See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/46456421

Stock Prices and the Macro Economy in China

Article · January 2009

Source: RePEc

CITATIONS READS
4 476

4 authors, including:

Nicolaas Groenewold
University of Western Australia
106 PUBLICATIONS   1,358 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Demand/Supply factors in the Chinese slowdown View project

All content following this page was uploaded by Nicolaas Groenewold on 17 December 2013.

The user has requested enhancement of the downloaded file.

 ECONOMICS

STOCK PRICES AND THE MACRO

ECONOMY IN CHINA

by

Jiangang Peng
Jie Cui
Fuyong Qin
Hunan University

and

Nicolaas Groenewold
Hunan University
and
The University of Western Australia

DISCUSSION PAPER 09.20

 Stock Prices and the Macro Economy in China

Jiangang Peng,a
Jie Cui,a
Nicolaas Groenewold,a b
and
Fuyong Qina

a
College of Finance,
Hunan University,
Changsha, Hunan Province, 410079
P. R. China
b
Economics Programme,
University of Western Australia,
Crawley, WA 6009
Australia

* Corresponding author; address: Nicolaas Groenewold, Economics Programme,

University of Western Australia M251, Crawley, WA 6009, Australia;
Phone: +61 (8) 6488 3345;
Fax: +61 (8) 6488 1016;
e-mail: nic.groenewold@uwa.edu.au

Abstract:
This paper analyses the relationship between stock prices and the Chinese macro
economy measured by the level of GDP. There are many possible channels of
influence between these two variables, channels which may operate in either direction.
There are also many theories relevant to these interrelationships. Rather than
explicitly testing theories, we focus on the empirical nature of this relationship which
we analyse in the context of a VAR/VEC model which allows for two-way influences
but is agnostic about the particular theoretical underpinnings. We apply tests for
stationarity and cointegration and find that there is a long-run, cointegrating
relationship between stock prices and GDP. We estimate a VEC model and use it to
analyse both short-run and long-run causality as well as to generate impulse response
functions (IRFs). We find that there is strong evidence of long-run causality from the
economy to the stock market but not vice versa. We also find modest but weaker
evidence of a similar short-run effect. These are borne out by the IRFs which show a
small and weak link from the stock market to the economy but a stronger and much
more substantial effect in the opposite direction. We rationalise our results in terms of
the relatively small size of China’s stock market.

Keywords: stock prices, output, China

JEL codes: E44, G10
 I Introduction

Recent sharp changes in stock prices prompt a host of questions about the

relationship between the stock market and the rest of the macroeconomy. At a broad

level there are questions about the benefits of a deregulated financial system for the

health of the economy while questions with a narrower focus are also common, such

as: do the fluctuations in stock prices reflect economic factors or are they simply

bubbles driven by (irrational) investor sentiment? On the other hand, do the stock

price movements spill over into the rest of the economy, via consumption, investment

or some other channel?

Given the importance of the stock market in the financial system of most

countries, it is not surprising that all of these questions have regularly exercised the

minds of policy-makers and have been the subject of a substantial amount of

empirical research. The broad question of the relationship between financial

development and economic development in general is the subject of a rapidly-growing

literature which has been recently surveyed in a wide-ranging paper by Levine (2005).

Measures of financial development in this literature include more than just those

relating to the stock market and range from assets and liabilities of banks and non-

bank financial intermediaries to the size of the bond market relative to the economy as

a whole, as well as stock market capitalisation and turnover variables. The focus of

the research in this area is long-run and typically uses large cross-country data bases

with variables measured as multi-year averages to capture the long-term nature of the

growth process although, more recently, panel data have also been used to enable the

analysis of questions of causality in the finance-growth nexus. Levine draws the

overall conclusion that there is a positive connection between measures of financial

1
development and economic growth and that stock market development makes a

significant contribution to this effect.

In addition to questions of long-run growth, the stock-market-macroeconomy

connection has also been analysed from a short-run perspective, focussing on the

relationship between stock prices and macroeconomic variables such as GDP,

consumption, investment, inflation, exchange rates and monetary policy measures.

Here relationships may run in both directions, from the macroeconomy to the stock

market and vice versa. While the evidence points to a positive effect of stock prices

on output there is mixed evidence on the sign of the effect in the opposite direction.

The present paper reports on the analysis of the relationship between stock

prices and the macroeconomy in China for the period since the establishment of the

Chinese stock market in the early 1990s. Our focus is on the short-run interaction

between stock prices and the macroeconomy, in contrast to the long-run emphasis of

the finance and growth literature. Moreover, we confine our attention to output as the

main measure of the macroeconomy.

Our motivation for this research is three-fold. First, the relationship between

output and stock prices is not clear; in particular, negative effects of output shocks on

stock prices have been reported in several papers such as Lee (1992) for the US,

Cheung and Ng (1998) for a set of five countries and Groenewold (2003) for Australia.

This is in contrast to other findings such as those by Gjerde and Saettem (1999) and

seems counterintuitive although Groenewold suggests an explanation which

distinguishes demand- from supply-driven output shocks. Analysis of this relationship

for different countries will provide more information on this important connection.

2
 Secondly, China’s stock market is relatively small although it is developing

rapidly and an analysis of this case will balance the predominance of developed

economy research.

Thirdly, little is known about the output-stock-price nexus in China. There is

a very limited Chinese-language literature and, to our knowledge, only two English-

language papers address this issue, namely Zhao (1999) and Liu and Sinclair (2008),

only the latter of which throws any light on the issue. Given the growing importance

of China in the world economy and in the international financial system, the

relationship between stock prices and macroeconomic variables is an important issue

in its own right and deserves a more thorough investigation.

The remainder of the paper is structured as follows. The next section presents

background material by way of a review of the relevant literature while section III

presents background material on the Chinese stock market and briefly puts it in the

perspective of the Chinese financial system. In section IV we describe our data and

report the results of tests for stationarity. Tests for cointegration, the estimated

VECM and the associated tests of causality (both short- and long-run) and impulse

response functions which we use as the main instruments to address the issue of this

paper are reported in section V. We present conclusions in section VI.

II Literature

There are various ways in which the short-run relationship between the stock

market and the macroeconomy has been modelled in the literature. One approach has

been from an asset-pricing perspective in which the Arbitrage Pricing Theory (APT)

or some other multi-factor asset-pricing model is used as a framework to address the

question of whether risk associated with particular macro variables is reflected in

3
expected asset returns. Examples include the original work by Chen, Roll and Ross

(1986) who applied the model to the US as did Kim and Wu (1987) and Chen and

Jordan (1993). There have been numerous applications to other countries’ stock

prices such as: Beenstock and Chan (1988), Clare and Thomas (1994), Cheng (1996),

Antoniou, Garrett and Priestley (1998) and Gunsel and Cukur (2007) for the UK;

Entorf and Jamin (2007) for Germany; Tsuji (2007) for Japan; Ariff and Johnson

(1990) for Singapore; Martikainen (1991) for Finland; Groenewold and Fraser (1997)

for Australia; Mateev and Videv (2008) for Bulgaria and Ihsan et al. (2007) for

Pakistan.

A closely-related analysis, based on intertemporal investor optimisation, is that

of the consumption-CAPM which concentrates on a single macro influence, the

growth of aggregate consumption; see, e.g., Breeden (1979) and Grossman and Shiller

(1981). Applications/tests have been reported in Breeden at al. (1989), Kocherlakota

(1997), Cashin and McDermott (1998) and Chen (2003).

