The aim of this paper is to examine the relationship between equity prices and
business investment, addressing the question of whether investment is influenced by
inefficient pricing in equity markets. It considers: whether share prices influence invest-
ment once some of the important macroeconomic determinants of investment are
controlled for; whether estimates of the deviation of share prices from their estimated
equilibrium values affect investment; and the behaviour of investment and share prices
in periods when share prices appear to have deviated widely from fundamentals. The
results suggest that, while there is a significant relationship between share prices and
business investment in some countries (the United States, Japan, the United Kingdom
and Canada), this largely reflects stock price correlation with, and anticipation of, other
macroeconomic developments. This suggests that pricing inefficiencies, to the extent
they are present, do not have a statistically or economically significant influence on
business investment.
There are a number of important caveats to bear in mind when considering the
analysis attempted in this paper. First, tests of stock market efficiency are joint tests of
efficiency and a model generating expected returns. Hence, the empirical evidence
presented in Section I and elsewhere cannot be used to reject the efficiency hypothesis
per se. Still, the accumulating weight of evidence suggests that economic policy should
not take efficiency for granted. Second, some of the tests presented in Section II
require estimates of the deviation of actual share prices from those that would be found
in an efficient market. Efficient market prices are not observable and must be controlled
for or proxied in some way. Therefore, a finding that deviations from these estimated
efficient prices affect investment may be due solely to an estimate of the equilibrium
price that omits the effects of certain important factors. Hence, these tests will be
biased towards finding that inefficient pricing in equity markets does affect investment.
Even with this bias, however, the results presented later do not strongly support such The paper is structured as follows. Section I examines the evidence on whether
equity markets price efficiently. The relationship between investment and stock prices
is then considered in Section II. Conclusions are provided in Section 111.
I. THE BEHAVIOUR OF EQUITY PRICES
The efficient markets hypothesis states that security prices should fully reflect all
available, relevant information. If this is the case then deviations of actual returns from
expected returns should be random -they ought, on average, to be zero and uncorre-
lated with information available to the market. To test whether prices satisfy these
conditions it is necessary to specify a model of the behaviour of expected returns and to
compare this with their actual performance. For this reason, tests of market efficiency
are joint tests of the efficiency hypothesis and the assumed model of expected return^.^
The most straightforward way to test efficiency is to assume that the expected rate
of return is constant. If this is the case, then changes in share prices should not be
serially correlated since the past history of share prices is the most readily available
piece of information in the market and any information in this history should already be
embedded in the current price. Price changes should only reflect new information
becoming available. Over short horizons (daily and weekly returns for example) this
appears to be the case (Fama, 1970). However, price changes in some markets have
been found to be serially correlated over longer horizons. A common feature of this
finding is that low-order price autocorrelations are positive but become negative over
longer lags. Fama and French (1988) identified such behaviour in stock prices in the
United States. This type of behaviour is also apparent in other countries (Poterba and
Summers, 1988). Figure 3 contains the correlogram of quarterly changes in stock
prices in the major seven OECD countries. This pattern, positive correlation at short
horizons and negative correlation at longer horizons, seems to occur in a number of
countries. In most cases, the hypothesis that the price changes are not serially corre-
lated can be rejected {Table 1). Cutler eta/. (1990a) show that this type of pattern is not
confined to stock markets. It appears in a wide range of asset markets across a number
of countries.
This joint hypothesis also implies that price changes should not be predictable
using other readily available information. Recent evidence shows that this may not be
the case. Simple measures of the deviation of the existing price from an estimate of the
equilibrium price seem to predict future price movements. Cutler eta/. (1990a) show
that the gap between a constant multiple of real dividends {their proxy for fundamental
influence on stock prices) and the current stock price helps predict future changes in
stock prices. The coefficient on this term tends to be positive, indicating that when
current prices are below the estimates of fundamentals, prices are more likely to rise
than to fall subsequently. This behaviour is also apparent, though to a lesser extent, in
other asset markets.
It has been suggested that these patterns indicate that the speculative behaviour
of market participants may drive prices away from equilibrium in the short run (hence
the positive serial correlation) but that over time prices slowly revert to equilibriumUnited States 25.24'
Japan 26.76"
Germany 34.69*'
Ffance 8.92
Italy 32.89'*
United Kingdom 20.94
Canada 20.67
*r) Significant at the ten (one) per cent level.
Q= Z a?"
N = number of observations
a = sample autocorrelations of lag i = 1 .... 12
Q is distributed as a chi-squared variable with 12 degrees of freedom.
