Excess liquidity and the money market in the euro area

Excess liquidity and the money market in the euro area
Renaud Beaupain∗and Alain Durré†
September 17, 2015
Abstract
This paper assesses the impact of the fixed-rate full-allotment procedure
implemented by the European Central Bank (ECB) in October 2008 on the
functioning of the interbank money market. More specifically, our work examines whether the excess liquidity positions of financial institutions observed
during the fixed-rate full-allotment period have altered the activity and the
liquidity of the overnight segment of the euro area money market. In normal times, the limited amount of excess liquidity makes the dynamics of the
money market insensitive to it. By contrast, the introduction of the fixed-rate
full-allotment procedure – as one prominent unconventional measure of the
ECB – has made the dynamics of the money market growingly dependent on
the time-varying level of excess liquidity in the euro area.
Keywords: Excess liquidity, fixed-rate full-allotment, monetary policy, money
market activity, money market liquidity.
JEL Classification: C32, E52, E58
∗
IESEG School of Management (LEM – CNRS), 3 rue de la Digue, F-59000 Lille, France.
Email: [email protected].
†
IESEG School of Management (LEM – CNRS), 3 rue de la Digue, F-59000 Lille, France.
Email: [email protected].
1
1
Introduction
Following the contagion from the subprime crisis in the summer of 2007 and then
from the collapse of large too-big-to-fail financial institutions in 2008 to the euro
area financial markets, the European Central Bank (ECB) has taken unprecedented measures to heal the confidence crisis rapidly spreading across financial
institutions. Among these measures, the ECB replaced on 13 October 2008 its
weekly variable-rate tenders by a fixed-rate full-allotment procedure (FRFA) for
the allocation of funds to financial institutions. Under the fixed-rate full-allotment
regime banks do effectively receive all the requested funds (i.e., a full allotment)
at a unique interest rate applicable to all financial institutions (i.e., the fixed rate
part). This procedure has allowed financial institutions to acquire liquidity beyond
liquidity needs at a very low (and decreasing) marginal cost. With the objective to
preserve the continuation of payment flows amid market malfunctioning, the ECB
eventually decided to deviate from the hands-off policy that it followed up to the
start of the financial crisis in August 2007 in order to promote the self-organisation
of the money market between its weekly auctions.
Against this background, this paper assesses whether and to what extent the
excess liquidity situation triggered by the fixed-rate full-allotment procedure of
the ECB has altered the functioning of the market for short-term funds in the
euro area. More specifically, this paper examines how the activity (mainly the
volume of transactions and the volatility of the market rate) and the liquidity
(i.e., the ability for financial institutions to balance their central bank reserves in
the interbank market when needed) of the money market react to excess liquidity.
This contribution to the literature is, to the best of our knowledge, novel.
Our findings can be summarised as follows. First, we show that, in normal
times as observed over the period of the variable-rate tenders, the amount of
excess liquidity held by financial institutions in the euro area is relatively low
and does not alter the dynamics of the interbank money market. By contrast,
the rise in excess liquidity that followed the introduction of the fixed-rate fullallotment procedure of the ECB directly influences the operation of this market
for unsecured funds. Specifically, our empirical results support the theoretical
developments of Hauck and Neyer (2014) and Heider, Hoerova and Holthausen
(2015) on the impact of liquidity hoarding on the interbank market of the euro area
during the financial crisis. Moreover, we show that this unconventional measure
of the ECB was successful in preserving the continuation of transactions and the
resilience of liquidity in this market amid destabilising tensions in the aftermath
of the Lehman collapse.
The remaining of this paper is organised as follows. The institutional background that led to a situation of excess liquidity in the euro area is presented in
Section 2. Section 3 discusses the data and the indicators used in this paper. Our
methodology and empirical results are reported in Section 4. Concluding remarks
and policy implications are finally provided in Section 5.
2
2
Institutional Framework
Under the rules of the Eurosystem’s operational framework for the implementation
of the monetary policy, weekly auctions were organised in the form of variablerate tender procedures prior to the start of the financial crisis on 9 August 2007.
At that time, a predefined amount of central bank reserves to be injected was
decided by the ECB with a view to match – on average over a pre-defined period
called the reserve maintenance period roughly equivalent to one calendar month
– the liquidity needs of credit institutions in the euro area. Eligible credit institutions accordingly submitted their requests for liquidity against adequate collateral
(meeting the eligibility criteria determined by the ECB) within the allotted benchmark with a corresponding interest rate in their request.1 The main policy rate
decided by the Governing Council of the ECB sets the floor for those variable-rate
tenders. In practice, this main policy rate is equally distant from the interest rates
applied to standing facilities (namely the rate applied to the deposit facility of the
ECB and the rate on its marginal lending facility). The conditions of the main
refinancing operations (MROs) are announced every Monday and bids are allotted
on Tuesdays (settlement occurs on Wednesdays). Bids are allotted as a decreasing
function of the interest rate specified by the bidding financial institutions, so that
higher interest rate bids have a greater probability of being allotted in full. This
procedure therefore creates an incentive for bidders to adequately assess their liquidity needs and to decide on the attached interest rate, which, if the request for
funds is executed, will jointly define the penalty paid by the bidder to the central
bank for receiving the funds. Outside of the weekly main refinancing operations,
the operational framework of the European Central Bank promotes a hands-off
policy from the perspective of the central bank which forces banks to turn to the
money market to balance their required reserves (see ECB (2012) for more details
on the operational framework).
Until recently (and certainly prior to August 2007), the bulk of short-term
funding for banks in the money market relied on uncollateralised transactions
between market participants (hence on the unsecured segment of the money market).2 During that period, confidence between market participants was of paramount
importance in this type of market (see, e.g., Raddant (2014) or Hatzopoulos, Iori,
Mantegna, Miccichè and Tumminello (2015)). In particular, market participants
attached specific attention to the ‘quality’ of their counterparties in their transactions to limit the risk of default (i.e., credit risk). And because market participants were overwhelmingly confident about the solvency of such counterparties
and about the efficiency of the money market, no one really considered the risk of
a reduced access to market liquidity (i.e., liquidity risk) as likely.
1
The so-called benchmark was estimated by the ECB on the basis of the autonomous factors
(namely the required reserves, the banknotes and the government deposits) and excess reserves
expected over the whole maintenance period. For further details, see Durré and Nardelli (2008)
or ECB (2012).
2
Specifically, this paper focuses on the interbank (deposit) market. Subsequent references to
the money market in the remaining of this article should be understood from this particular
perspective.
3
By altering the confidence between market participants, the first stage of the
financial crisis (i.e., the period from 9 August 2007 to 12 September 2008) shed
doubt on the certainty to have access to market liquidity at any time. Then concerns about the quality of the balance sheet of counterparties gradually emerged.
