This diploma thesis focuses on the development of the interbank market liquidity and interest rates in the Czech interbank market with special focus on the period of global crisis. We analyze determinants of the interbank interest rates and their development with respect to the key monetary policy rate. We explain the significant departure of the interbank interest rates from the key monetary policy rate (impairment of monetary policy transmission) during the global crisis by an increase in risk premia on interbank lending. The source of the risk premia is decomposed into the individual components such as liquidity risk, counterparty risk, foreign influence and other factors. Their contribution to the overall risk premia over time during the global crisis is analyzed. We find that the liquidity risk was the key determinant of tensions in the Czech interbank market in the beginning of the global crisis. However, its influence weakened over time while the role of counterparty risk increased. Keywords: interbank market, liquidity, interest rates, crisis, risk premia, credit risk, liquidity risk, counterparty risk JEL classification: G190, G210

Cette thèse étudie les conditions d’accès des banques européennes au financement interbancaire non sécurisé entre 2006 et 2012. Elle contient trois essais empiriques explorant des micro-données relatives aux transactions interbancaires. La première étude empirique adopte une approche en termes de paires banque prêteuse/banque emprunteuse et montre que, une fois le risque de contrepartie et les imperfections de marché contrôlées, les banques ayant un risque de liquidité plus élevé paient une prime de taux d’intérêt. Nous montrons également que cette prime est augmentée par les banques disposant d’excès de liquidités, sans doute motivées par la thésaurisation ou des stratégies de “short-squeezing” des banques en besoin de liquidité. Cette étude souligne finalement l’imperfection du marché interbancaire et l’importance des diverses interventions de la BCE qui ont cherché à réduire le risque de liquidité des banques au cours de la crise. La seconde étude, par le biais d’un model 2P-FRM, explore empiriquement l’impact des relations de clientèle entre banques sur la structure de maturité de la dette interbancaire. Les résultats dévoilent que l’accès aux prêts interbancaires longs et non sécurisés est facilité par les relations durables avant et durant les périodes de stress. Cependant, lors des moments aigus de la crise suivant la chute de la banque Lehman, ces effets positifs des variables bilatérales de relations fortes, calculées comme la concentration des actifs sur une banque emprunteuse, ne sont pas là. La deuxième partie de notre modèle montre que la part en volume des crédits à terme est plus faible pour les couples de banques partenaires. Finalement, notre variable unilatérale de relation interbancaire, qui mesure la concentration du réseau d’emprunt de la banque prêteuse, s’avère impacter négativement les prêts à terme post-Lehman. Cela confirme l’hypothèse que le propre risque de refinancement court du prêteur peut être l’origine du gel post-Lehman des prêts interbancaires à terme. Finalement, le troisième essai explore le lien entre la segmentation du marché interbancaire et le noeud de corrélation des risques souverains/bancaires. En utilisant les changements des primes des CDS souverains et bancaires, nous proposons une mesure originale de corrélation partielle des spillovers souverains-banques, qui permet d’attribuer une direction pays-banques à la contagion. Les résultats montrent que ces spillovers accentuent la segmentation du marché monétaire Italien lors de la phase critique de la crise des dettes souveraines. De plus, l’étude montre que, même si l’impact pays d’origine/banques est important, la contagion venant d’autres souverains en crise est loin d’être négligeable. / This thesis studies European banks’ terms to access to unsecured interbank funding during the period 2006 to 2012. It contains three empirical essays exploring micro-data on interbank transactions. The first empirical study adopts a bank pair panel approach evidencing that, once counterparty risk and other market imperfections are controlled for, banks with higher funding liquidity risk (liquidity-short banks) pay an interest rate premium. The bank pair level analysis also permits to show that this premium is charged by liquidity-long banks, probably motivated by strategic short-squeezing or prudential hoarding purposes during the crisis. This study emphasizes the imperfection of interbank markets and the importance of theECB’s emergency interventions dedicated to dampening banks’ funding risk concerns. The second essay explores empirically the impact of relationship lending on the interbank debt maturity structure of banks by mean of a two-part fractional response model. The findings show that durable bilateral liquidity partnerships can positively impact the probability of contracting term loans before and during periods of acute stress. The positive effects of the bilateral relationship lending variable measured as asset-side concentration, is however, not straightforward, especially after the Lehman default. The second part of our model shows that the post-Lehman maturity shift is pronounced for partner banks. Finally, we find that our unilateral (lender level) relationship variable impacts negatively long term lending confirming the rollover risk viewpoint of the term interbank market freeze. Finally, the third essay investigates the link between interbank market segmentation and bank–sovereign risk nexus. Using bank and country CDS spread changes it suggests an original partial correlation based measurement of sovereign/bank spillovers providing us with a direction of contagion. Empirical findings from this part of the thesis evidence that bank-sovereign risk correlation is a significant source of fragmentation during the most acute phase of the sovereign debt crisis. Moreover, the study shows that, even if home country/bank ties impact seriously interbank market integration, the risk from other distressed countries is far from negligible.

