You are on page 1of 4

Economics Letters 145 (2016) 258261

Contents lists available at ScienceDirect

Economics Letters
journal homepage: www.elsevier.com/locate/ecolet

Are financial markets less responsive to monetary policy shocks at the


zero lower bound?
Wenbin Wu
UCSD, 9500 Gilman Dr #0508, La Jolla, CA 92092-0508, United States

highlights
We study the time-varying effect of monetary policy shocks on financial markets.
The corporate bond market is highly responsive to monetary policy shocks at the zero lower bound.
The long-term Treasury bond market is highly sensitive to monetary policy shocks throughout 19902012.
The short-term Treasury bond market is severely constrained by the zero lower bound.
The stock market is less responsive to monetary policy shocks from 2008 to 2010.

article info abstract


Article history: This paper investigates the time-varying effect of monetary policy shocks on financial markets. We show
Received 8 October 2015 that the corporate bond market is highly responsive to monetary policy shocks throughout 20002012,
Received in revised form implying a high pass-through of policy-induced movements in Treasury yields to private yields even
9 December 2015
during the zero lower bound period. While the long-term Treasury bond market is highly sensitive to
Accepted 1 July 2016
Available online 7 July 2016
monetary policy shocks throughout almost the entire sample, the short-term Treasury bond market is
severely constrained by the zero lower bound. The stock market is less responsive from 2008 to 2010, but
JEL classification:
the responsiveness bounces back rapidly in 2011.
E44 2016 Elsevier B.V. All rights reserved.
E52
G12

Keywords:
Monetary policy
Zero lower bound
Financial market

1. Introduction policy instrument. At the zero lower bound (ZLB),1 the Fed turns
to other unconventional instruments (for example, large-scale
Many studies have documented that monetary policy shocks asset purchases, forward guidance, and operational twist).2
have an important impact on the stock market, Treasury yields, (2) An outstanding open question is whether or not monetary
and corporate yields (see, for example Thorbecke, 1997, Rigobon policies become less powerful over time, especially at the ZLB.
and Sack, 2004, Bernanke and Kuttner, 2005, Wright, 2012, Kiley,
2013, 2014 and Gilchrist et al., 2015). However, few of the
existing studies have further investigated the time-varying effect 1 The zero lower bound refers to the period during which the fed funds rate is set
of monetary policies on these variables. In this paper, we estimate at the range between 0 and 25 basis points.
the time-varying effect of monetary policy shocks on a range of 2 The Fed funds rate is no longer an effective tool at the ZLB. In order to
economic and financial variables using a similar approach to the lower long-term interest rates to give more stimulus to the economy, the Fed
one employed by Swanson and Williams (2014). conducted several rounds of large-scale asset purchases (LSAPs), where it purchased
a large amount of Treasury bonds, agency debt and mortgage backed securities
The consideration of the time-varying effect of monetary policy
(MBS), and other securities with medium to long maturity. The Fed also used
shocks is important because: (1) The way the Fed makes its move is other unconventional policy instruments to influence the economy, including: (1)
evolving over time. Conventionally, the fed funds rate serves as a forward guidance, where the Fed promised to keep the Fed funds target rate low
for a long period of time in order to affect the expectation of future rates; and
(2) operation twist, where the Fed sold a large amount of short-term bonds and
used the proceeds to buy long-term bonds in an effort to bring down their term
E-mail address: wew045@ucsd.edu. premiums.

http://dx.doi.org/10.1016/j.econlet.2016.07.001
0165-1765/ 2016 Elsevier B.V. All rights reserved.
W. Wu / Economics Letters 145 (2016) 258261 259