12
i=l
Note: The Q statistic tests whether the expected value of successive price changes are independent of all previous
changes (for an application see Dooley and Shafer, 1983). The critical value for Q indicates the probability of
rejecting this hypothesis. A rejection implies that there is autocorrelation.
Sample period: 196O:l-1991:4: United States, Germany, Italy, Canada.
1961 :4-1991:4: Japan.
1963:4-1991:4: France.
19621-1991:4: United Kingdom.
Sources: See Figure 1.
(hence the negative serial correlation at long horizons). Such patterns can be derived
from models in which some traders (sometimes called noise or feedback traders) base
their demand for aSsets on past price movements rather than the expected future
income streams (see Cutler eta/., 1990a and De Long et a/., 1990). A recent survey of
traders in the foreign exchange market would seem to confirm that trading decisions
are based on the past behaviour of prices. At least 90 per cent of those surveyed
placed some weight on analysis of past trends in prices when making trading decisions,
particularly in the short run (Taylor and Allen, 1992).
While these patterns do not fit the predictions of the simple efficiency hypothesis,
they do not necessarily imply that the behaviour of traders is irrational particularly when
market participants have short horizons (Froot et a/., 1992). In a market with both
rational speculators4 and feedback traders it may be optimal for the former to anticipate
the behaviour of the latter - buying when they expect some future buying by feedback
traders (De Long et a/., 1990). Thus, even informed investors could act to drive prices
away from fundamentals. One of the more important findings in this theoretical litera-
ture is that rational speculation need not ensure that prices reflect fundamentals in the
short run. Even though the expected return to arbitraging away mispricing (buying
underpriced stocks, for example) may be positive, it is not riskless. If the risk is
sufficiently large, then the mispricing will not be quickly eliminated.
The results in Table 2 add additional support to the view that speculative dynamics
may drive share prices away from their equilibrium in the short run but that prices
gradually revert to equilibrium over time. Table 2 reports the results of estimating an
Table 2. Tests of a return to fundamentals in share prices
Canada United
Kingdom United States Japan Germany France Italy
s (-1) -0.1 11 -0.062 -0.046 -0.099
i (-1) -0.010 -0.021 -0.005 -0.010
(0.004)** (0.006)** (0.005) (0,009)
(0.003)** (0.017)** (0.030) (0.047)’
PY (-1) 0.101 0.054 0.038 0.087
(0.030)’* (0.021)” (0,019)’ (0.047)*
(0.098)’ (0.101)** (0.103)‘ (0.137)
AS (-2) -0.033 -0.199 -0.138 -0.064
AS (-1) 0.218 0.382 0.204 -0.101
(0.092) (0.108)’ (0.088) (0.099)
AS (-3) -0.026 0.109 -0.1 11 -0.01 9
AS (-4)
(0.093) (0.125) (0.092) (0.132)
0.112 0.013 0.087 0.047
(0.075) (0.096) (0.076) (0.085)
Ai -0.023 -0.054 -0.029 -0.026
(0.009)” (O.OlO)*’ (0.019) (0.025)
Ai (-1) -0.017 -0.004 -0.007 -0.048
(0.012) (0.012) (0.023) (0.026)*
Ai (-2) -0.008 0.003 -0.023 -0.008
Ai (-3) 0.017 0.007 0.003 -0.004
(0.014) (0.010) (0.016) (0.020)
(0.010)* (0.01 1) (0.015) (0.020)
Ai (-4) 0.000 -0.004 -0.028 -0.044
(0.010) (0,009) (0.016)* (0.024)’
APY 0.688 0.476 1.463 -0.287
(0.529) (0.553) (0.421)’* (0.386)
APY (-1) -0.418 -0.685 0.349 0.003
(0.528) (0.476) (0.418) (0.457)
APY (-2) -0.629 0.067 -1.317 0.386
(0.554) (0.441) (0.418)” (0.619)
(0.514)’ (0.502) (0.567) (0.743)
APY (-4) -1.045 0.230 -0.980 -0.654
(0.515)’ (0.443) (0.429)’ (0.548)
(0.829)” (0.678)* (0.524)’ (I ,280)’
APY (-3) -0.916 0.089 -0.312 -0.549
ci -2.756 -1.634 -1.030 -2.370
S.E.E. 0.056 0.058 0.065 0.1 11
DW 2.0 1.98 1.95 2.05
R* 0.231 0.158 0.22 0.09 -
