what moves stock prices?
TRANSCRIPT
What moves stock prices? Moves in stock prices reflect something other than news about fundamental values.
David M . Cutler, James M . Poterba, and Lawrence! H . Summers
" F inancial economics has been enormously suc- ' CJ
! cessful in explaining the relative prices of different securities, a process facilitated by the powerful in- tuition of arbitrage. On the other hand, much less progress has been recorded in accounting for the ab- solute level of asset prices.
The standard approach holds that fluctuations in asset prices are attributable to changes in funda- menta I values. Voluminous evidence demonstrates that share prices react to announcements about cor- porate control, regulatory policy, and macroeconomic conditions that plausibly affect fundamentals. The stronger claim that only information affects asset val- ues is much more difficult to substantiate, however.
The apparent absence of fundamental eco- nomic news coincident with the dramatic stock mar- ket movements of late 1987 is particularly difficult to reconcile with the standard view. This paper explores whether the 1987 market crash is exceptional in this regard, or whether a large fraction of significant mar- ket moves are difficult to explain on the basis of in- formation.
Several recent studies of asset pricing have challenged the view that stock price movements are wholly attributable to the arrival of news. Roll (1988) show I; that it is difficult to account for more than one- third of the monthly variation in individual stock re- turns8 on the basis of systematic economic influences. Shiller's (1981) claim that stock returns are too variable to be explained by shocks to future cash flows or plausible variations in future discount rates argues
[I,
%
for other sources of movement in asset prices. French and Roll (1986) demonstrate that the variation in stock returns is larger when the stock market is open than when it is closed, even during periods of similar in- formation release about market fundamentals.
The difficulty of explaining returns on the basis of information is not confined to equity markets. Fran- kel and Meese (1987) report similar findings in the foreign exchange market. Roll (1984) finds that news about weather conditions, the principal source of var- iation in the price of orange juice, explains only a small share of the movement in orange juice futures prices.
This paper estimates the fraction of the varia- tion in aggregate stock returns that can be attributed to various types of economic news. The first section relates stock returns to the arrival of information about macroeconomic performance. We find that our news proxies can explain about one-third of the vari- ance in stock returns.
To examine the possibility that the stock mar- ket moves in response to information that does not enter our definition of news, the next section analyzes stock market reactions to identifi.able world news. While news regarding wars, the Presidency, or sig- nificant changes in financial policies affects stock prices, our results cast doubt on the view that "qual- itative news" can account for all the return variation that cannot be traced to macroeconomic innovations. This finding is supported by the observation that many of the largest market movemlents in recent years
DAV1:I) M. CUTLER is a graduate student and JAMES M. POTERBA is Professor of Economics at MIT in Cambridge (MA 02139). LAWRENCE H. SUMMERS is Nathaniel Ropes Professor of Political Economy at Harvard in Cambridge (MA 02138). The authors are grateful to Peter Bernstein, Eugene Fama, Kenneth French, and Andrei Shleifer for helpful comments, to the National Science Foundation for financial support, and to DRI for data assistance. This research is part of the NBER Programs in Economic Fluctuations and Financial Markets. A data appendix is available from the Interdisciplinary Con- sortium for Political and Social Research in Ann Arbor, MI.
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
have occurred on days when there were no major news events.
Our concluding section argues that further un- derstanding of asset price movements requires two types of research. The first should attempt to model price movements as functions of evolving consensus opinions about the implications of given pieces of information. The second should develop and test “propagation mechanisms” that can explain why shocks with small effects on discount rates or cash flows may have large effects on prices.
THE IMPORTANCE OF MACROECONOMIC NEWS
Here we seek to determine whether unex- pected macroeconomic developments can explain a significant fraction of share price movements. We analyze monthly stock returns for the 1926-1985 pe- riod, as well as annual returns for the longer 1871- 1986 period.
For each data set, our analysis has two parts. First, we estimate regression models relating each ma- croeconomic variable to its own history and that of the other variables. We use these models (vector au- toregressions) to identify the unexpected component of each time series and to consider the explanatory power of these news measures in explaining stock returns. Second, we adopt a less structured approach to the examination of macroeconomic news. After controlling for the influence of lagged economic fac- tors on prices, we measure the incremental explan- atory power of current and future values of our macroeconomic time series.
Structured Vector Autoregression Evidence
We begin by analyzing monthly stock returns for the 1926-1985 period, using seven measures of monthly macroeconomic activity, chosen to measure both real and financial conditions:’ 1. The logarithm of real dividend payments on the
value-weighted New York Stock Exchange port- folio, computed as nominal dividends from the Center for Research in Security Prices data base deflated by the monthly Consumer Price Index.
2. The logarithm of industrial production. 3. The logarithm of the real money supply (Ml). 4. The nominal long-term interest rate, measured as
5. The nominal short-term interest rate, measured as
6. The monthly CPI inflation rate. 7. The logarithm of stock market volatility, defined
following French, Schwert, and Stambaugh (1987) as the average squared daily return on the Stan- dard & Poor’s Composite Index within the month.
