Journal of International Money and Finance 108 (2020) 102156

Contents lists available at ScienceDirect

Journal of International Money and Finance

journal homepage: [Link]/locate/jimf

Macroprudential and monetary policies: The need to dance the

Tango in harmony
José David Garcia Revelo a, Yannick Lucotte a,b,⇑, Florian Pradines-Jobet b
a
Univ. Orléans, CNRS, LEO, FRE 2014, F-45067 Orléans, France
b
PSB Paris School of Business, Department of Economics, 59 rue Nationale, 75013 Paris, France

a r t i c l e i n f o a b s t r a c t

Article history: Considering a sample of 37 emerging and advanced economies from 2000Q1 to 2014Q4,
Available online 20 February 2020 we empirically assess how effective macroprudential policies are in curbing domestic
credit growth, and whether their effectiveness is affected by monetary policy conditions.
JEL codes: We obtain three important results. First our findings suggest in line with previous research
E43 that an overall tightening in macroprudential policies is associated with a reduction in
E58 credit growth. Second, we show that a restrictive monetary policy enhances the impact
G18
of macroprudential tightening on credit growth. Third, our results seem to suggest that
G28
monetary policy helps to reduce the transmission delay of macroprudential policy actions.
Keywords: Consequently our results confirm the need for coordination between the two policies.
Macroprudential policy Ó 2020 Elsevier Ltd. All rights reserved.
Monetary policy
Financial stability
Excessive credit growth
Policy synchronisation

1. Introduction

In the wake of the 2007–2008 global financial crisis, macroprudential policy has attracted considerable attention from
policymakers and researchers. A number of emerging countries were using macroprudential policy tools well before the cri-
sis, but substantial progress has subsequently been made in both emerging and industrialised economies in putting in place
dedicated institutional arrangements for macroprudential policy. The main objective of macroprudential policy is to safe-
guard the stability of the financial system as a whole by strengthening its resilience and preventing the build-up of systemic
risk. To ensure the achievement of this primary objective, the European Systemic Risk Board (ESRB/2013/1) defines five
intermediate objectives that macroprudential policy should aim to achieve: (i) mitigating and preventing excessive credit
growth and leverage; (ii) mitigating and preventing excessive maturity mismatch and market illiquidity; (iii) limiting direct
and indirect exposure concentrations; (iv) limiting the systemic impact of misaligned incentives with a view to reducing
moral hazard; and (v) strengthening the resilience of financial infrastructures. These intermediate objectives are seen as
transitional steps towards achieving robust financial stability.
Mitigating and preventing excessive growth in credit and leverage is particularly important for safeguarding financial sta-
bility. One important lesson of the global financial crisis is that financial imbalances largely developed because of the pro-
cyclical behaviour of the banking industry. Such behaviour tends to put upward pressure on asset prices and is often viewed

⇑ Corresponding author.
E-mail address: ylucotte@[Link] (Y. Lucotte).

[Link]
0261-5606/Ó 2020 Elsevier Ltd. All rights reserved.
2 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

as a key driver of banking crises and how severe they are. This is why several macroprudential tools have been designed to
curb excessive credit growth and mitigate the procyclicality of domestic credit. Such instruments include for instance the
countercyclical capital buffer and the loan-to-value ratio.
However, the implementation of macroprudential policies raises a number of challenges. The first of these is to evaluate
how effective the policies are. The empirical literature on this is still in infancy, but there is a growing body of it (see, e.g.,
Galati and Moessner, 2018). The findings seem especially to confirm that macroprudential tools are effective at containing
credit growth and housing prices. The second challenge more importantly concerns how macroprudential policy interacts
with other policies that also have an impact on financial stability, such as fiscal and monetary policies.
Most notably, macroprudential and monetary policies pursue different primary objectives that could be in conflict with
one another. As already noted, macroprudential policy primarily aims to promote financial stability, while the primary objec-
tive of monetary policy is to maintain price stability. Consequently, each policy can have ‘‘side effects” on the objective of the
other and make it more effective or less so. Side effects from monetary policy pose significant challenges for the conduct of
macroprudential policy if they are detrimental to financial stability.
Monetary policy can have detrimental side effects on financial stability through various channels (IMF, 2013). If the policy
interest rate is cut, monetary policy can worsen financial stability through two channels, the risk-taking channel and the
asset prices channel. The risk-taking channel means that a low interest rate environment may encourage banks to expand
their balance sheets and take on more risk, which in turn may contribute to an excessive expansion of credit and may in this
way amplify boom-bust cycles (Adrian and Shin, 2010; Borio and Zhu, 2012). It is often argued that these effects are worse if
the policy interest rate is held ‘‘too low for too long”. A low interest rate environment can also lead to sharp rises in asset
prices through the ‘‘financial accelerator” mechanism (Bernanke and Gertler, 1989; Bernanke and Gertler, 1995). Such a rise
in asset prices tends to intensify the financial cycle, which may lead to a bubble.
When the policy interest rate rises, monetary policy can affect financial stability negatively through three different chan-
nels: the balance sheet channel, the risk-shifting channel, and for small open economies, the exchange rate channel. The bal-
ance sheet channel means that a tightening of the monetary policy stance can hurt the capacity of borrowers to repay their
loans, which can lead to higher default rates and financial instability (see, e.g., Allen and Gale, 2001; Illing, 2007). The risk-
shifting effect operates through the balance sheets of banks. As banks typically take in short-term deposits and make long-
term loans, changes in the policy rate affect the interest rate applied to short-term deposits more than that for loans. A rise in
the policy rate then reduces intermediation margins, and leads financial intermediaries to seek more risk in order to main-
tain their profits. As a result, monetary tightening is expected to increase financial instability. Finally, monetary policy can
impact financial stability though the exchange rate channel, as the policy rate is an important determinant of capital inflows.
These inflows can in turn drive credit growth and, owing to the presence of exchange rate externalities, contribute to exces-
sive increases in leverage. The consequence is that, contrary to expectations, raising the policy rate may induce excessive
growth in credit, especially in emerging markets and small open economies. Of course the strength of these side effects
can vary with the financial cycle. As financial imbalances build up, monetary easing tends to reduce default rates, but can
induce banks to make riskier loans and increase their leverage. When the policy rate is raised close to the peak of the finan-
cial cycle, this can induce risk-shifting and borrower defaults.
More importantly, these side effects highlight the potential tradeoffs and complementarities between monetary policy
and macroprudential measures. Interactions between the two policies have been extensively studied by the recent theoret-
ical literature. Most work in this area uses New-Keynesian Dynamic Stochastic General Equilibrium (DSGE) models with
financial frictions (see, e.g., Loisel, 2014). These models usually consider two authorities that conduct their policies sepa-
rately and independently, focusing on the objective of each. Their results suggest that macroprudential and monetary poli-
cies are complements rather than substitutes, although the results vary for different types of shock. In the wake of a financial
shock, both policies should work in the same direction, even if the reaction in terms of macroprudential policy should be
larger. In the presence of productivity and demand shocks, results suggest that policy responses could differ depending
on the size and nature of the shocks. More recently, DSGE models have gone a step further by explicitly assessing the benefits
of coordination between macroprudential policy and monetary policy. To this end, they differentiate between two cases: the
perfect coordination of policy and the non-coordination of policy. They find that coordinating the two policies stabilises the
effect of real and financial shocks to the macro-environment and maximises social welfare.
Despite the apparent consensus emerging in the theoretical literature about the benefits of synchronisation between
macroprudential and monetary policies, little is known from an empirical perspective. Very few empirical studies have
addressed this issue and their results are far from conclusive (Bruno et al., 2017; Gambacorta and Murcia, forthcoming;
Zhang and Tressel, 2017). Moreover they only focus on a small sample of economies. Against this background, our paper aims
to fill this gap in the existing literature by investigating from a sample of 37 emerging and advanced countries whether the
effectiveness of macroprudential policy is conditional on monetary policy conditions.1
Our findings suggest macroprudential policy is more effective at curbing credit growth when macroprudential and mon-
etary policies are both working in the same direction in harmony. Considering different measures of the macroprudential
stance and using the Taylor gap as a measure of the monetary policy stance, we obtain two important results. First, we find

1
Our selection of sample countries is driven by data availability. Our sample contains countries listed in Table A1 in the Appendix A, and twelve euro area
countries, namely Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal and Spain.
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 3

that a restrictive monetary policy enhances the impact of macroprudential tightening on domestic credit growth. Second, our
results suggest that monetary policy helps to reduce the transmission delay of macroprudential policy actions. To the best of
our knowledge, it is the first empirical paper in the literature that formally confirms the benefits of synchronisation between
macroprudential and monetary policies.
The remainder of the paper is organised as follows. Section 2 reviews the existing empirical literature on the effectiveness
of macroprudential policies and the potential role of monetary policy. Section 3 presents the measures of macroprudential
policy stance and monetary policy stance that we consider. Section 4 presents some descriptive statistics, describes our
econometric approach and discusses our results. A battery of robustness checks is conducted in Section 5, while Section 6
concludes and gives some policy recommendations.

2. Literature review

In this section, we review the existing empirical literature that has analysed how macroprudential policies affect various
measures of financial vulnerability and stability by discussing whether these studies deal with the challenges discussed
above.2 Nabar and Ahuja (2011) and Lim et al. (2011) were the first to use a cross-country analysis to assess the effectiveness
of macroprudential policies. Considering a sample of 49 emerging and advanced economies, Nabar and Ahuja (2011) find that
loan-to-value caps have a negative effect on the growth in housing prices and mortgage lending, while debt-service-to-income
caps only reduce the growth in property lending. Considering a larger set of macroprudential instruments, Lim et al. (2011) find
that macroprudential policies are effective at reducing credit procyclicality. In the same vein, Cerutti et al. (2017a) investigate
whether a more developed macroprudential framework is associated with lower growth in credit and house prices. To this end,
they construct an aggregate macroprudential index designed to measure the number of instruments in place in a given country.
Their results confirm that macroprudential policies are effective at curbing credit growth, especially in developing and emerging
countries, but they do not appear to have a statistically significant effect on the growth in real housing prices.
One important drawback of studies cited above is the way of measuring macroprudential policy. As they focus only on the
existence of macroprudential instruments, these studies do not capture the direction of macroprudential policy actions and
the cross-country heterogeneity in macroprudential activism (Boar et al., 2017). To overcome this shortcoming, some recent
studies go a step further by considering how macroprudential policy evolves over time by being tightened or loosened
(Vandenbussche et al., 2015; Kuttner and Shim, 2016; Zhang and Zoli, 2016; Cerutti et al., 2017b; Akinci and Olmstead-
Rumsey, 2018; Altunbas et al., 2018; Carreras et al., 2018). For instance, using a cumulative macroprudential policy stance
indicator, Akinci and Olmstead-Rumsey (2018) find that macroprudential policy tightening is associated with lower growth
in bank credit, housing credit, and house prices. Their findings also suggest that borrower-targeted macroprudential instru-
ments tend to be more effective at curbing credit growth. Similar results are obtained by Zhang and Zoli (2016) for a sample
of Asian economies, and by Carreras et al. (2018) for the OECD countries.
Other studies focus specifically on how the stance of macroprudential policy affects the real estate market (see for
instance McDonald, 2015; Vandenbussche et al., 2015; Kuttner and Shim, 2016). Kuttner and Shim (2016) assess the relative
effectiveness of macroprudential and housing-related tax policies in curbing housing credit and house prices. They find that
both policies have a negative impact on the growth rate in housing credit. The results are more mixed when they consider
housing price growth as an endogenous variable. Changes in taxes still have a statistically significant impact on house prices,
but this is not the case for some of the macroprudential tools considered, such as the debt service ratio. As argued by Kuttner
and Shim (2016), this result can easily be explained because the cost of buying a house, and consequently demand and prices
in the real estate market, is directly affected by tax policies such as the deductibility of mortgage interest and property taxes,
but not so much by macroprudential tools. More importantly, the results of Kuttner and Shim (2016) confirm that one major
challenge for macroprudential policy is to interact with other policies in a way that fosters the effective conduct of this policy
in pursuit of its objective of financial stability.
A key issue in both the academic literature and the policy debate is the interaction between macroprudential policy and
monetary policy. As discussed in the introduction, each policy can have ‘‘side effects” on the objectives of the other. It is par-
ticularly widely recognised that monetary policy can have side effects on financial stability, for instance when policy rates
are held ‘‘too low for too long”. When monetary policy is very accommodative, there are greater incentives to borrow at low
interest rates that are difficult for macroprudential policy to contain fully. Consequently, an important empirical issue is to
assess how far the monetary policy stance affects how effective macroprudential policy is.
However, there is still very little empirical literature on this issue. To the best of our knowledge, only few studies try to
address this question (Bruno et al., 2017; Gambacorta and Murcia, forthcoming; Zhang and Tressel, 2017). A sample of 12
Asia-Pacific economies over the period 2004–2013 is used by Bruno et al. (2017) to investigate two supplementary issues.
First they distinguish between the pre and post-2007 periods and investigate whether macroprudential policies are synchro-
nised with changes in monetary policy rates.3 They find that before 2007 monetary policy usually changed in tandem with
macroprudential measures, but the opposite result is found after 2007. After 2007, there was a slight downward trend in the

2
For a comprehensive literature review on the effects of macroprudential policy, see Galati and Moessner (2018).
3
Akinci and Olmstead-Rumsey (2018) investigate a similar issue for a sample of emerging and industrialised economies. They find a relatively high
correlation between macroprudential measures and other policy actions.
4 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

average monetary policy rate in the region, while macroprudential measures were tightening slightly. Second, they assess how
effective macroprudential policy measures are in curbing growth in cross-border banking flows. They find that macroprudential
policies effectively reduced banking inflows over the period 2004–2007, but they were not effective after 2007. These findings
indirectly suggest that monetary and macroprudential policies tend to be more successful when they are pulling in the same
direction rather than when they act in opposite directions.
Gambacorta and Murcia (forthcoming) and Zhang and Tressel (2017) investigate this issue more specifically by
assessing whether the impact of macroprudential policies on credit growth depends on monetary policy conditions.
Using meta-analysis techniques and credit registry data for a sample of five Latin American countries, Gambacorta
and Murcia (forthcoming) find macroprudential tools to be made more effective at dampening credit cycles when mon-
etary policies are pushing in the same direction. Indeed, considering the change in the real money rate as a monetary
policy indicator, they find that a macroprudential policy tightening affects credit growth more when it is accompanied
by a countercyclical monetary policy. Zhang and Tressel (2017) adopt a similar approach to gauge this issue for euro
area countries. More precisely, they focus on the loan-to-value (LTV) ratio and assess whether macroprudential policy
is more effective in containing credit growth and housing prices when monetary policy is tightened. They do this by
interacting the LTV ratio with an interest rate gap computed using a Taylor rule. However, their results are relatively
mixed especially since the sign and the significance of estimated coefficients associated with the interaction term are
unstable depending on the lag order.
As we can see, empirical studies that assess the impact of monetary policy conditions on the effectiveness of macropru-
dential policies only focus on a small sample of economies, and their findings are relatively mixed. Against this background,
our paper contributes to the existing literature by explicitly investigating for a large sample of emerging and industrialised
countries whether synchronising macroprudential and monetary policies is a good way to strengthen the impact of macro-
prudential tools on domestic credit growth. Furthermore, in contrast to the existing literature, our empirical analysis con-
siders an extensive set of prudential tools to capture the overall stance of macroprudential policy. In this way, our paper
fills a gap in the literature and provides the first formal answer to an extensive academic and policy debate.