‘(“1 Sianificantlv different from zero at the ten (onel Der cent level
-0.080’
-0.005
(0.004)
0.042
(0.016)”
0.350
(0.1 05)”
0.073
(0.086)
0.027
(0.093)
0.143
(0.075)’
-0.030
(0.018)
0.017
0.01 1
(0.01 1)
-0.015
-0.019
-0.042
(0.527)
-0.424
(0.540)
-0.873
(0.466)
-0.175
(0.525)
-0.084
(0.625)
-1.309
(0.477)’’
0.090
1.924
0.24
(0.020)”
(0.01 1)
(0.013)
(0.01 0)’
-0.206
(0.056)”
-0.031
(0.007)*’
0.223
(0.056)**
0.270
-0.128
0.160
(0.088)’
-0.070
(0.084)
-0.051
(0.009)*’
0.01 0
(0.010)
0.003
(0.008)
0.012
(0,009)
-0.003
(0.010)
-0.068
10.377)
-0.308
(0.610)
0.384
(0.586)
0.838
(0.533)
-0.248
(0.529)
(1.41)**
0.072
2.02
0.37
(0.084)-
(0.099)
-5.583
-0.102
(0.035)”
-0.004
(0.005)
0.070
(0.024)**
0.195
(0.079)’
0.111
(0.087)
-0.109
0.127
(0.092)
-0.016
(0.016)
-0.022
(0.013)’
-0.008
0.01 1
-0.008
0.722
(0.540)
-0.595
(0.638)
-0.394
(0.583)
0.230
(0.615)
-0.805
(0.568)
-1.805
(0.616)”
0.065
2.03
0.16
(0.090)
(0.012)
(0.01 1)
(0.01 1)
. I.
defer to‘kquation[l] in the text.
Sources: Share prices: see Figure 1. Nominal GNP: OECD National Accounts. Long-term interest rates - OECD (except for
Japan): United States: 1 0-year government bonds; Japan: government bonds, benchmark 119th; Germany:
7-15 year public sector bonds; France: public and semi-public sector bonds; Italy: treasury bonds; United
Table 4. Contribution of stock prices to explaining investment
- - R2 from R* from
equation (a) equation (b)
Restriction:
Oriainal equation2 Auamented eauation3
United States
Japan
Germany
France
Italy
United Kingdom
Canada
0.344 0.355
0.317 0.346
0.019 0.055
0.071 0.076
0.158 0.164
0.023 0.034
0.118 0.142
0.3807 0.4398
0.0892' 0.0723'
0.1150 0.1740
0.7330 0.8877
0.6654 0.5821
0.5437 0.6896
0.2084 0.5224 I
*
1. The equations used to generate these statistics are:
Rejection of the null hypothesis at the ten per cent level.
2 2
Ah= m+ Z WAY,.!+ z pi r,.!+
i=l i=l
and:
2 2 4
i=l i=l i=3
Ah = a+ .Z AYi-i + Z pi rt-! + Z ytASt.i+ E( (b)
Where I, = log of real business fixed investment
Y, = log of real GNP
r, = long-term real interest rate
S, = log of share prices.
2. The statistics are the marginal significance levels for the null hypothesis that the share price terms could not be excluded
3. The marginal significance levels for the null hypothesis that the share price terms could not be excluded from the following
from equation (b).
equation:
2 2 2 4
i=l i=l i=l i=3
Ali=Q+ Z rnAYt.i+ Z pirt.i+ Z &iAl~..i+ .Z yiASt.i+tt
Sample period: see Table 1.
Sources: Share prices, real GNP, real business fixed investment as in other tables. Real interest rates - DAFFE (except for
Japan): United States: 10-year government bonds; Japan: government bonds, benchmark 119th; Germany:
7-15 year public sector bonds; France: public and semi-public sector bonds; Italy: treasury bonds; United
Kingdom: 10-year government bonds; Canada: over 10 years Federal government bonds.
with the other variables - does not qualitatively alter these results. In the latter case,
share prices become significant in a number of countries (Germany, the United
Kingdom and Canada) but the incremental R2 remain small, ranging from 0.019
(the United States) to 0.072 (the United Kingdom). The response of investment toshare
prices, even when statistically significant, does not appear to be economically impor-
tant. The coefficients on the share price terms suggest that a 10 percentage point rise
in the growth rate of stock prices will raise the growth rate of investment by 0.2 percent-
age points in Italy and by one percentage point in Germany, with the other countries
falling within this range. These results are similar to those of Morck eta/. (1990).l1 They
suggest that the stock market has little role in explaining investment over and above its
You must be logged in to post a comment.