The direction of influence underlying the asset-pricing literature is the

traditional one which is based on the notion that ultimately the share market reflects

the fundamental strengths and weaknesses of the aggregate economy so that the

direction of influence is from the economy to the share market. A similar focus is

found in the literature which explores the response of aggregate share prices to the

(expected) inflation rate in the spirit of the Fisher effect. Early work carried out in

this area is by Bodie (1976), Fama and Schwert (1977), Jaffe and Mandelker (1976),

Nelson (1976) and Gultekin (1983) whereas more recent applications include those by

Boudoukh and Richardson (1993), Balduzzi (1995), Graham (1996), Groenewold,

O’Rourke and Thomas (1997), Siklos and Kwok (1999), Crosby (2001) and Boucher

(2006).

4
 Related studies assess the response of the share market (often, but not always,

at an aggregate level) to other macro variables such as those which capture monetary

and fiscal policy shocks; e.g. Pearce and Roley (1985), Jain (1988), Aggarwal and

Schirm (1992), Singh (1993), Thorbecke (1997), Cassola and Morana (2004), Wong

et al. (2006) and Ioannidis and Kontonikas (2008).

An alternative, which looks at the influence in the opposite direction, is to

analyse the effects of share prices on the macroeconomy or selected macroeconomic

variables. A relationship of this nature which has received considerable attention in

the financial economics literature is that between share prices and investment (in the

economist’s sense of capital formation). Studies of this type start with Tobin’s q-

theory of investment (Tobin, 1969) and include Fischer and Merton (1984), Morck,

Schleifer and Vishny (1990), Blanchard, Rhee and Summers (1993), Chirinko and

Schaller (1996, 2001), Baker et al. (2003) and Gilchrist et al. (2004). The question in

that literature is whether firms, in making investment decisions, should or do pay any

heed to share prices or whether share prices are simply a veil which can be dispensed

with when making decisions about real variables such as investment.

Another route through which stock prices have been seen to influence the real

economy is via consumption, the most common theoretical basis being the wealth

effect in the consumption function although other channels such as increased

uncertainty (Romer, 1990), signalling effects (Poterba and Samwick,1995, Poterba,

2000), realised and unrealised wealth effects, liquidity-constraint effect and a stock-

option value effect (Ludwig and Sløk, 2002) have also been suggested. Empirical

work on the stock-price-consumption relationship has included cross-section, time-

series and mixed (panel) studies; examples are Mankiw and Zeldes (1991), Parker

(1999), Starr-McCluer (2002), Ludvigson and Steindel (1999), Shirvani and Wilbratte

5
(2000, 2002), Case et al. (2001), Edison and Sløk (2002), Bertaut (2002), Lettau and

Ludvigson (2004) and Sousa (2008).

In addition to the theoretically-informed analysis reviewed above, essentially

atheoretical empirical models have also been used to analyse the relationship between

the share market and the macroeconomy. These range from simple single-equation

ones of the types used by Chen (1991), Peiro (1996), Choi, Hauser and Kopecky

(1999) and Ioannidis and Kontonikias (2008) to more elaborate models which

recognise the two-way relationship between share prices and the economy as a whole.

However, unlike the models previously cited, they are not based on any particular

theoretical structure but seek simply to capture the empirical regularity between a

limited number of variables in a largely pragmatic way.1 The vector auto-regressive

(VAR) and vector error-correction (VEC) models have been particularly popular in

this area, given that they can be used as a framework for formal examination of inter-

relationships within a given data set without the need to specify a theoretical

framework a priori. Once estimated, the model can be used to simulate the effects of

shocks in a way that is consistent with the historical patterns in the data by the use of

impulse response functions and forecast-error-variance decompositions.

A relatively early application of the VAR model to the analysis of the

relationship between share prices and the macroeconomy is by Lee (1992) and more

recent ones can be found in Thorbecke (1997), Cheung and Ng (1998) and Gjerde and

Saettem (1999), Cassola and Morana (2004), McMillan (2005), Phylatkis and

Panazzolo (2005), Wong et al. (2006) and Pan et al. (2007).

1
An alternative approach which is theoretically-constrained is that based on the real-business-cycle
(RBC) approach to macroeconomics used by Canova and de Nicolo (1995) for the investigation of the
relationship between real activity and share prices. The extent to which RBC models are empirical is a
matter of some controversy. They are better seen as numerical simulation of theoretical models.

6
 An approach which is closely related to the VAR/VECM procedure is one

which is due to Campbell et al. – see Campbell and Shiller (1987, 1988) and

Campbell and Ammer (1993). More recent applications are by Lee et al. – see Lee

(1995, 1998), Chung and Lee (1998) and Hess and Lee (1999). While a VAR model

is used, the approach differs in at least two ways. First, the VAR model is a

constrained one where the constraints are derived from a linearised dividend-discount

model. It therefore has the advantage of a theoretical structure while at the same time

employing the dynamic flexibility of the VAR model. The second difference derives

from the first and is that the focus is on the relationship between share prices and

other financial variables such as the dividend yield rather than macroeconomic

variables such as output. This limits the usefulness of the approach for our purposes.

We conclude this section with a brief account of the limited literature on the

output-stock-price relationship in China. In the Chinese language literature there are

several papers which deal with our topic although there appears to be some confusion

between output and growth, with several papers claiming to be an analysis of stock

prices and economic growth but actually analysing the relationship between stock

prices and GDP (often both in levels) so that they are directly relevant to the work

reported in this paper. Examples are: Ran, Zhang and Wu (2003), Ran, Hu and Wang

(2005), Liang and Teng (2005) and Fan (2006). These papers generally test for

stationarity and cointegration in the (logs of) stock prices and macro variables,

principally output, and then go on to test for causality between them. Interesting

variations are recent papers by Wei and Yong (2007) and Han, Zhang and Wu

(2008), the latter of which focuses on inflation and stock prices and decomposes

inflation shocks into supply and demand-driven ones which, it is found, have different

effects on stock prices.

7
 Finally, two English-language papers are also related to our work, the first by

Zhao (1999) and the second by Liu and Sinclair (2008). The Zhao paper uses a single

equation framework to regress the rate of change of output (so growth rather than

output) on the rate of growth of stock prices. A distinction is made between total

growth and unexpected growth and regressors are entered contemporaneously. The

finding most relevant to our work is that total growth has a negative and significant

effect on stock returns but the unexpected component of growth has a positive and

significant effect. The reverse effect from stock prices to output growth is not tested.

Interestingly, the Liu and Sinclair paper purports to analyse the question of the

relationship between the stock market and economic growth (for Hong Kong and

Taiwan as well as for mainland China) but, in fact, analyses the relationship between

the log of stock prices and the log of output in a VECM framework, thus being more

closely related to our work than the Zhao paper is. They find short-run causality

running from stock prices to output but not vice versa but claim that output affects

stock prices in the long run, although they do not present test results for this

hypothesis.

The existing literature on the relationship between the stock market and the

economy as a whole in China is thus very limited and contradictory and considerably

more through-going analysis is necessary before the relationship is well understood.

We intend to contribute to such analysis.

In this paper we propose to use the VAR/VECM approach, given its flexibility

and the absence of any widely-accepted theoretical model of the share-market-

economy interrelationship. While the theoretically-restricted Campbell model is

attractive, its theoretical restrictions are not directly applicable to the relationship

between share prices and the macroeconomy and we therefore use an unrestricted

8
model. Before turning to the modelling framework and the empirical analysis, we

digress briefly to present some basic information about the Chinese stock market and

the financial system in general to provide background to the interpretation of our

results.

