Finance Research Letters xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Finance Research Letters

journal homepage: www.elsevier.com/locate/frl

The impact of monetary policy shocks on stock market bubbles:

International evidence
Petre Caraiani , Adrian Cantemir Cǎlin
⁎

Institute for Economic Forecasting, Romanian Academy, Bucharest University of Economic Studies, Bucharest, Romania

ARTICLE INFO ABSTRACT

Keywords: We extend previous research on monetary policy shocks and their impact on stock market
Stock markets bubbles, by considering a consistent data set of OECD countries in a time-varying BVAR fra-
Bubbles mework. We also take into account the zero lower bound. We further determine whether the
VAR measured impact is related to variables such as the degree of financial development, credit
Monetary policy
market conditions, or the business cycle indicators and consumer confidence.

1. Introduction

A major field of discussion in macroeconomics is whether monetary policy can and should respond to perceived bubbles (understood in
the sense of the deviations of asset prices from their fundamental value, i.e. dividends, see Cochrane (2005) or Gali and Gambetti (2015)).
In the aftermath of the 2008 global financial crisis, both members of academia and policy-making circles have argued that, eventually,
monetary policy should respond to accelerated deviations of asset prices from fundamentals. This came to be known as a ”leaning against
the wind” monetary policy, where the policy-maker responds to asset bubble formation by raising the interest rate. However, as Gali and
Gambetti (2015) argue, there is not yet a consensus, while the empirical evidence supporting either side is rather thin.
On the theoretical side, until recently, the arguments in favor of the leaning against the wind policy rested on the possibility that
raising the interest rate would negatively affect bubble formation. Before the above mentioned crisis, little attention (if any at all) had
been paid to designing monetary policy rules containing monetary or financial variables (Woodford, 2003). Furthermore, the leaning
against the wind monetary policy discussion has ignored the financial side (Taylor, 1998). Some arguments in favor of this type of
monetary policy have been developed by Gambacorta and Signoretti (2014). Using a DSGE model with balance sheets and a bank
lending channel for firms, they show that even if financial stability is not a priority, it is still more desirable in the case of supply
shocks. In contrast, as an argument opposing the leaning against the wind monetary policy, Gali (2014) develops a theoretical
framework, demonstrating that monetary policy shocks have a positive impact on bubbles.
In this paper, we extend previous results by building especially on the work by Gali and Gambetti (2015), who found empirical
support for Gali (2014). First, we consider a wider selection of countries. While Gali and Gambetti (2015) focused on the sole case of
the United States, we consider a wider range of countries that are part of the OECD. Another contribution of this paper is to
specifically address the unconventional monetary policy period in the aftermath of the 2008 global crisis. We use the newly con-
structed series of shadow rates of Wu and Xia (2016) to measure the extent to which the unconventional monetary policy impacted
the bubbles. A previous comment, see Caraiani and Cǎlin (2018), has also outlined the sensitivity of these results to the type of
interest rate used. Finally, this paper also considers the variety of impulse response functions (IRFs) for bubbles in the sample of
OECD countries in a quantitative framework. Using several key macro-financial variables, we appraise which factors are vital in

⁎
Corresponding author.
E-mail address: caraiani@ipe.ro (P. Caraiani).

https://doi.org/10.1016/j.frl.2019.08.016
Received 15 October 2018; Received in revised form 16 August 2019; Accepted 19 August 2019
1544-6123/ © 2019 Elsevier Inc. All rights reserved.

Please cite this article as: Petre Caraiani and Adrian Cantemir Cǎlin, Finance Research Letters,
https://doi.org/10.1016/j.frl.2019.08.016
P. Caraiani and A.C. Cǎlin Finance Research Letters xxx (xxxx) xxx–xxx

understanding the variety of bubble responses to monetary policy shocks in the sample considered.

2. A VAR model

In the following section, we outline the theoretical background for the VAR model. We focus on a time-varying Bayesian VAR. The
approach follows the earlier framework of Primiceri (2005).
The following time-varying specification is used for the autoregressive model:
x t = A0, t + A1, t xt 1 + …+Ap, t xt p + ut (1)
A0,t is a vector of time-varying intercepts, while the matrices Ai,t contain the time-varying coefficients. It is assumed that ut follows
a white noise Gaussian process characterized by a zero mean and a Σt covariance matrix. It is further considered that the innovations
in the reduced form are linear transformations of the structural shocks such that we can write: ut = St t . We also assume that
E { t t } = I and E { t t k} = 0, while St St = t .

