly market values derived (see Equation 2). (2) 6 These five institutions, which are publicly listed on the Jamaica Stock Exchange (JSE), account for 62.0 per cent of the total asset base of all deposit taking institutions in Jamaica. 7 Leverage ratios are computed as the ratio of the book value of assets of each institution to the book value of the institutions equity. 8 Adrian and Brunnermeier (2009) start with the concept that a firm s riskiness can be captured in the changes in the market value of its assets as perceived by the market. 4

5 By the same logic, the asset return of the system is given by, where. (3) Since weekly returns can exhibit unusually high or low levels of volatility due to idiosyncratic risks which are unrelated to market perceptions e.g. mergers, the weekly returns data are censored by eliminating returns which are above and below three standard deviations. The macroeconomic variables used in the analysis are four lags of Jamaica Stock Exchange Index weekly returns and four lags of the 30 day historical volatility of JSE weekly returns which are computed for daily index data available from the JSE. In addition, three other macro state variables are included in the analysis: quarterly total gross domestic product at current market prices (GDP), monthly consumer price index (CPI) and monthly loan to deposit spread (SPRD). The first two variables are obtained from the Statistical Institute of Jamaica (STATIN) and the last variable is obtained from prudential data submitted to the Bank of Jamaica. Each of these series, GDP, CPI and SPRD, are converted into weekly frequency using quadratic interpolation, cubic interpolation and simple averaging, respectively. The GDP series is converted into a monthly nominal growth rate by using equation (4) below: 4 / 4 (4) The CPI series is converted into monthly inflation, using the weekly CPI, and using equation (5) 4 / 4 (5) 3.0 Stylized Facts about the Evolution of Equity Market Capitalizations, the Co movement of Bank Returns and Leverage Ratio for Selected Banks ( ) The evolution of the market value of assets listed on the JSE can be characterized by three distinct phases over the period 2002 to Between March 2002 and end 2004, the market values of financial stocks were generally increasing (in some cases exponentially). Notably, between 2002 and end 2004, the market value for financial stocks increased by per cent to $171.1 billion. In the second period, between the first quarter of 2005 and the third quarter of 2008, market value recorded a decline of 12.5 per cent over the period, averaging $90.0 billion. Finally, between the third quarter of 2008 and the third quarter of 2010, market value declined by 32.7 per cent and averaged $74.3 billion (see Figure 2). 5

6 Figure 2. Market Value of Assets and Leverage Ratio for the Financial Sector Not surprisingly, the evolution of the equity valuations for financial stocks followed a similar trajectory over the review period which displayed a tight correspondence with movements of the overall JSE index. 9 There was also a noticeable amount of deleveragingg which took place over the review period. The overall leverage used by financial firms declined to 7.3 att end September 2010 relative to a leverage ratio of 10.0 at the beginning of the sample period. The decline in leverage used by financial firms also corresponded to the three periods outlined for the evolution of the market value of financial firms. Financial firms recorded average leverage ratios of 9.6 andd 8.7 for the periods 2002Q2 2004Q44 and 2005Q1 2008Q3, respectively. Not surprisingly, financial firms continued to deleverage following the intensification of the global financial crisis. This sector recorded an average leverage ratio of 8.2 for between 2008Q3 2010Q3. 9 There was a 92.0 per cent correlation between market capitalization of financial stocks and the overall JSE index between 2002 and This is not surprising given that financial stocks make up a significant proportion of the JSE index. 6