III The Chinese stock market and the Chinese financial system

The Chinese stock market consists of two exchanges, the Shanghai Securities

Exchange (SHSE) and the Shenzhen Securities Exchange (SZSE). The SHSE was

opened in December, 1990 and the SZSE in February, 1991. Since 1998, the market

has been supervised by the Chinese Securities Supervision Commission, before which

it was regulated by a State Council committee. While it has been subject to many

complicated regulations, including price limits from time to time, the trend is towards

cautious deregulation.

An interesting feature of the market for the first two decades of its existence is

various types of shares. The two main types are A and B shares.2 A shares are

denominated in the local currency (Renminbi or RMB) and are traded by domestic

residents and institutions – foreign individuals and institutions are not permitted to

buy and sell A shares. B shares are denominated in US dollars on the Shanghai

Exchange and Hong Kong dollars on the Shenzhen Exchange. They were originally

intended for trading by foreign investors but the restriction that only offshore

individuals and institutions are permitted to trade in B shares was lifted in 2001,

permitting domestic residents to trade in them but only in foreign currency. In

2
Qi, Wu and Zhang (2000) distinguish 5 types of shares by further sub-dividing the A and B share

according to ownership restrictions.

9
addition to A and B shares, some Chinese companies have shares listed on foreign

stocks exchanges such as H shares listed on the Hong Kong stock exchange.

Initially the scale of the market was very small with the SHSE listing 11

companies with a total value of RMB 500 million and the SZSE listing only 5

companies with a value of RMB 270 million. However, as shown by the data

reported in Tables 1 and 2, subsequent growth was rapid, with the number of

companies listing both A and B shares rising to 530 five years after the establishment

of the exchange with 85 of these companies listing only B shares. In addition, the

number of companies listing on the Hong Kong exchange has also risen steadily to

more than 150 at present.

[Tables 1 and 2 near here]

Table 2 provides some data on the size of China’s stock market since shortly

after its inception until the present. It shows strong growth in new financing, turnover

and market capitalisation. The capitalisation measure needs to be interpreted with

care, however. A hangover from state ownership of all enterprises is the continuing

high level of state ownership of listed shares which are effectively not tradable. This

is illustrated in Table 2 which distinguishes between total market value and tradable

market value with the difference between them being the shares held by government

and government-related bodies which are not traded. Clearly this is a sizable

proportion of total value – of the order of 75% of total shares by value are not tradable

– and the source of considerable market anxiety given that the government has on two

occasions attempted to begin unloading these shares with dramatic effects on share

prices leading to a rapid reversal of policy. As a measure of the size of the stock

market, the second figure is probably more appropriate than the first whereas the first

may be a better measure of the size of the listed corporate sector. Whichever measure

10
is used, however, it is clear that growth has been very rapid, although admittedly from

a low initial base.

While the stock market has expanded rapidly since its establishment, it is still

a relatively small part of the financial system. Table 3 provides some summary data

and shows that the size of the stock market relative to GDP is of the order of 40-45%

which compares to a value of around 150% for developed economies such as the US

and the UK.3 But even this is likely to grossly overstate the case for China – if we

include only tradable shares, the ratio fell to around 15% in 2008. Thus by the

standards of developing economies, the Chinese share market is very small relative to

the size of its economy. In contrast, the ratio of loans to GDP is about 15 times this

magnitude. If we focus on new financing, the gap is even larger. Data in Tables 2

and 3 suggest that new financing from the stock market is approximately 1.2% of

GDP while loans from banks, etc. is over 100% (for 2008). All in all, the stock

market, while growing rapidly, is still a relatively small part of the Chinese financial

system compared to other financial institutions, primarily the banks.

[Table 3 near here]

IV The data

Since the focus of the paper is on the interrelationships between output and

stock prices, we need only two series – one for stock prices and one for output. For

stock prices, we used the Composite Index for the Shanghai stock exchange (the

larger of the two exchanges) and for output we used GDP. Given that GDP is

available only at a quarterly frequency, we restricted our analysis to quarterly data.

3
Comparative international data are from the World Federation of Exchanges website: www.world-
exchanges.org.

11
The sample period used is 1992(1) to 2008(4), the start of the sample being dictated

by the availability of stock price data.

The stock price data were obtained from the GTA-CSMAR data base. Data in

this data base are reported on a daily basis and were averaged over the quarter to

obtain quarterly observations. This was done (in contrast to, say, taking a single

observation during the quarter) in order to match the GDP data best since GDP

applies to each quarter as a whole. Moreover, averaging over the quarter removes

very high frequency movements in stock prices which are hardly likely to respond to

GDP variations (or vice versa). The GDP data were taken from the Wind data base

for the years 1992 to 2002 and from the China State Statistical Bureau’s “Financial

Almanac”, various years, for the period since 2002.

Neither series was seasonally adjusted. This was particularly obvious for the

GDP data which has strong seasonal fluctuations. We experimented with various

methods of seasonal adjustment for the GDP data, some of which were clearly

unsatisfactory since there remained distinct seasonal movements in some years. We

eventually undertook the main analysis with data adjusted using the X12 procedure

available in EViews. Even this sophisticated method did not seem to completely

remove seasonal influences so we checked our results for sensitivity to this seasonal

adjustment method and briefly report the results of using an alternative method later

in the paper.

Before undertaking the analysis of the relationship between our two variables,

we tested them for stationarity using standard augmented Dickey-Fuller (ADF) tests.

They are tests of the hypothesis that α1 = 0 in equation (1):

k
∆yt = α 0 + α1 yt −1 + α 2t + ∑ γ i ∆yt −i + ε t (1)
i =1

12
where y is the variable of interest, ∆ denotes the first-difference operator, t is a time

trend and ε is a random error term assumed not to be autocorrelated. Two choices

need to be made before carrying out the test; first, which of the deterministic

components (the intercept and the time trend) to include and, second, the lag length

(the value of k). For our data these choices were not very important since the test

outcome was, in general, not sensitive to either choice.

We analyse both variables in logs, as is customary, so that the first difference

has the interpretation of a continuous rate of change: the rate of capital gains for stock

prices and the growth rate for GDP. In Table 4 we report the ADF test statistics

together with their p-values for lags 1 to 4 and various combinations of the

deterministic components.4

[Table 4 near here]

It is clear from the results reported in Table 4 that ly is non-stationary irrespective of

the nature of the deterministic components and the number of lags. The results for ls

are not so clear, however. If the trend is omitted from the testing equation, ls is

stationary if lags are set at 0 and 3 and when a trend is included it is stationary at lags

3 and 4. The trend terms is marginally significant so that we focus on the second line

of results where the non-stationary null hypothesis cannot be rejected for lags 0 to 2

but is clearly rejected for longer lags. The SIC indicates 3 lags but tests suggest that 1

lag is sufficient to remove autocorrelation and at 1 lag ls is clearly non-stationary. We

therefore conclude that, while the evidence is mixed, on balance there is evidence of

non-stationarity and we proceed under the assumption that both variables are non-

stationary and proceed to test for the stationarity of the first differences.

4
Note that we omit two deterministic cases – the one with no intercept in the levels test and the case
with trend in the first-difference test.

13
 The second part of the table suggests that first differences are stationary for

both variables, although there is some doubt about the log of GDP which is non-

stationary when the intercept is omitted from the ADF equation and the number of

lags is chosen by the SIC. Inspection of the estimated equation, however, shows that

the intercept is significant in this equation and, at SIC-determined lags, the first

difference in log GDP is stationary. We conclude, therefore, that both variables are

I(1).