The consequent market stress then exacerbated in the aftermath of the collapse
of the company Lehman Brothers, Ltd on 15 September 2008. This collapse cast
doubt on the ‘too big to fail’ market beliefs. The then-prevailing multi-dimensional
concerns significantly intensified market distortions in the various segments to lead
eventually to a sudden freeze of transactions in both the unsecured (deposit) and
secured (repo) segments in the money market in early October 2008. To address
this issue, the ECB decided on 8 October 2008 to adopt a full-allotment procedure
at a fixed price for its refinancing operations, the so-called fixed-rate full-allotment
procedure (FRFA).3 This decision took place in parallel to a significant cut in
its key policy rates coordinated with other major central banks. Over time the
ECB introduced other adjustments to its operational framework to implement its
monetary policy decisions but so far the FRFA remains the cornerstone of the
unconventional measures adopted since August 2007.4
Through the FRFA, the ECB aimed to reassure market participants about
their access to market liquidity, particularly important at a time when they were
inclined to liquidity hoarding. Such liquidity hoarding was motivated essentially
by three main elements.5 First, given the stress conditions and the market meltdown, banks became uncertain about their real liquidity needs, especially at the
consolidated level for large banking groups. Because of this uncertainty, they were
not sure to be able to absorb unexpected idiosyncratic liquidity shocks during the
period where they lent their cash to third parties (given market distortions). Second, they appeared reluctant to lend to counterparties because of increasing credit
risk, reflecting uncertainty about their solvency (Heider et al. 2015). Third, at this
stage of the crisis, market participants also faced uncertainty about how long the
replacement of financial intermediation by the European Central Bank would last.
These three elements therefore increased substantially the incentives to build up
precautionary liquidity reserves, whatever their opportunity costs.
In practice, eligible credit institutions could obtain a potentially unlimited
quantity of liquidity from the ECB at a fixed cost (corresponding to the interest
rate on the main refinancing operations, that is, the MRO rate). Through this
3
As recalled in Durré and Smets (2014), the ECB was already much less restrictive in providing
short-term liquidity already before October 2008. However, as the variable-rate tender procedure
was formally still in place, some uncertainty about the effective amount of liquidity ultimately
injected by the ECB still prevailed. By contrast, the decision made on 8 October 2008 (and
implemented on 13 October 2008) officially institutionalised the full-allotment procedure.
4
For a detailed review of the ECB unconventional measures (and their rationale) across the
various phases of the financial crisis, see Drudi, Durré and Mongelli (2012) or Durré, Maddaloni
and Mongelli (2014).
5
Although the discussion in this paper mainly focuses on the money market in the euro area,
liquidity hoarding is reported in money markets around the globe. See, e.g., Afonso, Kovner
and Schoar (2011) or Ashcraft, McAndrews and Skeie (2011) for the Federal Funds market or
Acharya and Merrouche (2013) for the money market in the United Kingdom.
4
procedure, the ECB moved from a regime of exogenous money supply (prior to
October 2008) to a regime where the money supply is endogenously determined
by banks. In light of the banks’ inclination to liquidity hoarding independently
of its cost, excess liquidity gradually emerged as it was mechanically permitted
by this change to the operational framework. In retrospect, the amount of those
excess reserves appeared more or less volatile in this period. Possible explanations
of this volatility include the time-varying concerns of market participants about
market access and about the credit risk of their counterparties.
3
Data and Definition of Measures
The data used in this paper is from the European Central Bank, which reports
on a regular basis official statistics for the euro area, including data on money
market conditions as well as on the balance sheet situation of monetary and financial institutions. Money market data is specifically from the Statistical Data
Warehouse (SDW) of the ECB. The data is checked for the presence of abnormal,
erroneous or missing records.6 Our sample covers the period from 10 March 2004
to 9 December 2014, which corresponds to 2,754 business days.7
Excess liquidity. As in ECB (2014), excess liquidity (EL) is measured as the
sum of excess reserves held by financial institutions and the net deposit facility of
the ECB. More specifically, excess reserves are measured as the difference between
the current accounts held by financial institutions at the central bank (available
at the end of each day) and their required reserves (defined on a monthly basis).
The net deposit facility corresponds to the difference between the deposit facility
and the marginal lending facility of the ECB, both available at a daily frequency.
As in Beaupain and Durré (2013), money market conditions are captured in
two dimensions. First, market activity is measured (a) by the volume traded in
the overnight money market and (b) by the volatility of the reference market rate.
Second, market liquidity is approximated by (a) market depth, (b) liquidity risk,
and (c) the speed of reversion to stable liquidity conditions (i.e., the resiliency
dimension of market liquidity from a microstructure perspective).
Market activity. Market activity is inferred from the transactions reported
by the panel of financial institutions participating in the determination of the
EONIA reference market rate. The average interest rate (i.e., the Euro Overnight
Index Average or EONIA) and the traded volume of these banks are officially
reported by the European Central Bank at the end of each business day.8 Two
6
Specifically, our filters removed 33 double entries from the data on daily liquidity conditions
provided by the ECB and excluded 1 day from the final sample due to missing EONIA volume.
7
Given the significant influence of the institutional setting on the dynamics of the money
market reported in the literature (see, e.g., Beaupain and Durré (2008) or Beaupain and Durré
(2013)), our sample starts on 10 March 2004 when the current version of the operational framework for the implementation of the monetary policy became applicable and it ends on 9 December
2014 when the duration of the reserve maintenance period changed from 4 weeks to 6 weeks.
8
Despite the over-the-counter nature of this market in the euro area, the transactions re-
5
indicators of market activity are extracted from this data:
Traded Volume: The daily total volume traded by the EONIA panel of financial
institutions (hereafter, VOLUME).
Volatility: We follow the literature on the volatility of the overnight market rate and rely on the conditional variance of the EONIA, extracted from an
EGARCH (1,1) specification (see, e.g., Nautz and Offermanns (2008), Moschitz
(2009) or Jardet and Le Fol (2010)). Our volatility indicator (hereafter, VOLATILITY) is the square root of the EGARCH conditional variance of the EONIA rate.
Market liquidity. This study relies on three complementary measures of
market liquidity, which provide a multi-dimensional view of liquidity conditions
in the money market.
Market depth: In deep markets, large traded volumes do not significantly move
market prices. By contrast, as market liquidity dries up, the market interest rate
becomes increasingly sensitive to traded sizes. Our measure of money market
illiquidity is Amihud’s (2002) ratio and is here computed as:
AM IHU Dt =
|∆rt |
V OLU M Et
where ∆rt is the daily change in the reference market rate (i.e., the EONIA
rate) and V OLU M Et is the total volume traded by the EONIA panel of banks
on the corresponding day. An increase in the value of this ratio points to a more
illiquid market following the deterioration of market depth.
Liquidity risk: Unstable market liquidity increases the uncertainty faced by
money market participants regarding the ability of this market to provide shortterm funds between the regular refinancing operations of the central bank. Against
this background, we define liquidity risk as the historical volatility of market liquidity conditions.9 Specifically, liquidity risk is the historical standard deviation
of Amihud’s (2002) illiquidity ratio, measured over a period of 20 days which represents the average duration of the reserve maintenance period in the euro area:
v
u
N
u1 ∑
t
LRISKt =
(AM IHU Dt−n − AM IHU Dt−1;t−N )2
N
n=1
where AM IHU Dt−1;t−N is the average market depth estimator over the N-day
ported by the EONIA panel capture the behaviour of market participants across all channels for
the provision of unsecured funds. Capturing all provision channels however takes on particular
importance over our sample period. Indeed, anecdotal evidence suggests that market participants switched to less transparent (i.e., bilateral) channels during the financial crisis, to avoid
displaying information to all market participants on more transparent channels (i.e., the e-MID
electronic platform in the case of the euro area). See Beaupain and Durré (2011) for a discussion
of the quality of alternative data sources.