Thesis (Ph. D.)--Monash University, 2005. / Includes bibliographical references (leaves 205-218).

Chapter 1 builds a two-bank bargaining model of the overnight interbank market in which, due to the commitment of the central bank to its interest rate target, bargaining between banks impacts loan sizes rather than interest rates (the converse of existing models). As a result, policy changes have a different impact to what is posited by existing models. The model is applied to a market where the commitment of the central bank is well documented (Australia). With reasonable parameter values, the model replicates five stylised facts of the Australian market. Moreover, the stylised facts are replicated without recourse to any asymmetries. Chapter 2 extends the two-bank model to incorporate a large number of heterogeneous banks. This model is able to replicate the asymmetric shape of banks' end-of-day central bank deposit distributions (despite symmetric initial distributions); a novel contribution to the literature. Moreover, after inputting recent changes in Australian central bank policy, this model produces percentage changes in interbank trading volumes that closely align with the data. Central banks typically supply more overnight deposits than banks desire to hold (in aggregate), but this aggregate is typically small relative to interbank lending. With commitment, this is not required for the central bank to achieve its interest rate target (the typical explanation in the literature). So, to explain this phenomenon, Chapter 3 builds a DSGE model that incorporates commitment and the results from the previous chapters. Due to asymmetric information, there may be stigma associated with borrowing from the central bank's overnight lending facility, which is costly. But while the central bank can reduce use of its lending facility, by increasing aggregate deposits, the resulting fall in interbank lending is also costly; because the interbank market helps banks monitor their counterparties. Therefore, low but positive aggregate deposits can be explained as the welfare-optimising point in the trade-off between stigma and monitoring costs (a novel contribution to the literature).

This thesis studies the emergence of financial exposures between banks and introduces a novel game of financial network formation. It shows empirically that governance structures influence how banks use the interbank market to manage liquidity and that strategic factors are additional drivers of interbank lending for private banks (Ch. 2). It further develops a model of optimal bank behaviour in the absence of liquidity shocks considering the effect of an exogenous bailout probability (Ch. 3), and introduces a model of endogenous liquidity co-insurance formation (Ch. 4). The first chapter, The Purpose of Interbank Markets, tests competing theories of interbank lending using 43 quarters (2002-2012) of confidential data on the German banking sector and interbank market. It shows that banks use the interbank market for liquidity co-insurance as traditionally assumed. However, the importance of the liquidity management function is higher for regionally-focused credit cooperatives and savings banks than for private commercial banks. A distinct effect for private banks is identified; for private banks, increases in interbank liabilities are shown to correlate with a proxy for the bailout probability of banks. The chapter thus offers empirical support for an emerging literature on strategic behaviour in interbank markets and highlights the need to extend the traditional model of liquidity co-insurance. The second chapter, The Emergence of Interbank Exposures, develops a model showing that, even in the hypothetical absence of liquidity shocks, under some conditions the presence of conditional liability guarantees can lead to interbank exposures as an equilibrium outcome. It shows that such an equilibrium is characterised by banks of different sizes and asymmetric bank behaviour. Some banks are active only as lenders with others investing in a productive technology while borrowing in the interbank market. An equilibrium interbank rate is derived which depends on parameters characterising the bailout probability, including different parameters of government behaviour. The third chapter, Coordination and Competition in the Formation of Financial Networks, introduces a generalisation and extension of the seminal work of Allen and Gale (2000). It studies liquidity co-insurance between deposit-taking banks in an n-region economy. Both a static and a dynamic model of the endogenous formation of interbank liquidity co-insurance links are examined. Using a novel approach to model liquidity co-insurance, it is shown that contrary to previous findings it is not possible for banks with limited information to insure optimally against liquidity shocks. However, in a dynamic formulation of the model with best-response dynamics and learning, socially optimal insurance is an evolutionary stable equilibrium. The chapter also studies an extension to the model that introduces non-zero bailout probabilities, which endogenously leads to interbank networks consistent with the structure of interbank exposure networks documented empirically.