V Results

Given our finding above that stock prices and output are both I(1), we need to

determine whether they are cointegrated before we can decide on whether to model

them as a VAR model in first differences or as a VECM. We proceed, therefore, to a

consideration of cointegration between the two variables.

We use the Johansen test to address the cointegration question and in this case

there are also numerous different possible specifications of the testing framework

depending on lag length and the specification of the deterministic components of the

VECM used for the test. An inspection of the data suggests that we should include an

intercept in the cointegrating equation since the levels of the series are quite different.

Also, a constant in the VECM seems appropriate since GDP in particular has a

marked positive trend and we should allow for drift over time. Since both variables

have a positive trend over the sample period, it is unlikely that a trend term will be

required in the error-correction term of the VECM although we will report some

results with a trend to gauge the sensitivity of the test outcomes to this assumption.

Cointegration test results for these three specifications for lag lengths from 1

to 5 are reported in Table 5.

14
 [Table 5 near here]

The test outcomes are clearly dependent on the specification, both on the nature of the

deterministic components and on the lag length so that a careful choice of each is

required.

We begin with lag length. Tests of lag exclusion both for a VAR model in the

first differences and for a VECM show that the fifth lag can be excluded but the

fourth cannot. Inspection of the estimated equations reveals a strongly significant

coefficient on the fourth lag of the first difference of the log of GDP in its own

equation (although lags 2 and 3 are generally insignificant) which may be an

indication of left-over seasonal effects in the seasonally-adjusted GDP series

mentioned in section 3, a matter we will return to later in the paper. Assessment of

the autocorrelation in the equations of the estimated models also shows that four lags

are necessary to remove autocorrelation, with the autocorrelation being particularly

strong at the fourth lag, confirming the regression results. We therefore choose four

lags.

An inspection of the cointegration test results for four lags in Table 5 shows

clear evidence of cointegration; this outcome does not depend on the deterministic

specification or on the type of test used. We therefore conclude that the log of GDP

and the log of the stock price index are cointegrated and proceed to a consideration of

the implications of this relationship.

We first look at the estimated cointegrating equation. A simple OLS

regression of the log of GDP (ly) on the log of stock prices (ls) produces the following:

lyt = 3.2713 + 0.9511 lst (2)

(4.65) (9.73)

where t-ratios appear in parentheses. Hence, a 1 percent increase in stock prices is

associated with a roughly equal percentage increase in GDP. Similar magnitudes of

15
the slope coefficient are obtained from more sophisticated estimates such as those

obtained from the Johansen procedure. If we estimate the equation within Johansen’s

VECM framework with four lags and an intercept in the cointegrating regression but

not in the VECM, we obtain:

lyt = -1.7519+ 1.6171 lst (2a)

(0.96) (6.58)

whereas if we also add a constant to the VECM equations, the cointegrating equation

becomes:

lyt = -1.3231 + 1.5905 lst (2b)

(6.53)

In both of these the slope coefficient is larger but of the same order of magnitude as in

the simple OLS estimate. All suggest that an increase in share prices of precent is

associated with an increase in real output of between 1 and 1.6 percent. Whether

these are plausible or not depends on the direction in which we imagine the causation

to run: a 1 percent increase in stock prices leading to a 1.6 percent increase in GDP

seems very large but a 1.6:1 relationship in the opposite direction is perhaps smaller

than many would expect. However, these equations cannot be used to infer causation,

something to which we now turn.

To address the direction and strength of causation between our two variables,

we must turn to the estimated VECM. Before we do, however, we briefly consider

the concept and measurement of causation in a cointegrated system of non-stationary

variables.

Testing for (Granger) causality is straightforward for pairs of stationary

variables: to test for causality from x to y, simply estimate an equation explaining y in

terms of lags of y and x and test for the joint significance of the lagged x variables.

For pairs of non-stationary non-cointegrated variables, it is also straightforward:

16
simply difference the variables until each is stationary and then apply the standard test

to the differenced variables. For pairs of non-stationary cointegrated variables we can

also difference and then apply the standard test to the stationary differenced series.

But this ignores the error-correction term which should be added to the VAR in the

differences (thus making it a VECM) and therefore carries out the test within a mis-

specified model. Moreover, and more importantly, it ignores the information in the

error-correction term and in particular that this terms also contains lagged x. It is

possible in this context to distinguish between long-run and short-run causality. This

distinction builds on the common interpretation of the VECM that the error-correction

term represents deviations from long-run equilibrium while the lagged first-

differenced terms capture short-run adjustment effects.

Using this distinction, testing for short-run causality is straightforward: to test

for causality from x to y test the joint significance of the lagged ∆x terms in the

VECM equation for ∆y.

Long-run causality tests are less common. We propose to use one based on

the work in the unpublished papers by Canning and Pedroni (1999, 2004). Although

they developed it for tests of causality in cointegrated panels, the test statistics are

actually derived in a single-equation context and we follow this derivation. At present

the test appears to be available for only two-variable models and we exposit it for this

case.

The intuition is simple. The error-correction term in the VECM consists of

deviations from the cointegrating vector which describes the long-run relationship

between the two variables. Since the two variables are governed by this long-run

equilibrium relationship, it must be the case that a change in one variable will be

associated, in the long run, with a change in the other in order to keep the relationship

17
satisfied. But this is not necessarily a causal relationship. It is possible, for example,

that an exogenous change in x will be followed by a change only in y or a change only

in x or, more likely, by a change in both x and y to ensure that equilibrium is re-

established. We can test this very simply using the significance of the coefficients of

the error-correction terms in the VECM equations: if a deviation from long-run

equilibrium caused by a change in x has a significant effect on dy (that is, the error-

correction term in the dy equation is significant), x causes y in the long run.

More formally, Canning and Pedroni say that x causes y in the long run if a

permanent shock to the x structural error has a permanent effect on y. In terms of the

earlier explanation: if a permanent shock to the x structural error is adjusted to in the

long run at least partly by a change in y then x causes y in the long run. As already

explained, the Canning-Pedroni test for this is based on the significance of the error-

correction term in the VECM equations but requires an additional restriction which is

that a shock to the y innovation has a permanent effect on y itself. In Canning and

Pedroni’s application, they derive this supplementary condition from the theory

underlying their model.5

The mechanics of the test may be briefly developed as follows. Consider a

two-variable VECM in x and y. Define z = (x,y)’ and ∆z = (∆x,∆y)’ where x and y are

both I(1) and cointegrated. We can write the stationary vector-moving average (VMA)

form of the two-equation model for x and y as

∆zt = F(L)εt (3)

where εt = (ε1t,ε2t)’ are the structural errors and F(L) is a (2x2) matrix of infinite-order

polynomials in the lag operator, L, Lnxt= xt-n. The (i,j) element of F(L) is given by:

5
This additional restriction appears to be ignored in many applications of the Canning-Pedroni
procedure, as does the restriction to two variables; see, for example, Basu, Chakraborty and Reagle
(2003), Christopoulos and Tsionas (2005), Narayan and Smyth (2008), Lee (2004) and Lee and Chang
(2008).

