Central Bank Liquidity and Market Liquidity: The Role of Collateral Provision on the French Government Debt Securities Market

We examine the effects of collateral provision as a potential channel between funding liquidity tensions and the scarcity of market liquidity. This channel consists in transferring the credit risk associated with refinancing operations between financial institutions to market participants that bear new liquidity risk on the market associated with collateral. In particular, we address the issue of the liquidity of the French government debt securities market, since these assets are used as collateral both in the open market operations of the ECB and on the interbank market. We use a time-varying transition probability (TVTP) VAR model considering both the monetary policy cycle and the cycle of French treasury auctions. We highlight the existence of a specific regime in which monetary policy neutrality is not verified on the market for French bonds. Moreover, the existence of conventional and unconventional regimes leads to asymmetries in monetary policy implementation.

Regarding the refinancing process operated by the central bank, the collateral requirements are essential for several reasons. Firstly, they concern marketable assets usually marked to market on a daily basis. Indeed, the value of collateral is provided by the market so that it is not constant throughout the duration of the loan. This does not expose the central bank to credit risk, but to market risk. One way to circumvent this problem is the mechanism implemented known as variation margins: if the value of the collateral varies, banks should compensate for potential losses.
If market risk is taken into account, market liquidity becomes a key factor in determining the value of collateral (Manning and Willison, 2006). Market liquidity is the ability to obtain a fair price for an asset given that enough agents are participating in market. It has been widely The question is whether tensions in the refinancing process of the banking system may turn credit risk into market liquidity risk via the extensive use of some types of collateral. For example, Green (2005) shows that the assets eligible as collateral provide lower rates of return than those not eligible by incurring an opportunity cost to owners. A potential risk to underscore in this paper is that the excessive use of an asset as collateral that is marked to market may ultimately create market inefficiencies. Through this mechanism, the increased credit 3 risk in the funding process is no longer borne by the lender but by all market participants.
This transfer of risk may come from various sources.
First, the increased risk associated with interbank refinancing may create a concentration on some types of eligible collateral, i.e. those of highest quality such as government bonds.
Second, this higher counterparty risk increases the haircuts on the value of the collateral so that larger amounts of collateral become necessary for the loan. Third, increased refinancing via the central bank raises the amount of collateral required. Fourth, the stepping-up of special refinancing operations, illustrating the tensions in the banking sector, results in more frequent aggressive trading of collateral. Finally, the longer maturities of refinancing operations also require larger amounts of collateral to provide to the lender. For example, in 2006, deposit collateral amounted to EUR 959 billion, while in September 2008 it stood at EUR 1,585 billion.
It should be noted that this channel is identified by Brunnermeier and Pedersen (2009), who show that haircut spirals in the funding process may destabilize the entire financial system and impact the value of the collateral itself. It was a concern of the ECB monetary policy framework right from the start not to impact other financial markets, such as government bond markets, too much and to remain neutral. For example, the choice to adopt a broad approach for types of collateral instead of a narrow one, as in the case of US Federal Reserve, is understandable, as mentioned by Cheun et al. (2009), since in the US the ratio of temporary operations to the size of the domestic government bond market before the crisis was 1/200, compared to 1/10 for the Eurosystem, which would lead to substantial constraints on collateral and have an impact on government bond markets. In the case of the Bank of England, this ratio is 1/9, which constituted a strong incentive to expand eligibility to include all euro area government bonds (with the presence of a rating threshold). However, given the particular circumstances prevailing since the onset of the crisis in 2008, and even if the collateral approach of the ECB was already broad, we wonder if market neutrality can always be achieved in the monetary policy stance. Chakravarty and Saskar (1999) also compare the different bond segments in terms of bid-ask spreads. Moreover, Dunne et al. (2002) and Dunne et al. (2007) show that, contrary to the prevailing market belief, the 10-year segment of the French bond market is a benchmark asset for the European bond market as a whole. In addition, this market has clearly developed over the last ten years to become an important market with international investors. In our analysis, we focus on the French debt market, which is used both by banks as collateral in open market operations and on the interbank market. Our approach is different from the papers cited above since we focus particularly on bonds as collateral, liquidity at the transaction level and the associated market risk using high-frequency data.
To analyze the role of collateral rules, we use two different maturities for French government securities: the rate on three-month Treasuries and the rate of 10-year notes. The analysis is based on high-frequency data identifying all quotations for on-the-run three-month and The main results are as follows. First, the stepping-up of special refinancing operations with high bid-to-cover ratios make more probable the appearance of an unconventional regime 5 in which liquidity, volatility and market segmentation between bonds occur. This is in line with the potential excessive use of some types of collateral on account of funding tensions.
Second, regime identification shows the potential asymmetry in the monetary policy stance between conventional and unconventional regimes, whereby the same decision (for example more frequent OMOs and loose liquidity provision) may have positive or negative effects depending on the regime markets are in. As a consequence, a regime in which non-neutrality of the monetary policy stance vis-à-vis the market for collateral is observed leads to higher associated risks in the funding process.
The paper is organized as follows. In the following section, we review the concepts of market liquidity and central bank liquidity focusing on the linkages between collateral rules and market dynamics. We present the recent developments on the French sovereign bond These diverse liquidity concepts are very close to each other but the potential channels between them are not so well understood. Here we investigate the potential channel between these three liquidity concepts constituted by the rules on collateral.