3. Data

Our methodology relies on two main data sets. The first supplies our VAR-type approach and consists of the following variables:
dividend yield, stock market index price, GDP, GDP deflator, interest rate and energy prices as a proxy for the dynamics of com-
modities. All the above variables cover the interval between the mid 1990s and 2017 and focus on a selection of eleven OECD
countries. A second dataset is introduced in Section 5 in order to deal with the dynamics of IRFs in the sample. Table A.1 in
Appendix A provides a synthetic presentation of the countries and associated samples.
Dividend yields and prices for each national index have been retrieved from Bloomberg. The next three have been assembled from
the OECD database, while the chosen commodity prices have been extracted from the World Bank database, before being converted
into national currencies. We employed dividend yields, index prices and GDP values in real terms. In addition to this, we incorporated
Wu-Xia shadow rates for the United States, United Kingdom and Euro Area members, (see Wu and Xia (2016)), which are available on
their website. Our country selection procedure is not arbitrary, but imposed by data availability for the above mentioned variables.
The lack of data regarding the dynamics of stock market dividends motivated the removal of several countries from the OECD sample
(e.g. Germany).

4. Results

4.1. Bayesian VAR estimates

We use the following variables in the given order: yt , pt , pte , it , qt , dt which denote log of output, log of the price level, the log of
commodity prices index, the central bank’s interest rate, the log of the main stock market index and the log of real dividends. The
vector of endogenous variables in VAR is given by: x t = [ yt , dt , pt , pte , it , t ]. A key issue is the identification of monetary policy
shocks. Since we extend the original contribution of Gali and Gambetti (2015) (and compare the results to those in this paper), we
follow their approach and set the identification of monetary policy shocks as in Christiano et al. (2005): monetary policy shocks do
not affect contemporaneously the Gross Domestic Product (GDP), dividends and inflation and, furthermore, the central bank does not
respond contemporaneously to innovations in real stock prices.
In this section we focus on analyzing the IRFs based on the estimated time-varying Bayesian VAR. The results of the Bayesian IRFs
are shown in the figures in Appendix B.
We first found that the responses were significant in each case. With respect to the sign of the impulse response functions, we
notice mixed results for the time-varying coefficients. In the short run, the responses of the bubble components are generally positive.
A considerable share of cases are characterized by positive responses initially, however, in the longer run, most of the responses of the
bubble components are negative.
We should note, however, that there is a lot of variation for these responses across OECD economies. Thus, the actual response of
the bubble component to monetary policy shocks seems to depend on the particular context of a certain economy.
In terms of responses around the crisis period, 2008–2009, we also found significant variation, with positive responses for countries
like Canada, Mexico, the UK and the United States, and negative responses, sometimes very strong, for the rest of the countries in the
sample (Chile, France, Ireland, Japan, Korea, Netherlands or Spain). This heterogeneity motivates our further analysis below.

4.2. Regression analysis

In this section, we expand our study by considering quantitative analyses of the estimated impact of monetary policy shocks on
the bubble component of the assets (as given by the difference between asset prices, i.e. the stock market index corresponding to each
country, and the fundamentals, i.e. real dividends). The exercise is motivated by the heterogeneity in the responses of bubbles to
monetary policy shocks and by the legitimate question on what drives these differences.
We consider the determinants of the impulse response functions of bubbles to a monetary policy shock. We use the impulse
response functions for the case of the estimated time-varying Bayesian VAR and construct time series, using them for the sample
period 2000–2017, and thus obtain a consistent number of observations. Since we are interested in looking at various determinants

2
P. Caraiani and A.C. Cǎlin Finance Research Letters xxx (xxxx) xxx–xxx

Table 1
Determinants of monetary policy impact at annual frequency.
Measure IRF short: 4 quarters IRF long: 20 quarters

Fd1 Not Significant Significant: negative

Fd2 Not Significant Not Significant
Fd3 Not Significant Significant: negative
Government Debt Significant: negative Not significant

Table 2
Determinants of monetary policy impact at quarterly frequency.
Measure IRFs: 1 quarter IRFs: 4 quarters IRFs: 20 quarters

BCI Significant: positive Not Significant Not Significant

CCI Significant: negative Not Significant Not significant
CLI Not Significant Not Significant Not significant
Sentix Not Significant Significant: positive Significant: positive
Liquidity Significant: negative Significant: negative Significant: negative