18
 F(L)ij = F0ij + F1ijL + F2ijL2 + …

which gives the time-path of the effects of a shock to εj on zi. Note that individual

elements of the Fτij sequence give the effects on ∆zi so that the sum gives the effect on

zi itself. The long run effect is just the sum of these effects from τ = 0 to ∞, that is

F(1)ij = F0ij + F1ij + F2ij + …

or just the accumulated effects on ∆zi. Write the matrix of F(1)ij elements as

 F (1)11 F (1)12 
F(1) =  
 F (1)21 F (1)22 

For the question of the causality running from x to y we are interested in knowing

whether F21 is non-zero. By the Granger representation Theorem (Engle and Granger,

1987) if x and y are I(1) and cointegrated, F(1) will contain a singularity of the form

F(1)λ = 0 where λ = (λ1, λ2)’ is the vector of coefficients of the error-correction terms

in the VECM equations. The condition F(1)λ = 0 has two equations, the second of

which contains the F(1) element of interest, F21 and is:

λ1F(1)21 + λ2F(1)22 = 0

The additional condition which they derived from their underlying theory that a shock

to the y innovation has a permanent effect on y itself implies that F22 ≠ 0. Considering

all possible combinations of signs of the λs and using the implication of the Granger

Representation Theorem that not both the λs are zero, it is simple to show that

F(1)21=0 if and only if λ2 = 0. Hence we can test the null hypothesis that F(1)21 = 0

(i.e., that x does not cause y in the long run) by testing the significance of the

coefficient of the error-correction term in the equation for ∆y; if we reject the null, we

conclude causation from x to y.

19
 Similarly we can test that y causes x in the long run by testing the significance

of the coefficient of the error-correction term in the ∆x equation provided we can

impose that the restriction that the long-run effect of x on itself is non-zero.

An alternative, less restrictive, approach to the testing for long-run causality is

based on the paper by Toda and Yamamoto (1995). It is applicable to variables which

may or may not be stationary and which may not be cointegrated (if non-stationary).

The test is based on a VAR model specified in the levels of the variables even though

they may be non-stationary. If this VAR model has known order of k and the highest

degree of integration of the variables is d, a VAR(k+d) is specified, estimated by OLS

and standard tests for causality are carried out, but using only the first k lags.

We now return to the estimation results. Setting lag length at 4 (on the basis

of previous tests), we found that the estimated VECM equations are not sensitive to

the specification of the deterministic components as long as the trend is excluded.

Thus, the slope coefficient in the cointegrating regression as well as the implications

for short-run and long-run causality do not depend whether there is an intercept in the

cointegrating regression or in the VECM equations. Moreover, all three cases are free

from autocorrelation and appear to be adequate representations of the system. In

Table 6 we report the intermediate case – with an intercept in the cointegrating

regression but not in the VECM equations.6 .

[Table 6 about here]

6
We also experimented with a trend term in the cointegrating equation. This resulted in the trend term
completely dominating the relationship between output and stock prices so that the latter was not
significant implying no long-run relationship between the two, despite the outcome of the Johansen test
to the contrary. Despite the lack of a relationship between the two variables, the error-correction term
was clearly significant in both VECM equations implying cointegration and the anomalous result that
stock prices adjust much more quickly to the gap of GDP from trend than GDP itself does. We
therefore did not proceed with this case.

20
 The implications of the VECM for short-run causality are straightforward to

derive and are reported in Table 7 for 4 lags and all three combinations of

deterministic components.

[Table 7 about here]

It is clear that there is no short-run causality from stock prices to output while there is

considerable evidence that there is causality in the opposite direction. The evidence

for the latter is strongest when there is an intercept in the cointegrating regression but

not in the VECM equations, marginal when there is no intercept in either and weak

when there is an intercept in both. As to the effects of varying lag length, the

evidence for short-run causality from output to stock prices is stronger with fewer lags

while the conclusion regarding causality in the opposite direction (no causality from

stock prices to output) is not sensitive lag length. We conclude that there is no

evidence of causality running from stock prices to output in the short run (irrespective

of the VECM specification) but considerable evidence for causality in the opposite

direction.

We turn now to examine the evidence for long-run causality. We focus on the

results using the Canning and Pedroni test. The results are reported in Table 8, again

for 4 lags and various combinations of deterministic components. The statistics are

just the estimated coefficients and corresponding t-statistics for the coefficients of the

error-correction terms in the VECM equations.

[Table 8 near here]

The results paint a very clear picture – there is strong evidence for causality from

output to stock prices and no evidence at all of causality in the opposite direction. For

the results reported in the table there is no sensitivity to the specification of the

deterministic components of the model. There is, however, sensitivity to the

21
exclusion of a trend from the cointegrating equation. If this is included, there is

evidence of causality in both directions at lags of 3 and 4. There is also some

sensitivity to lag length – at three lags, there is weak evidence of causality from stock

prices to output for one of the deterministic combinations (an intercept in the

cointegrating regression but none in the VECM). Given our earlier discussion of

specification, however, these are not important qualifications to the results reported in

Table 8 since the results are robust for our preferred lag length and for our preferred

deterministic configuration. We conclude, therefore, that there is convincing evidence

of long-run causality running from output to stock prices but none for causality in the

opposite direction.

The finding of one-way long-run causality conclusion is confirmed by the

application of the Toda-Yamamoto test discussed above. On the basis of a 4-lag

VECM, we would specify the VAR in log levels with 5 lags so that the Toda-

Yamamoto tests requires estimating a VAR in 6 lags and testing the significance of

the first five to assess long-run Granger causality. The results of the application to

our data are that there is no evidence of causality in either direction. However, it is

quite possible that this is because the long lags substantially reduce the power of the

test. This is borne out by the results obtained from shorter lags – the evidence for

causality from output to stock prices grows steadily as the lag length is reduced but

there is little change in the p-value for the test of causality in the opposite direction as

lags are dropped.

Finally, we can, of course, obtain some informal evidence of short-run and

long-run causality from simulations of the model to output and stock market shocks.

We capture the results of such simulations in impulse response function (IRFs).

These also provide information on the sign, magnitude and timing of the inter-

22
relationships between the variables. In Figure 1 we report the IRFs for our standard

model – one with 4 lags and an intercept in the cointegrating equation but not in the

VECM.

[Figure 1 near here]

The shocks are all unit shocks and not orthogonalised. Unit shocks were chosen so

that magnitudes could be compared across shocks. Orthogonalised shocks (using the

Cholesky approach) and generalised IRFs were also experimented with but were little

different to the ones reported, reflecting the low cross-equation correlation of the

residuals in the VECM (less than 0.01).

The IRFs bear out the conclusions we have built up so far. The two variables

are clearly non-stationary in that shocks to each has a permanent effect on the variable

itself, although the long-run effect is much smaller for share prices. In terms of

causality, in the short run there is a small positive effect running from stock prices to

output and a larger effect (initially negative) running from output to stock prices. The

initial negative effect of an output shock may explain the relatively weak evidence for

short-run Granger causality running in this direction despite the strong long-run

evidence.

In the long run, the effects of shocks are positive in both directions but there is

clearly a much larger effect running from output to stock prices than vice versa; a unit

shock to output has a long-run effect of about 3 on stock prices while a similar shock

to share prices has a negligible long-run effect on output. In terms of the earlier

discussion of the estimated cointegrating equation, most of any gap that appears

between stock prices and output is adjusted to by stock prices rather than output

before the long-run equilibrium is re-established.

23
 The change in the sign of the effect of an output shock on stock prices may

help reconcile apparently contradictory effects in the literature in which, as pointed

out earlier in this paper, some studies show a positive and others a negative effect of

output shocks on stock prices; our IRFs suggest that the effects may be negative for

the first two to three years before turning substantially positive. It is possible that

some of the differences arise from different time horizons, although there are

doubtless other influences as well.