Collateral provision and liquidity
The central bank provides liquidity to banks through several channels. The majority of these operations are main refinancing operations (MROs) with a weekly frequency. The central bank also uses long-term refinancing operations (LTROs), the more recent very long-term refinancing operations (VLTROs) and some other one-off fine-tuning operations (FTOs). A detailed description of this primary channel of central liquidity is discussed in Idier and Nardelli A second channel for providing liquidity is the use of standing facilities. Any bank may ask the central bank for refinancing at any time at a penalty rate. This penalty is such that few banks generally use these standing facilities. There is a "standing facility stigma" since banks are usually reluctant to use them: it is usually a signal of weakness to the market for the bank using the lending facility (even if it is theoretically confidential). However, the recent crisis has shown a clear increase in the deposit facility linked to an increase in risk aversion, and a reduction of the penalty associated with it.
All these operations managed by the central bank lead to some collateral immobilization, which we focus on here. To protect the ECB (and more generally any central bank) from losses due to open market operations, collateral is used to back the operations. In the event of credit failure, this collateral may be liquidated by the central bank to get its money back. The assets used as collateral must meet certain criteria to be eligible for the ECB in its refinancing operations (see ECB "The implementation of monetary policy in the euro area", November 2008). These rules were modified in early 2007 with the introduction of the single list of eligible assets. Notably, the ECB considers collateral eligibility for marketable and non-marketable assets.
Concerning marketable assets, euro-denominated debt instruments with high credit ratings traded on regulated markets 3 are eligible, provided that the issuer is an EU member or a G10 member. Non-marketable assets (credit claims and retail mortgage-backed debt instruments) must be issued by credit institutions located in the euro area, with high credit ratings and be denominated in euro. However, in the ECB's monetary policy framework, the collateral policy is generally restricted to marketable assets for outright transactions. 4 To ensure the quality of collateral the ECB uses two additional measures: the haircut and variation margins. An important aspect is that the value of the asset is marked to market so that the ECB is exposed to downward variations in the collateral's value during the loan period. As a consequence, counterparties (banks) must provide additional cash to maintain the value of the asset (margin-calls). Obviously, these variation margins are symmetrical and if the value of the asset rises above a certain level, the counterparty retrieves the corresponding cash.
The haircut is a percentage discount applied to the value of the collateral. The value of the collateral is thus calculated as the market value of the asset less the haircut applied to this category of assets.
The ECB has integrated this market dependency into its rules so that the level of the haircut is based on a liquidity criterion: the lower the liquidity, the higher the haircut. However, this criterion is quite rigid and may not completely hedge the ECB against market inefficiencies.
3 There are some exceptions in published lists of non-regulated markets accepted by the ECB. 4 Some of these criteria have been relaxed during the crisis episode but for a limited period of time.

8
There are five categories of liquidity for the assets used as collateral. The first category is considered to be the most liquid, with liquidity decreasing progressively across the four other categories. Table 1   In each of these categories, depending on the residual maturity and the coupon, the haircut varies from 0.5% to 20%. The best choice of collateral for participating to OMOs are thus the government debts instruments. They usually meet high credit standards, enjoy market liquidity and are traded on organized markets.
In addition, government bonds are also used as collateral in bilateral transactions on the interbank market to cover funding liquidity needs. Given the financial instability seen on the market during the last few months, the secured segment (on the short maturity) of the interbank market has served as a substitute for the unsecured segment due to the funding liquidity tensions. However, turnover in this market segment has slowed (a fall of 16% in comparison with 2004-2005 levels), due to the decline in the quality of the collateral usually used in bilateral repo transactions and the uncertainty related to counterparty risk. There is therefore also a clear incentive to ask for high quality collateral on the secured market.