9
See, e.g., Dionne and Chun (2013) where a similar measure captures liquidity risk in the US
corporate bonds market.
6
estimation period and N is set to 20 to encompass the average duration of one
reserve maintenance period.
Resiliency: The speed of reversion to stable market liquidity conditions is
quantified from Kempf, Mayston and Yadav’s (2009) framework. Following the
same line of reasoning as for our liquidity risk estimator, we rely on a 20-day
rolling window for resiliency estimations (hereafter, RESIL):
∗
∆AM IHU Dt = α − β × AM IHU Dt−1 +
L
∑
∆AM IHU Dt−l + εt
l=1
where β captures the strength of the mean-reversion in our depth estimator
and L∗ is the number of lags of the dependent variable which corresponds to
the optimal model specification based on the Schwarz information criterion.10 In
normal markets, our resiliency measure is accordingly expected to take values
from 0 (weakly-resilient period) to 1 (highly-resilient period). In stressful markets
where liquidity deteriorates markedly, the magnitude of the resiliency parameter
is however allowed to temporarily diverge from these boundaries.
All proxies for money market activity (VOLUME and VOLATILITY) and
money market liquidity (AMIHUD, LRISK and RESIL) are estimated daily to
match the frequency of our excess liquidity observations (EL). Table 1 shows
descriptive statistics for the above measures in the variable-rate tender period
(Panel A) and in the period of the fixed-rate full-allotment (Panel B). Excess
liquidity has increased markedly following the introduction of the fixed-rate fullallotment in the euro area. While the average excess liquidity amounted only
to 1.717 billion EUR until October 2008, it averages 244.767 billion EUR in the
period from October 2008 to December 2014. The test statistics reported in Panel
C of Table 1 lend statistical support to reject the null hypothesis of equal means,
medians and standard deviations for excess liquidity across the two periods. The
reported evidence also suggests that the fixed-rate full-allotment has altered the
dynamics of the money market. Market activity has deteriorated significantly after
October 2008: the daily traded volume has dropped from 42.539 billion EUR to
30.025 billion EUR and the volatility of the reference market rate has increased
from 9.10% to 11.02%. In the same vein, market liquidity appears significantly
lower in the fixed-rate full-allotment period: Amihud’s (2002) illiquidity ratio
moved from a daily average of 0.8597 to 1.3959 and liquidity risk jumped from
1.5300 to 2.7275 after October 2008. This deterioration of liquidity conditions
in the money market appears consistent with the survey evidence reported in
ECB (2015), where market participants report lower market liquidity after the
collapse of Lehman Brothers. The speed of reversion to stable liquidity conditions,
10
We include up to 5 lags (i.e., one week) of the dependent variable. Based on additional tests
of the robustness of our indicator not reported in the paper, our results are however insensitive
to alternative model specifications (i.e., our results are robust to alternative definitions of the
number of lags included in the model). The results presented in the remaining of this paper rely
on the optimal number of lags for resiliency estimations.
7
measured by our RESIL indicator, has nevertheless improved slightly in period 2,
where the market is found to absorb liquidity shocks faster than in period 1.
Pairwise correlations across our indicators are reported in Table 2 for the period
of the variable-rate tenders (Panel A) and for the period of the fixed-rate fullallotment (Panel B).11 Correlations are generally weak across all pairs of indicators
in the two periods, which alleviates concerns of potential collinearity among our
variables.
4
Empirical Evidence
4.1
Methodology
In this paper, we explore the dynamic effect, if any, of the excess liquidity held by
financial institutions on the functioning of the market for unsecured funds in the
euro area. We also consider the feedback effect that money market conditions may
have on market participants’ expectations, and thus on the amount of excess liquidity in the system. In fact, with the introduction of the FRFA, the ECB allowed
the supply of central bank reserves to become endogenous in practice, determined
in turn by concerns of market participants about their access to liquidity at the
aggregated level.
Specifically, we define a vector autoregressive (VAR) model which links the
evolution of our excess liquidity indicator and our money market activity and
liquidity proxies. The reduced-form VAR is defined as:
∗
Yt = α +
L
∑
βl Yt−l + et
l=1
where Yt is the vector of variables in the system at time t, that is, Yt = {∆ELt ,
VOLUMEt , VOLATILITYt , AMIHUDt , LRISKt and RESILt } and βl is the matrix
of coefficient estimates at lag l. The optimal number of lags in the model (L∗ ) is
chosen to minimise standard information criteria. We consider up to 20 lags of the
explanatory variables, which corresponds to the average duration of the reserve
maintenance period in the euro area. When information criteria lead to different
number of lags, we use the most parsimonious specification. The model is first
estimated over the period of the variable-rate tenders (from 10 March 2004 to 10
October 2008) to provide baseline estimates in a period of low excess liquidity. The
significant increase in excess liquidity observed during the fixed-rate full-allotment
period (from 13 October 2008 to 9 December 2014) leads to a separate estimation
of the model to capture its influence on the dynamics of the money market.12 The
11
Due to the nonstationary nature of our excess liquidity indicator, especially in the period of
the fixed-rate full-allotment procedure, all subsequent tests presented in this paper are based on
the first difference of this variable.
12
This choice is also guided by the evidence reported in Beirne (2012), where the excess liquidity
variable is shown to affect the money market more significantly in crisis times.
8
results are reported in Tables 3 and 4.
4.2
Results
Over the variable-rate tenders period, the results provided in Table 3 show the
independence of excess liquidity from money market conditions. Excess liquidity
is not affected by past market activity (the volume traded or the volatility of the
market rate) or by past market liquidity (market depth, liquidity risk or the speed
of reversion to stable liquidity conditions). Apart from a statistically significant
influence of excess liquidity on market depth, the reverse also holds: the activity
or the liquidity of the money market are not impacted by past changes in excess
liquidity. At the money market level, our results further highlight the interaction
between its activity and its liquidity. Trading volumes and market volatility react
positively to past levels and speeds of mean-reversion in market depth. Market
depth appears significantly responsive to market activity. Liquidity risk is positively linked to market illiquidity and depth resiliency reacts to market volatility.
Those relations receive further statistical support from the tests reported in Panel
A of Table 5 which confirm the Granger-causality that exists among those variables. Considering the results from the variable-rate tenders period as our baseline
period for our analysis, the reported evidence therefore suggests that in normal
times the money market is not affected by the excess liquidity positions of financial
institutions and that the activity and liquidity dimensions of the money market
interact to provide short-term funds to banks between the refinancing operations
of the central bank.