One way to look at this problem is to investigate the time-varying We study the impact of monetary policy shocks on three
responsiveness of the economic and financial variables to the markets (six variables): the corporate bond market (AAA yields
monetary policy shocks. and BAA yields), the Treasury bond market (2 year Treasury yields
Using the methodology developed in Swanson and Williams and 10 year Treasury yields), and the stock market (S&P 500
(2014), we show that the sensitivities of all these measures to index and VIX index).6 We next specify the steps to estimate
monetary policy shocks vary over time. The corporate bond market the time-varying sensitivity of a economic variable to monetary
remains highly responsive to monetary policy shocks throughout policy surprises. Following Swanson and Williams (2014), we first
the entire sample, implying that the Treasury yield changes estimate this sensitivity over a benchmark sample, 19902000,
induced by monetary policy shocks are largely passed through which is supposed to be free from the ZLB restriction. We
to private yields during the ZLB period. The long-term Treasury next estimate the rest of the sample, 20012012, which is then
bond market is highly responsive at the ZLB,3 but the short-term compared to the benchmark case to determine whether or not the
Treasury bond market is severely constrained by the ZLB. The stock power of monetary policy surprises decreases at the crisis or the
market exhibits weaker responses from 2008 to 2010 compared to ZLB.
the normal period (which will be clear in the next section), but Our model of measuring the sensitivity of an economic variable
the sensitivity bounces back quickly in 2011. ht to monetary policy shocks Mt takes the form of
Related Literature: The paper most relevant to mine is Swanson
1ht = + Mt + t (1)
and Williams (2014). They develop a new method of measuring
the time-varying sensitivity of interest rates to a range of where t indexes days and t is an error term.
macroeconomic announcements. We find that this methodology To measure the time-varying sensitivity i (i = 19902012),
is also useful in investigating the power of monetary policy shocks we run regressions year by year from 1990 to 2012.7 We estimate
at the ZLB. Kiley (2013) and Gilchrist et al. (2015) also examine the time-varying regression of the form
the pass-through from Treasury yields movement induced by
monetary policies to private yields. My work complements theirs 1ht = adi + di bMt + t (2)
by allowing the pass-through to vary over time. where a and are time-varying parameters, b is the constant
di di
The remainder of this paper is organized as follows: in Section 2, part of the sensitivity, i indexes years,8 and d indexes days within
we describe the data and present the methodology. In Section 3, we year i. Our focus is di , which measures the time-varying sensitivity
report the results. In Section 4, we present our conclusions. of ht to monetary policy surprises Mt . Note that in order to
separately identify di and b, we need to normalize di . Following
2. Data and methodology Swanson and Williams (2014), we normalize the average of di over
199020009 to be 1. In the subsequent periods, if di exceeds 1,
Kuttner (2001) and Grkaynak et al. (2005) show that economic variable h is more sensitive to monetary policy shocks compared
and financial variables only respond to unanticipated changes in to that of 19902000; if di is smaller than 1, variable h becomes
monetary policies. We therefore follow the convention by using less sensitive to monetary policy shocks.
Federal Open Market Committee (FOMC) announcements and In order to determine finer estimates of di , we follow Swanson
minutes4 as events for identifying monetary policy surprises. We and Williams (2014) by estimating daily rolling regressions as
first document the daily changes of 1, 2, 5, 7, and 10 year Treasury follows:
yields around these event dates. Next, we extract a factor from
rolling three year samples of these yield changes.5 The factor is
1ht = ad + d M
t + t (3)
then normalized to have 1 to 1 relationship with 2 year Treasuries where M t = bM
t and b is estimated from the regression (2). The
and used to measure the monetary policy surprises. The reason regression (3) estimates d for each day from Jan 1990 to the end
that we do not use short-end Treasuries or the fed funds rate is of sample over one-year rolling windows. Because d is estimated
that these interest rates essentially are constrained at the ZLB, at the second stage (b is estimated at the first stage), we also take
while the longer term interest rates remain very flexible. Therefore, into account this two-stage estimation error following Swanson
many recent studies use changes in long term interest rates to and Williams (2014) when the standard error is calculated.
measure the stance of monetary policy shocks in order to be able
to capture the variation of monetary policy shocks at the ZLB (see,
3. Estimation results
for example, Wright, 2012, and Kiley, 2014). The data used in this
study are downloadable from the website of the Federal Reserve
Table 1 reports the results for the regression (1) over the
Bank of St. Louis. The release dates of FOMC minutes (19962012)
normal sample from 19902000 (results are very similar for
can be acquired from the website of the Federal Reserve Board. We
the entire sample from 19902012). These results are robust to
pin down other release dates (19901995) by looking up news in
whether or not we add lags of Mt . Note again that a one unit
the Factiva Database.
increase in monetary policy shock is normalized to increase 2 year