7 Table 1. Average Weekly Asset Returns of Financial Firms Weekly Average Returns (Per cent) Ave. Weekly Weekly Min. Max. Rt_Sys Rt_Bank Rt_Bank Rt_Bank Rt_Bank 4 NA Rt_Bank Between 2002 and 2010, financial sector firms recorded average weekly asset returns of zero per cent. However, the fluctuations in the weekly returns were broadly positive prior to 2004 and negative between 2005 and Also, there was a significant amount of variability in the returns of financial stocks as reflected in the maximum weekly growth of 8.0 per cent and minimum weekly decline of 7.0 per cent (see Table 1). Table 2. Correlation between Weekly Asset Returns of Financial Firms SYS_RT Bank 1_RT Bank 2_RT Bank 3_RT Bank 4_RT Bank 5_RT SYS_RT Bank 1_RT Bank 2_RT Bank 3_RT Bank 4_RT Bank 5_RT Over the review period, Bank 1 recorded the highest correlation with overall financial sector weekly returns of 0.12, while the other financial institutions recorded low levels of correlation with the sector s returns. Bank 1 and Bank 5 also had the highest level of correlation, 0.33, between their weekly returns. This was followed by the co movement of returns between Bank 2 and Bank 1 and the correlation of returns between Bank 2 and Bank 5, which recorded correlations of 0.29 and 0.27, respectively (see Table 2) Estimation Procedure used to Determine VaR and CoVaR Measures Both the VaR and CoVaR measures are captured using quantile regressions. While many regression models are concerned with the assessment of the conditional mean of a dependent variable, risk management applications are primarily concerned with the lower tail of the distribution of the 10 As correlations between returns can be time varying the results from Table 2 should only be interpreted as being highly stylized. 7

8 dependent variable. 11 For example, for risk management purposes, one may be interested in describing how the 5 th percentile of the response variable is affected by regressor variables. Quantile regression models the quartiles of the dependent variable given a set of conditioning variables. Additionally, quantile regression does not require strong distributional assumptions about the functional form of the underlying distribution of the return. This methodology also allows for the easy introduction of state variables which can help avoid the omitted variable bias arising from failure to differentiate between systematic and idiosyncratic risk factors. Finally, even if the returns follow a fat tailed non normal distribution, estimates of VaR via quantile regression still possess all the desirable properties characteristic of a sample which is finite and normally distributed (see Koenker & Basset Jr.( 1978)). 12,13 The Quantile Regression Model Suppose there is a random variable X with a probability distribution function (6) So that for 0 1, the quantile of X may be defined as the smallest x satisfying : inf : (7) Given a set of n observations of X, the traditional empirical distribution function is given by: 1 (8) Where 1(z) is an indicator function that takes the value of 1 if the argument z is true and 0 otherwise. The associated empirical quantile is given by, inf : (9) or equivalently, in the form of a single optimization problem: 1 (10a) : : (10b) where 1 0 which weighs positive and negative values asymmetrically. 11 The paper also focuses on the lower tail of the distribution assuming that investors are taking long positions. This is a reasonable assumption given that the taking of short positions is quite rare and the derivatives market in Jamaica is still quite nascent. 12 That is, the methodology circumvents one of the chief criticisms of VaR, that if the return distribution is nonnormally distributed then the tail risk estimates will be ill described by the VaR measure. 13 Koenker and Basset (1978) show that quantile regression generates estimates that have comparable efficiency to least squares for Gaussian linear models while substantially out performing the least squares estimator over a wide class of non Gaussian error distributions. 8

9 Quantile regression extends this formulation to allow for regressors, denoted by the p vector M. A linear specification is assumed for the conditional quantile of the response variable X given values for M:, (11) where is the vector of coefficients associated with the th quantile. The analog to the unconditional quantile minimization above is the conditional quantile regression estimator: (12) The quantile regression estimator can be obtained as a solution to a linear programming problem. 14 Quantile Regression and Value at Risk Estimation The standard definition of VaR is the threshold value below which the historical market returns do not fall by more than some pre specified frequency or level of confidence,. That is, Pr VaR (13) The key contribution of Adrian and Brunnermeier (2009) is that, one can compute the VaR of the banking system either as an unconditional standard VaR, Pr VaR (14) or as a VaR conditional on the event that a specific bank has come under stress (i.e.. the bank s market value returns reaches its VaR level), which is dubbed the conditional VaR (Co VaR). Pr CoVaR (15) CoVar is, therefore, a measure of systemic risk, since it captures the spill over risk of one institution on the larger financial sector. When bank i contributes significantly to systemic risk, CoVaR would be very low, possibly a large negative number indicating a higher potential loss to the system with probability. The delta CoVaR,, captured by the difference between the CoVaR of an entity due to another institution being under stress and its unconditional stand alone VaR. This measure therefore captures the externality that the underlying bank imposes on other entities within the financial system (equation 16) as well as the externality the underlying bank imposes on the financial system (equation 17). (16) (17) 14 For more details see the Eviews 5 Manual. 9

