Advanced search
1 file | 2.39 MB Add to list

Liquidity provision in the interbank foreign exchange market

(2013)
Author
Promoter
(UGent)
Organization
Abstract
Market liquidity captures how easy it is to convert an asset into cash and is a key-variable of interest when trading on financial markets and when investigating them. Moreover, liquidity also determines the speed at which information about an asset can be processed and it affects as well the asset’s expected return. From a policy perspective, liquidity is an important factor for the stability of the global financial system. In this thesis, we study market liquidity by looking at the interaction amongst different types of participants on the Hungarian forint/ euro interbank foreign exchange market. In the first chapter we start from a very general level – in an international finance framework – by surveying the literature on exchange rate policy in Central and Eastern European Countries (CEEC’s). In 2004, a first wave of CEEC’s joined the European Union. As a result, these countries all have the common long-term goal of joining European Monetary Union. Joining the monetary union is, however, conditional on the realization of the Maastricht criteria, and these criteria include stability of the exchange rate inside the European Exchange Rate Mechanism (ERM II). Despite their common goal, the CEEC’s opted for different exchange rate policies. Furthermore, their exchange rate policies were subject to frequent changes and adjustments. In this chapter, we describe the official exchange rate arrangements in the CEEC’s, but we also consider the difference between de jure and de facto exchange rate regimes. Next, we survey the literature on exchange rate volatility and the link with exchange rate policy and monetary policy. Therefore, we consider switches between volatility regimes. A big difference between the timing of these switches and the dates of the respective policy changes may hint at a lack of credibility of the policy, including the unpeaceful exits from the pegs in the Czech Republic and Slovakia. Finally, we survey the literature on the influence of monetary authorities on the exchange rate. Here we look, amongst other things, to central bank intervention and central bank communication in CEEC’s. From the next chapter onwards we switch on the microscope, and look at the market microstructure of the interbank foreign exchange market. Throughout these chapters we use detailed data for the Hungarian forint/ euro market – which operates as an electronic limit order book – in 2003 and 2004. In the second chapter we investigate the link between news announcements, jumps (which are basically price discontinuities) and market liquidity. In a first stage we detect the intraday jumps, and show that they are prevalent and important: there is at least one price jump on 18.20% of the trading days contained in our sample period, and 42.59% of the price variation on these jump days can be attributed to the jumps. We also find that positive and negative jumps are symmetric in terms of both frequency and size. In a second stage, we try to link the intraday jumps with public news announcements. Here we consider both scheduled public news (e.g. GDP, PPI, trade balance information,…) and unscheduled public news (e.g. central bank interventions, polls, surveys, political changes,…). They can be respectively linked with 16% of the jumps and 30.4% of the jumps, which implies that more than half of the jumps cannot be explained by public information. Hence, we would like to take a closer look at the actual genesis of jumps: are they caused by (public or private) information inflow, noise trades or insufficient liquidity? We therefore study in a third stage the dynamics of liquidity in a two-hour window around the jumps. We look at liquidity as a multi-dimensional variable and distinguish the tightness dimension (the difference between the best bid and the best ask), the immediacy dimension (the amount of euro or forint traded), resiliency (the pace at which the price reverts to former levels after it changed in response to large order flow imbalances), the overall depth (the amount of euro or forint available in the limit order book) and the depth at the best quotes. As a result, we find that jumps do not happen when liquidity is unusually low, but rather when there is an unusually high demand for immediacy concentrated on one side of the order book. Moreover, this result is independent of whether the jump can be linked to a public news announcement or not, and our findings suggest that it is information inflow that causes the jump. Moreover, a dynamic order placement process emerges after a jump: more limit sell (buy) orders are added to the book subsequent to a positive (negative) jump. We attribute this to endogenous liquidity providers on the market. Attracted by the higher reward for providing liquidity, they submit limit orders at the side where it is needed the most. In a fourth and last