The introduction of the fixed-rate full-allotment procedure by the European
Central Bank in October 2008 has however altered the above dynamics. The
results presented in Table 4 show that excess liquidity becomes more responsive
to market liquidity conditions: excess liquidity reacts to the level (AMIHUD) and
to the standard deviation (LRISK) of market depth. Money market liquidity is
significantly affected by past changes in excess liquidity: increases in the level of
excess liquidity improve the liquidity of the money market. A positive change
in excess liquidity drives down market illiquidity (AMIHUD) and liquidity risk
(LRISK) and increases the speed of reversion to stable market liquidity conditions
(RESIL). Although, as our results suggest, in the period of the fixed-rate fullallotment, market activity remains strongly related to market liquidity, liquidity
risk becomes a significant driver of traded volumes and of our indicators of market
liquidity. The Granger-causality tests reported in Panel B of Table 5 confirm the
above relations in the fixed-rate full-allotment period. The reported evidence for
this period of excess liquidity therefore suggests that the fixed-rate full-allotment
of the ECB has altered the dynamics of the money market. Large excess liquidity
positions held by financial institutions influence the activity and the liquidity of
the money market, which itself becomes increasingly sensitive to the stability of
liquidity conditions (i.e., our liquidity risk indicator).
9
4.3
Impulse Response Functions
We further examine the reaction of the interbank market to excess liquidity shocks
by generating impulse response functions. Similarly, the reaction of the excess liquidity positions of financial institutions to money market shocks is considered. In
this paper, the impulse response functions are generated by a Cholesky decomposition of the covariance matrix of the innovations of the variables included in the
reduced-form VAR. By construction, this procedure is sensitive to the ordering of
the variables in the VAR (and hence, to their ordering in the covariance matrix).
Moreover, it imposes a recursive structure to the innovations in the variables. Indeed, the first variable in the ordering is assumed to influence contemporaneously
the innovations in all other variables in the VAR, but is assumed to be itself insensitive (contemporaneously) to the innovations in the other variables. The last
variable in the ordering is in turn assumed to be reactive to innovations in all
variables that precede it in the ordering, but will not (contemporaneously) influence the innovations in the preceding variables. The recursive structure of the
innovations is therefore as follows:
e1,t
e2,t
e3,t
e4,t
e5,t
e6,t
=
=
=
=
=
=
γ1,1 u1,t
γ2,1 e1,t + γ2,2 u2,t
γ3,1 e1,t + γ3,2 e2,t + γ3,3 u3,t
γ4,1 e1,t + γ4,2 e2,t + γ4,3 e3,t + γ4,4 u4,t
γ5,1 e1,t + γ5,2 e2,t + γ5,3 e3,t + γ5,4 e4,t + γ5,5 u5,t
γ6,1 e1,t + γ6,2 e2,t + γ6,3 e3,t + γ6,4 e4,t + γ6,5 e5,t + γ6,6 u6,t
where ei,t are the innovation terms from the reduced-form VAR and ui,t stands
for the structural shock to variable i at time t.
In the absence of a strong encompassing model in the theoretical literature linking the evolution of excess liquidity to the dynamics of the interbank market, the
ordering of the variables used in our analysis is based on the institutional setting
of this market and on empirical evidence reported in the market microstructure
literature. The robustness of the reported results is however checked in a series
of alternative variable orderings. Specifically, excess liquidity changes are placed
first in the variables ordering. In normal times, the excess reserves part of excess liquidity is critically influenced by the evolution of autonomous factors and
by the emergence of liquidity shocks for financial institutions. These elements
are exogenous to the dynamics of the money market. Similarly, the net recourse
to the deposit facility of the ECB does not react contemporaneously to money
market conditions. In crisis periods, the theoretical model developed in Heider
et al. (2015) suggests that, due to increased counterparty risk, banks may, under specific conditions, hoard liquidity which alters the dynamics of the interbank
market.13 Against this background, excess liquidity results in deteriorating money
13
The negative influence of liquidity hoarding on the interbank market receives further theoretical support from the model in Hauck and Neyer (2014), and is notably supported by empirical
evidence for the money market in the United Kingdom in Acharya and Merrouche (2013).
10
market conditions. Ordering the remaining variables for the Cholesky decomposition requires to decide on whether the activity or the liquidity conditions of the
interbank market react second and third. From an abundant literature on the subject, this clearly remains an empirical issue. In the specific case of the interbank
money market, we impose market liquidity to react first, then followed by market
activity. This choice is notably supported by the empirical evidence reported in
Angelini (2000) where the volume of transactions (market activity) is a function
of market liquidity conditions. Against this background, the ordering of our variables is ∆EL → Market Liquidity → Market Activity. Within the dimensions of
market liquidity examined in this paper, the level of market liquidity (AMIHUD)
is directly influenced by its stability over time (LRISK) and by its speed of convergence to equilibrium levels after being shocked (RESIL). Among the market
activity variables, the volatility of the reference market rate is assumed to be an
‘efficient market variable’ (in the sense of the efficient market hypothesis; see the
discussion in Sarno and Thornton (2004) for more details), and hence is assumed
to react contemporaneously to shocks in all the other variables. Our volatility
indicator is accordingly placed last in the ordering and is preceded by the traded
volume. The final ordering used for our Cholesky decomposition is therefore ∆EL
→ RESIL → LRISK → AMIHUD → VOLUME → VOLATILITY.
The impulse response functions reported in Figures 1 and 2 show the reaction
of the interbank market to a unit shock in excess liquidity. A confidence interval
of 2 standard deviations from the reaction function is also provided on the Figures.14 Over the variable-rate tender period (Panel A of Figures 1 and 2), a unit
shock in excess liquidity negatively affects the activity of the interbank market
(volatility increases slightly and traded volume drops). The statistically significant decrease in the volume of transactions is however short-lived, as it is fully
corrected after 2 days. Similarly, market liquidity deteriorates following an excess
liquidity shock: the market is less deep (AMIHUD increases), and its liquidity
is more unstable (LRISK increases and RESIL decreases, albeit not significantly
from a statistical viewpoint). Again, the deterioration of AMIHUD is fully corrected in 3 days. The situation however changes markedly during the period of
the fixed-rate full-allotment (Panel B of Figures 1 and 2). Over this period, excess
liquidity shocks do indeed alter the volatility of the market rate and the volume
traded in this market. In particular, VOLUME drops markedly and its reaction
to excess liquidity shocks is more persistent (the shock is not fully absorbed after
20 days). Shocks to the excess liquidity positions of financial institutions do also
interfere more significantly with the liquidity of the interbank market. Although
the depth of the market improves markedly over the 3 days following an excess
liquidity shock, market liquidity is however persistently more unstable (LRISK
remains negatively impacted 20 days after the shock). We conjecture that, due
the size of excess liquidity in this period, a positive shock in this parameter decreases the need for financial institutions to use the money market to balance their
central bank reserves: market activity declines and its liquidity becomes more un14
The confidence intervals reported in this paper are based on Monte Carlo standard errors,
generated with 1,000 replications.
11
stable. These findings therefore lend empirical support to the theoretical insights
developed in Heider et al. (2015).