10 The specification of bank I s asset return is given by (18) where is a vector of exogenous observed macro variables, and may be interpreted as the expected part of the asset s return. The set of state variables (M t ) used to estimate the time varying Co and VaR are: (i) JSE weekly returns (four lags) (ii) 30 day historical volatility of JSE weekly returns (four lags) (iii) monthly nominal growth in total gross domestic product (iv) monthly inflation (v) monthly loan to deposit spread (vi) monthly change in the 180 day treasury bill rate. The parameters and are estimated by quantile regression, and the fitted values are the measures for the stand alone VaRs. (19) (20) For the purposes of the study, the focus is on the 1 st quantile which corresponds to the 1.0 per cent VaR. Given these equations,. If one now includes in the information set at time t the as well as the, then the specification of the system s asset return can be written as: (21) whose fitted values evaluated at corresponds to the definition of CoVaR (22) 5.0 Empirical Findings 5.1 The Unconditional VaR of the Financial System and the Stand Alone VaRs for Financial Firms The unconditional 99.0 per cent weekly VaR of the financial system averaged 5.1 per cent (see Table 3). The largest weekly VaR, 10.6 per cent, was recorded on 27 February 2010 shortly after the end of the Jamaica Debt Exchange (JDX) transaction. This was complemented by a fairly high VaR of 8.2 per cent 10

11 recorded on January shortly after the commencement of the JDX. The year with the highest weekly annual average VaR of 5.7 per cent was 2004 which followed a year of significant macroeconomic volatility in This was followed by 2008 which recorded an annual average VaR of 5.6 per cent. Bank 1 recorded the lowest weekly average VaR of 8.5 per cent, while Bank 2 and Bank 4 recorded the highest stand alone VaRs of 16.5 per cent, each. Bank 5, in turn, recorded a stand alone VaR of 11.5 per cent, which was on average double that of the financial system s unconditional VaR. Further, the VaRs of individual financial institutions have trended upward over the period and in some cases have doubled particularly since With the exception of Bank 3, all institutions recorded an increase in their standalone VaRs (c.f. Table 3). Table 3. Average Weekly Stand Alone VaRs for Financial Firms and the Unconditional VaR of the Financial System Weekly Stand Alone VaRs (Per cent) Avg. Weekly Weekly Min. Max. VaR_Sys VaR_Bank VaR _Bank VaR _Bank VaR _Bank 4 NA VaR _Bank The Conditional VaR of the Financial System Bank 1 consistently recorded the lowest Co VaR indicating that it contributed least to the systemic risk exposure of the financial system even when it was under stress (see Table 4). In fact, in some periods Bank 1 contributed to an improvement in the overall risk exposure profile of the financial sector as indicated by an average positive Delta CoVaR (see Table 5). Similarly, Bank 5 also only contributed on average 0.1 of a percentage point to unconditional VaR of the financial system. In contrast, Banks 2 and 3 contributed to a 0.6 percentage point increase in the unconditional VaR of the financial system to an average of 5.7 per cent. In turn, Bank 3 added 0.5 of a percentage point on average to the downside risk of the financial system (c.f. Table 5). 11