To isolate the news component of these seven
Moody‘s AAA corporate bond yield.
the yield on three-month Treasury bills.
macroeconomic series, we fit vector autoregressions relating the current value of each to its own lagged values and those of the other six series. Each equation also includes a set of indicator variables for different months. We treat the residuals from these equations (denoted tit) as macroeconomic news and use them as explanatory variables for stock returns:
R = a0 + al*5 l̂t + a 2 * i t + a3*L + a4*L + a5*L -k a,*L + a,*L + Et. (1)
R, is the real, dividend-inclusive return on the value-weighted NYSE index, and the seven variables on the right-hand side are the macroeconomic news variables. The R* for Equation (1) measures the frac- tion of the return variation that can be explained by our right-hand side variables. In other words, it mea- sures the importance of these types of macroeconomic news in explaining stock price movements.’
Table 1 reports estimates of Equation (1) using
eral conclusions emerge from this table. First, ma- croeconomic news as we have defined it explains only about one-fifth of the movement in stock prices. In- creasing the number of lagged values included in the VARs does not substantially alter this finding. Sec- ond, most of the macroeconomic news variables affect returns with their predicted signs and statistically sig- nificant coefficients.’ For the full sample period, an unexpected 1% increase in real dividends raises share prices by about one-tenth of 1%, while a 1% increase in industrial production increases share values by about four-tenths of 1%. Both inflation and market volatility have negative and statistically significant ef- fects on market returns. An unanticipated 1% rise in volatility lowers share prices by slightly less than 0.025%, so a doubling of volatility would lower prices by about 2.5%. The other macroeconomic innovations appear to have a less significant effect on share prices.
We examine the robustness of our findings by performing similar tests for the 1871-1986 period. As monthly macroeconomic time series are unavailable for this extended period, we focus on annual returns. We measure R, as the January-to-January return on the Cowles/Standard & Poor’s stock price series. This series was developed by Robert Shiller and was used in Poterba and Summers (1988). Our macroeconomic variables include the logarithm of real dividend pay- ments during the year, the logarithm of real GNP from Romer (1988), the logarithm of real M1, the nominal long-term interest rate, the six-month commercial paper rate, and the inflation rate for the NNP deflator (all from Friedman and Schwartz, 1982), and the log- arithm of stock market volatility, defined as the sum of squared monthly returns on the Cowles/S&P Index within the year.
monthly data for both 1926-1985 and 1946-1985. Sev- 5
3 3
0 3 2 2 0,
3
2 3 2
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
TABLE 1
Restricted VAR Evidence on Macroeconomic News and Stock Returns
Coefficients on Macroeconomic News Variables
Lags Real Industrial Real Interest Rates in VAR. Dividends Production Money Long Short Inflation VolaElity R2
3
6
12
24
3
6
12
24 6 % E cl
8 e \o
2
3
5
0.081
0.094
0.116 (.014) 0.138 (.016)
(.OH)
(.012)
0.050
0.051 (.013) 0.068 (.016) 0.078
(.012)
(.020)
-0.024 (.180)
(. 186) -0.074
- 0.066
0.427
0.398 (.113) 0.373
0.382 (. 133)
(.112)
(. 121)
0.100 (. 166) 0.287 (. 186) 0.245 (.193) 0.073 (.235)
0.738‘ (.483) 0.875* (.450) 0.810‘
1926-1986 Sample (Monthly Data)
0.195 -2.64 - 0.682 (. 152) (1.57) (.638)
(. 158) (1.62) (. 654)
(. 165) (1.73) (. 079) 0.155 0.41 -1.340 (. 182) (2.02) (0.824)
0.074 -2.18 -0.586
0.066 - 1.91 -0.967
1946-1985 Sample (Monthly Data)
0.180 -2.15 - 1.23 (.355) (1.24) (.522) 0.081 -2.15 - 1.22 (.206) (1.31) (.546) 0.017 - 1.92 - 1.73 (.482) (1.42) (. 602)
-0.304 0.352 -2.21 (. 567) (1.83) (. 794)
1871-1986 Sample (Annual Data)
0.150 -0.021 -4.91 (.613) (3.83) (1.90) 0.235 0.175 -5.23 (. 639) (4.12) (2.10) 0.146 0.696 - 6.04
(.DO) (.530) (. 729) (5.07) (2.36)
- 0.079 (.071)
-0.123 (. 073)
-0.111 (. 079)
-0.138 (.088)
- 0.075 (.059)
-0.110 (. 062)
-0.114 ( .072)
(. 095) -0.148
-0.716 (.532)
(. 591) -0.418
(.671)
- 0.814
--0.022 ( .003)
-- 0.023 ( ,003)
(. 003)
(.004)
-. 0.023
-0.025
--0.017 (. 003)
(. 003) -0.017
(.003) -0.020
(.004)
--0.018
-- 0.006 (. 029)
(.030) 0.004 (.OB)
-- 0.004
0.185
0.186
0.188
0.187
0.149
0.144
0.155
0.126
0.064
0.064
0.022
~~ ~~
The dependent variable is the real return on value-weighted NYSE. Estimates correspond to Equation (l), with standard errors in parentheses The news variables are the logarithms of real dividends, industrial production, and real money supply, nominal long-term and short-term interest rates, inflation, and the logarithm of volatility. All VARs and the return equation include a time trend. “Industnal Production is real NNP for the long-term sample penod.