3 It reacted less in the Great Recession periods (20072008), during which the
6 At first, we also wanted to look at TIPS and breakeven inflation rates, but the
fed funds rate was higher than the ZLB.
4 We must thank the referee to point out that the sample size is small if lengths of these samples are too short.
7 As pointed out by Swanson and Williams (2014), this approach may deliver
only FOMC announcement dates are used. While the results are similar with or
without FOMC minute dates, it is meaningful to add them because the sample size volatile estimates because of the small sample problem. Swanson and Williams
becomes larger. As noted in Rosa (2013), FOMC minutes do contain important new (2014) deal with this small-sample problem by imposing a restriction that the
information about monetary policies. relative magnitude of sensitivity for different macroeconomic announcements are
5 The referee pointed out that it is problematic to extract the first principal constant over time. As discussed in footnote 3, we overcome the small sample
problem by including FOMC minute dates.
component using the covariance matrix of the data over the entire sample, because
8 i {1990, 1991, . . . , 2012}.
there is a sharp break in the correlation matrix before and after the ZLB (see Kiley,
2014). Therefore, we extract the first principal component from rolling three year 9 As noted by Swanson and Williams (2014), this period is supposed to be a
samples. It is worth noting that the results are quantitatively and qualitatively normal period during which monetary policies are not constrained by the zero
similar if we use the entire sample. lower bound.
260 W. Wu / Economics Letters 145 (2016) 258261

Table 1
Coefficient estimates from the linear regression (1) over 19902000.
Variables AAA yields BAA yields 2 yr Treasuries 10 yr Treasuries SP500 VIX

MP shock 0.5431 0.5723 1.1285 0.9031 4.4505 0.2119


(0.0860) (0.0699) (0.0475) (0.0686) (1.2886) (0.0773)
Obs. 5731 5731 5729 5729 5719 5713
Event dates 366 366 366 366 366 366
R2 0.02 0.02 0.05 0.03 <0.01 <0.01
Note: MP shock denotes monetary policy shock. It is measured by the first principle component of three-year rolling sample of the daily changes of 1, 2, 5, 7, 10 year Treasury
yields around the FOMC announcement and minute dates. It is normalized to have 1 to 1 relationship with 2 year Treasuries. Heteroskedasticity-consistent standard errors
in parentheses.

Table 2
Constant coefficient b estimates from the time-varying regression (2).
Variables AAA yields BAA yields 2 yr Treasuries 10 yr Treasuries SP500 VIX

MP shock 0.6048 0.5822 1.0812 0.9746 9.0091 0.2516


(0.0542) (0.0539) (0.0349) (0.0442) (3.0047) (0.1575)
Obs. 5731 5731 5729 5729 5719 5713
Event dates 366 366 366 366 366 366
R2 0.0306 0.0318 0.0659 0.0560 0.0112 0.0082
H0 : constant 0 0 0 0 0 0
Note: MP shock denotes monetary policy shock. It is measured by the first principle component of three-year rolling sample of the daily changes of 1, 2, 5, 7, 10 year Treasury
yields around the FOMC announcement and minute dates. It is normalized to have 1 to 1 relationship with 2 year Treasuries. Heteroskedasticity-consistent standard errors
in parentheses. constant is for the hypothesis that d = 1 for all years in the sample.