12 Figure 3: Weekly Asset Returns and VaRs for Each of the Five Financial Institutions and the System 12

13 Table 4. Average Weekly Conditional VaRs by Financial Institution Weekly Conditional VaRs (Per cent) Avg. Weekly Weekly Min. Max. VaR_Sys CoVaR_ CoVaR_ CoVaR_ CoVaR_4 NA CoVaR_ Table 5. Average Weekly Delta Conditional VaRs by Financial Institution Weekly Delta Conditional VaRs (Per cent) Avg. Weekly Weekly Min. Max. DCoVaR_ DCoVaR_ DCoVaR_ DCoVaR_4 NA DCoVaR_ From the foregoing analysis Bank 2, Bank 3 and Bank 4 presented increased systemic risk exposure to the financial system over the review period. This becomes more apparent when observing each institution s average stand alone VaR, their percentage contribution to risk on the financial system (DCoVaR) and the incremental dollar VaR (I VaR) of their presence in the financial system (see Figure 4). These institutions not only have fairly high stand alone VaRs but also have contributed to increased downside risk of the financial system. Bank 2, Bank 3 and Bank 4 contributed an average of J$781.9 million, J$568.4 and J$604.4 million to the riskiness of the financial system, respectively, over a 10 day horizon. There has also been considerable variability in both the stand alone VaRs as well as the delta CoVaRs for the individual financial institutions over the period as indicated by wide spread between the minimum and maximum observations. For this reason we also evaluate the relationship between the aforementioned factors at the 10 th percentile i.e. during periods of heightened risk. In these cases, Banks 2, 3 and 4 contributed J$3.8 billion, J$2.7 billion and $2.6 billion to the riskiness of the financial system, respectively, over a 10 day horizon (see Figure 5) In terms of the average market value of assets of the financial system these exposures where 4.5 per cent, 3.3 per cent and 3.2 per cent for Banks 2, 3 and 4, respectively. 13

14 Figure 4. Delta Conditional VaRs, Stand Alone VaRs and Incremental Dollar VaR by Financial Institution (Average) Figure 5. Delta Conditional VaRs, Stand Alone VaRs and Incremental Dollar VaR by Financial Institution (10 th Percentile) 14

15 5.3 Understanding the Dynamics of Risk Transfers during Periods of Heightened Volatility In order to understand the potential transmission of risks between banks within the system is evaluated during a period of heightened level of stress in the domestic financial system. The period chosen for this assessment was between September 2008 and December 2008, the period following the intensification of the impact of the global financial crisis. 16 It is clear that during this period, Bank 1 exhibited the distinct characteristics of too big to fail levels (in terms of the capital (TBTF). That is, in times of financial stress, it is capable of transmitting significant adequacy of other banks) of risk to other institutions within the financial system. Specifically, during this period, Bank 4 and Bank 5 would be critically affected should the downside risks emanating from Bank 1 s activities to materialize. Alternately, Bank 3 exhibited too connected to fail (TCTF) characteristics during this same period. Figure 6. Risk transmission network map during a time of heightened stress in Jamaican financial markets (September 2008 Decemb ber 2008) Figure 4 A large risk transfer is one thatt exceeds 10.0 per cent of the impacted bank s regulatory capital. The size of each node is scaled in proportion to the total value of financial institution s asset base while the direction of the arrow indicates the direction of the more prominent risk transfer. The thickness of the line is proportional to the value of risk exposure as a proportion of the capital base of the impacted bank. Red lines represent risk transfers in excess of 30.0 per of capital, and black lines indicate risk transfers betweenn 10.0 per cent and 30.0 per cent of capital. Dotted black lines represent risk transfers whichh are less than 10.0 per cent of capital, but greater than 5.0 per cent. Risk transfers which are less than 5.0 per cent of capital are not shown. The top number represents dollar value of the delta CoVaon the bottom represents the delta CoVaR in the opposite of pointed institution conditional on the event that the institute at the origin of the arrow is in distress. The number direction. 16 The disrup ption in inter bank activity post September 2008 reflected itself in increased periods of non trading in the interbank market as well as sharp and persistent increases in inter bank rates. Correspondingly, the maximum inter bank rate for 2008 was 41.0 per cent compared to 18.0 per cent in