The implications of the crisis for liquidity
We investigate the funding liquidity pressure impact on the liquidity and volatility of assets used as collateral. The 2008 crisis seriously undermined the interbank market. The ECB thus decided to provide huge amounts of liquidity to the banking system through regular and special OMOs.
With respect to regular open market operations, the ECB first increased the levels of allotments to meet liquidity needs through MROs and LTROs. Due to the high demand for liquidity, the ECB decided to use a fixed-rate tender with full allotment in order to completely satisfy bank liquidity needs. In a fixed-rate tender, the ECB gives the level of the rate applied to the MRO and banks are asked to give the corresponding amount of liquidity they are willing to obtain at this price. In this way, their needs are completely met given the level of the rate. This is in contrast to variable rate tenders in which banks provide ten rates and ten corresponding amounts of liquidity that they are willing to obtain with no guarantee of the final amount allotted. In this way, the ECB limits tensions on the interbank market, but provides large amounts of liquidity to the banking system, and this requires higher amounts of collateral. This has also been coupled with the stepping-up of special operations as shown in Figure 1. extended eligibility (with higher haircuts) to some other asset classes (asset backed securities, syndicated loans for a given period, Japanese, US and UK credit claims, for example).
All these tensions in the banking system, resulted in the interbank market in a reluctance for banks to lend to each other. However, when bilateral transactions occur, it seems that high quality collateral is required and French government bond securities belong to this category. This is illustrated in the OIS spread (see Figure 2), closely followed by market participants during the turmoil, and reflecting the credit risk associated with transactions on the interbank market.
Concerning bond market developments, there are some interesting issues. Typically, on the Due to governments' fiscal commitments aimed at tackling the crisis and a general rise in credit risk premia, risk aversion on bonds over the long term has increased. This is now reflected in the yields for long-term securities. However, this has not occurred for all government bonds within the euro area. On the one hand, bonds are suffering from a flight to quality phenomenon whereby investors shift trading to traditionally strong government debt securities (typically German or French ones). On the other hand, there is also a flight to liquidity issues with investors wishing to invest in liquid markets. In a period where refinancing is difficult on the interbank market, it is clear that banks are mitigating their risk by investing in markets where funds may be withdrawn rapidly. As a consequence, the liquidity of some bond markets has dried up (for example the Greek market) and trading has shifted to other bonds markets. This combination of flight to quality and flight to liquidity may have marked consequences for market efficiency.

Dataset and market indicators
Our dataset consists in high-frequency (on-the-run) quotes for French debt securities with 3-month and 10-year maturities from Reuters Data Tick History ranging from January 1st, 2003 to July 31st, 2009. We have around 3.5 millions quotes for the bonds in question. Some public holidays were removed from the sample due to the lack of trading (Christmas, New Year Eve, Easter, France's national holiday, etc.). Due to the greater dispersion of 3-month contracts, with more frequent adjudications, the number of quotes is lower than for the 10Y notes, for on-the-run contracts.

Liquidity indicators
The bid-ask spread reflects many factors (see Roll, 1984 . Under usual market conditions, the price is defined as the difference between the ask price and the bid price. We construct an average daily bid-ask spread for each rate. Since we are looking at the bid-ask spread for rates, which are inversely related to prices, the spread is defined as: for the i th transaction of day t. The daily liquidity indicator is then the mean over the day of all spreads: 6 Due to the partial information available on our dataset (e.g. no details about transaction volumes), we restrict our analysis to this standard indicator as suggested by Fleming (2003).

Volatility measures
Since Then, the day t volatility estimator is defined as: where r i and r i are subsequent returns for the considered subintervals of day t.  6 We also defined the daily spread as the median of spreads over a given day. This does not affect the results.
14 bond yields and computed on 15-minute-frequency returns.

Monetary policy and funding tension indicators
To investigate monetary policy, we construct several indicators representing the ECB's operational framework and tensions during the refinancing process. First of all, we consider a set of dummy variables for OMOs, and their type: MROs, LTROs or Other. We also use the allotted amount of OMOs and the bid-to-cover ratios of these operations. This bid-to-cover ratio is the supply-demand ratio for liquidity. It summarizes some of the tensions related to refinancing between banks. All these indicators give the intensity with which market participants are seeking refinancing from the central bank and how the collateral market, as a consequence, may be impacted by these operations.