This thesis examines both conventional and unconventional monetary policies in a DSGE model with an interbank market friction. The recent crisis during 2007-2009 affected economies worldwide and forced central banks to implement not just conventional monetary policies, but also direct interventions in financial markets. We investigate a DSGE model with financial frictions, to test conventional and unconventional monetary policies. The thesis starts by using the Gertler and Kiyotaki (2010)'s modelling framework, to examine eight different shocks under imperfect interbank market conditions. Unlike Gertler and Kiyotaki (2010) who consider the two extreme cases for the banking system, I firstly extend the analysis to a case in between the two extreme cases that they examined. The shocks considered include supply and demand shocks and also two shocks from the financial system itself (an interbank market shock and a shock to the deposit market). It is found that a negative shock to the interbank market has only a moderate impact to the banking system. However, a shock to the deposit market has a much stronger impact. Even though the impacts of these shocks are not large it is shown that thefinancial frictions magnify the effects of other shocks. The model is extended to include price stickiness. A modified Taylor rule is analysed to test how conventional monetary policy should respond to the shocks in the presence of financial frictions. Specifically the credit spread is added as a third term in the monetary policy rule. The stabilising properties of the policy rule are analysed and a welfare analysis is conducted. The model is further developed to include unconventional monetary policy in the form of direct lending to private sector firms from the central bank. A policy rule for unconventional policy is tested and its stabilising and welfare properties are analysed.

Instantaneous rate models, although theoretically satisfying, are less so in practice. Instantaneous rates are not observable and calibra- tion to market data is complicated. Hence, the need for a market model where one models LIBOR rates seems imperative. In this modeling process, we aim at regaining the Black-76 formula[7] for pricing caps and °oors since these are the ones used in the market. To regain the Black-76 formula we have to model the LIBOR rates as log-normal processes. The whole construction method means calibration by using market data for caps, °oors and swaptions is straightforward. Brace, Gatarek and Musiela[8] and, Miltersen, Sandmann and Sondermann[25] showed that it is possible to con- struct an arbitrage-free interest rate model in which the LIBOR rates follow a log-normal process leading to Black-type pricing for- mulae for caps and °oors. The key to their approach is to start directly with modeling observed market rates, LIBOR rates in this case, instead of instantaneous spot rates or forward rates. There- after, the market models, which are consistent and arbitrage-free[6], [22], [8], can be used to price more exotic instruments. This model is known as the LIBOR Market Model. In a similar fashion, Jamshidian[22] (1998) showed how to con- struct an arbitrage-free interest rate model that yields Black-type pricing formulae for a certain set of swaptions. In this particular case, one starts with modeling forward swap rates as log-normal processes. This model is known as the Swap Market Model. Some of the advantages of market models as compared to other traditional models are that market models imply pricing formulae for caplets, °oorlets or swaptions that correspond to market practice. Consequently, calibration of such models is relatively simple[8]. The plan of this work is as follows. Firstly, we present an em- pirical analysis of the standard risk-neutral valuation approach, the forward risk-adjusted valuation approach, and elaborate the pro- cess of computing the forward risk-adjusted measure. Secondly, we present the formulation of the LIBOR and Swap market models based on a ¯nite number of bond prices[6], [8]. The technique used will enable us to formulate and name a new model for the South African market, the SAFEX-JIBAR model. In [5], a new approach for the estimation of the volatility of the instantaneous short interest rate was proposed. A relationship between observed LIBOR rates and certain unobserved instantaneous forward rates was established. Since data are observed discretely in time, the stochastic dynamics for these rates were determined un- der the corresponding risk-neutral measure and a ¯ltering estimation algorithm for the time-discretised interest rate dynamics was pro- posed. Thirdly, the SAFEX-JIBAR market model is formulated based on the assumption that the forward JIBAR rates follow a log-normal process. Formulae of the Black-type are deduced and applied to the pricing of a Rand Merchant Bank cap/°oor. In addition, the corre- sponding formulae for the Greeks are deduced. The JIBAR is then compared to other well known models by numerical results. Lastly, we perform some computational analysis in the following manner. We generate bond and caplet prices using Hull's [19] stan- dard market model and calibrate the LIBOR model to the cap curve, i.e determine the implied volatilities ¾i's which can then be used to assess the volatility most appropriate for pricing the instrument under consideration. Having done that, we calibrate the Ho-Lee model to the bond curve obtained by our standard market model. We numerically compute caplet prices using the Black-76 formula for caplets and compare these prices to the ones obtained using the standard market model. Finally we compute and compare swaption prices obtained by our standard market model and by the LIBOR model. / Economics / D.Phil. (Operations Research)