3. Measuring the stance of macroprudential and monetary policies

3.1. Measuring the stance of macroprudential policy

To analyse how effective macroprudential instruments are at curbing the credit cycle, we first need to assess the overall
macroprudential policy stance. Efforts have been made recently in the academic literature to develop datasets that capture
the use of macroprudential policies in a large sample of emerging and industrialised economies.
Two types of dataset can be distinguished. First, some studies consider a large set of macroprudential tools to provide
information on the number of instruments adopted by countries. This lets them give a picture of the evolution of the macro-
prudential policy framework. For instance, the IMF’s 2017 Macroprudential Policy Survey (IMF, 2018) and national sources
are used by Cerutti et al. (2017a) to construct an aggregate macroprudential index in which each instrument considered is
coded as a simple binary variable, equal to 1 if the instrument is in place, and zero otherwise. Their results indicate the
increasing use of macroprudential measures across countries.
Other studies go a step further by providing data on the quarterly changes in macroprudential tools (Vandenbussche
et al., 2015; Kuttner and Shim, 2016; Cerutti et al., 2017b; Akinci and Olmstead-Rumsey, 2018; Alam et al., 2019). The main
objective of these datasets is to use information on easing and tightening of different macroprudential policy instruments to
reflect the policy direction.
In this paper, we use the database provided by Cerutti et al. (2017b), which is one of the most comprehensive datasets on
macroprudential policy actions. Using the same survey as Cerutti et al. (2017a,b) we consider five types of prudential instru-
ments across a sample of 64 countries over the period 2000Q1-2014Q4. The five types of instrument are capital buffers,
interbank exposure limits, concentration limits, loan-to-value ratio limits, and reserve requirements. More precisely, capital
buffers are divided into four sub-indexes: general capital requirements, specific capital buffers related to real estate credit,
specific capital buffers related to consumer credit, and other specific capital buffers. Reserve requirements are also divided
into two sub-indexes of reserve requirements on foreign currency-denominated accounts and reserve requirements on local
currency-denominated accounts.
Cerutti et al. (2017b) then record the number of easing and tightening measures for each type of macroprudential instru-
ment implemented by each country in each quarter. For a given instrument, a tightening action is coded +1 and a loosening
action is coded
1, while 0 means that no change occurs during the quarter. If multiple actions are taken within a given
quarter, the reported values correspond to the sum of all the changes recorded, so tightening and loosening actions taken
within the same quarter cancel each other out. An instrument that is not adopted by a given country is coded as missing
until it is applied by policymakers. Table 1 details the number of events for each macroprudential policy instrument, distin-
guishing between net tightening and net loosening events.4 As can be seen, reserve requirements on local and foreign
currency-denominated accounts and capital requirements are the most frequently used instruments.

4
Table A1 in the Appendix A gives a cross-country perspective of the number of events, by distinguishing between OECD and non-OECD countries.
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 5

Table 1
Macroprudential policy instruments: number of events.

Instruments Target No. of events No. of net tightening events No. of net loosening events
CB REC Lender 33 28 5
CB CC Lender 9 7 2
CB OS Lender 11 7 4
CAP REQ Lender 65 65 0
CONC Lender 15 14 1
IBEX Lender 16 16 0
LTV Borrower 47 33 14
RR FC Lender 56 33 23
RR LC Lender 108 47 61
Total events 360 250 110
[share] [17.86%] [12.41%] [5.46%]

Source: Cerutti et al. (2017b).

Note: CB REC: real estate credit related specific capital buffers; CB CC: consumer credit related specific capital buffers; CB OS: other specific capital buffers;
CAP REQ: capital requirements; CONC: concentration limits; IBEX: limits on interbank exposures; LTV: loan-to-value ratio; RR LC: reserve requirements for
deposit accounts denominated in local currency; RR FC: reserve requirements for deposit accounts denominated in foreign currency. The number of events
is based on our sample of 37 countries from 2000Q1 to 2014Q4.

Given these characteristics of the dataset provided by Cerutti et al. (2017b), we consider six different measures for assess-
ing the stance of macroprudential policies. Two of them, PruC and PruC2, were originally developed by Cerutti et al. (2017b),
and we also propose four alternative measures. These measures aim to give a better view of cross-country differences in
terms of macroprudential policy conduct.
PruC is a country index based on the sum of the quarterly changes of the nine instruments. It can take three different val-
ues:
1, 0, +1. Formally, PruC is defined as follows:
8 X
>
> þ1 if xa;i;t > 0
>
>
>
> a
X
<
PruC i;t ¼ 0 if xa;i;t ¼ 0 ð1Þ
>
> a
>
> X
>
>
:
1 if xa;i;t < 0
a

where subscripts i and t refer respectively to country and time period, while the subscript a represents a given macropru-
dential instrument from among the nine recorded in the database. It is important though to note that the number of instru-
ments considered can vary across countries depending on which instruments have or have not been adopted. As mentioned
above, the absence of legislation authorising the use of a particular macroprudential instrument by policymakers is coded in
the database as missing. xa;i;t reflects the orientation of instrument a in country i at time t. More precisely, for each instru-
ment, it corresponds to the difference between the number of tightening actions and the number of easing actions. Positive
values of xa;i;t indicate a net tightening of the macroprudential policy instrument a, while negative values indicate a net eas-
ing. So if PruC is equal to +1, the overall macroprudential policy framework has been tightened during the quarter, but if PruC
is equal to
1, then the framework has been loosened. PruC being equal to 0 can correspond to two cases: no change in any
instruments, or the same number of tightening and loosening actions during the quarter.
PruC2 is computed in a similar fashion to PruC. The only difference between these two country indexes is the way in
which the orientation of individual macroprudential instruments is recorded. The orientation is now bounded between

1 and +1. For a given quarter, an instrument takes the value +1 if the difference between tightening and loosening actions
is positive,
1 if this difference is negative, and 0 otherwise. PruC2 is computed as follows:
8 X
>
> þ1 if ya;i;t > 0
>
>
>
> a
X
<
PruC2i;t ¼ 0 if ya;i;t ¼ 0 ð2Þ
>
> a
>
> X
>
>
:
1 if ya;i;t < 0
a

where ya;i;t ¼ f
1; 0; þ1g summarises the orientation of the instrument a, in country i at time t. Unlike PruC; PruC2 gives the
same weight to each instrument adopted, whatever the number of tightening or loosening actions taken during a quarter for
any given instrument. PruC2 then corresponds to the difference between the number of tightened instruments and the num-
ber of eased instruments. PruC2 is equal to +1 if the number of tightened instruments during the quarter is higher than the
number of loosened instruments,
1 if the difference between tightened and loosened instruments is negative, and 0
otherwise.
6 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

In addition to the measures proposed by Cerutti et al. (2017b), we compute four alternative country indexes. First, to have
a more granular view of the macroprudential policy stance, we compute an overall index, called PruC3, which corresponds
for a given quarter to the difference between the sum of tightening actions and the sum of loosening actions. Formally, PruC3
is defined as follows:
X
PruC3i;t ¼ xa;i;t ð3Þ
a

where, as in Eq. (1), xa;i;t corresponds for each instrument a in country i at time t, to the difference between tightening and
loosening actions. A larger positive value of this index indicates a more restrictive macroprudential policy, while a larger
negative value reflects a more accommodative policy.
As with PruC and PruC2 proposed by Cerutti et al. (2017b), one shortcoming of the PruC3 index is that it does not take into
account that the number of instruments adopted can differ across countries. Indeed it would be expected that the number of
actions be partly driven by the number of instruments adopted, especially if all the instruments move in the same direction.
To address this issue, we compute an additional index, called PruC4, which is defined as follows:
PruC3i;t
PruC4i;t ¼ ð4Þ
ni;t
where ni;t corresponds to the number of instruments adopted in country i at time t. Like this, PruC4 captures the overall
direction of the macroprudential policy conditional on the number of tools implemented.
However, one potential drawback of the PruC4 index is that we do not distinguish between instruments that have actu-
ally been changed and those for which no action has been taken. To handle this we go a step further by computing an index
that reflects the macroprudential policy stance conditional on the number of instruments actually changed during a given
quarter. This index, called PruC5, is defined as follows:
PruC3i;t
PruC5i;t ¼ ð5Þ
ei;t
where ei;t corresponds to the number of instruments in country i at time t that have effectively been changed.
The last measure that we consider aims to distinguish between tightening and loosening actions. This measure, called
PruC6, is computed as follows:
P P
T xT;i;t L xL;i;t
PruC6i;t ¼ þ ð6Þ
Tighti;t Loosei;t

where xT;i;t corresponds to the recorded value of the macroprudential instrument T that is characterised by a net tightening
during the quarter, while xL;i;t corresponds to the recorded value of the instrument L that is characterised by a net loosening.
Tighti;t and Loosei;t stand for the number of net tightening instruments and the number of net easing instruments. PruC6 is
then complementary to the previous indexes described above, as it reflects both the macroprudential policy stance and the
more or less balanced path of the policy. In comparison to PruC5, the PruC6 index is notably better able to evaluate the stance
of macroprudential policy when some macroprudential tools move in opposite ways. A higher value of this index indicates a
more restrictive macroprudential policy.
To illustrate the pattern of our macroprudential indexes following policy changes, we consider four countries, Argentina,
Colombia, Poland and the Russian Federation, where different macroprudential policy actions have been taken during a given
quarter (see Table 2). As expected, the other macroprudential measures that we propose in this article tend to discriminate
the macroprudential policy stance across countries better than PruC and PruC2 do. Looking at Colombia, we can see that
PruC6 is higher in absolute terms than PruC5. This confirms the relevance of this index when the number of tightening
actions per instrument is lower than the number of loosening actions per instrument.

3.2. Measuring the stance of monetary policy

To assess the monetary policy stance, we need to differentiate between ‘‘rule-based” monetary policy and ‘‘discretionary”
monetary policy. We do this by following the existing literature (see, e.g., Bogdanova and Hofmann, 2012; Bruno et al., 2017)
and using the well-known Taylor rule (Taylor, 1993). The Taylor rule constitutes an approximation of the behaviour of a cen-
tral bank and has become popular in the academic literature for describing the monetary policy stance. The Taylor rule is a
reaction function that defines the central bank interest rate as a function of inflation and a measure of economic activity,
typically the output gap. Comparing the policy rate with the empirically estimated Taylor rate then gives an understanding
of how far policy rate setting has deviated from the Taylor rule.
In line with Bogdanova and Hofmann (2012) and using historical time series for each country in our sample, we estimate
the following reaction function:
it ¼ q it
1 þ ð1
qÞ½a þ bp ðpt Þ þ by ðyt
y
t Þ þ et ð7Þ
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 7

Table 2
Changes in the macroprudential indexes: country case studies.

Argentina Colombia Poland Russian Fed.

2002Q1 2008Q3 2012Q2 2008Q4
Macroprudential RR LC: +5 RR LC: +1 CB REC: +1 RR LC:
3
policy actions RR FC: +5 RR FC:
2 CB CC: +1 RR FC:
3
(+): tightening actions CB OS: +1
(
): loosening actions
No. of adopted inst. 8 9 7 7
No. of tightened inst. 2 1 3 0
No. of loosened inst. 0 1 0 2
PruC 1
1 1
1
PruC2 1 0 1
1
PruC3 10
1 3
6
PruC4 1.25
0.11 0.43
0.86
PruC5 5
0.5 1
3
PruC6 5
1 1
3

Source: Cerutti et al. (2017b) and authors’ calculations.

Note: RR LC: reserve requirements for deposit accounts denominated in local currency; RR FC: reserve requirements for deposit accounts denominated in
foreign currency; CB REC: real estate credit related specific capital buffers; CB CC: consumer credit related specific capital buffers; CB OS: other specific
capital buffers.

where it is the actual short-term policy rate of a given country, which is lagged one period on the right side of Eq. (7) to cap-
ture interest rate smoothing. As in the original Taylor rule, this assumes a gradual adjustment of policy rates to their bench-
mark level. pt is the contemporaneous inflation rate, yt
y
is the output gap, and et is the error term.5 A positive relationship
could be expected between the inflation rate, the output gap, and the policy rate, meaning bp > 0 and by > 0.
The central bank policy rates are taken from the database provided by the Bank for International Settlements (BIS). As
these data are collected on a monthly basis, we consider the end-of-quarter rates. The annual inflation rate comes from
the International Monetary Fund’s International Financial Statistics (IFS) database. Real GDP is taken from the OECD Statistics
for the OECD countries, and from the IFS for the others. The inflation and GDP data series are seasonally adjusted using the US
Census Bureau X-11-ARIMA method. Finally, the output gap corresponds to the difference between actual real GDP and its
trend, computed using the traditional Hodrick-Prescott filter.
Following Clarida et al. (2000) and the related literature, we estimate Eq. (7) using the Generalised Method of Moments
(GMM) estimator to overcome the issue of potential endogeneity. Furthermore, the time period for estimating Eq. (7) covers
the longest available data time span so as to give consistent estimates, and then differs across the countries in our sample. An
important consideration with such an approach is the selection of valid instruments. This selection is based on the overiden-
tification test developed by Hansen (1982). This implies that the set of instruments considered can be different for each
country. The Taylor rule estimates are reported in Table A2 in the Appendix A.
The Taylor gap then corresponds to the difference between the actual policy rate and the estimated Taylor rate (it
i^t ).
This gap reflects the monetary policy stance. A positive difference can be interpreted as a restrictive monetary policy, while a
negative difference can be understood as an accommodative monetary policy.
However, the use of the policy rate as the main monetary policy instrument can be challenged, and some industrialised
economies actually adopted unconventional monetary policies in the aftermath of the 2007–08 financial crisis. This means
that assessing the impact of the unconventional measures implemented and understanding the overall monetary policy
stance using the Taylor gap can be inappropriate. To address this issue, as is usual in the literature, we do not consider
the gap between the actual policy rate and the Taylor rate, but the difference between the shadow rate and the Taylor rate.
The shadow rate, first introduced by Black (1995), has recently been used by a number of papers to quantify the stance of
monetary policy in a ‘‘zero lower bound” environment (see, e.g., Krippner, 2013; Wu and Xia, 2016; Lombardi and Zhu,
2018). Indeed, when the zero lower bound is binding, the policy interest rate does not display meaningful variation and
so no longer conveys information about the monetary policy stance. The shadow rate meanwhile is not bounded and can
freely take on negative values to reflect unconventional monetary policy actions. Krippner (2015) and Wu and Xia (2016)
argue that the shadow rate can be used in place of the policy rate to describe the stance and effects of the monetary policy
in a ‘‘zero lower bound” environment. In this paper, we use the shadow rates provided by Krippner (2015) for the Euro area,
Japan, the United Kingdom and, the United States. These data are available on the website of the Reserve Bank of New
Zealand.