The reaction of the excess liquidity positions held by financial institutions to
the dynamics of the interbank market is assessed through the series of impulse
response functions reported in Figures 3 and 4. During the variable-rate tender
period (Panel A of Figures 3 and 4), excess liquidity appears insensitive to the evolution of the money market. All impulse response functions are indeed statistically
insignificant, confirming that in normal times, the decision to hold excess liquidity
is disconnected from the evolution of the interbank money market. Again, the
situation changes over the fixed-rate full-allotment period (Panel B of Figures 3
and 4). Specifically, although excess liquidity is similarly insensitive to shocks
in market activity, it becomes strongly reactive to market liquidity shocks. Excess liquidity positions increase markedly following unit shocks in AMIHUD and
in LRISK. One possible explanation is the liquidity hoarding observed over the
period: after observing a deterioration in the level (AMIHUD) and in the stability (LRISK) of market liquidity, financial institutions hoard more liquidity. This
therefore suggests that financial institutions hoard liquidity not only as a reaction
to increased counterparty risk (Heider et al. 2015), but also due to deteriorating
liquidity conditions in the interbank market (i.e., market liquidity). These shocks
are nevertheless absorbed quickly: they are fully absorbed in 2 (unit shock to
LRISK) or 3 (unit shock to AMIHUD) days.
The robustness of the above findings is finally checked in a series of additional
tests. First, splitting the fixed-rate full-allotment period into two sub-periods
(from 13 October 2008 to 10 August 2011, when the sovereign debt crisis escalated
in Europe, and from 11 August 2011 to 9 December 2014) lends further support to
the reported bidirectional relation between market liquidity and excess liquidity
changes. This therefore lowers potential concerns that our results are driven by
the heightened instability of period following the collapse of Lehman, rather than
by the institutional adjustments made by the ECB in the implementation of its
monetary policy (mainly, the introduction of the FRFA). Second, changing the lag
length15 or removing the more correlated variables16 in our VAR leads to qualitatively similar patterns for our impulse response functions. Third, assuming that
market activity leads market liquidity in the ordering of our Cholesky decomposition17 confirms the robustness of the impulse response functions presented in this
paper. Although these tests support the robustness of our findings, we note that
our results remain significantly influenced by the recursive identification scheme
used in this paper. In line with the observation in Holthausen and Pill (2010),
future research is indeed needed on the identification of structural VARs for the
dynamics of the interbank money market.
15
We alternatively consider the inclusion of 5, 10 and 15 lags of our variables.
We estimate separate models where LRISK is excluded due to a relatively high correlation
between LRISK and AMIHUD shocks (0.52 in the period of the variable-rate tenders and 0.63
in the fixed-rate full-allotment period).
17
The alternative ordering is accordingly ∆EL → VOLUME → VOLATILITY → RESIL →
LRISK → AMIHUD.
16
12
4.4
Interpretations
As mentioned above, the introduction of the FRFA by the ECB intervened in
the aftermath of the Lehman collapse that intensified the stress in all segments
of the money market. In fact, to the liquidity crisis that erupted as from August
2007 joined solvency concerns in the banking system with the impression that
no single financial institution was immune from an unexpected collapse like that
of Lehman. Therefore, banks did appear not only concerned about the risk of
unexpected adverse liquidity shocks (during the period over which they lent their
cash surplus to counterparties) but they also appeared concerned about the risk
of not being reimbursed at all when reversing money market transactions (even
at the overnight maturity). All these concerns among market participants rapidly
intensified which led to a sudden stop of transactions in most segments of the
interbank (deposit) market at the end of September 2008. The FRFA procedure
that the ECB implemented following these developments aimed to reassure market
participants about their access to liquidity in all circumstances. In a context of
liquidity hoarding, this unconventional monetary policy measure by the ECB led
not surprisingly to the emergence of growing excess liquidity. Together with an
interest rate cycle converging to the zero lower bound, the opportunity cost of
such liquidity hoarding (i.e., holding excess liquidity) diminished drastically.
When assessing the impact of excess liquidity on the dynamics of the money
market, our results point out that it is essential to distinguish the market activity
indicators from liquidity tools elaborated by the microstructure literature. In fact,
although the excess liquidity in the system put downward pressures on the level
of interest rates, volatility of the EONIA and money market trading volume –
hence reducing market activity – liquidity in the microstructure sense improved
after the introduction of the FRFA allowing excess liquidity. Such results could
probably be understood through the lens of growing financial fragmentation in
the euro area. In fact, financial institutions with reduced (or no) access to the
money market became more dependent on the liquidity provision of the ECB – the
so-called ‘addicted banks’ named by the ECB (Bini Smaghi 2009) – while financial institutions having recourse to the ECB’s deposit facility were those having
regained access to the market. The latter institutions – because of their restored
good and sound reputation – were also those collecting cash in euros from institutions outside the euro area (not eligible to the ECB’s standing facilities). As
a result, ‘red-lined’ institutions were removed from the market to ECB’s operations, thereby increasing incentives to trade in the money market within a de
facto smaller group of counterparties for which the risk of default substantially
decreased, if not disappeared.
5
Concluding Remarks and Policy Implications
In October 2008, the European Central Bank replaced its variable-rate tender
procedure by a fixed-rate full-allotment procedure to provide central bank re-
13
serves to financial institutions of the Eurosystem against eligible collateral. This,
coupled with a decreasing monetary policy rate of the central bank, led to a situation of excess liquidity in the euro area. Given the central role played by the
interbank money market in providing unsecured liquidity to banks between the
regular refinancing operations of the central bank, we examine whether and how
excess liquidity has altered its dynamics. We specifically focus on the activity (the
traded volume and the volatility of the reference market rate) and on the liquidity
(market depth, liquidity risk and the speed of mean-reversion to stable liquidity
conditions) of this market.
Over the variable-rate tenders period (from 10 March 2004 to 10 October 2008),
we report evidence of a disconnection between the amount of excess liquidity held
by financial institutions and the dynamics of the interbank money market. This
appears in line with the willingness of the central bank not to interfere with the
functioning of the money market in normal times, that is, the so-called handsoff policy of the ECB. This situation has however changed markedly after the
introduction of the fixed-rate full-allotment by the ECB (from 13 October 2008
to 9 December 2014) where excess liquidity becomes a significant driver of the
activity and of the liquidity of this market. With the fixed-rate full-allotment,
banks are in a position to hold liquidity in excess of their needs and this excess
liquidity directly interferes with the dynamics of the money market.
The empirical evidence reported in this paper sheds light on the question in
relation to what central banks can and cannot do, at least for the money market. Given the close natural link between the money market and the operational
framework of the ECB, the reported evidence shows the ability of the central bank
to influence (the ‘can’) at all times the dynamics of the money market by altering
the design of the tools used to implement monetary policy decisions. In crisis
times, the ECB can deviate from its hands-off policy and support the market
more actively. Through the FRFA, the ECB has offered liquidity insurance to
market participants precisely at a stage of the financial crisis where they appeared
supportive of liquidity hoarding. However, the FRFA removed the control by the
ECB of the size of its balance sheet (the ‘cannot’), thereby reducing de facto its
control of the dynamics of the money market through unstable excess liquidity,
reflecting time-varying stress conditions in the market as this implied in a first
phase a lower trading volume and a higher volatility of interest rates.
6
Acknowledgements
The authors are thankful to William Lastrapes (the Editor), Daniel Thornton (the
Guest Editor) and to an anonymous referee for useful comments and suggestions.
14
References
Acharya, V. V. and Merrouche, O.: 2013, Precautionary hoarding of liquidity and
interbank markets: Evidence from the subprime crisis, Review of Finance
17(1), 107–160.