‘The results for the longer sample period, pre- sented in the bottom panel of Table 1, are similar to those for the post-1926 period. When two lagged val- ues of the annual series are used in defining news components, the R’ in the returns equation is 0.064. Longer. lags in the first stage reduce the extent to which the news can explain returns; with five lagged values, the Rz declines to 0.022. Using annual data for the post-1925 period, the R2 for the two-lag equa- tion is -0.003, and that for the regression including five lags is -0.061. The estimated coefficients on the macroeconomic surprises for the 1871-1985 period re- semble those for the post-1925 monthly return sam- ple, adjusted for the annual rather than monthly span of the dependent variable, with one notable excep- tion: the real dividend innovation has a negative co- efficieint for the long sample, although its large standard error also permits a wide range of positive value!;
Urirestricted Regression Evidence
The foregoing method of defining macroecon- omic news suffers from three potential problems.
First, it does not capture new information about fu- ture macroeconomic conditions that is revealed in pe- riod t but not directly reflected in that periods variables. Second, if the models foir measuring news are misspecified, our estimated residuals may not re- flect new information accurately. If market partici- pants operate with an information !set larger than the one we have considered, our residuals may overstate the news content of contemporaneous series. Finally, there are timing issues associated with the release of macroeconomic information. The Consumer Price In- dex for month t, for example, is announced during month t + 1, but market participants may have some information about this variable during month t. These considerations motivate our less-structured approach to identifying the importance of macroeconomic news.
We implement such an approach by first re- gressing stock returns on the lagged values of our macroeconomic time series and th.en including cur- rent and future values of these time series in the regres- sions. The incremental R2 associated with these additional variables measures the importance of ma-
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
croeconomic news in explaining stock returns. This approach is not without shortcomings. It
may understate the true explanatory power of news, because we still omit changes in expectations about the distant future that are not reflected in macroe- conomic variables in period t or the near future. Con- versely, if stock market movements attributable to variables outside our information set affect future ma- croeconomic activity, our approach of including fu- ture macroeconomic realizations will overstate the role of expectational revisions.
Table 2 presents results using different num- bers of lagged and led values of the macroeconomic variables for the 1926-1985 sample of monthly data. The findings are supportive of the results using the more structured VAR approach. Lagged values of the macroeconomic variables we consider can explain less than 5% of the variance of returns. Including the con- temporaneous values of the seven macroeconomic time series significantly raises the explanatory power of these equations. With only one lagged value of the series included, the R2 rises to 0.14, and with twenty- four lags of each variable the R' is 0.29. Including the one- and two-period led values of the macro variables raises the Rz even more, to 0.29 when only one lagged value of the series is included and as high as 0.39 when the longer lags are included. Results for the postwar period, presented in the middle panel of Ta- ble 2, are consistent with those for the longer sample
TABLE 2
Unrestricted VAR Evidence on Macro News and Stock Returns
R2 for Equations Including: Number of Lags Lagged in Specification Lagged Lagged and Current Current, and Led
1 3 6
12 24
1 3 6
12 24
1926-1985 Sample (Monthly Data)
0.005 0.139 0.292 0.010 0.192 0.333 0.018 0.208 0.343 0.034 0.250 0.360 0.035 0.289 0.393
1946-1985 Sample (Monthly Data)
0.060 0.194 0.318 0.087 0.254 0.332 0.080 0.259 0.327 0.065 0.267 0.327 0.136 0.355 0.396
1871-1986 Sample (Annual Data)
0.078 0.210 0.515 0.122 0.149 0.509 0.113 0.162 0.511 0.124 0.102 0.534
Each entry reports the RZ from a regression of the real value- weighted NYSE return (Cowles return in annual data) on k lagged values, k lagged values and the current value, or k lagged, two led, and the current value, of the seven macroeconomic series noted in Table 1. Column 1 reports k. For the annual data, only one led value is included.
period. The lagged regressors have somewhat greater explanatory power in the more recent period.
We also applied our less structured approach to the 1871-1986 annual data sample. The explanatory power of the regressions with only lagged values of macroeconomic variables is greater for annual than for monthly data, ranging from 0.078 with one lag of each variable to a high of 0.124 with five lags. Adding the contemporaneous values of the macroeconomic series again raises the R2, with the largest gain an increase from 0.078 to 0.210 when only one lagged value is included. These results are similar to those obtained using monthly data.