5
Because our sample includes inflation targeting and non-inflation targeting countries, we do not consider the inflation gap in Eq. (7). In most non-inflation
targeting countries, the central bank does not publicly announce a numerical inflation target or the horizon of this target. However, under the assumption that
the target is constant over time, it is captured in the constant term a.
8 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

4. Empirical analysis

Using the insights from the existing literature and the arguments presented above, our empirical analysis aims to gauge
the effectiveness of macroprudential policies in curbing credit growth and whether the monetary policy stance drives this
effectiveness. Given data availability, we do this by considering a sample of 37 advanced and emerging economies over
the period 2000Q1-2014Q4. However, before turning to the econometric analysis, this section presents some descriptive
statistics and preliminary findings.

4.1. Preliminary findings

We start our empirical investigation by analysing in Fig. 1 the evolution of the macroprudential policy stance in our sam-
ple of countries. Panels (A) and (B) focus on the PruC3 index, which is the most explicit indicator for giving a clear picture of
the overall evolution of macroprudential policies. Panel (A) represents the cross-sectional average value of the PruC3 index
for each quarter. The blue bars indicate positive values, while the red bars indicate negative values. As would be expected, it
appears that a broad range of macroprudential instruments have been significantly tightened in the aftermath of the global
financial crisis. However, this does not mean that all the economies in our sample conducted their macroprudential policy in
the same direction. To illustrate this we split in panel (B) our sample of countries into two categories: countries with a net
tightened macroprudential policy stance during a given quarter, and those with a net loosened macroprudential policy
stance. The blue bars correspond to the average value of the PruC3 index for the first category of countries, while the red
bars indicate the average value of the PruC3 index for the second category of countries. A loose macroprudential policy
can be observed in some countries since mid-2009. Panel (C) goes a step further by providing a picture of macroprudential
activism. It represents the number of countries in which the macroprudential policy stance changed over a given quarter,
and it also distinguishes the orientation of this change. The red bars correspond to the number of countries with a net loos-
ened macroprudential policy stance, while the blue bars correspond to the number of countries with a net tightened macro-
prudential policy stance. It shows that macroprudential activism tended to increase over the period considered. Finally,
panel (D) presents the frequency of quarterly observations with a net tightened or a net loosened macroprudential policy
stance. It shows that net tightening actions are more than twice as common as net loosening actions in our sample.
In Fig. 2 we represent the country-by-country cross-correlation coefficient between different lags of the PruC3 index and
the residuals of the annual growth rate of total credit to the private non-financial sector from banks. The residuals are
obtained by regressing credit growth on the annual GDP growth lagged by one period. They capture the part of credit that
is not driven by real economic activity and so can be viewed as a proxy for excess credit growth. Panels (A), (B), (C) and (D)
consider 1, 2, 3 and 4 lags for the PruC3 index respectively. As expected, the cross-correlation coefficient is negative for most
countries, suggesting that a net tightened macroprudential stance is associated with lower credit growth. Furthermore, we
can observe that the number of countries characterised by a negative correlation increases with the lag order and that a
tightening in macroprudential policy is associated with a larger reduction in annual credit growth after one year.
In Fig. 3, we give an overview of the synchronisation of macroprudential and monetary policies by comparing changes in
macroprudential policy with the monetary policy stance. In panels (A), (B) and (C), grey bars represent the number of coun-
tries in each period where both policies exhibit the same stance. Not surprisingly, panel (C) shows that monetary policy
tended to desynchronise from macroprudential policy in the aftermath of the global financial crisis. During this period, most
central banks around the world conducted accommodative monetary policies, and many countries strengthened their
macroprudential framework at the same time [see panel (A)]. This picture is confirmed in panel (D), which represents the
trend of the cross-sectional correlation between the PruC3 index and the Taylor gap for each quarter.6
Finally, in Fig. 4 we analyse whether the monetary policy stance drives the country-by-country cross-correlation between
the PruC3 index and credit growth. To this end, we compute the partial correlation by controlling for the Taylor gap. The
partial correlation coefficients are illustrated by the red points, while the blue points indicate the correlation coefficients
reported in Fig. 2. Overall, the partial correlation coefficients are higher than the pairwise correlation coefficients regardless
of the lag order considered. When the coefficient is negative, this means that taking account of the monetary policy stance
reduces the correlation between the macroprudential policy stance and credit growth. These preliminary findings then con-
firm the importance of monetary policy for the effectiveness of macroprudential policy in curbing credit growth. This issue is
investigated in more detail in the next sub-section.

4.2. Econometric approach

Our empirical analysis proceeds in two steps. First we follow the existing literature by reinvestigating how our different
measures of the macroprudential policy stance affect credit growth. To this end we consider two alternative measures of
domestic credit, which are total credit to the private non-financial sector from banks and total credit to households and
non-profit institutions serving households. These data are taken from the BIS.

6
See, for instance, Borio and Shim (2007) and Bruno et al. (2017) for further evidence. See also the Table A1 in the Appendix A for a cross-country comparison
of the pairwise correlation between the PruC3 index and the Taylor gap.
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 9

Fig. 1. Descriptive statistics on macroprudential policy stance. Source: Cerutti et al. (2017b) and authors’ calculations. Note: All panels are based on our
sample of 37 countries. Panel (A) presents the cross-sectional average value of the PruC3 index for each quarter. In panel (B), the blue bars correspond to the
average value of the PruC3 index of countries with a net tightened macroprudential policy stance, and the red bars correspond to the average value of the
PruC3 index of countries with a net loosened macroprudential policy stance. Panel (C) presents the number of countries in which the macroprudential
policy stance changed over a given quarter by distinguishing between tightened and loosened stances. Panel (D) presents the share of quarterly
observations with a net tightened or a net loosened macroprudential policy stance. No action corresponds to no change in all instruments or to the same
number of tightening and loosening actions during a given quarter.

The baseline model that we estimate is the following:

X
4
DCrediti;t ¼ a þ bk MaP i;t
k þ g X i;t
1 þ h Crisist þ li þ i;t ð8Þ
k¼1

where DCrediti;t is the yearly growth of our different measures of credit, and MaPi;t
k corresponds to our alternative macro-
prudential policy stance indexes, for which we include four lags (see, e.g., Kuttner and Shim, 2016; Zhang and Tressel, 2017),
as some macroprudential actions may be delayed in their effect. X i;t
1 represents the vector of control variables. Following
the existing literature, we consider two control variables, which are the annual GDP growth rate and the change in the nom-
inal monetary policy rate. These two variables are lagged one period and are taken from the IFS database and the BIS respec-
tively. A negative relationship should be expected between the change in policy rate and the growth of credit, while higher
GDP growth should be associated with higher credit growth. The GDP growth rate is included to control for the procyclicality
of credit (Athanasoglou et al., 2014), and this then allows us to capture the part of credit that is not driven by real economic
activity, which is excess credit growth. Crisist is a dummy variable capturing a potential drop in credit growth during the
recent crisis period. It is equal to 1 from 2008Q3 to 2012Q4, and 0 otherwise. Country-fixed effects li allow for cross-
country differences in average credit growth, and i;t is the error term. We could expect bk < 0, meaning that a more restric-
tive macroprudential policy helps to curb domestic credit growth. In the second step, we extend our previous baseline model
to assess whether a tighter macroprudential policy is more likely to curb domestic credit growth when it is accompanied by a
restrictive monetary policy, giving a positive Taylor gap. More precisely, the equation that we estimate is the following:
P
4 P
4  
DCrediti;t ¼ a þ bk MaP i;t
k þ ck MaPi;t
k  TGi;t
k  Di;t
k þ g X i;t
1 þ h Crisist þ li þ i;t ð9Þ
k¼1 k¼1
10 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

Fig. 2. Correlation between the macroprudential stance and credit growth. Source: Cerutti et al. (2017b), Bank for International Settlements and authors’
calculations. Note: All panels are based on our sample of 37 countries. They represent the country-by-country cross-correlation coefficient between the
PruC3 index and the residuals of the growth rate of total credit to the private non-financial sector from banks. The residuals are obtained by regressing
credit growth on GDP growth lagged by one period. Panels (A), (B), (C) and (D) consider 1, 2, 3 and 4 lags for the PruC3 index respectively.

where TGi;t
k corresponds to the Taylor gap described in the previous section and Di;t
k is a dummy variable equal to 1 when
macroprudential and monetary policies are both restrictive in a given quarter, and 0 otherwise. Consequently, the interaction
 
term MaP i;t
k  TGi;t
k  Di;t
k captures the additional effect of macroprudential policies on credit growth conditional on
the stance of monetary policy.7
As we are primarily interested in assessing whether the monetary policy stance is an important determinant of the effec-
tiveness of macroprudential policies, we focus particularly on the marginal effect of our alternative macroprudential stance
indexes on credit growth. Formally, this marginal effect can be derived from Eq. (9) as follows:
d DCrediti;t  
¼ bk þ ck TGi;t
k  Di;t
k ð10Þ
dMaP i;t
k
If we find that bk < 0 and ck < 0, this means that a more restrictive monetary policy reinforces the effect of macroprudential
policies on credit growth. We can also expect the case where bk is not statistically significant at the conventional levels
(bk ¼ 0) and ck < 0. Such a result indicates that credit growth cannot be contained through macroprudential policy alone,
but that this needs the support of monetary policy. In other words, macroprudential tightening actions are more likely to
reduce credit growth if they are implemented in tandem with a restrictive monetary policy.
Finally, we re-estimate Eq. (9) by considering the first difference of the Taylor gap as an alternative measure of the mon-
etary policy stance. The Taylor gap reflects whether a monetary policy is accommodative or restrictive, while its first differ-
ence captures the monetary policy orientation, meaning whether monetary policy has been tightened or loosened. Formally,
the equation that we estimate is the following:

7
It is justified to consider the dummy variable Di;t
k because the macroprudential indexes and the Taylor gap can take positive and negative values. In this
 
case, the estimated coefficient associated with the interaction term MaP i;t
k  TGi;t
k cannot be interpreted properly. Indeed if both variables are positive, a
negative coefficient would be expected, but if both variables are negative, a positive coefficient would be expected. This then justifies the use of a three-way
interaction term to assess whether a tighter macroprudential policy is more likely to curb domestic credit growth when it is accompanied by a restrictive
monetary policy.
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 11

Fig. 3. Synchronisation of the stances of macroprudential and monetary policies. Source: Cerutti et al. (2017b) and authors’ calculations. Note: All panels
are based on our sample of 37 countries. Panels (A) and (B) represent the number of countries with a net tightened and a net loosened macroprudential
policy stance respectively for each quarter. Panel (C) represents the number of countries in which the macroprudential policy stance changed over the given
quarter, whatever the direction of the macroprudential policy. For each of these panels, the grey bars correspond to the number of cases where
macroprudential and monetary policies move in the same direction. Panel (D) represents the trend of the cross-sectional correlation between the PruC3
index and the Taylor gap for each quarter. The trend is obtained using the Hodrick-Prescott filter.

P
4 P
4  
DCrediti;t ¼ a þ bk MaP i;t
k þ ck MaPi;t
k  DTGi;t
k  Ii;t
k þ g X i;t
1 þ h Crisist þ li þ i;t ð11Þ
k¼1 k¼1

where DTGi;t
k is the first difference of the Taylor gap, and Ii;t
k is a dummy variable equal to 1 if the macroprudential
index considered and the first difference of the Taylor gap are both positive in a given quarter, and 0 otherwise. As above,
if monetary policy is an important driver of macroprudential policy effectiveness, we would expect ck < 0.8

4.3. Results

Our results are reported in Tables 3 to 5. In Table 3, we report the results when we consider PruC and PruC2 as alternative
measures of the macroprudential policy stance, Table 4 displays the results obtained with PruC3 and PruC4, while Table 5
displays the results obtained with PruC5 and PruC6. To give a better view of how important the monetary policy stance is
for the conduct of macroprudential policy, we present the results of the baseline and extended models side by side. For each
macroprudential index considered, the first column displays the results obtained when we consider only the effects of
macroprudential policy stance on credit growth. The next two columns report the results when we take the monetary policy
stance into account, which is proxied using two alternative measures, the Taylor gap and its first difference.
We find three important results. First, in line with the recent literature on macroprudential policy, our empirical findings
suggest that an overall tightening in macroprudential policies is associated with a reduction in credit growth. Other than for
two specifications (columns [2.1] and [2.7]), we find a negative and statistically significant relationship between our macro-
prudential indexes and domestic credit growth. Furthermore, and as might be expected, macroprudential policy actions take

8
Please note that we also estimate Eq. (11) by considering the policy interest rate variation as an additional measure of the monetary policy orientation.
Detailed results are available upon request.
12 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

Fig. 4. Partial correlation between the macroprudential stance and credit growth. Source: Cerutti et al. (2017b), Bank for International Settlements and
authors’ calculations. Note: All panels are based on our sample of 37 countries. Red points represent the country-by-country partial cross-correlation
coefficient between the PruC3 index and the residuals of the growth rate of total credit to the private non-financial sector from banks. The residuals are
obtained by regressing credit growth on the one period lagged GDP growth. The partial correlation corresponds to the correlation between the PruC3 index
and credit growth when controlling for the Taylor gap. Blue points are the cross-correlation coefficients reported in Fig. 2. Panels (A), (B), (C) and (D)
consider 1, 2, 3 and 4 lags for the PruC3 index, respectively.

time to curb domestic credit growth effectively. For most specifications, we can see that the coefficients associated with
macroprudential indexes are only significant at the third and fourth order lags.
Second, the results when we add the interaction term in the baseline model show that the monetary policy stance matters
for the effectiveness of macroprudential policy. The coefficients associated with the interaction term are negative and sta-
tistically significant. When we consider the Taylor gap in the interaction term, this negative sign means that a restrictive
monetary policy actually enhances the impact of macroprudential tightening actions on credit growth. Equally, the results
from considering the first difference of the Taylor gap show that the marginal effect on credit growth of tightening macro-
prudential instruments is affected by whether the prevailing monetary policy stance is tight or loose. The benefits of syn-
chronisation between macroprudential and monetary policies are also confirmed in columns [2.2], [2.3], [2.8] and [2.9].
While PruC3 and PruC4 did not appear statistically significant in the baseline specification (columns [2.1] and [2.7]), we
can now observe a significant marginal effect of both indexes on credit growth when macroprudential and monetary policies
complement each other.
Finally, we find evidence that monetary policy helps to reduce the transmission delay of macroprudential policy actions
on private sector credit growth, as the coefficients associated with the interaction term are negative and significant at the
first, second and third order lags. The results are more mixed when we consider the growth of credit to households as a
dependent variable.
In sum, even though monetary and macroprudential policies pursue different primary objectives, our empirical analysis
confirms that the two policies are complementary. Our results particularly emphasise the importance of implementing a
monetary policy that supports the macroprudential policy by moving in the same direction, and then attenuating its poten-
tial side effects on financial stability.
Table 3
Results obtained with PruC and PruC2.