(with some being available only at annual frequencies), we analyze the impulse response functions at both annual and quarterly
frequencies.
We first describe the econometric approach. The following regression is employed1:
irfj, t = 0 + 1 f j, t + j, t (2)

Here irfj,t stands for the estimated impact of a monetary policy shock on the bubble for country j at time t, while β0 stands for the
constant. The variable fj,t is the financial or monetary variable used in the regression and β1 is the coefficient attached to this variable.
As in Ma and Lin (2016), we focus on the fixed effects specification.
Following Belke and Beckmann (2015), we look at the determinants of the impact of monetary policy on bubbles by performing a
bivariate analysis. The motivation behind this approach is that the number of potential explanatory variables is small. Furthermore,
presenting various combinations of explanatory variables would have made the presentation of the full results even more difficult
given the limited available space.
In this section, we rely on a second data set, used along the determined measures for IRFs in a battery of regressions. For this
purpose we firstly computed the initial response, the short-term (the mean of the impulse response function over a 4 quarters horizon)
and the long-term impact (the mean over 20 quarters) for the IRFs. We further used three measures for business cycles, namely the
OECD Composite Leading Indicators, (CLI), the OECD Business confidence index, (BCI), and the OECD Consumer confidence index,
(CCI). These were followed by a government debt to GDP ratio and three measures of financial development, computed on the logic
put forward by Ma and Lin (2016), on World Bank Data. These measures are fractions of Stock Market Capitalization, Domestic Credit
and the sum of the two, to GDP (denoted by fd1, fd2 and fd3 respectively). Furthermore, our data series incorporate specifications for
sentiment indices. The sentiment components (Sentix Indices) have been retrieved from Bloomberg. We employ specific Sentiment
Indices for every geographical area.
We first analyze the data at annual level, see Table 1. We follow this approach in order to use the variables related to financial
development, as proposed earlier in Ma and Lin (2016). We use three different variables of financial development, namely fd1, fd2
and fd3, defined as in Ma and Lin (2016). In addition to this, we incorporate a measure of government debt. The latter is justified
given that a number of countries were either directly or indirectly affected by the European 2012 sovereign debt crisis. An additional
argument points to the recent literature that shows consistent effects of a large debt on macroeconomic outcomes.
We found significant effects of financial development only for the long term impact of monetary policy shocks: the more fi-
nancially developed the country, the more chances of a negative impact of monetary policy on bubbles. This makes sense, since
financially developed markets imply that investors are able to better interpret the signals given by the Central Bank (this rather
reinforces the conventional view regarding the impact of monetary policy shocks on bubbles).
In terms of the impact of government debt, there is some evidence regarding the short run effects of having more debt. The larger
the debt, the stronger the negative response of bubbles to positive shocks in the interest rate. In light of the fact that the sample
includes many European countries affected by the sovereign debt crisis in 2012, this is rather expected.
The second analysis performed focuses on quarterly data, ensuring a larger number of observations. We look at several key
variables: liquidity (considered also by Belke and Beckmann (2015)), business cycle indicators, consumer confidence, consumer
leading indicators and sentiment indices (we report the findings for which enough data were available). The results are presented in
Table 2 below.
In terms of confidence related variables, we find significant effects only for the initial response for CCI and BCI, as well as for the
medium-term response of Sentix indices. The evidence is conflicting, positive for BCI and Sentix, but negative for CCI. For the CCI, the

1
The left hand side is the impact on the bubble from a monetary policy shock. This is a measure of bubble reaction to monetary policy shocks.

3
P. Caraiani and A.C. Cǎlin Finance Research Letters xxx (xxxx) xxx–xxx

evidence signifies that monetary policy shocks lead to a more pronounced decline in bubble size. This implies that the better the
mood of consumers, the better the perceived expectations on the stock market that, however, raise the prospects of hikes in the
interest rates. The prospect of future increasing the interest rate has thus the effect of convincing investors that the gap between asset
prices and their fundamentals will shrink.
The positive responses for BCI and Sentix can also be reconciled with a theoretical framework. More confidence on the side of
firms and investors leads to better expectations regarding the market. If the central bank raises the interest rate, it might also signal
that it expects further increases in asset prices, thus paradoxically increasing (but only in the short run) confidence in the stock
market performance and continuing to fuel the bubble formation.
We also found a negative impact of liquidity: the higher the liquidity, the more negative the impact of monetary policy shocks on
bubbles. In an increased volatility environment, as in times of crisis or uncertainty, a rising interest rate leads to a more negative
impact on bubbles, which is a close to what one would expect. For this particular case, the evidence is robust across the different time
horizons considered.