Afonso, G., Kovner, A. and Schoar, A.: 2011, Stressed, not frozen: The Federal
Funds market in the financial crisis, Journal of Finance 66(4), 1109–1139.
Amihud, Y.: 2002, Illiquidity and stock returns: Cross-section and time-series
effects, Journal of Financial Markets 5(1), 31–56.
Angelini, P.: 2000, Are banks risk averse? Intraday timing of operations in the
interbank market, Journal of Money, Credit and Banking 32(1), 54–73.
Ashcraft, A., McAndrews, J. and Skeie, D.: 2011, Precautionary reserves and the
interbank market, Journal of Money, Credit and Banking 43(7), 311–348.
Beaupain, R. and Durré, A.: 2008, The interday and intraday patterns of the
overnight market: Evidence from an electronic platform, European Central
Bank Working Paper No. 988 .
Beaupain, R. and Durré, A.: 2011, Inferring trading dynamics for an OTC market:
The case of the euro area overnight money market, Quantitative Finance
11(9), 1285–1295.
Beaupain, R. and Durré, A.: 2013, Central bank reserves and interbank market
liquidity in the euro area, Journal of Financial Intermediation 22(2), 259–
284.
Beirne, J.: 2012, The EONIA spread before and during the crisis of 2007-2009:
The role of liquidity and credit risk, Journal of International Money and
Finance 31(3), 534–551.
Bini Smaghi, L.: 2009, An ocean apart? Comparing transatlantic responses to
the financial crisis, Taking stock: Global implications of transatlantic differences, Banca d’Italia, Bruegel and the Peterson Institute for International
Economics.
Dionne, G. and Chun, O. M.: 2013, Default and liquidity regimes in the
bond market during the 2002-2012 period, Canadian Journal of Economics
46(4), 1160–1195.
Drudi, F., Durré, A. and Mongelli, F. P.: 2012, The interplay of economic reforms
and monetary policy: The case of the euro area, Journal of Common Market
Studies 50(6), 881–898.
Durré, A., Maddaloni, A. and Mongelli, F.: 2014, Financial fragmentation and
ECB monetary policy, Comparative Economic Studies 56(3), 396–423.
15
Durré, A. and Nardelli, S.: 2008, Volatility in the euro area money market: Effects
from the monetary policy operational framework, International Journal of
Finance and Economics 13(4), 307–322.
Durré, A. and Smets, F.: 2014, Sovereign debt and monetary policy in theory and
practice: The case of the euro area, in J. S. Chadha, A. Durré and L. Sarno
(eds), New developments in macro-finance yield curve modelling, Cambridge
University Press.
ECB: 2012, The implementation of monetary policy in the euro area: General
documentation on Eurosystem monetary policy instruments and procedures,
European Central Bank.
ECB: 2014, Recent developments in excess liquidity and money market rates,
Monthly Bulletin January, 69–82.
ECB: 2015, Euro Money Market Study 2014, European Central Bank.
Hatzopoulos, V., Iori, G., Mantegna, R. N., Miccichè, S. and Tumminello, M.:
2015, Quantifying preferential trading in the e-MID interbank market, Quantitative Finance 15(4), 693–710.
Hauck, A. and Neyer, U.: 2014, A model of the Eurosystem’s operational framework and the euro overnight interbank market, European Journal of Political
Economy 34, S65–S82.
Heider, F., Hoerova, M. and Holthausen, C.: 2015, Liquidity hoarding and interbank market spreads: The role of counterparty risk, Journal of Financial
Economics in press.
Holthausen, C. and Pill, H.: 2010, The forgotten markets: How understanding
money markets helps us to understand the financial crisis, ECB Research
Bulletin 9, 2–5.
Jardet, C. and Le Fol, G.: 2010, Euro money market interest rate dynamics and
volatility: How they respond to recent changes in the operational framework,
International Journal of Finance and Economics 15(4), 316–330.
Kempf, A., Mayston, D. and Yadav, P. K.: 2009, Resiliency in limit order book
markets: A dynamic view of liquidity, Mimeo .
Moschitz, J.: 2009, Monetary policy implementation and the euro area money
market, Applied Financial Economics 19(1), 39–57.
Nautz, D. and Offermanns, C. J.: 2008, Volatility transmission in the European money market, North American Journal of Economics and Finance
19(1), 23–39.
Raddant, M.: 2014, Structure in the Italian overnight loan market, Journal of
International Money and Finance 41, 197–213.
16
Sarno, L. and Thornton, D. L.: 2004, The efficient market hypothesis and identification in structural VARs, Federal Reserve Bank of St. Louis Review
86(1), 49–60.
17
Panel A - Variable-Rate Tender Period
Panel B - Fixed-Rate Full-Allotment Period
Response of VOLATILITY to Excess Liquidity Shock
Response of VOLATILITY to Excess Liquidity Shock
0.012
0.06
0.04
0.008
0.02
0.004
0.00
0.000
-0.02
-0.04
-0.004
-0.008
-0.06
1
2
3
4
5
6
7
8
9
10 11
12 13 14
15 16
17 18
-0.08
19 20
1
2
3
4
Response of VOLUME to Excess Liquidity Shock
200
0
0
-200
-200
-400
-600
-400
-800
-600
-1,000
-1,200
-800
-1,400
-1,000
1
2
3
4
5
6
7
8
9
10 11
12 13 14
15 16
5
6
7
8
9
10 11
12 13 14
15 16
17 18
19 20
17 18
19 20
Response of VOLUME to Excess Liquidity Shock
17 18
19 20
1
2
3
4
5
6
7
8
9
10 11
12 13 14
15 16
Figure 1: Impulse response functions – Response of market activity to excess
liquidity shock. This Figure shows the reaction of our indicators of market activity
to a unit shock in excess liquidity in the variable-rate tender period (Panel A) and
in the period of the fixed-rate full-allotment (Panel B).
Panel A - Variable-Rate Tender Period
Panel B - Fixed-Rate Full-Allotment Period
Response of AMIHUD to Excess Liquidity Shock
Response of AMIHUD to Excess Liquidity Shock
0.30
0.4
0.2
0.25
0.0
-0.2
0.20
0.15
-0.4
-0.6
0.10
0.05
-0.8
-1.0
0.00
-0.05
1
2
3
4
5
6
7
8
9
-1.2
10 11 12 13 14 15 16 17 18 19 20
1
2
3
4
Response of LRISK to Excess Liquidity Shock
0.02
0.05
0.00
0.04
-0.02
0.03
-0.04
0.02
-0.06
0.01
-0.08
0.00
7
8
9
10 11 12 13 14 15 16 17 18 19 20
-0.10
1
2
3
4
5
6
7
8
9
-0.12
10 11 12 13 14 15 16 17 18 19 20
1
2
3
4
Response of RESIL to Excess Liquidity Shock
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Response of RESIL to Excess Liquidity Shock
0.02
0.01
0.06
0.04
0.00
-0.01
0.02
-0.02
-0.03
0.00
-0.02
-0.04
-0.05
-0.06
6
Response of LRISK to Excess Liquidity Shock
0.06
-0.01
5
-0.04
1
2
3
4
5
6
7
8
9
-0.06
10 11 12 13 14 15 16 17 18 19 20
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Figure 2: Impulse response functions – Response of market liquidity to excess
liquidity shock. This Figure shows the reaction of our indicators of market liquidity
to a unit shock in excess liquidity in the variable-rate tender period (Panel A) and
in the period of the fixed-rate full-allotment (Panel B).