PruC PruC2
Credit to private sector Credit to households Credit to private sector Credit to households
(1.1) (1.2) (1.3) (1.4) (1.5) (1.6) (1.7) (1.8) (1.9) (1.10) (1.11) (1.12)
[Link] 0.972 0.931 0.743
0.377
0.557
0.697 0.980 0.948 0.758
0.416
0.583
0.724

J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156
(1.184) (0.959) (1.032) (1.638) (1.403) (1.437) (1.199) (0.973) (1.044) (1.651) (1.418) (1.449)
[Link]
0.658
0.861
1.007
2.377
2.556⁄
2.645⁄
0.721
0.914
1.060
2.450
2.616⁄
2.704⁄
(1.132) (0.937) (0.958) (1.679) (1.471) (1.467) (1.162) (0.966) (0.987) (1.705) (1.497) (1.489)
[Link]
2.029
2.285⁄
2.475⁄
4.030⁄⁄
4.207⁄⁄
4.356⁄⁄
2.082
2.324⁄
2.513⁄⁄
4.073⁄⁄
4.234⁄⁄
4.382⁄⁄
(1.388) (1.256) (1.227) (1.914) (1.703) (1.657) (1.382) (1.258) (1.229) (1.911) (1.708) (1.662)
[Link]
3.073⁄
3.670⁄⁄
3.675⁄⁄
5.769⁄⁄
5.987⁄⁄⁄
6.058⁄⁄⁄
3.053⁄
3.632⁄⁄
3.637⁄⁄
5.727⁄⁄
5.927⁄⁄⁄
5.998⁄⁄⁄
(1.775) (1.473) (1.425) (2.472) (1.993) (1.968) (1.768) (1.468) (1.420) (2.481) (2.003) (1.977)
L. (MaP  TG  D)
1.156⁄⁄⁄
0.896
1.154⁄⁄⁄
0.891
(0.231) (0.604) (0.231) (0.606)
L2. (MaP  TG  D)
0.914⁄⁄⁄
0.816⁄
0.911⁄⁄⁄
0.811⁄
(0.188) (0.433) (0.189) (0.435)
L3. (MaP  TG  D)
0.824⁄⁄⁄
0.771⁄
0.821⁄⁄⁄
0.767⁄
(0.198) (0.387) (0.198) (0.388)
L4. (MaP  TG  D) 0.215
0.336 0.214
0.337
(0.175) (0.515) (0.175) (0.514)
L. (MaP  D TG  I)
1.078⁄⁄⁄
0.822
1.077⁄⁄⁄
0.818
(0.209) (0.545) (0.209) (0.547)
L2. (MaP  D TG  I)
0.931⁄⁄⁄
0.833⁄⁄
0.928⁄⁄⁄
0.829⁄⁄
(0.136) (0.328) (0.136) (0.330)
L3. (MaP  D TG  I)
0.856⁄⁄⁄
0.806⁄⁄⁄
0.854⁄⁄⁄
0.802⁄⁄⁄
(0.111) (0.271) (0.111) (0.271)
L4. (MaP  D TG  I) 0.135
0.382 0.134
0.382
(0.169) (0.497) (0.169) (0.496)
L.D GDP 2.460⁄⁄⁄ 2.155⁄⁄⁄ 2.142⁄⁄⁄ 2.725⁄⁄⁄ 2.442⁄⁄⁄ 2.429⁄⁄⁄ 2.458⁄⁄⁄ 2.153⁄⁄⁄ 2.140⁄⁄⁄ 2.719⁄⁄⁄ 2.437⁄⁄⁄ 2.423⁄⁄⁄
(0.256) (0.241) (0.234) (0.296) (0.363) (0.347) (0.256) (0.242) (0.234) (0.294) (0.365) (0.348)
L.D Policy rate
0.793 2.415 2.452
2.406
0.340
0.306
0.790 2.408 2.446
2.403
0.355
0.318
(0.722) (1.628) (1.539) (2.173) (4.511) (4.315) (0.725) (1.631) (1.541) (2.177) (4.516) (4.318)
Crisis dummy
4.246⁄⁄⁄
4.707⁄⁄⁄
4.692⁄⁄⁄
5.568⁄⁄⁄
6.119⁄⁄⁄
6.075⁄⁄⁄
4.237⁄⁄⁄
4.698⁄⁄⁄
4.682⁄⁄⁄
5.548⁄⁄⁄
6.098⁄⁄⁄
6.054⁄⁄⁄
(1.442) (1.291) (1.293) (1.802) (1.620) (1.612) (1.442) (1.293) (1.295) (1.801) (1.622) (1.614)
Constant 5.974⁄⁄⁄ 7.356⁄⁄⁄ 7.333⁄⁄⁄ 8.480⁄⁄⁄ 9.737⁄⁄⁄ 9.671⁄⁄⁄ 5.981⁄⁄⁄ 7.360⁄⁄⁄ 7.338⁄⁄⁄ 8.493⁄⁄⁄ 9.743⁄⁄⁄ 9.679⁄⁄⁄
(1.173) (0.957) (0.942) (1.210) (1.171) (1.124) (1.172) (0.957) (0.942) (1.208) (1.175) (1.128)
Observations 2,015 2,015 2,011 1,950 1,950 1,946 2,015 2,015 2,011 1,950 1,950 1,946
Number of countries 37 37 37 37 37 37 37 37 37 37 37 37
Adjusted R-squared 0.248 0.267 0.270 0.249 0.262 0.263 0.248 0.267 0.270 0.249 0.262 0.263

Note: Robust standard errors are reported in parentheses. ⁄, ⁄⁄, and ⁄⁄⁄ denote statistical significance at the 10%, 5% and 1% levels respectively. MaP is the different macroprudential policy indexes considered, TG
corresponds to the Taylor gap, D TG corresponds to the first difference of the Taylor gap, and D and I correspond to the alternative dummy variables capturing the stance of macroprudential and monetary policies.

13
 14
Table 4
Results obtained with PruC3 and PruC4.

PruC3 PruC4
Credit to private sector Credit to households Credit to private sector Credit to households
(2.1) (2.2) (2.3) (2.4) (2.5) (2.6) (2.7) (2.8) (2.9) (2.10) (2.11) (2.12)
[Link]
0.025 0.343 0.345
0.985
0.879
0.855 0.017 2.462 2.459
7.708
7.098
6.932

J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156
(0.719) (0.718) (0.736) (0.993) (1.082) (1.092) (5.470) (5.471) (5.625) (7.828) (8.364) (8.455)
[Link]
0.682
0.502
0.480
1.795
1.596
1.555
4.499
3.103
2.962
13.076
11.516
11.221
(0.662) (0.685) (0.682) (1.130) (1.201) (1.200) (4.889) (4.981) (4.970) (8.719) (9.122) (9.128)
[Link]
1.387
1.083
1.075
2.430⁄
2.108
2.080
10.072
7.906
7.870
18.254⁄
15.966
15.774
(0.825) (0.973) (0.969) (1.247) (1.430) (1.427) (6.189) (7.130) (7.099) (9.717) (10.848) (10.839)
[Link]
1.526
2.020⁄
1.983⁄
3.441⁄
3.332⁄
3.285⁄
11.690
14.994⁄
14.726⁄
26.257⁄
25.305⁄
24.966⁄
(1.269) (1.148) (1.135) (1.941) (1.679) (1.670) (9.217) (8.449) (8.348) (14.269) (12.639) (12.572)
L. (MaP  TG  D)
0.099⁄⁄⁄
0.059
0.779⁄⁄⁄
0.457
(0.024) (0.062) (0.192) (0.504)
L2. (MaP  TG  D)
0.079⁄⁄⁄
0.060
0.634⁄⁄⁄
0.490
(0.015) (0.038) (0.120) (0.306)
L3. (MaP  TG  D)
0.075⁄⁄⁄
0.065⁄
0.602⁄⁄⁄
0.515⁄
(0.014) (0.033) (0.107) (0.263)
L4. (MaP  TG  D) 0.023
0.025 0.175
0.211
(0.017) (0.045) (0.137) (0.369)
L. (MaP  D TG  I)
0.092⁄⁄⁄
0.055
0.726⁄⁄⁄
0.429
(0.022) (0.057) (0.179) (0.460)
L2. (MaP  D TG  I)
0.074⁄⁄⁄
0.058⁄
0.601⁄⁄⁄
0.468⁄
(0.013) (0.034) (0.105) (0.271)
L3. (MaP  D TG  I)
0.071⁄⁄⁄
0.062⁄⁄
0.570⁄⁄⁄
0.492⁄⁄
(0.012) (0.029) (0.090) (0.231)
L4. (MaP  D TG  I) 0.021
0.023 0.160
0.196
(0.016) (0.041) (0.128) (0.338)
L.D GDP 2.466⁄⁄⁄ 2.209⁄⁄⁄ 2.204⁄⁄⁄ 2.706⁄⁄⁄ 2.504⁄⁄⁄ 2.493⁄⁄⁄ 2.466⁄⁄⁄ 2.209⁄⁄⁄ 2.204⁄⁄⁄ 2.710⁄⁄⁄ 2.509⁄⁄⁄ 2.498⁄⁄⁄
(0.267) (0.237) (0.234) (0.297) (0.343) (0.336) (0.267) (0.236) (0.233) (0.298) (0.343) (0.336)
L.D Policy rate
0.733 1.574 1.633
2.362
1.199
1.114
0.766 1.544 1.602
2.384
1.241
1.157
(0.834) (1.508) (1.454) (2.471) (4.191) (4.082) (0.827) (1.524) (1.469) (2.473) (4.242) (4.132)
Crisis dummy
4.294⁄⁄⁄
4.735⁄⁄⁄
4.684⁄⁄⁄
5.738⁄⁄⁄
6.173⁄⁄⁄
6.101⁄⁄⁄
4.303⁄⁄⁄
4.744⁄⁄⁄
4.692⁄⁄⁄
5.742⁄⁄⁄
6.178⁄⁄⁄
6.106⁄⁄⁄
(1.507) (1.309) (1.304) (1.836) (1.601) (1.582) (1.508) (1.311) (1.306) (1.835) (1.601) (1.582)
Constant 5.969⁄⁄⁄ 6.975⁄⁄⁄ 6.931⁄⁄⁄ 8.479⁄⁄⁄ 9.227⁄⁄⁄ 9.166⁄⁄⁄ 5.959⁄⁄⁄ 6.966⁄⁄⁄ 6.922⁄⁄⁄ 8.465⁄⁄⁄ 9.214⁄⁄⁄ 9.152⁄⁄⁄
(1.138) (0.955) (0.951) (1.179) (1.125) (1.109) (1.135) (0.953) (0.949) (1.178) (1.134) (1.118)
Observations 2,015 2,015 2,011 1,950 1,950 1,946 2,015 2,015 2,011 1,950 1,950 1,946
Number of countries 37 37 37 37 37 37 37 37 37 37 37 37
Adjusted R-squared 0.246 0.261 0.261 0.245 0.252 0.252 0.246 0.261 0.261 0.246 0.253 0.252

Note: Robust standard errors are reported in parentheses. ⁄, ⁄⁄, and ⁄⁄⁄ denote statistical significance at the 10%, 5% and 1% levels respectively. MaP is the different macroprudential policy indexes considered, TG
corresponds to the Taylor gap, D TG corresponds to the first difference of the Taylor gap, and D and I correspond to the alternative dummy variables capturing the stance of macroprudential and monetary policies.
Table 5
Results obtained with PruC5 and PruC6.