Bond market and funding liquidity tensions
The amount of the negotiable debt for the French government almost doubled between 1998 and 2008, reaching EUR 988 billion at the end of September 2008. This upward trend was made possible by the introduction of marketable products grouped into three categories based on their initial maturities. The first category comprises the short-term bond class with maturities less than one year. In this category, three-month maturity bonds are typically issued weekly and respond to short-term financing needs. The second category includes bonds with two or five-year maturities with a new adjudication per month. The last category concerns long-term bonds with maturity from seven to 50 years with one adjudication per month.
After these regular pre-scheduled auctions, securities are actively traded on the secondary market, where transactions are not centralized. This secondary market is an over the counter market (OTC) and bilateral transaction details are partially known.
One main development in this market is its internationalization. An increasing share of the negotiable French debt is held by foreign investors : by the end of 1998, it represented 18.8% of negotiable debt compared with 62% by mid-2008. This internationalization could be a vector of increasing liquidity on the market with a wider pool of market participants. Figure 3 shows the daily changes in bond rates. While the short-term rate is anchored to changes in the minimum bid rate of the ECB, 10-year rates are more independent of the rise in interest rates that occured from the end of 2005. As a consequence, the bond spread shrank until spring 2008. From September 2008, the financial crisis and the ECB's decision to cut interest rates clearly increased this bond rate spread, with a huge drop in short-term maturity rates (see Figure 2). With respect to liquidity, Figure 4 presents the bid-ask spread for the two bonds between 2003 and July 2009. The short-term maturity bond is less liquid than the longer-term one over the sample. The average bid-ask spread for three-month maturity rates is about 3.7 bp, while it falls by 0.8 bp for the 10-year one. 7 Broadly on the sample, the bid-ask spread for short-term maturity fell except during the crisis of 2008 where it jumped twice in September and October. Moreover, volatility also surged during this period. 8 We note that the impact is stronger for the three-month maturity than for 10-year bonds. Indeed, volatility for long-term bonds rose but did not soar.
There are several things to investigate in this area. First, liquidity and volatility may not have the same interactions depending on the market segment. In particular, the monetary policy framework may impact these indicators, as we mentioned earlier, so that some markets may be more vulnerable than others, even if they belong to the same liquidity class of collateral as defined by the ECB. Let consider a VAR model with P lags [VAR(P)] as : where X t is the vector of variables of interest as yield variations (∆r 3,t , ∆r 120,t ), volatilities (σ 3,t , σ 120,t ) and liquidities (S 120,t , S 3,t ), and OMO t a dummy variable having the value of 1 on OMOs announcement days and zero otherwise.
However, at daily frequency, it is hard to consider a homogeneous linear model for market dynamics. Market dynamics are not usually linear, or even piece linear, but usually governed by several coexisting regimes. In line with Hamilton's (1994) findings, we use a Markov switching VAR to capture the dynamics of the variable of interest. 9 To limit the number of parameters to be estimated and to circumvent identification and estimation problems, since a VAR model is estimated for each regime, we limit our analysis to two states (S = 2). So, conditional on fX t g t=1:N , the history of past variables, we consider a two state model for s t = 1, 2 similar to Krolzig (1997) such that and with u t follows a Gaussian distribution with zero mean. Since the volatility of the rates is not constant over the sample (as presented in Figure 5), we also consider heteroskedasticity in this MSVAR through the Σ(s t ) state dependent variance-covariance matrix u t . By considering the state dependent variance, we obtain a mixture of two Gaussian distributions making it possible to replicate the skewed and leptokurtotic distribution of bond yield variations (see Idier et al., 2008).
The state process is generated by a homogeneous Markov chain with two states so that the transition probabilities are defined by: The model is a stationary VAR since no cointegration relationship was validated by the usual tests.

20
for i, j = f1, 2g. In this setting, the Markov chain transition matrix T is such that: However, the FTP-model only identifies regimes without explicitly considering the information or factors leading to such regimes. For example, the transition matrix may display jumps during specific days because some variables lead to greater persistence of some regimes. It thus seems more relevant to identify factors which can drive changes of the transition probabilities.