The overall character of the IRFs is quite robust to alterative specifications. If

we vary the deterministic specification by also omitting the intercept from the

cointegrating regression, the IRFs look almost identical while if we include an

intercept in both the cointegrating regression and the VECM equations, the pattern of

effects is much the same although adjustment to the long-run position is quicker and

the magnitudes of the long-run effects are smaller. The inclusion of a trend in the

cointegrating equation produces anomalous effects: “own-effects” which are still

positive for output, zero for share prices and cross-effects are negative in both

directions in the long run.

If we reduce lag length to three, the overall conclusions are little changed: the

long-run effect of an output shock on stock prices is still positive and large while the

effect of a shock in the opposite direction is small (but now negative).

We conclude that our IRFs confirm the results of earlier Granger-causality

testing. Moreover, the magnitude of the effect of an output shock on share prices is

positive in the long run and at least an order of magnitude larger than the effect in the

opposite direction. The results are, as before, reasonably robust to alternative

specifications, although not to the inclusion of a trend in the cointegrating regression.

24
 Before rationalising our results and comparing them to the limited existing

literature, we return briefly to the matter of the seasonal adjustment of the output

series. It was noted above that there were often highly significant coefficients at four

lags in the VECM while lags 1, 2 and 3 were generally insignificant. This suggested

some residual seasonality after the X12 adjustment process. An inspection of the raw

data shows very distinct seasonal fluctuations as one would expect for quarterly

unadjusted GDP. There appears to have been a noticeable change in the seasonal

pattern in the middle of the sample period. In the first part of the sample the

amplitude of the seasonal fluctuations grow steadily as one would expect but around

2000 they decrease suddenly and markedly and then increase steadily again until the

end of the sample period. The X12 procedure, not surprisingly perhaps, has trouble

with this and there seems to be some residual seasonal movement in the middle of the

sample after adjustment. We therefore experimented with alternative adjustment

procedures and found that the use of the “Tramo/Seats” procedure available on

EViews produced an adjusted series without the above problems.7 The use of this

series resulted in outcomes reassuringly similar to those reported above – (i) the two

series are still clearly non-stationary, (ii)they are cointegrated, (iii)generally four lags

were required in the VECM but all lags were generally significant, (iv)the evidence

for short-run Granger causality is weak but less so from output to stock prices than

vice versa, (v)there is clear evidence of long-run causation from output to stock prices

but none for long-run causation in the opposite direction and (vi)the IRFs are similar

in shape to those in Figure 1. We capture the thrust of these supplementary results in

7
The EViews manual explains that “Tramo” is an acronym for “Time series regression with ARIMA
noise, missing observations and outliers” and “seats” for “signal extraction in ARIMA time series”.
Thank heaven for acronyms! It also claims that “it is a commonly used alternative to the Census X12
program”.

25
the IRFs for a VECM with four lags, and an intercept in the cointegrating equation but

not in the VECM. They are presented in Figure 2.

[Figure 2 near here]

We can summarise our results very simply. First, the logs of stock prices and

output are cointegrated with a significant positive long-run relationship between them.

Second, there is reasonable evidence of short-run Granger causality from output to

stock prices but no evidence at all of causality running from stock prices to output in

the short run. Third, there is strong and robust evidence of long-run causality from

output to stock prices but no causality in the opposite direction. Fourth, the long-run

effect on stock prices of a shock to output is at least several times larger than the

effect of a stock-price shock on output. And, finally, these results are generally robust

to reasonable variations in the model (both lag length and deterministic specification)

and to the method of seasonal adjustment of output.

In the light of the features of the Chinese financial system described briefly in

section 3 of the paper, these results are not surprising. We saw that while the Chinese

stock market has grown rapidly over the past two decades, it is still small relative to

the economy as a whole and provides only minor financing for new investment while

banks provide by far the bulk of financial resources for investment. Thus, it is

plausible that stock price fluctuations do not affect the rest of the economy through

either of the usual channels of investment and consumption. It is also consistent with

the results of one of the Chinese-language papers reviewed above, Liang and Teng

(2005), that shows output to be sensitive to variations in banking financial variables

but not to stock market fluctuations.

26
 Our results are somewhat at odds with the two existing English-language

papers in the area. The paper by Zhao (1999) found a strong contemporaneous effect

on stock prices of output shocks, both total and unexpected output changes. However,

the former had a negative effect and the latter a positive effect. The negative effect is

consistent with our short-run effect although we do not distinguish between expected

and unexpected shocks. Moreover, Zhao does not report any analysis of the effects in

the opposite direction – not surprisingly, since only contemporaneous variables were

used. Hence the results are likely to be confounded by reverse causation and it

therefore difficult to compare them to ours.

The results obtained by Liu and Sinclair (2008) are, however, more clearly

comparable to ours since they use a similar framework. While their conclusions

regarding long-run causation are similar to ours, their short-run results are the

opposite of ours – they find that short-run causation runs from stock prices to output

and not vice versa. They do not, however, report IRFs so it can not be seen how

quickly these are reversed over time as they must be if there is long-run causality

running in the opposite direction. Finally, Liu and Sinclair use real GDP data with

nominal stock prices which seems inconsistent and it is possible that the short-run

effects we have detected are mainly on the price component in the short run but on the

real component in the long run which would go some way to reconciling our results to

theirs.

27
 VI Conclusions

In this paper we have set out to analyse the relationship between GDP and

stock prices in China since the establishment of the Chinese stock market in the early

1990s. We did so using time-series techniques applied to quarterly, seasonally

adjusted GDP and the Composite Index for the Shanghai Stock Exchange. Both

series were used in log form.

We found both series to be non-stationary in the (log) levels but stationary in

the first differences. Cointegration tests pointed clearly to the existence of a positive

long-run relationship between the two variables.

In our analysis of Granger causality we distinguished between short-run and

long-run causality. The evidence for short-run causality was modest and favoured the

causality running from output to stock prices with no evidence for causality in the

opposite direction.

Evidence for long-run causality was much more clear-cut, with output clearly

causing stock prices in the long run. No evidence was found to support long-run

causation running in the other direction.

Simulations of the model confirmed these findings with output shocks having

an effect on stock prices which was much larger than any effect stock prices might

have on output.

We argued that this was consistent with the relative immaturity of the Chinese

stock market and its relatively minor role in the financial system as a whole. While it

is plausible for stock prices to respond to news of changes in the economy, it is not

surprising to find that changes in stock prices have at best minor repercussions on the

rest of the economy.