Table 2 also reports the R2 for annual equations including lagged, contemporaneous, and one Zed value of the macroeconomic data series. 'The R' ex- ceeds 0.50, but this almost surely overstates the effect of macroeconomic news on share prices, because it also includes the effect of higher share prices on eco-
and Merton (1984) show that stock returns in year t can explain more than half of the variation in GNP growth in year t + 1, suggesting a strong correlation between returns and subsequent economic activity. While the same problem arises in our monthly anal- ysis, the possibility of large feedback from the market to the economy is substantially greater with annual
nomic outcomes within the following year.4 Fischer 7
3 i
2 U
8 r data. 2
Our results are broadly consistent with earlier 8
3 P 2 studies, such as Fama (1981): a substantial fraction of
return variation cannot be explained by macroecon- omic news. The central question in interpreting this evidence is whether the unexplained return move- ments are due to omitted macroeconomic news var- iables and other information about future cash flows and discount rates, or to other factors that may not affect rational expectations of these variables. Below we present some evidence designed to distinguish these views.
BIG NEWS A N D BIG MOVES: ARE THEY RELATED?
The foregoing analysis excludes a variety of important sources of information, besides macroe- conomic developments, that could affect share prices. Political developments that affect future policy ex- pectations and international events such as wars that affect risk premiums should also be important in asset pricing.
This section examines the importance of these other factors in two ways. First, we study the stock market reaction to major non-economic events such as elections and international conflicts. Neiderhoffer (1971) conducted a similar investigation for a wider sample of events during the 1960s. Second, we ana-
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
lyze the largest stock market movements of the last fifiy years and review coincident news reports to iden- tify, where possible, the proximate causes of these moves.
We begin by analyzing stock market reactions to non -economic events. We identified a sample of such events using the ”Chronology of Important World Events” from the World Almanac. We first ex- cluded events that we thought were unlikely to affect the stock market. We narrowed our set of events still further by considering only those events that the New York Times carried as the lead story, and that the New York Times Business Section reported as having af- fected stock market participants. Winnowing the events in this way biases our sample toward those news items that are likely to have had the largest impact on stock prices. This should bias our results toward finding a large stock market reaction to the forty-n me political, military, and economic policy
Table 3 lists these forty-nine events along with 8 events in our sample. z n
VI - z 5 % TABLE 3
Major Events and Changes in the S&P Index, 1941-1987
Event Percent
Date Channe
Japanese bomb Pearl Harbor US declares war against Japan
Roosevelt defeats Dewey
Rooseve’t dies
Atomic bombs dropped on Japan: Iiiroshima bomb Nagasaki bomb; Russia declares war Japanese surrender
‘Truman defeats Dewey
North Korea invades South Korea Truman to send US troops
Eisenhower defeats Stevenson
Eisenhower suffers heart attack
Eisenhower defeats Stevenson
U-2 shoi down; US admits spying
Kennedy defeats Nixon
Bay of Pigs invasion announced; Details ieleased over several days
Cuban missile crisis begins: Kennedy announces Russian buildup Soviei letter stresses peace Fonriiila to end dispute reached
Kennedy assassinated; Orderly transfer of power to Johnson
US fires on Vietnamese ship
Johnson defeats Goldwater
Johnson withdraws from race, halts Vietnamese raids, urges peace talks
Dec. 8, 1941 -4.37 Dec. 9, 1941 -3.23
NOV. 8, 1944 -0.15
Apr. 13, 1945 1.07
Aug. 6, 1945 0.27 Aug. 9, 1945 1.65 Aug. 17, 1945 -0.54
NOV. 3, 1948 -4.61
June 26, 1950 -5.38 June 27, 1950 -1.10
Nov. 5, 1952 0.28
Sep. 26, 1955 -6.62
NOV. 7, 1956 -1.03
May 9, 1960 0.09
Nov. 9, 1960 0.44
Apr. 17, 1961 0.47 Apr. 18, 1961 -0.72 Apr. 19, 1961 -0.59
Oct. 23, 1962 -2.67 Oct. 24, 1962 3.22 Ob. 29, 1962 2.16
NOV. 22, 1963 -2.81 Nov. 26, 1963 3.98
Aug. 4, 1964 -1.25
NOV. 4, 1964 -0.05
Apr. 1, 1968 2.53
Robert Kennedy assassinated June 5, 1968 -0.49
Nixon defeats Humphrey Now. 6, 1968 0.16
Nixon imposes price controls, requests Federal tax cut, strengthens dollar Aug. 16, 1971 3.21
Nixon defeats McGovern Nov. 8, 1972 0.55
Haldeman, Ehrlichman, and Dean resign Apr. 30, 1973 -0.24
Dean tells Senate about Nixon cover-up June 25, 1973 - 1.40
Agnew resigns Oct. 10, 1973 -0.83
Carter defeats Ford NOV. 3, 1976 -1.14
Volcker appointed to Fed July 25, 1979 1.09
Fed announces major policy changes
Soviet Union invades Afghanistan
Attempt to free Iranian hostages fails
Reagan defeats Carter
Reagan shot, NYSE closes early; Reopens next day
US Marines killed in Lebanon
US invades Grenada
Reagan defeats Mondale
House votes for Tax Reform Act of 1986
Chernobyl nuclear reactor meltdown; Details released over several days
Senate Committee votes for tax reform
Greenspan named to replace Volcker
Important Events