16 Heightened levels of risk of Bank 3 would lead to significant deteriorations in the down side risk of Bank 5 and considerable deterioration in the stability of Bank 2 as well as have some, albeit, much smaller impact on Bank 3 and Bank 4. This analysis also brings to the fore the consideration of contagion effects. The most deleterious critical path for the propagation of risks in the financial system is clearly from the Bank 1 to the Bank 5 which is also large and whose demise would have significant implications for the stability of the financial system. Finally, alternate scenarios are also possible. If several smaller firms are simultaneously affected, as is common during periods of financial distress when correlations rise, the net effect could lead to the demise of an institution which is deemed TBTF via the interbank bank market. For example, if banks 2, 3 and 4 simultaneously experience stress then this may be sufficient to lead to significant levels of risk transfers to Bank 5, which could be inimical to its survival (see Figure 6). 5.4 Understanding the Risk Transmission Mechanisms in the Jamaican Financial Sector Using Panel Regression Data The objective of the exercise to examine whether various balance sheet and P&L factors can explain the risk transmission mechanism, as captured by the measure, within the financial sector. These predictive regressions allow preemptive macro prudential policy and ex ante regulation of systemic risk contribution. The panel regression is of the form where i represents all other peer financial institutions which are not bank j, is the average monthly marginal risk contributions coming from peer institutions i and is the binary variable for each institution i. is the peer bank characteristics which include total loans (LOAN), interbank assets (INTER_A), interbank deposits (INTERDEP), interbank loans (INTERLOAN), retained earnings (RETEARN), deposit liabilities (STLIAB), leverage ratio (LR) and the market to book value ratio (MVBV). Total loans are used as a proxy for the size of the peer institution to which bank j is exposed. INTER_A, INTERDEP and INTERLOAN variables are intended to capture activity in the interbank market which is deemed, a priori, to be a critical transmission channel for risks within the banking system. On the other hand, RETEARN is intended to reflect the solvency and crisis absorption ability of various financial institutions. Deposit liabilities (STLIAB) are used as a proxy for each bank s liquidity position. The leverage ratio is intended to capture an entity s appetite for risk taking while the MVBV reflects the market s assessment of the future strength of the entity. Lagged values of and VAR i are included in the regressions so as to control for the persistence of systemic risk contribution. 16

17 Table 6: Fixed Effect Regression Results for Bank 1 Bank 5 17 Regressions of CoVaR of the Following Banks Bank Sheet Variables System Bank_1 Bank_2 Bank_3 Bank_4 Bank_5 Total loans (lagged) (0.94) (0.20) (0.03) Interbank loans (lagged) (4.11)*** (0.61) (0.56) 0.63 (3.67)*** (3.32)*** Interbank deposits (lagged) (1.75)* 1.08 (1.73)* 1.47 (1.52) 0.64 Retained Earnings (lagged) * (0.42) (1.36) Deposit Liabilities (lagged) (0.18) 0.64 (0.68) (0.04) Interbank assets (lagged) ** (1.24)** 1.73* (1.10) 3.47*** (0.52) CoVaR (lagged) ** (0.31) (0.66) (1.06) VaR i (lagged) * ** Leverage Ratio (lagged) (0.53) 0.86 Market to book (lagged) ** ** *** Constant (0.99) (0.30) (0.43) 0.07 (0.50) R Number of observations With regard to the results for the financial system the downside risk deteriorates as a result of the activities of entities which with higher leverage, higher levels of inter bank deposits and loans which is partially off set by the activity of entities with higher market to book ratios (see Table 6). These results are broadly consistent with apriori expectations as well as the findings of Brunnermeier and Adrian (2008) for internationally active banks. Turning to the regression results for the individual banks, it is worth noting that of each impacted bank is sensitive not only to different balance sheet variables but also to different degrees. For instance, Bank 1 s is primarily determined by the retained earnings of peer banks which is not the case for the other banks. The coefficient on retained earnings is positive and statistically significant which indicates that if peer banks are solvent and possess the ability to absorb shocks then they should help decrease the risk of Bank 1. This result is intuitive. The interbank channel was found to be statistically significant for Banks 4 and 5 via inter bank loans and for Bank 2 via inter bank deposits. That 17 A lag of 3 months was used in these panel regressions. 17