Treasuries by the same amount. Over the period of 19902000, Panel C of Fig. 1 presents the results for 2 year Treasuries. Two
a 1% increase in monetary policy shock causes AAA yields, BAA year Treasury yield is very insensitive to monetary policy shocks
yields, 2 year Treasuries and 10 year Treasuries to increase by at the ZLB, reflecting a large constraint of the ZLB on the short-
0.5431%, 0.5723%, 1.1285% and 0.9031%, respectively, all of which term bond yields. However, the ZLB poses little constraints on the
are significant. As expected, it increases the VIX index by 0.2119%, yields of long term bonds as can be seen from Panel D of Fig. 1.12
but lowers the stock market return by 4.4505%. At the ZLB, the traditional policy instrument fed funds rate
was no longer effective. Thus, the Fed conducted unconventional
The estimation results in Table 2 show the coefficient estimates
monetary policies during this period to bring down long-term
b from the regression (2). Results are similar to those shown in
interest rates. That is the reason that long-term interest rates are
Table 1. The estimates for the sensitivity of the stock return and
still quite responsive to monetary policy shocks at the ZLB, while
VIX index increase. We also report the R squares and the p value short-term interest rates are largely constrained.
of testing whether di is constant over time.10 Consistent with the Fig. 2 presents the results for the VIX index and the S&P 500
model, the hypotheses that di is constant over time are rejected index, which strongly contrast with those seen in Fig. 1. The
for all specifications. Great Recession and the ZLB appear to seriously constrain the
Figs. 1 and 2 report the time-varying sensitivity coefficients di power of monetary policies to affect the stock market. Although
from the regression (3). The blue solid line in each plot represents the estimates are quite volatile, they largely remain over 1
the estimated value of di on each date d. The dotted lines represent before the ZLB for both the VIX index and the S&P 500 index.
the 95% confidence intervals along time. The horizontal line at 1 During 20082010, the sensitivity drops immensely, implying the
is drawn for each panel in order to contrast with the benchmark limitation of monetary policies on affecting the stock market.
period. The lightly shaded regions (yellow) depict periods during However, the sensitivity bounces back quickly after 2010 when the
which di is significantly below 1 but significantly larger than economy starts to recover.
zero. The red shaded regions denote periods during which di is
4. Conclusions
significantly below 1 but does not differ significantly from zero.
Panel A of Fig. 1 depicts the sensitivity of AAA yields to
Have monetary policy shocks become less powerful over the
monetary policy shocks. It shows that the sensitivity does not period of 20002012, especially during the ZLB period? This paper
fall significantly below 1 at the ZLB though it varies over attempts to answer this question by studying the time-varying
time. The sensitivity actually stays mostly above 1 during the effect of monetary policy shocks. We find that the corporate
ZLB period. Wright (2012) also found that corporate bond bond market stays highly sensitive to monetary policy shocks
yields responded strongly to unconventional monetary policy throughout the Great Recession and the ZLB. While the 2 year
shocks. We complement Wright (2012) by pointing out that this Treasury yield is severely constrained by the ZLB, the 10 year
responsiveness is also as strong as in the normal period 19902000. Treasury yield remains highly responsive to monetary policy
The results for BAA yields are reported in Panel B of Fig. 1. It is shocks over the entire sample except a short period in the Great
clear that the sensitivity of BAA yields is similar to AAA yields. It Recession. The stock market becomes less responsive to monetary
is worth noting that although the confidence intervals get larger policy shocks from 2008 to 2010. But this sensitivity recovers
at the ZLB, the response of BAA yields remains quite sensitive to quickly after 2011. Overall, our findings imply that monetary
monetary policy shocks even at 2012.11 policies still have large power even during the ZLB period, although
their effects on the stock market and the short-term Treasury bond
market are qualified.

10 We do not need to test the hypothesis that relative b is constant over time
because we only have one regressor. 12 The long-term Treasury yields react less in the Great Recession periods
11 During 20022003, BAA yields are quite insensitive to monetary policy shocks. (20072008), during which the fed funds rate was higher than the ZLB.
W. Wu / Economics Letters 145 (2016) 258261 261

Fig. 1. Time-varying sensitivity coefficients d . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2. Time-varying sensitivity coefficients d . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Acknowledgments Kiley, Michael T., 2014. The response of equity prices to movements in long-term
interest rates associated with monetary policy statements: Before and after the
zero lower bound. J. Money Credit Bank. 46 (5), 10571071.
I am grateful to the editor and the referee for helpful comments. Kuttner, Kenneth N., 2001. Monetary policy surprises and interest rates: Evidence
from the fed funds futures market. J. Monetary Econ. 47 (3), 523544.
Rigobon, Roberto, Sack, Brian, 2004. The impact of monetary policy on asset prices.
References J. Monetary Econ. 51 (8), 15531575.
Rosa, Carlo, 2013. The financial market effect of fomc minutes. Econom. Policy Rev.
Bernanke, Ben S., Kuttner, Kenneth N., 2005. What explains the stock markets 19 (2).
reaction to federal reserve policy? J. Finance 60 (3), 12211257. Swanson, E.T., Williams, J.C., 2014. Measuring the effect of the zero lower bound on
Gilchrist, Simon, Lpez-Salido, David, Zakrajek, Egon, 2015. Monetary policy and medium-and longer-term interest rates. Amer. Econ. Rev. 104 (10).
real borrowing costs at the zero lower bound. Amer. Econ. J.: Macroecon. 7 (1), Thorbecke, Willem, 1997. On stock market returns and monetary policy. J. Finance
77109. 52 (2), 635654.
Grkaynak, Refet S., Sack, Brian, Swanson, Eric T., 2005. Do actions speak louder Wright, Jonathan H., 2012. What does monetary policy do to long-term interest
than words? the response of asset prices to monetary policy actions and rates at the zero lower bound?*. Econom. J. 122 (564), F447F466.
statements. Int. J. Cent. Bank..
Kiley, Michael T., 2013. Monetary policy statements, treasury yields, and private
yields: before and after the zero lower bound.

You might also like