Oct. 6, 1979
Clec. 26, 1979
Apr. 26, 1980
hlov. 5, 1980
Mar. 30, 1981 Mar. 31, 1981
Oct. 24, 1983
Ckt. 25, 1983
Nov. 7, 1984
1)ec. 18, 1985
Apr. 29, 1986 Apr. 30, 1986
May 8, 1986
June 2, 1987
- 1.25
0.11
0.73
1.77
-0.27 1.28
0.02
0.29
1.09
- 0.40
-1.06 - 2.07
- 0.49
- 0.47
Average Absolute Return Standard Deviation of Returns
All Days Since 1941
1.46 2.08
Average Absolute Return Standard Deviation of Returns
0.56 0.82
the associated percentage changes in the Standard & Poor’s Composite Stock Index. Some of the events are clearly associated with substantial movements in the aggregate market. On the Monday after President Ei- senhower’s heart attack in September 1955, for ex- ample, the market declined by 6.62%. On the Monday after the Japanese attack on Pearl FIarbor, the market fell 4.37%. The orderly presidential transition after President Kennedy was assassinated coincided with a 3.98% market uptick, while the actual news of the assassination reduced share values by nearly 3%. On the two days in 1985 and 1986 when passage of the Tax Reform Act of 1986, the most :significant tax leg- islation in three decades, became imuch more likely, aggregate market reactions were leiss than one-half of 1%.5 For the set of events we analyze, the average absolute market move is 1.46% in contrast to 0.56% over the entire 1941-1987 period.
These findings suggest a surprisingly small ef- fect of non-economic news, at least of the type we
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
have identified, on share prices. The standard devia- tion (variance) of returns on the news days we have identified is 2.08% (4.33%), compared with the daily average of 0.82% (0.67%) for the post-1941 period. This implies that the return on a typical event day in Table 3 is as variable as the cumulative return on 6.40 (4.33067) ”ordinary” days. If every day involved as much news as the forty-nine days in this sample, the standard deviation of annual returns would be 32% instead of the actual 13%. As most days do not wit- ness information release as important as that on the days in Table 3, it may be difficult to explain the “missing variation” in stock returns with events of this kind.
An alternative strategy for identifying the im- portance of news is to examine large changes in share prices and related news developments. Table 4 lists the fifty largest one-day returns on the Standard & Poor’s Composite Stock Index since 1946, along with the New York Times account of fundamental factors that affected prices.
It is difficult to link major market moves to release of economic or other information. On several of these days, the New York Times actually reported that there were no apparent explanations for the mar- ket’s rise or decline. At the other extreme, some of the days clearly mark important information releases; the 1948 election outcome, President Eisenhower’s heart attack, and the announcement of President Ken- nedy’s success in rolling back the 1962 steel price in- crease are examples. On most of the sizable return days, however, the information that the press cites as the cause of the market move is not particularly important. Press reports on subsequent days also fail to reveal any convincing accounts of why future profits or discount rates might have changed. Our inability to identify the fundamental shocks that ac- counted for these significant market moves is difficult to reconcile with the view that such shocks account for most of the variation in stock returns.
CONCLUSIONS
Our results suggest the difficulty of explaining as much as half of the variance in aggregate stock prices on the basis of publicly available news bearing on fundamental values. The results parallel Roll’s (1988) finding that most of the variation in returns for individual stocks cannot be explained using readily available measures of new information. Of course, it is possible that we have failed to consider some type of news that actually accounts for a significant fraction of asset price volatility. Although the hypothesis that stock prices move in response to news that is observed by market participants but not by investigators study- ing the market is irrefutable, we are skeptical of this
possibility. News important enough to account for large swings in the demand for corporate equities would almost surely leave traces in either official eco- nomic statistics or media reports about market move- ments.
The problem of accounting for price changes on the basis of fundamental values is not confined to the overall stock market. Studies of price behavior in settings where fundamental values can be measured directly have similar trouble in explaining prices. The classic example is closed-end mutual funds, discussed by Malkiel and Firstenberg (1978). These funds have traded at both discounts and premiums relative to their net asset value during the last twenty years. At any moment, the cross-sectional dispersion in dis- counts is substantial and difficult to link to funda- mental factors. The widely documented patterns in stock returns over weekends, holidays, and different calendar periods, summarized in Thaler (1987a, 1987b), are also difficult to attribute to news about fundamentals, because fundamental values are not likely to move systematically over these periods.