stage, we provide some further analyses and apply a probit model that shows that none of the liquidity variables offers predictive power for a jump occurrence (consistent with what we find for the dynamics of liquidity around jumps) or for the magnitude of the jump. In addition, we find that more limit orders relative to market orders are submitted to the book after the jump, and that the post-jump order flow is in general less informative than in normal trading periods. Overall, our results provide insight into the origin of jumps and map the impact of endogenous liquidity provision on this market without designated market makers. In the last two chapters, we zoom in on the process of endogenous liquidity provision. We focus in these chapters on the link between the tightness dimension/ bid-ask spread and the cost of providing liquidity. We distinguish respectively order processing costs (the operational costs of providing market making services, such as wages of traders, floor space rent, fees that have to be paid to the platforms,…), inventory holding costs (the cost of holding an unwanted inventory, which results from accommodating incoming orders) and adverse selection costs (the cost of engaging in a transaction with a market participant who has superior information). In the third chapter we provide evidence using an established, structural model that allows us to split up the spread into these different cost components. We find that over the two years, 40.09% of the bid-ask spread can be explained by inventory holding costs, 38.34% can be explained by order processing costs and 21.57% can be explained by adverse selection costs. Our results differ in some ways from previous results for the foreign exchange market where the same methodology was used, and are to some extent more intuitive. In comparison with the existing studies, the tier of the market we analyze, the completeness of the data, the size of the market and institutional differences between markets seem to play an important role. Furthermore, we find that the estimated spread on large trades is over the whole dataset 32.35% higher than the spread on small trades. We show that this higher spread is caused by a higher combined inventory holding and adverse selection cost. In the fourth chapter, we follow a novel direction. Here we study the bid-ask spreads using an empirical spread decomposition model and specify the individual spread components explicitly. The combined inventory holding and adverse selection cost is here modeled as an option premium. This is very intuitive, and has the advantage that the risk can be quantified using option valuation techniques. We provide the first complete forex results for this type of model, and show that the combined component accounts for 52.52% of the bid-ask spread. Furthermore, we provide evidence for an endogenous tick size of 0.05 HUF/ EUR and we also estimate the number of liquidity providers based on the results for the risk component. In addition, the empirical approach we follow in this chapter allows us to examine two interesting spread patterns: the stylized difference in spreads between peak-times and non-peak times and the spread pattern around a speculative attack against the Hungarian forint in the beginning of 2003. First, we confirm the stylized difference in spreads between peak-times and non-peak times. As a matter of fact, during non-peak times the spread is more than double as high as during peak-times. We find that this is caused by an increase in the risk component, and if we elaborate on the origin of it we show that it is not only the calculated option premium that increases but also the sensitivity to this option premium. Clearly, the increase in the premium still underestimates the actual increase in risk for the liquidity provider. We explain this by the increased probability that the liquidity provider will have to keep his position overnight. Second, we map the spread pattern around the speculative attack. Prior to the attack, the spread decreases until it reaches a level below the endogenous tick size. This decrease is caused by a strong decrease in the risk component. During the speculative attack, the spread increases massively, as a result of the rising risk component. The order processing component, on the other hand, decreases at the same time. This pattern is consistent with increased competition amongst liquidity providers who are well aware of the increased risk that their activity during this period of high speculation involves. After the attack, both the order processing component and risk component increase. Consequently, the tightness of this market is much lower than before the attack. Overall, this chapter demonstrates the relevance of an option based decomposition approach for understanding how liquidity is provided on the interbank foreign exchange market.
Keywords
Forex, Inventory, Exchange Rate, Jumps, Hungary, Liquidity, News Announcements, Spreads, Microstructure, Adverse Selection

