Measuring the Reaction of Monetary Policy
to the Sto ck Market
1
Rob erto Rigob on
Sloan School of Management, MIT
and NBER
and
Brian Sack
Board of Governors of the
Federal Reserve System
April 19, 2001
1
The authors would like to thank Olivier Blanchard, James Clouse, Rudi Dornbusch, William English,
William Nelson, and participants at the Macroeconomics Faculty lunch and the International seminar at
MIT for valuable comments.
Comments are welcome
to bsack@frb.gov or rigobon@mit.edu. The opinions
expressed are those of the authors and do not necessarily reect the views of the Board of Governors of the
Federal Reserve System or other members of its sta.
Abstract
Movements in the stockmarket can have a signicant impact on the macro economy and are there-
fore likely to be an imp ortant factor in the determination of monetary p olicy. However, little is
known about the magnitude of the Federal Reserve's reaction to the stock market. One reason is
that it diÆcult to estimate the p olicy reaction b ecause of the simultaneous resp onse of equity prices
to interest rate changes. This paper uses an identication technique based on the heteroskedastic-
ity of sto ck market returns to identify the reaction of monetary policy to the sto ck market. The
results indicate that monetary policy reacts signicantly to sto ck market movements, with a 5%
rise (fall) in the S&P 500 index increasing the likelihoo d of a 25 basis p oint tightening (easing) by
ab out a half. This reaction is roughly of the magnitude that would be expected from estimates of
the impact of sto ck market movements on aggregate demand. Thus, it app ears that the Federal
Reserve systematically resp onds to sto ck price movements only to the extent warranted by their
impact on the macro economy.
JEL Classication Numbers: E44, E47, E52
Keywords: Monetary Policy, Sto ck Market, Identication, Heteroskedasticity
Rob erto Rigob on Brian Sack
Sloan Scho ol of Management Division of Monetary Aairs
Massachusetts Institute of Technology Federal Reserve Board of Governors
50 Memorial Drive, E52-447 20th and C Streets, Mail Stop 73
Cambridge, MA 02142-1347 Washington, DC 20551
and NBER bsack@frb.gov
rigob on@mit.edu
1 Intro duction
In December 1996, Federal Reserve Chairman Alan Greenspan sho ok global nancial markets when
he raised the p ossibility of \irrational exub erance" distorting equity prices. His concern, it app ears
from the text of his sp eech, was determining the appropriate monetary p olicy resp onse in such
situations. However, the central bank likely has a broader concern ab out equity prices, in that equity
price movements, through their inuence on the macro economy,may b e an important determinant
of the appropriate stance of monetary p olicy. Indeed, the Chairman mentioned the impact of rising
sto ck prices on household wealth or sp ending in every single one of his semi-annual testimonies to
Congress over the subsequentfouryears.
This impact of the sto ckmarket on the macro economy comes primarily through twochannels.
The rst, as suggested by the Chairman's testimonies, is that movements in sto ck prices inuence
aggregate consumption through the wealth channel. The total nancial wealth of U.S. households
sto od at over $35 trillion as of the end of 1999, of which over a third was in the form of equity
holdings.
1
More sp ecically, households are estimated to have had ab out $8.5 trillion of direct
equity holdings and another $5.4 trillion in indirect holdings of equities, including holdings through
mutual funds, life insurance, p ension funds, and trusts and estates. Because of the magnitude of
these holdings, sto ck price movements are an important determinant of household wealth. During
the run-up in sto ck prices from 1995 through 1999, capital gains on equity holdings added over
$12.7 trillion to the wealth of U.S. households. Moreover, the decline in share prices in 2000 lopped
over $2 trillion o of household nancial wealth.
Second, sto ck price movements can also aect the cost of nancing to businesses. In 1999,
for example, U.S. non-nancial corporations raised a gross amount of $110 billion through equity
oerings. Higher share prices at that time probably made credit more accessible to various segments
of the market, as indicated by the surge in technology-related IPOs realized that year. In addition,
rms raised ab out $55 billion in venture capital funds in 1999, whichmayhave also b een encouraged
by the prospect of high share prices.
2
As stock prices slump ed in the second half of 2000, the pace
of IPOs fell back signicantly.
Because of their p otential impact on the macro economy, sto ck market movements are likely
to be an important determinant of monetary policy decisions. Despite this potential imp ortance,
there has b een little empirical evidence measuring the magnitude of the Federal Reserve's reaction
to the sto ck market. The primary reason is that it is diÆcult to empirically estimate the monetary
p olicy reaction due to the simultaneous resp onse of the sto ck market to policy decisions. Indeed,
the p olicy reaction cannot b e identied using traditional approaches for addressing the simultaneity
problem, including exclusion restrictions or instrumental variables. For example, it is diÆcult to
1
The data are taken from the FlowofFunds accounts produced by the Federal Reserve Board. The gures include
the equity holdings of non-prot organizations, which are not separated from those of households in the data.
2
Data on gross equity issuance are from Securities Data Company, while that on venture capital are based on a
survey conducted byVenture Economics.
1
nd any instruments that would aect the stockmarket without b eing correlated with interest rate
movements.