The view that movements in stock prices reflect something other than news about fundamental values is consistent with evidence on the correlates of ex post returns. If prices were periodically driven away from fundamental values by something other than news but ultimately returned to fundamentals, one would
market is high relative to some indicator of funda- mental value, and high when the market is low rel- ative to fundamental value. Such patterns emerge from studies of ex post returns that use the past level of prices, earnings, and dividends as indicators of fundamental value.6
Our results underscore the problem of account- ing for the variation in asset prices. Throwing up one’s hands and simply saying that there is a great deal of irrationality that gives rise to “fads” is not construc- tive. Two more concrete lines of attack strike us as potentially worthwhile. First, volatility may reflect changes that take place in average assessments of given sets of information regarding fundamental val- ues as investors re-examine existing data or present new arguments. This view is suggested by French and Roll’s (1986) finding that return volatility is greater when the market is open than when it is closed.
Second, it may be fruitful in accounting for volatility to explore propagation mechanisms that could cause relatively small shocks to have large ef- fects on market price^.^ “Informational freeloading” on observed asset prices may have something to do with market volatility. In a world where most inves- tors accept prices as indicators of fundamental value, small changes in the supply of or demand for secu-
9
5 q
3 o =i
2
i-
$ z
expect a tendency for returns to be low when the 8 ~
2 3 2
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
TABLE 4
Fifty Largest Post-War Movements in S&P Index and Their "Causes"
Percent Percent Change New York Times Explanation* Date Change New York 'rimes Explanation*
- 20.47 Worry over dollar decline and trade 28 Nov. 1, 1978 3.97 Steps by Carter to strengthen deficit, fear of US not supporting dollar. dollar. 29 Oct. 22, 1987 -3.92 Iranian attack 13ri Kuwaiti oil
Interest rates continue to fall; terminal; fall. in markets overseas; deficit talks in Washington; analysts predict lower prices. bargain hunting. 30 Oct. 29, 1974 3.91 Decline in short-term interest rates;
ease in future monetary policy; calls; reaction to falling foreign lower oil prices. stocks. 31 Nov. 3, 1982 3.91 Relief over small Democratic
-6.73 "No basic reason for the assault on victories in House. prices." 32 Feb. 19, 1946 -3.70 Fear of wage-price controls
- 6.68 Kennedy forces rollback of steel lowering coi:porate p-rofits; labor price hike. unrest.
- 6.62 Eisenhower suffers heart attack. 33 Jun. 19, 1950 -3.70 Korean War continues; fear of long
-5.38 Outbreak of Korean War.
9.10
- 8.28 Fear of budget deficits; margin
war.
delay in coal contract approval; fear of new mid-East war.
analysts express optimism.
prospects for price decontrol.
stock split.
previous w'eek's decline.
prime rate [ a h .
retail sales low.
decline in short-term interest rates.
42 Dec. 9, 1946 3.44 Coal strike ends; railroad freight rates increase.
43 Jun. 29, 1962 3.44 "Stock prices advanced strongly chiefly because they had gone down so lclng and so far that a rallv was due."
34 Nov. 18, 1974 -3.67 Increase in unemployment rate; 5.33 Investors looking for "quality stocks."
-5.24 Labor unrest in maritime and trucking industries.
-5.16 Fear of trade deficit; fear of higher interest rates; tension with Iran.
policy. "The stock surge happened for no fundamental reason." 38 Jun. 4, 1962 -3.55 Profit taking; continuation of
lower interest rates; crackdown 39 Aug. 20, 1982 3.54 Congress passes Reagan tax bill; on triple witching announced.
4.76 Interest rates decline. 40 Dec. 3, 1987 - 3.53 Computerized selling; November
4.65 Optimistic brokerage letters;
35 Apr. 22, 1980 3.64 Fall in short-term interest rates;
36 Oct. 31, 1946 3.63 Increase in commodity prices;
37 Jul. 6, 1955 3.57 Market optimism triggered by GM 5.02 Rumors of change in economic
-4.81 Foreign governments refuse to
institutional and corporate 41 Sep. 19, 1974 3.50 Treasury Secretary Simon predicts buying; suggestions of tax cut.
-4.61 Truman defeats Dewey.
interest rates.
over past week.
4.60 Ford to reduce inflation and
- 4.57 Weakness in economic indicators
4.49 Eisenhower urges confidence in
Date
1 Ocl:. 19, 1987
2 Oct. 21, 1987
3 Oct. 26, 1987
4 Sep. 3, 1946
5 May 28, 1962
6 Sep. 26, 1955 7 Jun. 26, 1950 8 Oct. 20, 1987
9 Sep. 9, 1946
10 Oct. 16, 1987
11 May 27, 1970
12 Sep. 11, 1986
13 Aug. 17, 1982 14 May 29, 1962
15 Nov. 3, 1948 16 Oci:. 9, 1974
15' Fetl. 25, 1946
18 Oc':. 23, 1957
19 Oct. 29, 1987
20 Nov. 5, 1948
2l Ncv. 6, 1946
22 Oct. 7, 1974
23 Nov. 30, 1987 2,4 Jd. 12, 1974
25 Oct. 15, 1946
26 Oct. 25, 1982
2.7 Nov. 26. 1963
economy.
durable goods orders increase; rallies overseas.