PruC5 PruC6
Credit to private sector Credit to households Credit to private sector Credit to households
(3.1) (3.2) (3.3) (3.4) (3.5) (3.6) (3.7) (3.8) (3.9) (3.10) (3.11) (3.12)
[Link]
0.014 0.384 0.372
1.318
1.175
1.156
0.015 0.399 0.386
1.323
1.175
1.156

J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156
(1.008) (0.998) (1.023) (1.347) (1.371) (1.388) (1.009) (0.995) (1.020) (1.347) (1.368) (1.385)
[Link]
1.409⁄
1.255
1.240
3.003⁄⁄
2.822⁄
2.776⁄
1.416⁄
1.255
1.240
3.029⁄⁄
2.845⁄
2.799⁄
(0.830) (0.851) (0.858) (1.379) (1.424) (1.431) (0.829) (0.844) (0.851) (1.388) (1.426) (1.434)
[Link]
2.526⁄⁄
2.270⁄
2.288⁄
4.237⁄⁄⁄
3.952⁄⁄
3.940⁄⁄
2.552⁄⁄
2.302⁄⁄
2.319⁄⁄
4.287⁄⁄⁄
4.008⁄⁄
3.996⁄⁄
(1.032) (1.145) (1.143) (1.521) (1.642) (1.643) (1.031) (1.135) (1.133) (1.523) (1.631) (1.634)
[Link]
2.733⁄
3.226⁄⁄
3.182⁄⁄
5.518⁄⁄
5.343⁄⁄
5.297⁄⁄
2.791⁄
3.292⁄⁄
3.247⁄⁄
5.587⁄⁄
5.418⁄⁄
5.371⁄⁄
(1.611) (1.485) (1.467) (2.300) (2.017) (2.006) (1.594) (1.478) (1.460) (2.278) (2.003) (1.992)
L. (MaP  TG  D)
0.199⁄⁄⁄
0.124
0.200⁄⁄⁄
0.125
(0.048) (0.126) (0.048) (0.126)
L2. (MaP  TG  D)
0.157⁄⁄⁄
0.121
0.157⁄⁄⁄
0.121
(0.031) (0.078) (0.031) (0.078)
L3. (MaP  TG  D)
0.150⁄⁄⁄
0.127⁄
0.150⁄⁄⁄
0.127⁄⁄
(0.025) (0.063) (0.025) (0.062)
L4. (MaP  TG  D) 0.041
0.059 0.042
0.058
(0.035) (0.099) (0.035) (0.099)
L. (MaP  D TG  I)
0.186⁄⁄⁄
0.117
0.186⁄⁄⁄
0.117
(0.044) (0.114) (0.044) (0.115)
L2. (MaP  D TG  I)
0.149⁄⁄⁄
0.117⁄
0.150⁄⁄⁄
0.117⁄
(0.026) (0.068) (0.026) (0.068)
L3. (MaP  D TG  I)
0.143⁄⁄⁄
0.122⁄⁄
0.143⁄⁄⁄
0.122⁄⁄
(0.020) (0.054) (0.020) (0.053)
L4. (MaP  D TG  I) 0.036
0.055 0.037
0.055
(0.033) (0.091) (0.033) (0.091)
L.D GDP 2.485⁄⁄⁄ 2.225⁄⁄⁄ 2.218⁄⁄⁄ 2.732⁄⁄⁄ 2.525⁄⁄⁄ 2.514⁄⁄⁄ 2.488⁄⁄⁄ 2.228⁄⁄⁄ 2.222⁄⁄⁄ 2.740⁄⁄⁄ 2.532⁄⁄⁄ 2.521⁄⁄⁄
(0.269) (0.238) (0.234) (0.300) (0.343) (0.335) (0.271) (0.238) (0.235) (0.303) (0.342) (0.334)
L.D Policy rate
0.657 1.763 1.820
2.295
1.086
0.997
0.661 1.767 1.824
2.304
1.090
1.001
(0.794) (1.570) (1.503) (2.418) (4.339) (4.212) (0.789) (1.569) (1.502) (2.418) (4.338) (4.211)
Crisis dummy
4.194⁄⁄⁄
4.631⁄⁄⁄
4.583⁄⁄⁄
5.623⁄⁄⁄
6.069⁄⁄⁄
6.001⁄⁄⁄
4.198⁄⁄⁄
4.634⁄⁄⁄
4.587⁄⁄⁄
5.634⁄⁄⁄
6.079⁄⁄⁄
6.011⁄⁄⁄
(1.514) (1.326) (1.319) (1.870) (1.648) (1.626) (1.517) (1.327) (1.321) (1.873) (1.650) (1.628)
Constant 6.025⁄⁄⁄ 7.060⁄⁄⁄ 7.018⁄⁄⁄ 8.604⁄⁄⁄ 9.386⁄⁄⁄ 9.326⁄⁄⁄ 6.021⁄⁄⁄ 7.058⁄⁄⁄ 7.016⁄⁄⁄ 8.597⁄⁄⁄ 9.381⁄⁄⁄ 9.321⁄⁄⁄
(1.131) (0.950) (0.946) (1.163) (1.127) (1.110) (1.131) (0.950) (0.946) (1.163) (1.126) (1.108)
Observations 2,015 2,015 2,011 1,950 1,950 1,946 2,015 2,015 2,011 1,950 1,950 1,946
Number of countries 37 37 37 37 37 37 37 37 37 37 37 37
Adjusted R-squared 0.250 0.265 0.265 0.252 0.259 0.259 0.250 0.265 0.265 0.253 0.260 0.260

Note: Robust standard errors are reported in parentheses. ⁄, ⁄⁄, and ⁄⁄⁄ denote statistical significance at the 10%, 5% and 1% levels respectively. MaP is the different macroprudential policy indexes considered, TG
corresponds to the Taylor gap, D TG corresponds to the first difference of the Taylor gap, and D and I correspond to the alternative dummy variables capturing the stance of macroprudential and monetary policies.

15
16 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

Fig. 5. Robustness checks: OECD countries sub-sample – Taylor gap.

5. Robustness checks

We check the robustness of our previous findings in five ways. First, we address the fact that our sample includes
advanced and emerging countries by checking whether our initial findings remain unchanged when we focus only on OECD
economies. Second, we check how our results are sensitive to alternative sets of control variables. Third, we take into account
the potential sensitivity of the interest rate gap to the Taylor rule specification. Fourth, we check whether our initial results
are robust when we consider alternative shadow rate measures. Finally, we address the potential endogeneity of macropru-
dential measures.
Sub-sample of OECD countries. Our sample contains 29 OECD countries and 8 non-OECD countries. We then re-estimate
our different specifications by considering only the OECD countries in our sample. Figs. 5 and 6 summarise the results that
we obtain for the alternative macroprudential variables and the interaction terms.9 As we can see, results are very similar to
those obtained on the whole sample. In particular, we still find that the estimated coefficients associated with the interaction
term between the macroprudential policy indexes and our two measures of monetary policy stance is negative and statistically
significant at the conventional levels. This then suggests that the coordination between macroprudential and monetary policies
is especially important in OECD countries.
Alternative sets of control variables. First, we replace the change in the nominal policy rate as a control variable by the
change in the bank lending rate. Indeed, the transmission of monetary policy to credit markets takes time and transmission
delays can differ considerably across countries. A number of empirical studies show that the pass-through of monetary pol-
icy shocks to bank lending rates is usually sluggish and incomplete, and depends on a number of determinants (see, e.g.,
Gigineishvili, 2011; Leroy and Lucotte, 2016; Horvath, 2018). As Illes et al. (2019) argue, there are two main reasons why
bank lending rates diverge from policy rates. First, the policy rate is a very short-term rate, while the lending rates to firms
and households normally reflect longer-term loans. The spread between the lending and policy rates therefore reflects the
maturity risk premium alongside other factors that determine the transmission of monetary policy shocks to bank lending
rates. Second, the policy rate does not represent the marginal cost of funds for banks. Banks obtain funds from a variety of
sources including retail deposits, senior unsecured or covered bond markets and the interbank market, and these differ in
nature from policy rates since they comprise a range of liabilities of different maturities and risk characteristics.
All these reasons certainly explain why our initial findings did not indicate a statistically significant relationship between
the change in the policy rate, lagged one period, and the domestic credit growth. Consequently, it is important to check
whether our initial findings are robust when we replace the nominal policy rate by the bank lending rate as a control vari-
able. The main issue with the use of bank lending rates is the lack of harmonised data for a large sample of advanced and
emerging economies. Then, as it is usual in the bank interest rate pass-through literature (see, e.g., Gigineishvili, 2011),
we use data on bank lending rates from different sources. For the euro area countries, we use the average of the composite
cost-of-borrowing indicators computed by the European Central Bank (ECB). Four basic categories of lending rates are con-
sidered by the ECB: short-term and long-term lending rates both to non-financial corporations and to households for house

9
Detailed results are available upon request.
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 17

Fig. 6. Robustness checks: OECD countries sub-sample – First difference of the Taylor gap.

purchase. For the non-euro area countries, we use the IFS database and, depending on data availability, we complete the
database using data provided by national central banks.
To save space, we do not report the results, but detailed results are available upon request. We find that our main results
are not impacted by the use of this alternative control variable. We still find a negative and significant relationship between
the interaction terms and our measures of domestic credit growth, confirming that a restrictive monetary policy enhances
the impact of macroprudential tightening actions on domestic credit growth. Concerning the bank lending rate, similarly to
the nominal policy rate, we find that the estimated coefficients associated with the change in the bank lending rate are not
statistically significant at the conventional levels.
Second, we add in the set of control variables two alternative measures of capital controls. Indeed, capital controls are a
potential important driver of domestic credit growth. As documented by Ehlers and McGuire (2017), international financial
inflows can induce more rapid credit growth through two main channels. The first refers to the direct cross-border lending,
whereby foreign banks provide credit directly to firms and households, in particular through their branches and subsidiaries.
The second refers to the cross-border interbank credit, whereby domestic banks inside the country finance their domestic
credit activity by funding it with cross-border liabilities. In both cases, the funding that supports the credit to the non-
financial sector is sourced outside the borrower country.
Capital inflows can also drive credit demand by the private sector. As shown by Bremus and Neugebauer (2018) for a large
sample of European small and medium-sized enterprises, credit inflows tend to reduce the financing costs of firms, which
should induce a higher credit demand. Furthermore, Lane and McQuade (2014) argue that a higher level of financial inflows
can also increase domestic credit growth by pushing up domestic asset prices and raising the level of domestic demand in
goods markets, thereby encouraging greater investment and financial acquisitions.
The two alternative measures of financial openness that we consider are the KAOPEN index developed by Chinn and Ito
(2006) and the KAI index of capital controls on inflows developed by Fernández et al. (2016). These two measures have been
widely used in the empirical literature on capital flows. The KAOPEN and the KAI indexes are de jure measures of financial
openness that aim to traduce the degree of restrictions on cross-border financial transactions. They are based on information
regarding restrictions in the International Monetary Fund’s Annual Report on Exchange Arrangements and Exchange Restrictions
(AREAER). These two indexes are normalised between 0 and 1. The higher the value of these indexes is, the more open is the
country to cross-border capital transactions.
We include alternatively the first lag of these two indexes as an additional control variable. To save space, we do not
report the tables of results. Detailed results are available upon request. We find that the estimated coefficients associated
with the KAOPEN and the KAI indexes are not statistically significant at the conventional levels. Nonetheless, this result
could be explained by the low degree of cross-country heterogeneity in terms of financial openness and capital controls.
Indeed, amongst the economies that we consider in our sample, most of them are OECD member countries. The degree of
financial openness in this category of countries is relatively high.10 More importantly, results that we obtain when we consider
these two additional control variables are very similar to our initial findings. In most specifications, we find that the interaction

10
For instance, if we consider the KAOPEN index, 60% of the observations in our sample are equal to 1, which corresponds to a fully liberalized capital account
regime.
18 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

Table 6
Alternative measures of the monetary policy stance.

Benchmark Definition of the benchmark

 
Taylor (1) it ¼ 0:9it
1 þ 0:1 rr t þ p
þ 1:5ðpt
p

Þ þ 0:5y~t
Taylor (2) it ¼ rr t þ p

þ 1:5ðpt
p
Þ þ 0:5y
~t
Taylor (3) it ¼ 1:5ptþ12 þ 0:5y ~
   t 
Taylor (4) it ¼ 0:9it
1 þ 0:1 rr t þ p
þ 1:5ðptþ12
p
Þ þ 0:5y~t
  
Taylor (5) it ¼ it
1 þ Dit , with Dit ¼ 0:5ðptþ12
p
Þ þ 0:5Dy ~t
Interest trend (6) it ¼ HPðit Þ
Equilibrium real rate rr t ¼ Dyt , with yt ¼ HPðyt Þ

Source: Colletaz et al. (2018).

 
~t ¼ yt
yt , with yt ¼ HPðyt Þ. HPðxÞ means Hodrick-Prescott Filter
Note: y
applied to variable x. All measures of the monetary policy stance are computed as
the difference between the actual interest rate it and the corresponding bench-
mark it . p

corresponds to mean inflation over the sample period.

Fig. 7. Robustness checks: results obtained with the median of the alternative Taylor gaps. Note: The results reported are obtained by estimating Eq. (9) and
correspond to the coefficient estimates associated with the alternative macroprudential indexes and interaction terms. All significant coefficients have the
expected negative sign.

term between the macroprudential policy indexes and our two measures of monetary policy stance is negative and statistically
significant at the conventional levels. This then confirms that monetary policy stance matters for the effectiveness of macropru-
dential policy.
Finally, we test whether our initial findings are robust to the inclusion of the VIX index as an additional control variable.
Created by the Chicago Board Options Exchange (CBOE), the VIX is a real-time market index that represents the market’s
expectation of 30-day forward-looking volatility. Derived from the price inputs of the S&P 500 index options, the VIX index
can be viewed as a contrarian sentiment indicator that helps to determine when there is too much optimism or fear in the
market, and is usually used in the empirical literature as a proxy for global risk aversion. We take the VIX index from the
Federal Reserve Economic Data of the Federal Reserve Bank of St. Louis.
As shown by Akinci (2013), real and financial conditions in small open economies are highly correlated with global risk
conditions that are exogenous to these countries. Consequently, including the VIX index as a right-hand side variable aims to
control for global cycles in financial conditions, and then allows to disentangle the effect of macroprudential policies on
domestic credit growth (Akinci and Olmstead-Rumsey, 2018). More precisely, following Akinci and Olmstead-Rumsey
(2018), we include the logarithm of the VIX index as an additional control variable. One would expect a negative relationship
between the VIX index and credit growth. Results that we obtain, available upon request, suggest that the VIX index nega-
tively and significantly impacts the growth rate of total credit to the private non-financial sector, while the estimated coef-
ficients associated with the VIX index appear not statistically significant at the conventional levels when we consider the
growth rate of total credit to households as the dependent variable. This result then suggests that credit to households is
mainly driven by domestic economic and financial conditions. More importantly, our initial findings are not impacted by
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 19

Fig. 8. Robustness checks: results obtained with the first difference of the median of the alternative Taylor gaps. Note: The results reported are obtained by
estimating Eq. (11) and correspond to the coefficient estimates associated with the alternative macroprudential indexes and interaction terms. All
significant coefficients have the expected negative sign.

the inclusion of the VIX index as an additional control variable. We still find that monetary policy stance matters for the
effectiveness of macroprudential policy.
Alternative Taylor rule specifications. We take into account the potential sensitivity of the interest rate gap to the Taylor
rule specification by considering six alternative Taylor rules (see Table 6) and compute the median of the resulting Taylor
gaps.
Figs. 7 and 8 summarise the results that we obtain by considering this alternative measure of the Taylor gap when we
estimate Eqs. (9) and (11).11 Like with the results discussed in the previous section, we find that the tightening of macropru-
dential policy tools leads to a reduction in domestic credit growth, even if macroprudential policy actions seem to take time to
curb credit growth effectively. As before, the estimated coefficients associated with macroprudential indexes are only significant
at the third and fourth order lags for most specifications. More importantly, our results confirm the importance of the monetary
policy stance for the effectiveness of macroprudential policy. Regardless the measure of monetary policy stance considered, we
find that the interaction term is negative and statistically significant at the conventional levels.
Alternative shadow rate measures. Different empirical approaches have been developed to estimate the shadow rate and
gauge the overall monetary policy stance in a zero lower bound environment.12 Most papers in this literature use a shadow
rate term structure model à la Black (1995) to assess the effective stance of monetary policy in the context of unconventional
measures (see, e.g., Christensen and Rudebusch, 2014; Krippner, 2015; Bauer and Rudebusch, 2016; Kortela, 2016; Wu and Xia,
2016; Lemke and Vladu, 2017). However, while all these papers build on the same theoretical approach, they often provide very
different results in practice (Lombardi and Zhu, 2018). As shown by Christensen and Rudebusch (2014) and Krippner
(forthcoming), this is explained by the fact that shadow short rate estimates are very sensitive to model specification, like
the number of factors used for modeling the term structure, the type of approximations chosen, or the way to count for the
lower bound.
Given those sensitivities, we consider different shadow rate measures to capture the stance of monetary policy in coun-
tries where non-standard monetary policy measures have been implemented. As there is no consensus in the literature on
the best shadow rate to be used, we employ an agnostic approach and consider all shadow rates for which we obtained data
from the author(s). In total, we have eight different shadow rate measures. However, as the shadow rates are usually esti-
mated for one specific central bank, the number of shadow rate series considered is different for each central bank. Table A3
in the Appendix A details the shadow short rate measures that we consider, while Fig. A1 in the Appendix A plots the dif-
ferent shadow rate series.
We then compute the median of the different shadow rate series considered, and use this median to calculate the Taylor
gap.13 More precisely, we consider two different Taylor gap series, depending on whether we consider the estimated Taylor rate
to compute the Taylor gap (it
i^t ), or the median of the six alternative Taylor rates described above.