In this pap er, we apply a new identication pro cedure developed by Rigob on (1999) to solve this
problem. The identication is based on the heteroskedasticity of sho cks to sto ckmarket returns. In
particular, shifts in the imp ortance of sto ckmarket sho cks relativetomonetarypolicyshocks, and
the estimated changes in the covariance between the shocks that result, allow us to measure the
reaction of interest rates to changes in sto ck market prices. The results suggest that an unexpected
increase in the S&P 500 index by 5% increases the federal funds rate exp ected after the next FOMC
meeting by ab out 14 basis p oints. Translating this into discrete p olicy moves, a 5% rise in the S&P
500 index increases the probabilityofa25basispointtightening by just over a half.
3
Because the
mo del is symmetric, a 5% decline in sto ck prices has similar implications for policy easing.
It is imp ortant to note upfront that this result do es not imply that the Federal Reserve is
targeting sto ck prices or reacting to p erceived misalignments in sto ck prices.
4
In fact, Chairman
Greenspan has stated that central banks should remain fo cused on achieving price stability and
maximum sustainable growth, suggesting that p olicymakers should only react to sto ck prices to the
extent that they aect the economic outlo ok. The ndings in this paper are consistent with this
view. Using rough calculations, the estimated policy resp onse is approximately of the magnitude
needed to oset the exp ected pass-through of equity market sho cks to aggregate demand. Thus,
it app ears that the Federal Reserve systematically responds to sto ckpricemovements only to the
extentwarranted by their impact on the macro economy.
The pap er is organized as follows. Section 2 discusses the problem of identication and demon-
strates why other widely used identication methods are inappropriate in this context. Section 3
develops the metho d for identifying the system through sto ck market heteroskedasticity. Section 4
presents the results and evaluates whether the magnitude of the estimated p olicy resp onse is sen-
sible. Section 5 investigates the robustness of the results and explores whether the ndings could
be driven by alternative explanations. Section 6 concludes.
2 The Endogeneity Problem
Although movements in the sto ck market may imp ortantly aect monetary p olicy decisions, iden-
tifying the monetary policy resp onse to the stockmarket is diÆcult. The problem is that the sto ck
market endogenously resp onds to monetary p olicy decisions at the same time that p olicy is reacting
to the sto ck market. The simultaneous determination of interest rates and sto ck prices is depicted
in Figure 1. Holding everything else equal, higher interest rates are asso ciated with lower stock
3
That is, if the probabilityofapolicytightening were 30% under the existing economic situation, a 5% rise in
stock prices would increase the probabilityoftightening to 80%. If the probabilityofaneasingwere 10%, the rise in
stock prices would result in a 40% chance of tightening.
4
Cecchetti, Genberg, Lipsky,andWadhwani (2000) argue that monetary p olicymakers should react to p erceived
misalignments in asset prices to reduce the likelihoo d of asset price bubbles forming.
2
market prices, given the higher discountrate for the exp ected stream of dividends.
5
At the same
time, the Federal Reservemay react to higher sto ck prices by raising interest rates. Realizations of
sto ck prices and interest rates are determined by the intersection of these twoschedules and do not
provide a clear reading of whether the p olicy reaction function is upward sloping in sto ck prices.
The rst hint that the endogeneity of the sto ck market resp onse may be an imp ortant consid-
eration comes from the simple correlation b etween movements in short-term interest rates and the
sto ck market. The correlation between daily changes in the three-month Treasury bill rate and
daily changes in the S&P 500 index, shown later in Figure 3, is typically negative over the perio d
since 1985|the opp osite of the sign that would have b een exp ected from the reaction of monetary
p olicy.
6
Of course, the correlation between interest rates changes and stock market returns could be
inuenced byanumber of factors. To add more structure, wecharacterize the dynamic interaction
between the sto ck market and interest rates using a Vector Autoregression (VAR). Assume the
dynamics of the short-term interest rate and stockmarket returns can be written as follows:
i
t
=
s
t
+
x
t
+
"
t
;
(1)
s
t
=
i
t
+
x
t
+
t
;
(2)
where
i
t
is the three-month Treasury bill rate and
s
t
is the daily return on the S&P 500 index.
The data are daily, and the sample runs from March 1985 to December 1999. The variable
x
t
consists of 5 lags of the stock market return and the interest rate, as well as observable macro e-
conomic sho cks. These macro economic sho cks are measured by the monthly releases of ma jor
macro economic variables, including the core consumer price index (CPI), the National Asso ciation
of Purchasing Managers survey (NAPM), non-farm payrolls (NFPAY), the core pro ducer price in-
dex (PPI), and retail sales (RETL). Each of these variables are measured by the dierence b etween
the released value and the exp ected value, where those exp ectations are taken from the Money
Market Services survey ab out a week b efore the release.
Equation (1) in the VAR can be interpreted as a high frequency policy reaction function for
the Federal Reserve. Of course, it is more common to estimate a reaction function using lower
frequency data. But the use of daily data is imp ortant in this pap er because it allows us to more
accurately dene the heteroskedasticity of the shocks, as will become apparent b elow. Note that
the three-month Treasury bill rate is used in the daily reaction function rather than the federal
funds rate. While the federal funds rate is adjusted only every six weeks or so, the three-month
Treasury bill rate will adjust daily according to changes in exp ectations of monetary p olicy over
5
Of course, the impact on sto ck prices likely depends on the source of the interest rate movement. On average,
however, it app ears from the evidence b elow that higher interest rates cause stock prices to fall.
6
Note also that the correlation exhibits rich patterns, often b ecoming p ositive during p eriods when the volatility
of the stockmarket increases. These patterns are the basis for the identication pro cedure used, as discussed shortly.
3