28
References

Aggarwal, R and Schirm, D. C. (1992), ‘Balance of Trade Announcements and Asset

Prices: Influence on Equity Prices, Exchange Rates and Interest Rates’, Journal
of International Money and Finance, 11, 80-95.
Antoniou, A., Garrett, I. and Priestley, R. (1998), ‘Macroeconomic Variables as
Common Pervasive Risk Factors and the Empirical Content of the Arbitrage
Pricing Theory, Journal of Empirical Finance, 5, 221-240.
Ariff, M. and Johnson, L. W. (1990), ‘Ex Ante Risk Premia on Macroeconomic
Factors in the Pricing of Stocks: An Analysis Using Arbitrage Pricing Theory’,
Ariff, M. and Johnson, L. W. (eds), Securities Markets and Stock Pricing,
Longman, Singapore; 194-204.
Baker, M., Stein, J. and Wurgler, J. (2003), ‘When Does the market Matter? Stock
Prices and the Investment of Equity Dependent Firms’, Quarterly Journal of
Economics, 118, 969-1006.
Balduzzi, P. (1995), ‘Stock Returns, Inflation and the ‘Proxy Hypothesis’: A New
Look at the Data’, Economics Letters, 48, 47-53.
Basu, P., Chakraborty, C. and Reagle, D. (2003), ‘Liberalization, FDI and Growth in
Developing Countries: A Panel Cointegration Approach’, Economic Inquiry,
41, 510-516.
Beenstock, M. and Chan, K. (1988),’Economic Forces in the London Stock Market’,
Oxford Bulletin of Economics and Statistics, 50, 27-39.
Bertaut, C. C. (2002), ‘Equity Prices, Household Wealth and Consumption Growth in
Foreign Industrial Countries: Wealth Effects in the 1990s’, International
Finance Discussion Paper Number 724, Board of Governors of the Federal
Reserve System.
Blanchard, O. J., Rhee, C. and Summers, L. (1993), ‘The Stock Market, Profit and
Investment’, Quarterly Journal of Economics, 108, 115-136.
Bodie Z., (1976), ‘Common Stocks As a Hedge Against Inflation’, Journal of Finance,
31,. 459-470.
Boucher, C. (2006), ‘Stock Prices-Inflation Puzzle and the Predictability of Stock
Returns’, Economics Letters, 90, 205-212.
Boudoukh, J. and Richardson, M. (1993), ‘Stock Returns and Inflation: A Long-
Horizon Perspective’, American Economic Review, 83, 1346-1355.
Breeden, D. (1979), ‘An Intertemporal Asset Pricing Model with Stochastic
Investment and Consumption Opportunities’, Journal of Financial Economics,
7, 265-296.
Breeden, D. T, Gibbons, M. R. and Litzenberger, R. H. (1989), ‘Empirical Tests of
the Consumption-Oriented CAPM’, Journal of Finance, 44, 231-262.
Campbell, J. Y. and Shiller, R. J. (1987), ‘Cointegration and Tests of Present Value
Models’, Journal of Political Economy, 95, 1062-1088.
Campbell, J. Y. and Shiller, R. J. (1988), ‘Stock Prices, Earnings and Expected
Dividends’, Journal of Finance, 43, 661-676.
Campbell, J. Y. and Ammer, J. (1993), ‘What Moves the Stock and Bond Markets? A
Variance Decomposition for Long-Term Asset Returns’, Journal of Finance,
48, 3-37.
Canning, D. and Pedroni, P. (1999), ‘Infrastructure and Long Run Economic Growth’,
Discussion Paper, Indiana University.
Canning, D. and Pedroni, P. (2004), ‘The Effect of Infrastructure on Long Run
Economic Growth’, Discussion Paper, Williams College.

29
Canova, F. and De Nicolo, G. (1995), ‘Stock Returns and Real Activity: A Structural
Approach’, European Economic Review, 39, 981-1015.
Case, K. E., Quigley, J. M. and Shiller, R. J. (2001), ‘Comparing Wealth Effects: The
Stock Market versus the Housing Market’, Wellesley College Working Paper
2001-09, Department of Economics, Wellesley College.
Cashin, P. and McDermott, C. J. (1998), ‘Testing the Consumption-CAPM in
Developing Equity Markets’, International Journal of Finance and Economics,
3, 127-141.
Cassola, N. and Morana, C. (2004), ‘Monetary Policy and the Stock Market in the
Euro Area’, Journal of Policy Modeling, 26, 387-399.
Chen, M.-H. (2003), ‘Risk and return: CAPM and CCAPM’, Quarterly Review of
Economics and Finance, 43, 369-393.
Chen, N. (1991), ‘Financial Investment Opportunities and the Macroeconomy’,
Journal of Finance, 46, 529-554.
Chen, N., Roll, R and Ross, S. A. (1986), ‘Economic Forces and the Stock Market’,
Journal of Business, 59, 383-403.
Chen, S. J. and Jordan, B. D. (1993), ‘Some Empirical Tests of the Arbitrage Pricing
Theory: Macro Variables vs. Derived Factors’, Journal of Banking and
Finance, 17, 65-89.
Cheng, A. C. S. (1996),’The UK Stock Market and Economic Factors: A New
Approach’, Journal of Business Finance and Accounting, 22, 129-142
Cheung, Y.-W. and Ng, L. K. (1998), ‘International Evidence on the Stock Market
and Aggregate Economic Activity’, Journal of Empirical Finance, 5, 281-296.
Chirinko, R. S. and Schaller, H. (1996), ‘Bubbles, Fundamentals and Investment: A
Multiple Equation Testing Strategy’, Journal of Monetary Economics, 38, 47-
76.
Chirinko, R. S and Schaller, H. (2001), ‘Business Fixed Investment and ‘Bubbles’:
The Japanese Case’, American Economic Review, 31, 663-680.
Chung, H. and Lee, B. S. (1998), ‘Fundamental and Nonfundamental Components in
Stock Prices of Pacific-Rim Countries’, Pacific-Basin Journal of Finance, 6,
321-346.
Choi, J. J., Hauser, S. and Kopecky, K. J. (1999), ‘Does the Stock Market Predict Real
Activity? Time Series Evidence from the G-7 Countries’, Journal of Banking
and Finance, 23, 1771-1792.
Christopoulos, D. K. and Tsionas, E.. G. (2005), ‘Productivity Growth and Inflation in
Europe: Evidence from Panel Cointegration Tests’, Empirical Economics, 30,
137-150.
Clare, A. D. and Thomas, S. N. (1994),’Macroeconomic Factors, the APT and the UK
Stock Market’, Journal of Business Finance and Accounting, 21, 309-330.
Crosby, M. (2001), ‘Stock Returns and Inflation’, Australian Economic Papers, 40,
156-165.
Edison, H. and Sløk, T. (2002), ‘Stock Market Wealth Effects and the New Economy:
A Cross-Country Study’, International Finance, 5, 1-22.
Entorf, H. and Jamin, G. (2007), ‘German Exchange Rate Exposure at DAX and
Aggregate Levels, International Trade and the Role of Exchange Rate
Adjustment Costs’, German Economic Review, 8, 344-374.
Fama, E. F. and Schwert, G. W. (1977), ‘Asset Returns and Inflation’, Journal of
Financial Economics, 5, 115-146.
Fan, X. J. (2006), ‘Financial Development and Economic Growth: An Empirical
Analysis from China’, Financial Research, 3, 57-66 (in Chinese).