Downloads

  • PhD Frederick Van Gysegem Inside.pdf
    • full text
    • |
    • open access
    • |
    • PDF
    • |
    • 2.39 MB

Citation

Please use this url to cite or link to this publication:

MLA
Van Gysegem, Frederick. “Liquidity Provision in the Interbank Foreign Exchange Market.” 2013 : n. pag. Print.
APA
Van Gysegem, F. (2013). Liquidity provision in the interbank foreign exchange market. University Press, Ghent.
Chicago author-date
Van Gysegem, Frederick. 2013. “Liquidity Provision in the Interbank Foreign Exchange Market”. Ghent: University Press.
Chicago author-date (all authors)
Van Gysegem, Frederick. 2013. “Liquidity Provision in the Interbank Foreign Exchange Market”. Ghent: University Press.
Vancouver
1.
Van Gysegem F. Liquidity provision in the interbank foreign exchange market. [Ghent]: University Press; 2013.
IEEE
[1]
F. Van Gysegem, “Liquidity provision in the interbank foreign exchange market,” University Press, Ghent, 2013.
@phdthesis{4197286,
  abstract     = {Market liquidity captures how easy it is to convert an asset into cash and is a key-variable of interest when trading on financial markets and when investigating them. Moreover, liquidity also determines the speed at which information about an asset can be processed and it affects as well the asset’s expected return. From a policy perspective, liquidity is an important factor for the stability of the global financial system. In this thesis, we study market liquidity by looking at the interaction amongst different types of participants on the Hungarian forint/ euro interbank foreign exchange market. In the first chapter we start from a very general level – in an international finance framework – by surveying the literature on exchange rate policy in Central and Eastern European Countries (CEEC’s). In 2004, a first wave of CEEC’s joined the European Union. As a result, these countries all have the common long-term goal of joining European Monetary Union. Joining the monetary union is, however, conditional on the realization of the Maastricht criteria, and these criteria include stability of the exchange rate inside the European Exchange Rate Mechanism (ERM II). Despite their common goal, the CEEC’s opted for different exchange rate policies. Furthermore, their exchange rate policies were subject to frequent changes and adjustments. In this chapter, we describe the official exchange rate arrangements in the CEEC’s, but we also consider the difference between de jure and de facto exchange rate regimes. Next, we survey the literature on exchange rate volatility and the link with exchange rate policy and monetary policy. Therefore, we consider switches between volatility regimes. A big difference between the timing of these switches and the dates of the respective policy changes may hint at a lack of credibility of the policy, including the unpeaceful exits from the pegs in the Czech Republic and Slovakia. Finally, we survey the literature on the influence of monetary authorities on the exchange rate. Here we look, amongst other things, to central bank intervention and central bank communication in CEEC’s.  From the next chapter onwards we switch on the microscope, and look at the market microstructure of the interbank foreign exchange market. Throughout these chapters we use detailed data for the Hungarian forint/ euro market – which operates as an electronic limit order book – in 2003 and 2004. In the second chapter we investigate the link between news announcements, jumps (which are basically price discontinuities) and market liquidity. In a first stage we detect the intraday jumps, and show that they are prevalent and important: there is at least one price jump on 18.20% of the trading days contained in our sample period, and 42.59% of the price variation on these jump days can be attributed to the jumps. We also find that positive and negative jumps are symmetric in terms of both frequency and size. In a second stage, we try to link the intraday jumps with public news announcements. Here we consider both scheduled public news (e.g. GDP, PPI, trade balance information,…) and unscheduled public news (e.g. central bank interventions, polls, surveys, political changes,…). They can be respectively linked with 16% of the jumps and 30.4% of the jumps, which implies that more than half of the jumps cannot be explained by public information. Hence, we would like to take a closer look at the actual genesis of jumps: are they caused by (public or private) information inflow, noise trades or insufficient liquidity? We therefore study in a third stage the dynamics of liquidity in a two-hour window around the jumps. We look at liquidity as a multi-dimensional variable and distinguish the tightness dimension (the difference between the best bid and the best ask), the immediacy dimension (the amount of euro or forint traded), resiliency (the pace at which the price reverts to former levels after it changed in response to large order flow imbalances), the overall depth (the amount of euro or forint available in the limit order book) and the depth at the best quotes. As a result, we find that jumps do not happen when liquidity is unusually low, but rather when there is an unusually high demand for immediacy concentrated on one side of the order book. Moreover, this result is independent of whether the jump can be linked to a public news announcement or not, and our findings suggest that it is information inflow that causes the jump. Moreover, a dynamic order placement process emerges after a jump: more limit sell (buy) orders are added to the book subsequent to a positive (negative) jump. We attribute this to endogenous liquidity providers on the market. Attracted by the higher reward for providing liquidity, they submit limit orders at the side where it is needed the most. In a fourth and last stage, we provide some further analyses and apply a probit model that shows that none of the liquidity variables offers predictive power for a jump occurrence (consistent with what we find for the dynamics of liquidity around jumps) or for the magnitude of the jump. In addition, we find that more limit orders relative to market orders are submitted to the book after the jump, and that the post-jump order flow is in general less informative than in normal trading periods. Overall, our results provide insight into the origin of jumps and map the impact of endogenous liquidity provision on this market without designated market makers. In the last two chapters, we zoom in on the process of endogenous liquidity provision. We focus in these chapters on the link between the tightness dimension/ bid-ask spread and the cost of providing liquidity. We distinguish respectively order processing costs (the operational costs of providing market making services, such as wages of traders, floor space rent, fees that have to be paid to the platforms,…), inventory holding costs (the cost of holding an unwanted inventory, which results from accommodating incoming orders) and adverse selection costs (the cost of engaging in a transaction with a market participant who has superior information). In the third chapter we provide evidence using an established, structural model that allows us to split up the spread into these different cost components. We find that over the two years, 40.09% of the bid-ask spread can be explained by inventory holding costs, 38.34% can be explained by order processing costs and 21.57% can be explained by adverse selection costs. Our results differ in some ways from previous results for the foreign exchange market where the same methodology was used, and are to some extent more intuitive. In comparison with the existing studies, the tier of the market we analyze, the completeness of the data, the size of the market and institutional differences between markets seem to play an important role. Furthermore, we find that the estimated spread on large trades is over the whole dataset 32.35% higher than the spread on small trades. We show that this higher spread is caused by a higher combined inventory holding and adverse selection cost. In the fourth chapter, we follow a novel direction. Here we study the bid-ask spreads using an empirical spread decomposition model and specify the individual spread components explicitly. The combined inventory holding and adverse selection cost is here modeled as an option premium. This is very intuitive, and has the advantage that the risk can be quantified using option valuation techniques. We provide the first complete forex results for this type of model, and show that the combined component accounts for 52.52% of the bid-ask spread. Furthermore, we provide evidence for an endogenous tick size of 0.05 HUF/ EUR and we also estimate the number of liquidity providers based on the results for the risk component. In addition, the empirical approach we follow in this chapter allows us to examine two interesting spread patterns: the stylized difference in spreads between peak-times and non-peak times and the spread pattern around a speculative attack against the Hungarian forint in the beginning of 2003. First, we confirm the stylized difference in spreads between peak-times and non-peak times. As a matter of fact, during non-peak times the spread is more than double as high as during peak-times. We find that this is caused by an increase in the risk component, and if we elaborate on the origin of it we show that it is not only the calculated option premium that increases but also the sensitivity to this option premium. Clearly, the increase in the premium still underestimates the actual increase in risk for the liquidity provider. We explain this by the increased probability that the liquidity provider will have to keep his position overnight. Second, we map the spread pattern around the speculative attack. Prior to the attack, the spread decreases until it reaches a level below the endogenous tick size. This decrease is caused by a strong decrease in the risk component. During the speculative attack, the spread increases massively, as a result of the rising risk component. The order processing component, on the other hand, decreases at the same time. This pattern is consistent with increased competition amongst liquidity providers who are well aware of the increased risk that their activity during this period of high speculation involves. After the attack, both the order processing component and risk component increase. Consequently, the tightness of this market is much lower than before the attack. Overall, this chapter demonstrates the relevance of an option based decomposition approach for understanding how liquidity is provided on the interbank foreign exchange market.},
  author       = {Van Gysegem, Frederick},
  keywords     = {Forex,Inventory,Exchange Rate,Jumps,Hungary,Liquidity,News Announcements,Spreads,Microstructure,Adverse Selection},
  language     = {eng},
  pages        = {188},
  publisher    = {University Press},
  school       = {Ghent University},
  title        = {Liquidity provision in the interbank foreign exchange market},
  year         = {2013},
}