18
Panel A - Variable-Rate Tender Period
Panel B - Fixed-Rate Full-Allotment Period
Response of Excess Liquidity to VOLATILITY Shock
Response of Excess Liquidity to VOLATILITY Shock
800
3,000
400
2,000
1,000
0
0
-400
-1,000
-800
-1,200
-2,000
1
2
3
4
5
6
7
8
9
10 11
12 13 14
15 16
17 18
19 20
-3,000
1
2
3
4
Response of Excess Liquidity to VOLUME Shock
1,200
4,000
1,000
3,000
800
2,000
600
6
7
8
9
10 11
12 13 14
15 16
17 18
19 20
17 18
19 20
1,000
400
0
200
-1,000
0
-2,000
-200
5
Response of Excess Liquidity to VOLUME Shock
1
2
3
4
5
6
7
8
9
10 11
12 13 14
15 16
17 18
19 20
-3,000
1
2
3
4
5
6
7
8
9
10 11
12 13 14
15 16
Figure 3: Impulse response functions – Response of excess liquidity to market
activity shock. This Figure shows the reaction of excess liquidity to a unit shock
in our indicators of market activity in the variable-rate tender period (Panel A)
and in the period of the fixed-rate full-allotment (Panel B).
Panel A - Variable-Rate Tender Period
Panel B - Fixed-Rate Full-Allotment Period
Response of Excess Liquidity to AMIHUD Shock
Response of Excess Liquidity to AMIHUD Shock
1,200
12,000
10,000
800
400
8,000
6,000
0
4,000
2,000
-400
0
-2,000
-800
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
-4,000
1
2
3
4
Response of Excess Liquidity to LRISK Shock
400
200
0
-200
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
6,000
5,000
4,000
3,000
2,000
1,000
0
-1,000
-2,000
-3,000
1
2
3
4
Response of Excess Liquidity to RESIL Shock
1,000
800
600
400
200
0
-200
-400
-600
-800
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Response of Excess Liquidity to LRISK Shock
600
-400
5
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Response of Excess Liquidity to RESIL Shock
3,000
2,000
1,000
0
-1,000
-2,000
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
-3,000
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Figure 4: Impulse response functions – Response of excess liquidity to market
liquidity shock. This Figure shows the reaction of excess liquidity to a unit shock
in our indicators of market liquidity in the variable-rate tender period (Panel A)
and in the period of the fixed-rate full-allotment (Panel B).
19
20
Market Activity
VOLUME
VOLATILITY
42,539.04
0.0910
40,755.00
0.0616
10,745.87
0.1649
82,340.00
3.4051
17,133.00
0.0519
1157
1157
Market Activity
VOLUME
VOLATILITY
30,025.44
0.1102
28,368.50
0.0603
9,644.17
0.4892
62,893.00
16.5224
5,781.00
0.0521
1578
1578
AMIHUD
1.3959
0.2547
3.8164
73.3437
0.0000
1578
AMIHUD
0.8597
0.2042
2.0283
24.1895
0.0000
1157
Market Liquidity
LRISK
2.7275
2.4428
2.5338
16.3829
0.0787
1578
Market Liquidity
LRISK
1.5300
1.1813
1.1884
5.7522
0.1815
1157
RESIL
0.7389
0.7404
0.5019
3.5195
-8.1533
1578
RESIL
0.6332
0.6767
0.5910
3.6114
-6.2788
1157
Excess Liquidity
Market Activity
Market Liquidity
EL
VOLUME
VOLATILITY
AMIHUD
LRISK
RESIL
Mean Equality
1777.2279∗∗∗
986.4190∗∗∗
2.1087
22.4853∗∗∗
271.0932∗∗∗ 24.1898∗∗∗
Median Equality
1419.7969∗∗∗
749.0504∗∗∗
14.8288∗∗∗
113.2987∗∗∗ 115.9255∗∗∗ 48.5456∗∗∗
∗∗∗
∗∗∗
Variance Equality
863.3401
16.7361
1.5120
13.6966∗∗∗
200.0994∗∗∗ 11.9689∗∗∗
This Table reports descriptive statistics for all indicators of excess liquidity, money market activity and money
market liquidity. The variable-rate tender period is from 10 March 2004 to 10 October 2008. The fixed-rate
full-allotment period is from 13 October 2008 to 9 December 2014. EL and VOLUME are in millions of EUR.
AMIHUD is resized (multiplied by 1,000,000). Mean, median and variance equality across periods is checked by
means of Welch, van der Waerden and Brown-Forsythe nonparametric test statistics, respectively. ***, **, and *
denote statistical significance at the 1%, 5%, and 10% levels, respectively.
Panel C – Equality Tests across Periods
Mean
Median
Standard Deviation
Maximum
Minimum
Observations
Excess Liquidity
EL
244,767.10
173,323.00
228,056.61
811,857.00
-134,833.00
1578
Panel B – Fixed-Rate Full-Allotment Period
Mean
Median
Standard Deviation
Maximum
Minimum
Observations
Excess Liquidity
EL
1,717.53
916.00
17,985.08
224,834.00
-67,739.00
1157
Panel A – Variable-Rate Tender Period
Table 1: Descriptive Statistics
Table 2: Pairwise Correlations across Periods
Panel A – Variable-Rate Tender Period
VOLUME
VOLATILITY
AMIHUD
LRISK
RESIL
∆EL
-0.0360
0.0329
0.0979
0.0079
-0.0302
VOLUME
VOLATILITY
AMIHUD
LRISK
0.0649
0.0778
0.0125
0.1121
0.4251
0.1571
-0.0941
0.1892
-0.2126
-0.0542
Panel B – Fixed-Rate Full-Allotment Period
∆EL VOLUME VOLATILITY AMIHUD LRISK
VOLUME
-0.0004
VOLATILITY
0.1397
0.0804
AMIHUD
0.0521
-0.0789
0.1820
LRISK
0.0202
0.3330
0.0945
0.2762
RESIL
-0.0430
0.1116
-0.0130
-0.0675
0.0273
This Table reports pairwise correlations for all indicators of excess liquidity,
money market activity and money market liquidity in the variable-rate tender
(Panel A) and fixed-rate full-allotment (Panel B) periods.