-4.40 Further reaction to Truman victory over Dewey.
- 4.31 Profit taking; Republican victories in elections presage deflation.
4.19 Hopes that President Ford would announce strong anti-inflationary measures.
4.46 Deficit reduction talks begin;
-4.18 Fear of dollar fall. 4.08 Reduction in new loan demands;
4.01 Meat prices decontrolled; prospects lower inflation previous month.
of other decontrols.
Reserve's failure to cut discount rates.
Kennedv assassination.
-4.00 Disappointment over Federal
3.98 Confidence in Johnson after
44 Sep. 5, 1946
45 Oct. 30, 1987
46 Jan. 27, 1975
47 Oct. 6. 1982
48 Jul. 19, 1948
49 Nov. 30, 1982
50 Oct. 24, 1962
3.43 "Replacement buying" after earlier fall.
3.33 Dollar stabilizes; increase in prices abroad.
3.27 IBM wins appeal of antitrust case; short-term interest rates decline.
3.27 Interest rates fall;; several large companies announce increase in profits.
Berlin; possibility of more price controls.
3.22 "Analysts were at a loss to explain why the Dow jumped so dramatically in the last two hours."
decisions on Cuban Missile Crisis; calls for US-Soviet summit.
-3.26 Worry over liussian blockade of
3.22 Krushchev promises no rash
-
*Per the financial section or front page.
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
rities can have large effects on prices. Suppose, for example, that all investors desired
to hold the market portfolio in order to achieve op- timum diversification, except for one investor who wishes to concentrate holdings on a single security regardless of its price. The equilibrium price of this security would be infinite. This example, while ex- treme because speculators would intervene to sell an irrationally demanded stock well before its price ap- proached infinity, makes a n important point. If many investors accept market prices as indicators of value and so do not trade on the basis of their own assess- ment of values, market values will be more suscep- tible to those who trade on the basis of their own opinions.
The possibility that many investors do not for- mulate their own estimates of fundamental value is consistent with trading patterns surrounding the sharp stock market decline of October 1987.’ Despite the market’s dramatic drop, the vast majority of shares were not traded. This is only explicable if investors rely on market prices to gauge values, or if investors received information that led to significant downward revisions in fundamental values. It seems difficult to identify the information that would sup- port the second explanation.
Most of the monthly data series were drawn from the Data Resources, Inc., data base. Money supply data prior to 1960 come from Friedman and Schwartz (1963). More recent data are from various Federal Reserve Bulletins. Moody‘s corporate bond yield is from the Board of Governors of the Federal Reserve System, Banking and Monetary Statistics: 1914-41 and 1941-70, and various issues of the Federal Reserve Bulletin.
We report R2 because it is a measure of goodness of fit that corrects for the expected explanatory power of additional regressors. While adding irrelevant regressors to an equa- tion will raise-the equation’s R2, it will not affect the expected value of the R2 = (T- l)/(T-K)R2 - (K- l)/(T-K), where T is the total number of observations, and K the number of degrees of freedom used in estimation.
A related investigation by Chen, Roll, and Ross (1986) showed that various macroeconomic ”factors” have positive prices. Their study is concerned with explaining the ex ante return on different securities, however, while ours consid- ers the ex post movements in prices that result from ma- croeconomic innovations.
The future dividend variable is the major source of the im- pressive fit when led values are included. The link between these series, however, is likely to be much stronger than would be the case if it reflected only information about t + 1 dividends that was released (and incorporated in prices) at t. In a model where dividends adjust to lagged share prices, as in Marsh and Merton (1987), future dividends are associated with current prices, but the principal causality is reversed.
Cutler (1988) examines the events leading up to the Tax Reform Act in greater detail. The small aggregate market reaction on these days is matched by little abnormal cross- sectional variation in stock returns, despite the substantial differences in the law’s likely impact across firms.
Campbell and Shiller (1988), Cutler, Poterba, and Summers (1988), Fama and French (1988b), Poterba and Summers (1988), and Shiller (1984) find evidence consistent with this view. Models that explain the predictability of returns on the basis of trading by uninformed ”noise traders” have been discussed by Black (1986) and DeLong, Shleifer, Sum- mers, and Waldman (1987).
Mandelbrot (1966) presents a rational model in which ap- parently small news releases can trigger large revaluations in expected future profits.
Frankel (1989) suggests a number of stylized facts regarding foreign exchange markets, such as the short-term focus of most traders, that are consistent with the absence of in- dependent assessment of fundamentals by most investors.