11
Detailed results are available upon request.
12
See Comunale and Striaukas (2017) for an extensive discussion of the different shadow short rate measures developed in the literature.
13
Please note that, for the Japan, only the shadow rate estimates provided by Krippner (2015) are available.
20 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

Fig. 9. Robustness checks: Taylor gap computed as the difference between the alternative shadow rate series and the estimated Taylor rate.

Fig. 10. Robustness checks: first difference of the Taylor gap computed as the difference between the alternative shadow rate series and the estimated
Taylor rate.

Figs. 9–12 summarise the results that we obtain by considering alternative shadow rate series to compute the Taylor
gap.14 Figs. 9 and 10 correspond to the results obtained when we consider the estimated Taylor rate to compute the Taylor
gap, while Figs. 11 and 12 refer to the results obtained when we consider the median of the six alternative Taylor rules to com-
pute the Taylor gap. As we can see, results are very similar to our initial findings with the Krippner (2015) shadow rate series. In
most specifications, we find that the interaction term is negative and statistically significant at the conventional levels. Overall,
the different robustness checks conducted in this section confirm that monetary policy stance matters for the effectiveness of
macroprudential policy, and reinforce the idea that macroprudential and monetary policies complement each other.
Endogeneity issue. Macroprudential tools can depend on domestic credit conditions. Indeed, as Akinci and Olmstead-
Rumsey (2018) argue, one would expect that economies that experienced rapid credit growth are more likely to tighten their
macroprudential policy. The GDP growth and the monetary policy stance are also subject to a similar endogeneity bias. To
address this issue, following Cerutti et al. (2017a) and Akinci and Olmstead-Rumsey (2018), we consider a dynamic panel

14
Detailed results are available upon request.
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 21

Fig. 11. Robustness checks: Taylor gap computed as the difference between the alternative shadow rate series and the median of Taylor rates.

Fig. 12. Robustness checks: first difference of the Taylor gap computed as the difference between the alternative shadow rate series and the median of
Taylor rates.

specification that we estimate using the system GMM estimator (Arellano and Bover, 1995; Blundell and Bond, 1998). This
estimator addresses the two sources of endogeneity bias that we face in our case, namely the correlation between the lagged
dependent variables and country-fixed effects, and the endogeneity of the other right hand side variables.
More precisely, the equation that we estimate is the following:

P
3 P
4 P
4  
DCrediti;t ¼ a þ qk DCrediti;t
k þ bk MaPi;t
k þ ck MaPi;t
k  TGi;t
k  Di;t
k þ g X i;t
1 þ h Crisist þ li þ i;t
k¼1 k¼1 k¼1

ð12Þ

Except the crisis dummy, all the covariates are considered as endogenous. Similarly to the baseline estimates, we alterna-
tively consider the Taylor gap and its first difference to measure the stance of monetary policy. Results that we obtain are
reported in Tables A4 and A5 in the Appendix A. More precisely, in Table A4, we report the results when we consider
PruC; PruC2 and PruC3 as alternative measures of the macroprudential policy stance, while Table A5 displays the results
obtained with PruC4; PruC5 and PruC6. As we can see, correcting for the potential endogeneity bias does not impact our
22 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

initial findings. Indeed, results tend to confirm that monetary policy stance matters for the effectiveness of macroprudential
policy.
More importantly, the reliability of the system GMM estimator largely depends on the validity of the instruments and on
the absence of serial correlation in the residuals. We check the validity of the instruments by using the Hansen (1982) J-test.
The row for the Hansen J-test reports the p-values for the null hypothesis of the validity of the overidentifying restrictions,
i.e. instruments are exogenous. In all specifications, we do not reject the null hypothesis, indicating that our instruments are
valid. The values reported for AR(1) and AR(2) are the p-values for first and second order autocorrelated disturbances in the
first difference equation. As we can see, there is high first order autocorrelation, and no evidence for significant second order
autocorrelation. The results of these tests suggest that our model is correctly specified.

6. Conclusion

Since the 2007–2008 global financial crisis, the conduct of macroprudential policy has raised several important issues.
One of them is the interaction of macroprudential and monetary policies, and this issue is currently at the heart of the aca-
demic and policy debate. It is well-known that monetary policy can have detrimental side effects on financial stability, while
financial stability is the primary objective of macroprudential policy. This means that monetary policy can make macropru-
dential policy less effective at achieving its objective, and this suggests the need for synchronisation.
A growing number of theoretical studies address this issue and confirm the benefits of coordination between the two
policies, but little is known from an empirical standpoint. Our paper fills this gap in the existing literature by providing
the first empirical evidence for a large sample of economies on how monetary policy conditions impact the effectiveness
of macroprudential policy.
More specifically, we obtain two important results. First, we find that a restrictive monetary policy enhances the impact
of macroprudential tightening actions on domestic credit growth. Second, our results suggest that monetary policy helps to
reduce the transmission delay of macroprudential policy actions. Our findings then confirm the complementarities between
the two policies and the potential benefits of coordination highlighted by the theoretical literature.
To translate this result into a policy recommendation, a crucial open question concerns what the appropriate institutional
framework and governance structure for conducting macroprudential policy should be. There is no clear-cut consensus
among economists about this issue and in practice countries have implemented different macroprudential policy frame-
works. While some countries have assigned macroprudential mandates to an independent council, some other countries
have delegated macroprudential regulation to the central bank (see, e.g., Masciandaro, 2018; Masciandaro and Romelli,
2018; Edge and Liang, 2019). This choice of assigning a leading role in macroprudential policy to the central bank is usually
justified by the argument that it will facilitate policy coordination between the two policies. It can also ensure that macro-
prudential policy draws on the expertise of the monetary authority in financial and macroeconomic analysis, and is expected
to facilitate analysis of the side effects of each policy. Finally, as most of the central banks around the world are independent
of the government, it would be expected that such an institutional arrangement would help to protect the macroprudential
policy function from political pressure. This suggests it would be interesting to investigate empirically whether the institu-
tional framework and the governance structure of macroprudential policy are the key drivers of its effectiveness. We leave
this issue for further research.

CRediT authorship contribution statement

José David Garcia Revelo: Conceptualization, Methodology, Software, Formal analysis, Writing - original draft, Writing -
review & editing. Yannick Lucotte: Conceptualization, Methodology, Software, Formal analysis, Writing - original draft,
Writing - review & editing. Florian Pradines-Jobet: Conceptualization, Methodology, Software, Formal analysis, Writing -
original draft, Writing - review & editing.

Acknowledgments

We would like to thank the two anonymous referees and Lawrence Christiano, Sebastian Edwards, Paul Hubert, Michele
Lenza, Grégory Levieuge and Patrick Villieu for their comments and suggestions. We also thank the participants of the 5th
HenU/INFER Workshop on Applied Macroeconomics, the 21st INFER Annual Conference, the 5th International Conference on
Applied Theory, Macro and Empirical Finance, the International Conference TIMTED 2019, the 23rd International Conference
on Macroeconomic Analysis and International Finance, the 2019 Vietnam Symposium in Banking and Finance, the 13th
South-Eastern European Economic Research Workshop, the 4th PANORisk Conference, and of the internal seminar at the
Bank of Lithuania for their useful feedbacks. We are very grateful to Christina Bräuning, Mariarosaria Comunale, Jan Willem
van den End, Hibiki Ichiue, Tomi Kortela, Leo Krippner, Wolfgang Lemke, Marco Lombardi, Renske Maas, Christiaan Pat-
tipeilohy, Jonas Striaukas, Yoichi Ueno, Andreea Vladu, Jing Cynthia Wu, Fan Dora Xia, and Feng Zhu for providing us with
their shadow rate series. This paper was finalised while Yannick Lucotte was a visiting researcher at the Bank of Lithuania.
He would like to thank the Bank of Lithuania for its hospitality and financial support. The views expressed in this paper are
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 23

those of the authors and do not necessarily represent the official views of the Bank of Lithuania or the Eurosystem. Any
remaining errors are ours.

Appendix A

See Tables A1–A5.

Table A1
Macroprudential policy conduct: cross-country comparison.

Countries No. of MaP No. of net No. of net Correlation

events tightening events loosening events PruC3 - TG PruC3 - D TG
OECD countries
Australia 8 5 3
0.01 0.14
Austria 4 2 2
0.01 0.02
Belgium 4 2 2
0.16
0.01
Canada 8 6 2
0.09
0.10
Chile 3 1 2
0.17
0.05
Czech Republic 4 2 2 0.02
0.07
Denmark 6 4 2
0.03
0.03
Finland 5 3 2
0.11
0.07
France 8 6 2 0.02 0.07
Germany 8 6 2
0.07
0.01
Greece 4 2 2
0.16
0.03
Hungary 9 3 6
0.06
0.01
Ireland 7 5 2
0.01
0.05
Israel 10 10 0
0.02 0.05
Italy 6 4 2
0.08 0.04
Japan 2 2 0
0.16
0.05
Korea, Rep. 16 12 4 0.20 0.08
Luxembourg 6 3 3 0.02 0.38⁄
Mexico 4 4 0
0.41⁄
0.38⁄
Netherlands 10 8 2
0.31⁄
0.05
New Zealand 2 2 0
0.06
0.02
Poland 12 9 3 0.01 0.03
Portugal 5 3 2
0.04
0.15
Spain 7 3 4
0.09 0.02
Sweden 8 8 0 0.02
0.09
Switzerland 8 8 0 0.06 0.01
Turkey 38 29 9
0.02
0.01
United Kingdom 3 3 0
0.14 0.06
United States 2 2 0
0.30⁄ 0.14
Mean - OECDy 7.48 5.41 2.07
0.04
0.01
Median - OECD 6.00 4.00 2.00 – –
St. Dev - OECD 6.68 5.34 1.96 – –
Non-OECD countries
Argentina 52 26 26 0.71⁄ 0.62⁄
Brazil 36 22 14 0.38⁄ 0.29⁄
Colombia 9 6 3 0.09
0.02
India 42 26 16 0.40⁄ 0.39⁄
Indonesia 11 9 2 0.14 0.16
Russian Federation 39 22 17
0.23
0.21
South Africa 2 2 0
0.18
0.01
Thailand 11 8 3 0.28⁄ 0.29⁄
Mean - Non-OECDy 25.25 15.13 10.13 0.45⁄ 0.41⁄
Median - Non-OECD 23.50 15.50 8.50 – –
St. Dev - Non-OECD 18.94 9.82 9.40 – –

Source: Cerutti et al. (2017b) and authors’ calculation.

Note: The number of events is based on our sample of 37 countries from 2000Q1 to 2014Q4. The number of MaP events corresponds to the number of
quarters during which the overall macroprudential policy framework has been tightened or loosened. TG corresponds the Taylor gap and D TG corresponds
to the first difference of the Taylor gap. ⁄ indicates statistical significance at the 5% level.
y: The pairwise correlations reported for the OECD and non-OECD countries are computed by considering all the observations in each sub-sample, and do
not refer to the average of individual correlations.
24 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

Table A2
Results of the Taylor rule estimates.