A time-varying transition probability model
As a consequence, we also propose a time-varying transition probability model (as in Filardo and Gordon, 1998) to explicitly consider both the monetary policy cycle and government debt issuing cycles. In order to ensure some good convergence properties, the complexity of this model calls for a more parsimonious model than the previous one. We thus consider the 10-year debt market and the three-month market rate separately in the following formulae: ∆r 120,t ln(σ 120,t ) S 120,t ∆r 3,t ∆r 120,t ln(σ 120,t ) S 120,t ∆r 3,t and 0 21 with s t = 1, 2 and with ε t and u t follow a Gaussian distribution with zero mean and state dependent variance-covariance matrices Σ ε (s t ) and Σ u (s t ) respectively. Even if we have to separate models, we consider the direct linkages between rate variations by introducing ∆r 3,t in eq. 5 and ∆r 120,t in eq. 6.
To identify the regimes detected in the model, we consider an endogenous transition prob- We thus consider both dummy variables and the bid-to-cover ratios resulting from the corresponding adjudications to disentangle two possible effects: 1. We assume that the frequency of open market operations responds to a disequilibrium in cash demand from banks and smooths funding liquidity tensions. In the same vein, the scheduled debt market cycle is assumed to provide liquidity to the corresponding bond market segment.
2. However, if the higher frequency of OMOs is associated with higher bid-to-cover ratios, these operations may have an impact on markets, especially on the market for collateral, thus revealing tensions in the funding process. For the same reasons, high bid-to-cover ratios for debt auctions may reveal excess demand for this class of assets.
In this context, the Markov chain transition matrix T is such that for the rate with maturity 22 m = f3; 120g: ln( f (x t j x t 1, x t 2 , ...x 1 )).
The set of parameters Ω comprises the mean equation parameters and the parameters used for time varying transition probabilities. The Gaussian density f (x t j x t 1, x t 2 , ...x 1 ) considers the different states of the Markov switching process as with f (x t j s t =s) the density conditional on the state and Pr(s t =s j x t 1, x t 2 , ...x 1 ) the state probabilities, elements of the (1, S) vector Π t updated for each date as with * the Hadamard product, ι a (1 S) vector of ones, T the transition matrix defined in equation 7 and f (x t ) the vector (1, S) with elements f (x t j s t =s). For the robustness checks below, we apply likelihood based tests to compare the different models to one another.

Model robustness and estimates
The . Several alternative specifications are tested. In particular we present here the different steps concerning the restricted datasets X 1 t =(∆r 3,t , σ 3,t , S 3,t , ∆r 120,t ) and X 2 t =(∆r 120,t , σ 120,t , S 120,t , ∆r 3,t ). The first step was to estimate a simple stationary VAR model. The next step was to consider Markov switching VAR models for these sets of variables, and finally consider the TVTP-VAR models as presented in equations 5 and 6.
On the basis of the estimated likelihoods, we consider Vuong's (1989) model selection tests. Let us consider two competing models with densities f and f 0 given a respective set of parameters Ω and Ω 0 . The test is defined as σ T is the heteroskedastic and autocorrelated adjusted variance of the test defined aŝ The test indicates that introducing Markov switching models always improves the fit of the models. For the TVTP models, it improves the fit for 10 year rates, while the FTP and TVTP models for the 3 month rates are only equivalent.
In addition, we also apply the RCM statistics from Ang and Bekeart (2002) to check if the two regimes in the TVTP-VAR models are clearly identified compared to a FTP-VAR. This statistic is defined for two states as so that it is bounded between 0 and 100 and indicates that the lower the statistics, the better

Market inefficiencies, volatility and liquidity premia in the two regimes
The impulse response functions are obtained from the estimates of the two TVTP-MSVAR models (see Appendix 6).
In the standard regime (regime 2), rate comovements are significant and positive, both from long to short-term rates and conversely. This illustrates French sovereign bond market dynamics as whole. An initial development that appears to reveal a non standard regime on the bond market is the increasing segmentation of the bond market.
A second element is that volatility premia in the standard regime become liquidity premia, especially for the long-term rate. The liquidity premium is a factor that has proved to be of investors in the three-month market, since they are closer to the liquidation of their contracts and have the option of rolling or liquidating them at a shorter horizon than for the 10-year rate.
A last noteworthy fact is the spiral between market liquidity and volatility in the two regimes (see Figures 11 and 12). Market illiquidity leads to stronger volatility for the 10 year rate and conversely, higher volatility leads to lower market liquidity. These market dynamics are reinforced in non standard regimes. This may strengthen the links mentioned in the early part of this paper. At some point in time, due to the large impact of monetary policy and tensions in the funding process, the monetary policy stance may not guarantee neutrality with respect to markets, especially the market for collateral. Via this channel, there is a regime in which funding tensions lead to market liquidity tensions when the frequency of OMOs is high. Therefore, due to the use of this asset class as collateral, it increases the risk of some adjustments for banks exposed to collateralized credit with margin calls, since the operation itself has an impact on the market used to determine the value of the collateral. This risk is thus incompatible with the conventional monetary stance based on market neutrality.
To clarify the links between the monetary policy stance and possible market dynamics, below we discuss the impact of the monetary policy cycle in regime identification.