Management of systemic risk in financial markets is traditionally associated with setting (higher) capital requirements for market participants. There are indications that while equity ratios have been increased massively since the financial crisis, systemic risk levels might not have lowered, but even increased (see ECB data 1 ; SRISK time series 2 ). It has been shown that systemic risk is to a large extent related to the underlying network topology of financial exposures. A natural question arising is how much systemic risk can be eliminated by optimally rearranging these networks and without increasing capital requirements. Overlapping portfolios with minimized systemic risk which provide the same market functionality as empir- ical ones have been studied by Pichler et al. (2018). Here we propose a similar method for direct exposure networks, and apply it to cross-sectional interbank loan networks, consisting of 10 quarterly observations of the Austrian interbank market. We show that the suggested framework rearranges the network topol- ogy, such that systemic risk is reduced by a factor of approximately 3.5, and leaves the relevant economic features of the optimized network and its agents unchanged. The presented optimization procedure is not intended to actually re-configure interbank markets, but to demonstrate the huge potential for systemic risk management through rearranging exposure networks, in contrast to increasing capital requirements that were shown to have only marginal effects on systemic risk (Poledna et al., 2017). Ways to actually incentivize a self-organized formation toward optimal network configurations were introduced in Thurner and Poledna (2013) and Poledna and Thurner (2016). For regulatory policies concerning financial market stability the knowledge of minimal systemic risk for a given economic environment can serve as a benchmark for monitoring actual systemic risk in markets.

Instantaneous rate models, although theoretically satisfying, are less so in practice. Instantaneous rates are not observable and calibra- tion to market data is complicated. Hence, the need for a market model where one models LIBOR rates seems imperative. In this modeling process, we aim at regaining the Black-76 formula[7] for pricing caps and °oors since these are the ones used in the market. To regain the Black-76 formula we have to model the LIBOR rates as log-normal processes. The whole construction method means calibration by using market data for caps, °oors and swaptions is straightforward. Brace, Gatarek and Musiela[8] and, Miltersen, Sandmann and Sondermann[25] showed that it is possible to con- struct an arbitrage-free interest rate model in which the LIBOR rates follow a log-normal process leading to Black-type pricing for- mulae for caps and °oors. The key to their approach is to start directly with modeling observed market rates, LIBOR rates in this case, instead of instantaneous spot rates or forward rates. There- after, the market models, which are consistent and arbitrage-free[6], [22], [8], can be used to price more exotic instruments. This model is known as the LIBOR Market Model. In a similar fashion, Jamshidian[22] (1998) showed how to con- struct an arbitrage-free interest rate model that yields Black-type pricing formulae for a certain set of swaptions. In this particular case, one starts with modeling forward swap rates as log-normal processes. This model is known as the Swap Market Model. Some of the advantages of market models as compared to other traditional models are that market models imply pricing formulae for caplets, °oorlets or swaptions that correspond to market practice. Consequently, calibration of such models is relatively simple[8]. The plan of this work is as follows. Firstly, we present an em- pirical analysis of the standard risk-neutral valuation approach, the forward risk-adjusted valuation approach, and elaborate the pro- cess of computing the forward risk-adjusted measure. Secondly, we present the formulation of the LIBOR and Swap market models based on a ¯nite number of bond prices[6], [8]. The technique used will enable us to formulate and name a new model for the South African market, the SAFEX-JIBAR model. In [5], a new approach for the estimation of the volatility of the instantaneous short interest rate was proposed. A relationship between observed LIBOR rates and certain unobserved instantaneous forward rates was established. Since data are observed discretely in time, the stochastic dynamics for these rates were determined un- der the corresponding risk-neutral measure and a ¯ltering estimation algorithm for the time-discretised interest rate dynamics was pro- posed. Thirdly, the SAFEX-JIBAR market model is formulated based on the assumption that the forward JIBAR rates follow a log-normal process. Formulae of the Black-type are deduced and applied to the pricing of a Rand Merchant Bank cap/°oor. In addition, the corre- sponding formulae for the Greeks are deduced. The JIBAR is then compared to other well known models by numerical results. Lastly, we perform some computational analysis in the following manner. We generate bond and caplet prices using Hull's [19] stan- dard market model and calibrate the LIBOR model to the cap curve, i.e determine the implied volatilities ¾i's which can then be used to assess the volatility most appropriate for pricing the instrument under consideration. Having done that, we calibrate the Ho-Lee model to the bond curve obtained by our standard market model. We numerically compute caplet prices using the Black-76 formula for caplets and compare these prices to the ones obtained using the standard market model. Finally we compute and compare swaption prices obtained by our standard market model and by the LIBOR model. / Economics / D.Phil. (Operations Research)

We propose a statistical model for weighted temporal networks capable of measuring the level of heterogeneity in a financial system. Our model focuses on the level of diversification of financial institutions; that is, whether they are more inclined to distribute their assets equally among partners, or if they rather concentrate their commitment towards a limited number of institutions. Crucially, a Markov property is introduced to capture time dependencies and to make our measures comparable across time. We apply the model on an original dataset of Austrian interbank exposures. The temporal span encompasses the onset and development of the financial crisis in 2008 as well as the beginnings of European sovereign debt crisis in 2011. Our analysis highlights an overall increasing trend for network homogeneity, whereby core banks have a tendency to distribute their market exposures more equally across their partners.