Country q a bp by Hansen test p-value Observations Period of estimation

Argentina 0.554⁄⁄⁄ 9.155⁄⁄⁄
0.171 0.392 0.844 85 1993Q2:2015Q3
Australia 0.667⁄⁄⁄ 1.862⁄⁄⁄ 1.247⁄⁄⁄ 0.818 0.128 161 1976Q2:2017Q3
Brazil 0.841⁄⁄⁄ 4.032 1.345⁄⁄ 2.634⁄⁄⁄ 0.569 81 1996Q2:2017Q3
Canada 0.957⁄⁄⁄
2.432 2.249⁄⁄⁄ 4.054⁄⁄ 0.160 184 1970Q1:2017Q3
Chile 0.532⁄⁄⁄ 3.016⁄⁄⁄ 0.225 0.731⁄⁄⁄ 0.162 54 2003Q1:2017Q3
Colombia 0.692⁄⁄⁄ 1.702⁄⁄⁄ 0.902⁄⁄⁄ 1.047⁄⁄⁄ 0.105 65 2000Q2:2017Q3
Czech Republic 0.924⁄⁄⁄
0.482 0.903⁄⁄⁄ 1.035⁄⁄ 0.140 82 1996Q1:2017Q3
Denmark 0.880⁄⁄⁄
0.845 1.367⁄⁄⁄ 0.880⁄⁄⁄ 0.699 86 1995Q1:2017Q3
Euro Area 0.767⁄⁄⁄ 0.450 0.831⁄⁄⁄ 0.626⁄⁄⁄ 0.169 71 1999Q1:2016Q4
Hungary 0.936⁄⁄⁄
0.555 0.738⁄⁄⁄ 2.647⁄⁄ 0.326 81 1995Q1:2017Q3
India 0.873⁄⁄⁄ 6.768⁄⁄⁄ 0.0381 1.179⁄⁄⁄ 0.679 75 1996Q3:2017Q3
Indonesia 0.894⁄⁄⁄ 7.538⁄⁄⁄
0.0767 1.117⁄⁄⁄ 0.303 80 2005Q3:2017Q3
Israel 0.880⁄⁄⁄ 1.418⁄⁄⁄ 0.723⁄⁄⁄ 0.047 0.283 86 1995Q1:2017Q3
Japan 0.869⁄⁄⁄ 0.178⁄⁄ 0.016 0.087 0.309 72 1994Q1:2013Q1
Korea, Rep. 0.928⁄⁄⁄ 1.070 0.526 2.584⁄⁄⁄ 0.144 69 1999Q2:2017Q3
Mexico 0.561⁄⁄⁄ 2.593⁄⁄⁄ 0.763⁄⁄⁄ 0.521⁄⁄⁄ 0.726 75 1998Q4:2017Q3
New Zealand 0.741⁄⁄⁄ 3.960⁄⁄⁄ 0.805⁄ 1.529⁄⁄ 0.791 117 1988Q1:2017Q3
Poland 0.676⁄⁄⁄ 1.681⁄⁄⁄ 1.275⁄⁄⁄ 0.104 0.228 86 1995Q1:2017Q3
Russian Federation 0.665⁄⁄⁄ 7.356⁄⁄⁄ 0.217⁄⁄⁄ 0.164⁄⁄⁄ 0.254 47 2003Q2:2017Q2
South Africa 0.696⁄⁄⁄ 7.206⁄⁄⁄ 0.464⁄⁄⁄ 1.170 0.240 147 1980Q4:2017Q3
Sweden 0.929⁄⁄⁄ 1.705⁄ 0.914⁄⁄⁄ 0.873 0.437 146 1980Q1:2017Q3
Switzerland 0.917⁄⁄⁄ 0.437 0.763⁄⁄⁄ 0.513⁄⁄ 0.164 186 1970Q1:2017Q3
Thailand 0.897⁄⁄⁄ 0.513 0.847⁄⁄
0.280 0.360 66 2000Q2:2016Q4
Turkey 0.839⁄⁄⁄ 0.771 0.877⁄⁄⁄ 1.368⁄⁄⁄ 0.198 61 2002Q1:2017Q3
United Kingdom 0.900⁄⁄⁄ 2.086⁄⁄ 0.705 1.943⁄⁄ 0.290 113 1989Q1:2017Q3
United States 0.828⁄⁄⁄
1.036 1.735⁄⁄⁄ 0.920⁄⁄ 0.175 188 1970Q1:2017Q3

Note: ⁄, ⁄⁄, and ⁄⁄⁄ denote statistical significance at the 10%, 5% and 1% level respectively.

Table A3
Summary of shadow rate estimates.

Paper Methodology Central Period

bank
Bauer and Mutlifactor shadow rate term structure model, considering both ‘‘yields-only” and macro- ECB 2004Q4-2014Q4
Rudebusch finance shadow-rate models
(2016)
Ichiue and Ueno Vector autoregressive (VAR) model based on survey forecasts of macroeconomic variables. FED 2000Q1-2014Q4
(2018)
Kortela (2016) Multifactor shadow rate term structure model with a time-varying lower bound. ECB 2000Q1-2014Q4
Krippner (2015) Two-factor term structure model estimated using yield curve data. ECB, FED, 2000Q1-2014Q4
BoE, BoJ
Lemke and Vladu Multifactor shadow rate term structure model including the possibility of an occasionally ECB 2000Q1-2014Q4
(2017) changing and possibly negative effective lower bound.
Lombardi and Dynamic factor model based on a large set of variables reflecting unconventional monetary FED 2000Q1-2014Q4
Zhu (2018) policy actions: interest rates, monetary aggregates, Federal Reserve balance sheet.
Pattipeilohy et al. Factor model based on yield curve data. ECB 2004Q4-2014Q4
(2017)
Wu and Xia Multifactor shadow rate term structure model based on a simple analytical representation for ECB, FED, FED, BoE:
(2016) bond prices. BoE 2000Q1-2014Q4
ECB: 2004Q4-
2014Q4

Note: ECB: European Central Bank; FED: Federal Reserve; BoE: Bank of England; BoJ: Bank of Japan. The Bauer and Rudebusch (2016) shadow rate series are
obtained from Comunale and Striaukas (2017).
Table A4
Robustness checks: results obtained with a system GMM estimator.

PruC PruC2 PruC3

Credit to private sector Credit to households Credit to private sector Credit to households Credit to private sector Credit to households
L.D Credit 0.859⁄⁄⁄ 0.929⁄⁄⁄ 0.919⁄⁄⁄ 0.971⁄⁄⁄ 0.858⁄⁄⁄ 0.931⁄⁄⁄ 0.920⁄⁄⁄ 0.980⁄⁄⁄ 0.987⁄⁄⁄ 0.934⁄⁄⁄ 0.997⁄⁄⁄ 0.986⁄⁄⁄
(0.054) (0.032) (0.051) (0.034) (0.052) (0.033) (0.050) (0.031) (0.040) (0.033) (0.035) (0.033)
L2.D Credit
0.098
0.058
0.140 0.023
0.096⁄⁄
0.066
0.120
0.033
0.098
0.033
0.029
0.024
(0.150) (0.078) (0.143) (0.086) (0.039) (0.083) (0.136) (0.070) (0.077) (0.078) (0.079) (0.076)
L3.D Credit
0.005
0.093 0.031
0.172⁄
0.011
0.084 0.004
0.117
0.066
0.132
0.138
0.137
(0.154) (0.077) (0.132) (0.093) (0.016) (0.081) (0.124) (0.075) (0.070) (0.079) (0.088) (0.082)
[Link] 2.076 1.247⁄ 10.157⁄⁄ 0.666 6.417 1.248⁄ 9.698⁄⁄ 0.624 0.818 0.798⁄ 0.560 0.507

J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156
(4.960) (0.645) (4.055) (0.691) (4.737) (0.645) (4.240) (0.694) (1.363) (0.437) (0.482) (0.484)
[Link]
10.111⁄⁄⁄ 0.113
12.271⁄⁄
0.392
10.802⁄⁄ 0.067
12.460⁄⁄
0.446 0.985 0.279 0.101 0.041
(3.052) (0.762) (5.649) (0.878) (5.278) (0.790) (5.559) (0.899) (1.108) (0.538) (0.623) (0.653)
[Link]
1.385
1.083 3.211
1.330
2.591
1.047 3.841
1.389⁄ 1.472
0.659
0.687
0.753
(2.455) (0.814) (3.986) (0.863) (5.964) (0.769) (3.839) (0.807) (1.221) (0.586) (0.685) (0.655)
[Link]
14.379⁄⁄⁄
2.470⁄⁄⁄
14.536⁄⁄⁄
2.800⁄⁄⁄
12.084⁄⁄⁄
2.474⁄⁄⁄
14.037⁄⁄⁄
2.972⁄⁄⁄
1.166⁄⁄
1.319⁄⁄
1.592⁄⁄
1.632⁄⁄
(4.783) (0.739) (3.949) (0.967) (1.871) (0.730) (3.863) (0.889) (0.471) (0.623) (0.680) (0.659)
L. (MaP x TG x D)
1.364⁄⁄
1.465⁄⁄
1.685⁄⁄
1.487⁄⁄
0.030
0.024⁄
(0.668) (0.645) (0.643) (0.716) (0.043) (0.014)
L2. (MaP x TG x D)
0.926⁄
0.818⁄
0.993⁄⁄
0.841
0.034⁄⁄
0.036⁄⁄
(0.529) (0.447) (0.451) (0.503) (0.014) (0.017)
L3. (MaP x TG x D)
0.910⁄
0.989⁄⁄
0.837⁄⁄
1.024⁄⁄
0.036⁄⁄
0.007
(0.526) (0.478) (0.404) (0.500) (0.018) (0.027)
L4. (MaP x TG x D) 1.134⁄⁄⁄ 0.345 0.960⁄⁄⁄ 0.310 0.097⁄⁄⁄ 0.091⁄⁄⁄
(0.248) (0.387) (0.225) (0.405) (0.012) (0.012)
L. (MaP x D TG x I)
0.472⁄⁄
0.340
0.470⁄⁄
0.344
0.040⁄⁄⁄
0.025
(0.176) (0.297) (0.178) (0.305) (0.010) (0.015)
L2. (MaP x D TG x I)
0.283
0.172
0.284
0.201
0.046⁄⁄⁄
0.034⁄
(0.203) (0.261) (0.206) (0.297) (0.011) (0.018)
L3. (MaP x D TG x I)
0.062 0.022
0.061 0.021
0.022
0.011
(0.203) (0.264) (0.203) (0.257) (0.017) (0.023)
L4. (MaP x D TG x I) 0.760⁄⁄⁄ 0.657⁄⁄ 0.763⁄⁄⁄ 0.674⁄⁄ 0.095⁄⁄⁄ 0.085⁄⁄⁄
(0.172) (0.248) (0.173) (0.263) (0.010) (0.013)
L.D GDP 0.367 0.227⁄ 0.048 0.069 0.282 0.227⁄ 0.036 0.057
0.022 0.226⁄ 0.052 0.070
(0.220) (0.122) (0.248) (0.112) (0.249) (0.122) (0.252) (0.105) (0.114) (0.125) (0.104) (0.103)
L.D Policy rate 4.506⁄⁄ 1.424 1.885 0.553 4.519⁄⁄⁄ 1.416 1.918 0.613 1.091 0.704
0.484
0.184
(2.088) (1.079) (1.395) (1.919) (1.174) (1.085) (1.353) (1.894) (0.736) (0.561) (1.120) (1.165)
Dummy crisis
0.767
1.045⁄⁄
1.656⁄
1.052⁄⁄⁄
0.993
1.036⁄⁄
1.647⁄
1.090⁄⁄⁄
1.220⁄⁄⁄
1.162⁄⁄⁄
1.249⁄⁄⁄
1.228⁄⁄⁄
(0.876) (0.403) (0.897) (0.378) (1.054) (0.402) (0.922) (0.375) (0.413) (0.387) (0.371) (0.373)
Constant 3.307⁄⁄⁄ 1.633⁄⁄⁄ 3.352⁄⁄⁄ 1.914⁄⁄⁄ 3.351⁄⁄⁄ 1.605⁄⁄⁄ 3.438⁄⁄⁄ 1.826⁄⁄⁄ 1.481⁄⁄ 1.666⁄⁄⁄ 1.729⁄⁄⁄ 1.744⁄⁄⁄
(1.146) (0.444) (0.857) (0.539) (0.733) (0.459) (0.896) (0.471) (0.566) (0.447) (0.470) (0.466)
Observations 2,012 2,008 1,935 1,931 2,012 2,008 1,935 1,931 2,012 2,008 1,935 1,931
Number of countries 37 37 37 37 37 37 37 37 37 37 37 37
AR(1) test - [Link] 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
AR(2) test - [Link] 0.609 0.160 0.108 0.145 0.238 0.190 0.135 0.161 0.198 0.110 0.172 0.161
Hansen test - [Link] 0.137 0.353 0.192 0.350 0.257 0.265 0.286 0.566 0.246 0.272 0.306 0.301
Number of instruments 25 14 24 14 30 13 24 17 22 13 14 14

Note: Robust standard errors are reported in parentheses. ⁄, ⁄⁄, and ⁄⁄⁄ denote statistical significance at the 10%, 5% and 1% levels respectively. MaP is the different macroprudential policy indexes considered, TG
corresponds to the Taylor gap, D TG corresponds to the first difference of the Taylor gap, and D and I correspond to the alternative dummy variables capturing the stance of macroprudential and monetary policies.

25
 26
Table A5
Robustness checks: results obtained with a system GMM estimator.

PruC4 PruC5 PruC6

Credit to private sector Credit to households Credit to private sector Credit to households Credit to private sector Credit to households
L.D Credit 0.952⁄⁄⁄ 0.934⁄⁄⁄ 0.984⁄⁄⁄ 0.985⁄⁄⁄ 0.886⁄⁄⁄ 0.933⁄⁄⁄ 0.959⁄⁄⁄ 0.982⁄⁄⁄ 0.871⁄⁄⁄ 0.933⁄⁄⁄ 0.960⁄⁄⁄ 0.981⁄⁄⁄
(0.040) (0.033) (0.048) (0.033) (0.049) (0.034) (0.047) (0.033) (0.053) (0.034) (0.052) (0.033)
L2.D Credit
0.112
0.029
0.188⁄
0.018
0.104⁄⁄⁄
0.013
0.187 0.001
0.179
0.013
0.186 0.002
(0.075) (0.077) (0.111) (0.087) (0.036) (0.091) (0.118) (0.079) (0.138) (0.092) (0.118) (0.080)
L3.D Credit
0.042
0.136⁄ 0.041
0.142
0.016
0.156 0.064
0.164 0.065
0.155 0.065
0.165
(0.060) (0.077) (0.102) (0.096) (0.015) (0.099) (0.109) (0.097) (0.149) (0.101) (0.109) (0.098)
[Link] 34.065 6.229⁄
20.579⁄⁄ 3.896 0.565 0.849 2.737 0.363 0.802 0.873 3.236 0.406