The role of the monetary policy cycle in regime identification
If The impact of both the frequency of OMOs and the tensions resulting from these operations may be considered in different ways. On the one hand, the frequency of OMOs is supposed to respond to the funding needs of the banking system so that their frequency smooths possible disruptions to the financial system. On the other hand, if these operations are combined with a high bid-to-cover ratio, two effects may be expected: (i) a high bid-to-cover ratio may reveal the lack of funding liquidity and the fact that the ECB is not responding sufficiently to liquidity needs; (ii) however, it may also reveal the need for the central bank to limit its role in funding markets and to only meet efficient demand ('efficient' would mean what is calculated as benchmark supply by the central bank), so as not to replace the interbank market.
Looking at the impact of monetary and bond cycles in these regimes (see Appendices 4 and 5), the impact of OMOs is two-fold.
In regime 2, which is the standard one, the impact of the frequency of OMOs and the bidto-cover ratio of these operations are insignificant for the 10-year rate, thus preserving the market neutrality of monetary policy implementation. However, this is not true for the three-11 This followed implementation of the new operational framework in March 2004. 29 month rate, the less liquid asset: the more frequent the OMOs, the higher the probability of switching from the standard to the non standard regime. Moreover, when the bid-to-cover ratio is low during these more frequent OMOs (i.e. the ECB largely satisfies all bids posted by banks during the auction), the persistence of the standard regime is lower. This tends to bear out the hypothesis mentioned in the early part of the paper: by stepping up OMOs and ensuring higher allotment, the central bank may be encouraging the switch from a standard regime to a non standard one by affecting the market used as collateral and triggering market inefficiencies.
However, our findings concerning the impact of monetary policy on bond market dynamics, in the non-standard regime, indicate some asymmetric results. When this regime occurs, more frequent OMOs associated with low bid-to-cover ratios (i.e. a loose liquidity policy) appear to limit the persistence of this crisis regime.
This result is crucial since it introduces an asymmetry in the conduct of monetary policy: on the one hand, policy makers should limit OMOs and supply limited liquidity to prevent a switch from the standard to the non standard regime. On the other hand, if a crisis occurs, by stepping up OMOs and minimizing the bid-to-cover ratio of these operations, they reduce the persistence of the crisis regime.
To summarize, regime identification highlights the difficulty of managing monetary policy since identical measures may have "good" or "bad" effects depending on the regime markets are in. To limit the switch from a standard to a crisis regime, OMOs should not be stepped up and the supply of liquidity should be limited. However, to limit the persistence of the crisis regime, OMOs should be stepped up and satisfy demand for funding. This regime is characterized by higher liquidity-volatility feedback and market segmentation between the three-month rate and the 10-year rate. Moreover, the persistence of this unconventional regime may be reduced by stepping up OMOs and ensuring low bid-to-cover ratio (i.e. a loose liquidity policy). This policy recommendation is however asymmetric since the same monetary policy stance, i.e. more OMOs and loose liquidity, in the conventional regime increases the probability of switching from the conventional to the unconventional regime.
In particular, the fact that the monetary policy stance is based on the market neutrality hypothesis poses a new risk in the funding process when the unconventional regime occurs.
There is, in this regime, the potential for monetary policy to impact on some markets whose assets are used as collateral. Therefore, banks are exposed to higher risk with collateralized credits if margin calls are required or haircuts increased.
This highlights the difficulty for central banks in implementing an optimal liquidity policy due to these asymmetries in expected effects. The detection of which regime is prevailing in order to determine the appropriate monetary policy is challenging. It is even more challenging now in 2010, since the question remains as to whether the prevailing regime is a conventional or unconventional one. A return to market neutrality may therefore constitute an exit strategy for monetary policy.
32   ( ) Indicates significance at 5% Student statistics are provided below the estimated coefficients Appendix 6: Within regime Impulse response functions response of ∆r 120,t to ∆r 3,t response of ∆r 3,t to ∆r 120,t