21
22
Excess Liquidity
Market Activity
Market Liquidity
∆ELt
V OLU M Et
V OLAT ILIT Yt AM IHU Dt
LRISKt
RESILt
Intercept
−2518.4639
12792.0413∗∗∗
−0.0147
−0.9665∗∗∗
0.0342
0.2367∗∗∗
(−1.2922)
(12.8661)
(−0.8167)
(−4.1448)
(0.7783)
(3.7928)
∆ELt−1
−0.1555∗∗∗
0.0026
−0.0000
0.0000∗∗
0.0000
−0.0000
(−5.3122)
(0.1727)
(−1.1302)
(2.0958)
(0.0797)
(−0.3851)
V OLU M Et−1
0.0629
0.6777∗∗∗
0.0000∗∗∗
0.0000∗∗∗
0.0000
−0.0000
(1.4798)
(31.2682)
(4.2963)
(7.8464)
(0.9971)
(−0.2182)
V OLAT ILIT Yt−1
−945.7172
654.2715
−0.1284
−1.5403∗∗∗
−0.0901
0.2736∗∗∗
(−0.3071)
(0.4164)
(−4.5148)
(−4.1800)
(−1.2955)
(2.7747)
AM IHU Dt−1
216.1018
226.1103∗
0.0446∗∗∗
0.3724∗∗∗
0.0360∗∗∗
0.0088
(0.8508)
(1.7451)
(19.0114)
(12.2535)
(6.2747)
(1.0765)
LRISKt−1
−118.3578
76.5938
−0.0007
−0.0445
0.9455∗∗∗
0.0146
(−0.3056)
(0.3877)
(−0.1976)
(−0.9596)
(108.1660)
(1.1757)
RESILt−1
219.4412
847.7306∗∗
0.0130∗
0.0112
−0.0215
0.5630∗∗∗
(0.2789)
(2.1122)
(1.7930)
(0.1193)
(−1.2092)
(22.3492)
Adjusted R2
0.0217
0.4727
0.2648
0.1704
0.9156
0.3081
This Table reports vector autoregressive estimates in the variable-rate tender period. t-statistics are reported in the
parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.
Table 3: Vector Autoregressive Estimates – Variable-Rate Tender Period
23
Excess Liquidity
Market Activity
Market Liquidity
∆ELt
V OLU M Et
V OLAT ILIT Yt AM IHU Dt
LRISKt
RESILt
Intercept
−8334.8873∗∗
3677.9857∗∗∗
−0.0664
−0.2381
−0.0470
0.2119∗∗∗
(−2.2260)
(7.4107)
(−1.5503)
(−0.7287)
(−0.7450)
(5.5045)
∆ELt−1
−0.5354∗∗∗
−0.0032
0.0000
−0.0000∗∗∗
−0.0000∗∗
0.0000∗∗∗
(−20.9083)
(−0.9340)
(0.4194)
(−6.9433)
(−2.3717)
(3.1977)
∆ELt−2
−0.1797∗∗∗
0.0025
−0.0000
0.0000
0.0000
−0.0000∗∗
(−6.9886)
(0.7199)
(−0.1653)
(0.4232)
(0.2113)
(−2.2351)
V OLU M Et−1
0.2950
0.6259∗∗∗
0.0000∗∗∗
−0.0000
0.0000
0.0000
(1.4778)
(23.6561)
(3.1301)
(−0.1003)
(1.1219)
(1.6167)
V OLU M Et−2
−0.0924
0.2175∗∗∗
−0.0000
0.0000∗∗
0.0000
−0.0000
(−0.4618)
(8.1966)
(−1.0524)
(2.2443)
(0.5058)
(−0.5375)
V OLAT ILIT Yt−1
149.0777
217.1061
−0.0040
−0.2488
−0.0141
0.0157
(0.0674)
(0.7401)
(−0.1564)
(−1.2882)
(−0.3783)
(0.6910)
V OLAT ILIT Yt−2
1041.1009
−18.4612
0.0146
−0.0838
0.0006
0.0017
(0.4883)
(−0.0653)
(0.5992)
(−0.4503)
(0.0179)
(0.0766)
AM IHU Dt−1
3700.6698∗∗∗
238.4259∗∗∗
0.0505∗∗∗
0.3964∗∗∗
0.0113∗
0.0019
(9.4520)
(4.5943)
(11.2652)
(11.6028)
(1.7126)
(0.4636)
AM IHU Dt−2
693.8246
−3.1663
−0.0134∗∗∗
−0.0424
0.0078
−0.0030
(1.9428)
(−0.0669)
(−3.2733)
(−1.3618)
(1.2997)
(−0.8063)
LRISKt−1
−7021.5950∗∗∗
901.3360∗∗∗
−0.0170
−0.5103∗∗∗
1.0381∗∗∗
−0.0087
(−3.6782)
(3.5621)
(−0.7795)
(−3.0634)
(32.2648)
(−0.4415)
LRISKt−2
5534.3535∗∗∗
−713.3235∗∗∗
0.0110
0.5217∗∗∗
−0.0930∗∗∗
0.0120
(2.9675)
(−2.8855)
(0.5162)
(3.2053)
(−2.9597)
(0.6246)
RESILt−1
243.3798
367.3356
0.0372
−0.0086
0.0064
0.4684∗∗∗
(0.1000)
(1.1387)
(1.3369)
(−0.0404)
(0.1551)
(18.7168)
RESILt−2
−384.6747
−158.8128
−0.0402
0.0332
0.0019
0.1418∗∗∗
(−0.1586)
(−0.4938)
(−1.4486)
(0.1570)
(0.0462)
(5.6849)
Adjusted R2
0.2916
0.6954
0.1175
0.1569
0.9287
0.3232
This Table reports vector autoregressive estimates for the fixed-rate full-allotment period. t-statistics are reported
in the parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.
Table 4: Vector Autoregressive Estimates – Fixed-Rate Full-Allotment Period
Table 5: Granger-Causality Tests
Panel A – Variable-Rate Tender Period
∆EL
VOLUME
VOLATILITY
AMIHUD
LRISK
RESIL
Excess Liquidity
∆EL
−
2.1899
0.0943
0.7239
0.0934
0.0778
Market Activity
VOLUME VOLATILITY
0.0298
1.2774
−
18.4579∗∗∗
0.1734
−
3.0454∗
361.4331∗∗∗
0.1503
0.0390
4.4616∗∗
3.2148∗
AMIHUD
4.3922∗∗
61.5653∗∗∗
17.4723∗∗∗
−
0.9209
0.0142
Market Liquidity
LRISK
RESIL
0.0064
0.1483
0.9943
0.0476
1.6784
7.6989∗∗∗
39.3723∗∗∗
1.1588
−
1.3824
1.4620
−
Panel B – Fixed-Rate Full-Allotment Period
Excess Liquidity
Market Activity
Market Liquidity
∆EL
VOLUME VOLATILITY AMIHUD
LRISK
RESIL
∆EL
−
2.7440
0.3626
68.3154∗∗∗
8.2128∗∗
29.7536∗∗∗
VOLUME
3.8483
−
16.5780∗∗∗
14.0394∗∗∗
7.3241∗∗
4.4506
VOLATILITY
0.2416
0.5546
−
1.8357
0.1438
0.4812
AMIHUD
127.7568∗∗∗
24.7599∗∗∗
128.5940∗∗∗
−
7.5786∗∗
0.6746
LRISK
19.7341∗∗∗
18.2761∗∗∗
1.5962
10.3114∗∗∗
−
0.8546
RESIL
0.0254
1.3192
2.5254
0.0275
0.0483
−
This Table reports test statistics of the null hypothesis that the variable in the row Granger-causes the
variable in the column in the variable-rate tender period (Panel A) and in the fixed-rate full-allotment
period (Panel B). In each period, the tests are based on the optimal VAR specification that minimises
standard information criteria. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels,
respectively.
24