J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156
(22.833) (3.105) (9.183) (3.493) (3.137) (0.664) (4.086) (0.719) (3.572) (0.658) (4.305) (0.709)
[Link]
13.239 2.691
4.166 0.943
4.359
0.024
4.326
0.375
4.993
0.041
4.453
0.391
(14.199) (4.067) (10.942) (5.184) (4.738) (0.793) (4.425) (0.913) (5.152) (0.799) (4.672) (0.917)
[Link]
3.910
4.932
2.026
5.880 0.557
0.949 3.916
1.235 1.781
0.990 3.265
1.289
(5.006) (4.350) (10.217) (4.924) (2.939) (0.730) (2.321) (0.806) (1.989) (0.748) (2.316) (0.822)
[Link]
9.093⁄⁄
9.915⁄⁄
25.845
12.361⁄⁄
10.348⁄⁄⁄
1.437
10.479⁄
2.071
10.503
1.452
10.124⁄
2.082
(3.967) (4.356) (18.867) (4.754) (2.193) (1.190) (5.885) (1.344) (6.589) (1.185) (5.869) (1.343)
L. (MaP x TG x D)
0.723⁄⁄
0.577
0.437⁄⁄
0.346⁄
0.484⁄⁄
0.326
(0.295) (0.608) (0.186) (0.187) (0.217) (0.196)
L2. (MaP x TG x D) 0.003
0.940⁄
0.294⁄⁄⁄
0.229⁄⁄
0.325⁄⁄⁄
0.204⁄⁄
(0.286) (0.475) (0.079) (0.089) (0.094) (0.097)
L3. (MaP x TG x D)
0.007
0.514
0.292⁄⁄⁄
0.277⁄⁄
0.356⁄⁄
0.249⁄
(0.137) (0.320) (0.101) (0.130) (0.151) (0.134)
L4. (MaP x TG x D) 0.868⁄⁄⁄ 0.330 0.254⁄⁄⁄ 0.134 0.164 0.137
(0.068) (0.244) (0.032) (0.088) (0.105) (0.084)
L. (MaP x D TG x I)
0.318⁄⁄⁄
0.191
0.082⁄⁄⁄
0.055⁄⁄
0.083⁄⁄⁄
0.056⁄⁄
(0.075) (0.129) (0.018) (0.024) (0.018) (0.025)
L2. (MaP x D TG x I)
0.363⁄⁄⁄
0.274⁄
0.090⁄⁄⁄
0.068⁄
0.090⁄⁄⁄
0.068⁄
(0.092) (0.143) (0.020) (0.037) (0.020) (0.036)
L3. (MaP x D TG x I)
0.173
0.083
0.049
0.025
0.048
0.024
(0.138) (0.193) (0.034) (0.048) (0.034) (0.048)
L4. (MaP x D TG x I) 0.751⁄⁄⁄ 0.680⁄⁄⁄ 0.164⁄⁄⁄ 0.146⁄⁄⁄ 0.165⁄⁄⁄ 0.146⁄⁄⁄
(0.073) (0.099) (0.017) (0.016) (0.017) (0.016)
L.D GDP 0.086 0.221⁄ 0.063 0.069 0.058 0.234⁄
0.068 0.078 0.152 0.236⁄
0.045 0.080
(0.114) (0.124) (0.159) (0.103) (0.170) (0.124) (0.168) (0.101) (0.179) (0.125) (0.172) (0.103)
L.D Policy rate 1.067 0.708 1.708
0.193 5.704⁄⁄⁄ 0.653 2.470⁄
0.122 5.313⁄⁄⁄ 0.658 2.227
0.117
(0.919) (0.565) (1.135) (1.213) (1.161) (0.469) (1.325) (1.025) (1.276) (0.470) (1.428) (1.026)
Dummy crisis
1.287⁄⁄⁄
1.162⁄⁄⁄
1.296⁄⁄
1.240⁄⁄⁄
1.449⁄⁄
1.145⁄⁄⁄
1.691⁄⁄⁄
1.226⁄⁄⁄
1.662⁄⁄
1.145⁄⁄⁄
1.640⁄⁄⁄
1.232⁄⁄⁄
(0.433) (0.391) (0.539) (0.375) (0.620) (0.393) (0.600) (0.368) (0.739) (0.395) (0.600) (0.370)
Constant 1.647⁄⁄⁄ 1.667⁄⁄⁄ 2.304⁄⁄ 1.760⁄⁄⁄ 3.583⁄⁄⁄ 1.743⁄⁄⁄ 2.928⁄⁄⁄ 1.908⁄⁄⁄ 3.569⁄⁄⁄ 1.743⁄⁄⁄ 2.800⁄⁄⁄ 1.913⁄⁄⁄
(0.473) (0.449) (0.861) (0.525) (0.617) (0.447) (0.822) (0.453) (0.971) (0.450) (0.901) (0.457)
Observations 2,012 2,008 1,935 1,931 2,012 2,008 1,935 1,931 2,012 2,008 1,935 1,931
Number of countries 37 37 37 37 37 37 37 37 37 37 37 37
AR(1) test - [Link] 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.000 0.005 0.000 0.002 0.000
AR(2) test - [Link] 0.302 0.101 0.494 0.213 0.584 0.171 0.650 0.184 0.805 0.175 0.660 0.186
Hansen test - [Link] 0.288 0.281 0.104 0.148 0.228 0.277 0.142 0.554 0.126 0.276 0.237 0.556
Number of instruments 19 13 24 12 28 13 22 17 11 13 16 17

Note: Robust standard errors are reported in parentheses. ⁄, ⁄⁄, and ⁄⁄⁄ denote statistical significance at the 10%, 5% and 1% levels respectively. MaP is the different macroprudential policy indexes considered, TG
corresponds to the Taylor gap, D TG corresponds to the first difference of the Taylor gap, and D and I correspond to the alternative dummy variables capturing the stance of macroprudential and monetary policies.
 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156 27

See Fig. A1.

Fig. A1. Comparison of shadow rate series. Note: For Japan, for the periods when no target interest rate was adopted, the policy rate corresponds to the
interest rate applied on excess reserves.

Appendix B. Supplementary material

Supplementary data associated with this article can be found, in the online version, at [Link]
2020.102156.

References

Adrian, T., Shin, H.S., 2010. Chapter 12 – Financial intermediaries and monetary economics. In: Friedman, B.M., Woodford, M. (eds.), Handbook of Monetary
Economics. Elsevier, vol. 3, pp. 601–650.
Akinci, O., 2013. Global financial conditions, country spreads and macroeconomic fluctuations in emerging countries. J. Int. Econ. 91 (2), 358–371.
Akinci, O., Olmstead-Rumsey, J., 2018. How effective are macroprudential policies? An empirical investigation. J. Financial Intermed. 33, 33–57.
Alam, Z., Alter, A., Eiseman, J., Gelos, G., Kang, H., Narita, M., Nier, E., Wang, N., 2019. Digging deeper – evidence on the effects of macroprudential policies
from a new database. IMF WP/19/66. International Monetary Fund.
Allen, F., Gale, D., 2001. Bubbles and crises. Econ. J. 110 (460), 236–255.
Altunbas, Y., Binici, M., Gambacorta, L., 2018. Macroprudential policy and bank risk. J. Int. Money Finance 81, 203–220.
Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation of error-components models. J. Econometrics 68 (1), 29–51.
Athanasoglou, P.P., Daniilidis, I., Delis, M.D., 2014. Bank procyclicality and output: Issues and policies. J. Econ. Bus. 72, 58–83.
Bauer, M.D., Rudebusch, G.D., 2016. Monetary policy expectations at the zero lower bound. J. Money Credit Bank. 48 (7), 1439–1465.
Bernanke, B., Gertler, M., 1989. Agency costs, net worth, and business fluctuations. Am. Econ. Rev. 79 (1), 14–31.
Bernanke, B.S., Gertler, M., 1995. Inside the black box: the credit channel of monetary policy transmission. J. Econ. Perspect. 9 (4), 27–48.
Black, F., 1995. Interest rates as options. J. Finance 50 (5), 1371–1376.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. J. Econometrics 87 (1), 115–143.
Boar, C., Gambacorta, L., Lombardo, G., da Silva, L.A.P., 2017. What are the effects of macroprudential policies on macroeconomic performance? BIS Quart.
Rev.
Bogdanova, B., Hofmann, B., 2012. Taylor rules and monetary policy: a global ‘‘Great Deviation”?. BIS Quart. Rev. September:37–49.
Borio, C., Shim, I., 2007. What can (macro-)prudential policy do to support monetary policy? BIS Working Papers No 242, Bank for International Settlements.
28 J.D. Garcia Revelo et al. / Journal of International Money and Finance 108 (2020) 102156

Borio, C., Zhu, H., 2012. Capital regulation, risk-taking and monetary policy: a missing link in the transmission mechanism?. J. Financial Stabil. 8 (4), 236–
251.
Bremus, F., Neugebauer, K., 2018. Reduced cross-border lending and financing costs of SMEs. J. Int. Money Finance 80, 35–58.
Bruno, V., Shim, I., Shin, H.S., 2017. Comparative assessment of macroprudential policies. J. Financial Stabil. 28, 183–202.
Carreras, O., Davis, E.P., Piggott, R., 2018. Assessing macroprudential tools in OECD countries within a cointegration framework. J. Financial Stabil. 37, 112–
130.
Cerutti, E., Claessens, S., Laeven, L., 2017a. The use and effectiveness of macroprudential policies: New evidence. J. Financial Stabil. 28, 203–224.
Cerutti, E., Correa, R., Fiorentino, E., Segalla, E., 2017b. Changes in prudential policy instruments – a new cross-country database. Int. J. Cent. Bank. 13 (2),
477–503.
Chinn, M.D., Ito, H., 2006. What matters for financial development? capital controls, institutions, and interactions. J. Dev. Econ. 81 (1), 163–192.
Christensen, J.H.E., Rudebusch, G.D., 2014. Estimating shadow-rate term structure models with near-zero yields. J. Financial Econometrics 13 (2), 226–259.
Clarida, R., Gali, J., Gertler, M., 2000. Monetary policy rules and macroeconomic stability: evidence and some theory. Quart. J. Econ. 115 (1), 147–180.
Colletaz, G., Levieuge, G., Popescu, A., 2018. Monetary policy and long-run systemic risk-taking. J. Econ. Dyn. Control 86, 165–184.
Comunale, M., Striaukas, J., 2017. Unconventional monetary policy: interest rates and low inflation. a review of literature and methods. CORE Discussion
Papers 2017/26, Université catholique de Louvain.
Edge, R.M., Liang, J.N., 2019. New financial stability governance structures and central banks. Finance and Economics Discussion Series 2019–019, Board of
Governors of the Federal Reserve System (US).
Ehlers, T., McGuire, P., 2017. Foreign banks and credit conditions in EMEs. BIS Papers No 91, Bank for International Settlements.
Fernández, A., Klein, M.W., Rebucci, A., Schindler, M., Uribe, M., 2016. Capital control measures: a new dataset. IMF Econ. Rev. 64 (3), 548–574.
Galati, G., Moessner, R., 2018. What do we know about the effects of macroprudential policy?. Economica 85 (340), 735–770.
Gambacorta, L., Murcia, A., mbacorta and Murcia, forthcoming. The impact of macroprudential policies in Latin America: an empirical analysis using credit
registry data. J. Financial Intermed. forthcoming.
Gigineishvili, N., 2011. Determinants of interest rate pass-through - do macroeconomic conditions and financial market structure matter? IMF Working
Papers 11/176, International Monetary Fund.
Hansen, L.P., 1982. Large sample properties of Generalized Method of Moments estimators. Econometrica 50 (4), 1029–1054.
Horvath, R., 2018. Financial market fragmentation and monetary transmission in the Euro area: what do we know?. J. Econ. Policy Reform 21 (4), 319–334.
Ichiue, H., Ueno, Y., 2018. A survey-based shadow rate and unconventional monetary policy effects. IMES Discussion Paper Series 18-E-05, Bank of Japan.
Illes, A., Lombardi, M.J., Mizen, P., 2019. The divergence of bank lending rates from policy rates after the financial crisis: the role of bank funding costs. J. Int.
Money Finance 93, 117–141.
Illing, G., 2007. Financial stability and monetary policy – A framework. CESifo Working Paper Series No. 1971, Ifo Institute for Economic Research.
IMF, 2013. The interaction of monetary and macroprudential policies. Technical report, International Monetary Fund.
IMF, 2018The IMF’s annual macroprudential policy survey - objectives, design, and country responses. Technical report, International Monetary Fund.
Kortela, T., 2016A shadow rate model with time-varying lower bound of interest rates. Research Discussion Papers 19/2016, Bank of Finland.
Krippner, L., 2013. Measuring the stance of monetary policy in zero lower bound environments. Econ. Lett. 118 (1), 135–138.
Krippner, L., 2015. Zero lower bound term structure modeling: a practitioner’s guide. Palgrave-Macmillan.
Krippner, L., ippner, forthcoming. A note of caution on shadow rate estimates. J. Money Credit Bank. forthcoming.
Kuttner, K.N., Shim, I., 2016. Can non-interest rate policies stabilize housing markets? evidence from a panel of 57 economies. J. Financial Stabil. 26, 31–44.
Lane, P.R., McQuade, P., 2014. Domestic credit growth and international capital flows. The Scand. J. Econ. 116 (1), 218–252.
Lemke, W., Vladu, A.L., 2017. Below the zero lower bound: a shadow-rate term structure model for the euro area. Working Paper Series 1991, European
Central Bank.
Leroy, A., Lucotte, Y., 2016. Structural and cyclical determinants of bank interest-rate pass-through in the Eurozone. Comparative Econ. Stud. 58 (2), 196–
225.
Lim, C.H., Costa, A., Columba, F., Kongsamut, P., Otani, A., Saiyid, M., Wezel, T., Wu, X., 2011. Macroprudential policy: What instruments and how to use
them? Lessons from country experiences. IMF WP/11/238. International Monetary Fund.
Loisel, O., 2014. Discussion of ”Monetary and macroprudential policy in an estimated DSGE model of the euro area”. Int. J. Cent. Bank. 10 (2), 237–247.
Lombardi, M., Zhu, F., 2018. A shadow policy rate to calibrate US monetary policy at the zero lower bound. Int. J. Cent. Bank. 14 (5), 305–346.
Masciandaro, D., 2018. Central banks and macroprudential policies: Economics and politics. BAFFI CAREFIN Working Papers 1878, Universita Bocconi.
Masciandaro, D., Romelli, D., 2018. Central bankers as supervisors: do crises matter?. Eur. J. Polit. Econ. 52, 120–140.
McDonald, C., 2015. When is macroprudential policy effective? BIS Working Papers 496, Bank for International Settlements.
Nabar, M., Ahuja, A., 2011. Safeguarding banks and containing property booms; Cross-country evidence on macroprudential policies and lessons from Hong
Kong SAR. IMF WP/11/284. International Monetary Fund.
Pattipeilohy, C., Bräuning, C., van den End, J.W., Maas, R., 2017. Assessing the effective stance of monetary policy: a factor-based approach. DNB Working
Papers 575, Netherlands Central Bank.
Taylor, J.B., 1993. Discretion versus policy rules in practice. Carnegie-Rochester Conference Series on Public Policy 39, 195 –214.
Vandenbussche, J., Vogel, U., Detragiache, E., 2015. Macroprudential policies and housing prices: a new database and empirical evidence for Central, Eastern,
and Southeastern Europe. J. Money Credit Bank. 47 (S1), 343–377.
Wu, J.C., Xia, F.D., 2016. Measuring the macroeconomic impact of monetary policy at the zero lower bound. J. Money Credit Bank. 48 (2–3), 253–291.
Zhang, L., Zoli, E., 2016. Leaning against the wind: Macroprudential policy in Asia. J. Asian Econ. 42, 33–52.
Zhang, Y., Tressel, T., 2017. Effectiveness and channels of macroprudential policies: lessons from the Euro area. J. Financial Regul. Compliance 25 (3), 271–
306.