REFERENCES
Black, Fischer. ”Noise.” Journal of Finance 41 (July 1986), pp. 529- 542. Campbell, John Y., and Robert Shiller. “Stock Prices, Earnings, and Expected Dividends.” journal of Finance 43 (July 1988), pp. 661- 676. Chen, Nai-Fu, Richard Roll, and Stephen Ross. “Economic Forces and the Stock Market.” Journal of Business 59 (July 1986), pp. 383- 404. Cutler, David M. “Tax Reform and the Stock Market: An Asset Price Approach.” American Economic Review 78 (December 1988). Cutler, David M., James M. Poterba, and Lawrence H. Summers. “Why Do Dividend Yields Forecast Stock Returns?’ Massachusetts Institute of Technology, 1988. DeLong, J. Bradford, Andrei Shleifer, Lawrence Summers, and Robert Waldman. “The Economics of Noise Traders.” Harvard Uni- versity, 1987. Fama, Eugene. ”Stock Returns, Real Activity, Inflation, and Money.” American Economic Review 71 (1981), pp. 545-565. Fama, Eugene, and Kenneth French. “Dividend Yields and Ex- pected Stock Returns.” Forthcoming in Journal of Financial Econom- ics, 1988. -. “Permanent and Transitory Components in Stock Prices.” Journal of Political Economy 96 (April 1988), pp. 246-273. Fischer, Stanley, and Robert Merton. “Macroeconomics and Fi- nance: The Role of the Stock Market.” Carnegie-Rochester Conference Series in Public Policy 21 (1984), pp. 57-108. Frankel, Jeffrey. “Fixed Exchange Rates: Experience Versus Theory.” Journal of Portfolio Management, Winter 1989, pp. 45-54. Frankel, Jeffrey, and Richard Meese. “Are Exchange Rates Exces- sively Variable?” in s. Fischer, ed., NBER Macroeconomics Annual 1987. Cambridge: MIT Press, 1987, pp. 117-152. French, Kenneth, and Richard Roll. “Stock Return Variances: The Arrival of Information and the Reaction of Traders.” journal of Fi- nancial Economics 17 (September 1986), pp. 5-26. French, Kenneth, G . William Schwert, and Robert Stambaugh. ”Expected Stock Returns and Stock Market Volatility.” Journal of Financial Economics 19 (September 1987), pp. 3-30. Friedman, Milton, and Anna Schwartz. A Monetary History of the United States, 2867-1 960. Princeton: Princeton University Press, 1963. - . Monetary Trends in the United States and the United Kingdom. Cambridge: Cambridge University Press, 1982.
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.
Malkiel, Burton G., and Paul B. Firstenberg. “A Winning Strategy for an Efficient Market.” Journal of Portfolio Management, Summer
Mandelbrot, Benoit. “Forecasts of Future Prices, Unbiased Mar- kets, and Martingale Models.” Journal of Busine.ss, 39 (Special Sup- plement, January 1966), pp. 242-255. Marsh, ‘Terry, and Robert C. Merton. “Dividend Behavior and the Aggregate Stock Market.” Journal of Business 60 (January 1987), pp.
Neiderhoffer, Victor. “The Analysis of World Events and Stock Prices.” Journal of Business 44 (April 1971), pp. 193-219. Poterba, James, and Lawrence Summers. “Mean Reversion in Stock Prices: Evidence and Implications.” Forthcoming in Journal of Fi- nancial Economics, 1988. Roll, Richard. “Orange Juice and Weather.” American Economic Re- v im 74 (December 1984), pp. 861-880. -- . “It2.” journal of Finance 43 (July 1988), pp. 541-566.
1978, p ~ l . 20-25.
1-40.
Romer, Christina. “The Prewar Business Cycle Reconsidered: New Estimates of Gross National Product, 1869-:L908.” Forthcoming in Journal of Political Economy. Schwert, William. “Why Does Stock Market Volatility Change Over Time?” Working Paper GPB-8711, William Simon Graduate School of Management, University of Rochester, 1!387. Shiller, Robert. ”Do Stock Prices Move Too Much to be Justified by Subsequent Dividends?” American Econ,mzic R e v h 71 (1981),
- . “Stock Prices and Social Dynamics.” I3rookings .Papers on Eco- nomic Activity 1984:2, pp. 457-498. Thaler, Richard. ”Anomalies: The January Effect.” journal of Eco- nomic Perspectives 1 (Summer 1987), pp. 197-201. - . “Anomalies: Weekend, Holiday, Turn of the Month, and Intraday Effects.” Journal of Economic Perspectives 3 (Fall 1987), pp.
pp. 421-436.
169-178.
12
The
Jou
rnal
of
Port
folio
Man
agem
ent 1
989.
15.3
:4-1
2. D
ownl
oade
d fr
om w
ww
.iijo
urna
ls.c
om b
y PR
INC
ET
ON
UN
IVE
RSI
TY
on
10/0
9/13
.It
is il
lega
l to
mak
e un
auth
oriz
ed c
opie
s of
this
art
icle
, for
war
d to
an
unau
thor
ized
use
r or
to p
ost e
lect
roni
cally
with
out P
ublis
her
perm
issi
on.