18 is, the more loans and deposits given to other peer banks the higher of the bank under examination. This result is consistent with a priori expectations which indicate that activity in the interbank market can be a significant driver of systemic risk within the financial system. Inter bank assets were found to be a statistically significant determinant for risk transfers for Bank 4 and Bank 2. The positive sign on the coefficient suggests that interacting with larger peer institutions in the interbank market helped to mitigate its own risks during times of financial stress. The size of the total loan of peer banks does not seem to be an important driver of risks for financial institutions Conclusion Assessing the level of contribution to systemic risk and the financial linkage of such risk between various financial institutions can serve as additional tool for bank supervisors to employ in determining policy regarding bank regulation, especially for institutions which are considered too big to fail or too connected to fail. In practice, this framework can promulgate a change of the supervisory and regulatory approach which will focus on encouraging systemically important institutions to internalize the externalities that their risk taking activities impose on the financial system. In this regard, the empirical findings of this paper highlight the critical importance of the timely amendment to the BOJ Act which will allow the central bank to monitor systemically important institutions. 18 Given that banks operate in an increasingly intricate network, it is highly likely that a large but plausible idiosyncratic shock may transmit across the network and impact other institutions and the system as a whole. It is in this context that this systemic risk must be measured and mitigated via enhanced macroprudential supervision. As such, a financial institution that generates more externalities via risk spillovers than other institutions would warrant closer prudential supervision (Roengpitya & Rungchareonkitkul (2010)). Frameworks, such as the one presented in this paper, will therefore play a critical role in framing financial sector regulation in the future, in such a way as to encourage such institutions to internalize these externalities via higher standards of risk management, disclosure of information as well as more stringent capital adequacy requirements. 18 See Appendix B. 18

19 Bibliography Adrian, T., & Brunnermeier K., M. (2009, August 1). CoVaR. Retrieved February 18, 2011, from Federal Reserve Bank of New York: Bank of Jamaica. (2008). Bank of Jamaica Annual Report. Kingston : Bank of Jamaica. Bank of Jamaica. (2010). Bank of Jamaica Annual Report. Kingston: Bank of Jamaica. Bank of Jamaica. (2011). Financial Stability Report Kingston: Bank of Jamaica. Brunnermeier A., M., Croket, A., Goodhart, C., Perssaud, A., & Shin, H. (2009). The Fundamental Principals of Financial Regulation: 11th Geneva Report of the World Economy. Geneva. Brunnermeier K., M., & Adrian, T. (2008, April 29). Febraban. Retrieved March 21, 2011, from Febraban: Jorion, P. (2007). Value at Risk: The New Benchmark for Managing Financial Risk. New York: McGraw Hill. Koenker, R., & Basset Jr., G. (1978, January). Regression Quantiles. Econometrica, 46(1), Pearson, D. N. (2002). Risk Budgeting: Portfolio Problem Solving with Value at Risk. In D. N. Pearson, Risk Budgeting (pp. 3 13). New York: John Wiley & Sons, Inc. Quantitative Micro Software, LLC. (2004). Eviews 5: User's Guide. Roengpitya, R., & Rungchareonkitkul, P. (2010, February 26). Measuring Systemic Risk and Financial Linkages in the Banking System. Retrieved February 18, 2011, from Bank of International Settlements (BIS): 19

20 Appendix A Figure 3: CoVaR Estimation Results 20

21 Appendix B Amendment to the Bank of Jamaica Act In December 2010 Cabinet approved the decision for the Bank of Jamaica to be assigned institutional responsibility for the stability of Jamaica s financial system. This decision is consistent with the response of most jurisdictions to locate this function within the respective central banks, following the recent global financial crisis. This reform also comprises a set of legislative reforms that underpin the current Stand-By Arrangement with the IMF. The amendments to the Act will: (i) outline the mandate of the Bank of Jamaica in relation to its role of maintaining financial system stability; (ii) mandate the establishment of a Financial System Stability Committee to coordinate the activities pursuant to the objective of financial system stability; expand the regulatory oversight of the Bank of Jamaica to financial institutions whose operations are deemed to be of systemic importance; (iv) grant the necessary powers to the Bank of Jamaica to obtain information from these financial institutions that will allow for the assessment of risks to the financial system (including the powers to initiate inspection and powers to demand information); (v) give the Bank of Jamaica the necessary powers to direct and impose measures to mitigate and control these risks (including the extension of liquidity; and powers to issue Prescriptive Rules, Standards and Codes pertinent to this oversight of the stability of the financial system). (vi) mandate the establishment of a Central Financial System database: and (vii) mandate the publication of a financial stability report within three (3) months after the end of each financial year. Excerpt taken from Bank of Jamaica Annual Report (2010) 21