5. Conclusion

In this paper, we revisit this key issue of whether monetary policy shocks do or do not have a negative impact on the bubble
component of asset prices (Table A.1). Our results point to heterogeneous responses for the recent period, post 2000 (Figs. B.1–B.11).
When taking a closer look at the determinants of the impact of shocks on bubbles, we also found that a number of factors have
explanatory power (the level of financial development, consumer confidence or liquidity).

Appendix A. Dataset

Table A1
Data sample.
Country BVAR model Regression Frequency

Canada 1993:1 – 2017:2 2001:2 – 2017:2 Annual/Quarterly

Chile 1996:1 – 2017:3 2003:2 – 2017:3 Annual/Quarterly
France 1993:2 – 2017:3 2001:2 – 2017:3 Annual/Quarterly
Ireland 1997:1 – 2017:3 2004:2 – 2017:3 Annual/Quarterly
Japan 1994:1 – 2017:3 2001:2 – 2017:3 Annual/Quarterly
Korea 1996:4 – 2017:3 2001:2 – 2017:3 Annual/Quarterly
Mexico 1994:3 – 2017:3 2004:1 – 2017:3 Annual/Quarterly
Netherlands 1993:2 – 2017:3 2001:4 – 2017:3 Annual/Quarterly
Spain 1995:1 – 2017:3 2002:2 – 2017:3 Annual/Quarterly
UK 1993:2 – 2017:3 2001:3 – 2017:3 Annual/Quarterly
US 1993:1 – 2017:3 2000:2 – 2017:3 Annual/Quarterly

Appendix B. The impact of monetary policy shocks using Bayesian VARs: time-varying responses

Fig. B1. Canada.

4
P. Caraiani and A.C. Cǎlin Finance Research Letters xxx (xxxx) xxx–xxx

Fig. B2. Chile.

Fig. B3. France.

Fig. B4. Ireland.

5
P. Caraiani and A.C. Cǎlin Finance Research Letters xxx (xxxx) xxx–xxx

Fig. B5. Japan.

Fig. B6. Korea.

Fig. B7. Mexico.

6
P. Caraiani and A.C. Cǎlin Finance Research Letters xxx (xxxx) xxx–xxx

Fig. B8. Netherlands.

Fig. B9. Spain.

Fig. B10. United Kingdom.

7
P. Caraiani and A.C. Cǎlin Finance Research Letters xxx (xxxx) xxx–xxx

Fig. B11. United States.

References

Belke, A., Beckmann, J., 2015. Monetary policy and stock prices cross-country evidence from cointegrated VAR models. J. Bank. Finance 54 (C), 254–265.
Caraiani, P., Cǎlin, A.C., 2018. The effects of monetary policy on stock market bubbles at zero lower bound: revisiting the evidence. Econ. Lett. 169 (C), 55–58.
Christiano, L.J., Eichenbaum, M., Evans, C.L., 2005. Nominal rigidities and the dynamic effects of a shock to monetary policy. J. Politic. Econ. 113 (1), 1–45.
Cochrane, J., 2005. Asset Pricing. Princeton University Press.
Gali, J., 2014. Monetary policy and rational asset price bubbles. Am. Econ. Rev. 104 (3). 721–52
Gali, J., Gambetti, L., 2015. The effects of monetary policy on stock market bubbles: some evidence. Am. Econ. J. Macroecon. 7 (1). 233–57
Gambacorta, L., Signoretti, F.M., 2014. Should monetary policy lean against the wind? J. Econ. Dyn. Contr. 43 (C), 146–174.
Ma, Y., Lin, X., 2016. Financial development and the effectiveness of monetary policy. J. Bank. Finance 68 (C), 1–11.
Primiceri, G.E., 2005. Time varying structural vector autoregressions and monetary policy. Rev. Econ. Stud. 72 (3), 821–852.
Taylor, J.B., 1998. An Historical Analysis of Monetary Policy Rules. Working Paper. National Bureau of Economic Research.
Woodford, M., 2003. Interest and Prices. Princeton University Press.
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.