30
Fischer, S. and Merton, R. C. (1984), ‘Macroeconomics and Finance: The Role of the
Stock Market’, Carnegie-Rochester Conference Series on Public Policy, 21,
57-108.
Gilchrist, S., Himmelberg, C. P. and Huberman, G. (2004), ‘Do Stock Price Bubbles
Influence Corporate Investment?’, Federal Reserve Bank of New York Staff
Report No. 177.
Gjerde, O. and Saettem, F. (1999), ‘Causal Relations Among Stock Returns and
Macroeconomic Variables in a Small, Open Economy’, Journal of
International Financial Markets, Institutions and Money, 9, 61-47.
Graham, F. C. (1996), ‘Inflation, Real Stock Returns and Monetary Policy’, Applied
Financial Economics, 6, 29-35.
Groenewold, N. and Fraser, P. (1997), ‘Share Prices and Macroeconomic Factors’
Journal of Business Finance and Accounting, 24, 1367-1383.
Groenewold, N., O’Rourke, G and Thomas, S. (1997), ‘Stock Returns and Inflation: A
Macro Analysis’, Applied Financial Economics, 7, 127-136.
Grossman, S. and Shiller, R. (1981), ‘The Determinants of the Variability of Stock
Market Prices’, American Economic Review, 71, 222-227.
Gultekin, N.B. (1983), ‘Stock Market Returns and Inflation: Evidence from Other
Countries’, Journal of Finance, 38, 49-65.
Gunsell, N. and Cukur, S. (2007), ‘The Effects of Macroeconomic Factors on the
London Stock returns: A Sectoral Approach’, International Research Journal
of Finance and Economics, 10, 140-152.
Han, X. H., Zhang, Y. Y. and Wu, C. M. (2008), ‘An Explanation on the Relationship
between Stock Market and Inflation:1992-2007’, Financial Research, 4, 21-
（
36. in Chinese）
Hess, P. J. and Lee, B. S. (1999), ‘Stock Returns and Inflation with Supply and
Demand Disturbances’, The Review of Financial Studies, 12, 1203-1218.
Ihsan, H., Ahmad, E. Ihsan ul Haq, M. and Sadia, H. (2007), ‘Relationship of
Economic and Financial Variables with Behaviour of Stock Returns’, Journal
of Economic Cooperation, 28, 1-24.
Ioannidis, C. and Kontonikas, A. (2008), ‘The Impact of Monetary Policy on Stock
Prices’, Journal of Policy Modeling, 30, 33-53.
Jaffe, J. and Mandelker, G. (1976), ‘The ‘Fisher Effect’ for Risky Assets: An
Empirical Investigation’, Journal of Finance, 31, 447-458.
Jain, P. C. (1988), ‘Response of Hourly Stock Prices and Trading Volume to
Economic News’, Journal of Business, 61, 219-231.
Kim, M. K. and Wu, C. (1987), ‘Macro-Economic Factors and Stock Returns’,
Journal of Financial Research, 10, 87-98.
Kocherlakota, N. R. (1997), ‘Testing the Consumption CAPM with Heavy-Tailed
Pricing Errors’, Macroeconomic Dynamics, b, 551-557.
Lai, M., Peng, S. and Bao Q., 2006. ‘Technology Spillovers, Absorptive Capacity and
Economic Growth’, China Economic Review, 17, 300-320.
Lee, B. S. (1992), ‘Causal Relations Among Stock Returns, Interest Rates, Real
Activity and Inflation’, Journal of Finance, 47, 1591-1603.
Lee, B. S. (1995), ‘The Response of Stock Prices to Permanent and Temporary
Dividend Shocks’, Journal of Financial and Quantitative Analysis, 30, 1-22.
Lee, B. S. (1998), ‘Permanent, Temporary and Non-Fundamental Components of
Stock Prices’, Journal of Financial and Quantitative Analysis, 33, 1-32.
Lee, C. (2005), ‘Energy Consumption and GDP in Developing Countries: A
Cointegrated Panel Analysis’, Energy Economics 27, 415-427.

31
Lee, C. and Chang, C. (2008), ‘Energy Consumption and Economic Growth in Asian
Economies: A More Comprehensive Analysis Using Panel Data’, Resource
and Energy Economics, 30, 50-65
Lettau, M. and Ludvigson, S. C. (2004), ‘Understanding Trend and Cycle in Asset
Values: Reevaluating the Wealth Effect on Consumption’, American
Economic Review, 94, 276-299.
Levine, R. (2005), ‘Finance and Growth: Theory and Evidence’, Aghion, P. and
Durlauf, S. N. (eds), Handbook of Economic Growth, Volume 1A, Elsevier,
865-934.
Liang, Q. and Teng, J. Z (2005), ‘Stock Market, Banking and Economic Growth:
Evidence from China’, Financial Research, 10, 9-19 (in Chinese).
Liu, X. and Sinclair, P. (2008), ‘Does the linkage between Stock market Performance
and Economic Growth Vary across Greater China?’, Applied Economics
Letters, 15, 505-508.
Ludvigson, S. and Steindel, C. (1999), ‘How Important is the Stock Market Effect on
Consumption?’, FRBNY Economic Policy Review, July 1999, 29-51.
Ludwig, A. and Sløk. T. (2002), ‘The Impact of Changes in Stock Prices and House
Prices on Consumption in OECD Countries’, IMF Working Paper WP/02/1,
Research Department, International Monetary Fund.
Mankiw, N. G. and Zeldes, S. P. (1991), ‘The Consumption of Stockholders and
Nonstockholders’, Journal of Financial Economics, 29, 97-112.
Martikainen, T. (1991), ‘On the Significance of the Economic Determinants of
Systematic Risk: Empirical Evidence with Finnish Data’, Applied Financial
Economics, 1, 97-104.
Mateev, M. and Videv, A. (2008), Multifactor Asset Pricing Model and Stock Market
in Transition: New Empirical Tests’, Eastern Economic Journal¸34, 223-237.
McMillan, D. G. (2005), ‘Time Variation in the Cointegrating Relationship between
Stock Prices and Economic Activity’, International Review of Applied
Economics, 19, 359-368.
Morck, R., Schleifer, A. and Vishny, R. W. (1990), ‘The Stock Market and
Investment: Is the Market a Sideshow?’, Brookings Papers on Economic
Activity 2, 157-215.
Narayan, P. K and Smyth, R. (2008), ‘Energy Consumption and Real GDP in G7
Countries: New Evidence from Panel Cointegration with Structural Breaks’,
Energy Economics, 30, 2331-2341.
Nelson, C. R., (1976), ‘Inflation and Rates of Return on Common Stock’, Journal of
Finance, 31, 471-483.
Pan, M.-S., Fok, R. C. and Liu, Y. A. (2007), ‘Dynamic Linkages between Exchange
and Stock Prices: Evidence from East Asian Markets’, International Review of
Economics and Finance, 16, 503-520.
Parker, J. A. (1999), ‘Spendthrift in America? On Two Decades of Decline in the U.S.
Saving Rate’, NBER Macroeconomics Annual, 317-370.
Pearce, D. K. and Roley, V. V. (1985), ‘Stock Prices and Economic News’, Journal of
Business, 58, 49-67.
Peiro, A. (1996), ‘Stock Prices, Production and Interest Rates: Comparison of Three
European Countries with the USA’, Empirical Economics, 21, 221-234.
Phylatkis, K. and Panazzolo, F. (2005), ‘Stock Prices and Exchange Rate Dynamics’,
Journal of International Money and Finance, 24, 1031-1053.
Poterba, J. M. (2000), ‘Stock Market Wealth and Consumption’, Journal of Economic
Perspectives, 14, 99-118.

32
Poterba, J. S. and Samwick, A. A. (1995), ‘Stock Ownership Patterns, Stock Market
Fluctuations and Consumption’, Brookings Papers on Economic Activity,
2:1995, 295-357.
Qi, D., Wu, W. and Zhang, H. (2000), ‘Shareholding Structure and Corporate
Performance of Partially Privatised Firms: Evidence from Listed Chinese
Companies’, Pacific-Basin Finance Journal, 8, 687-610.
Ran, M. S., Hu, G. P. and Wang, B. (2005), ‘An Empirical Analysis of Linkages
between Stcok Exchange Index and Economic Growth’, Journal of Chongqing
University (Natural Science Edition), 4, 157-159 (in Chinese).
Ran, M.S., Zhang, Z. Y and Wu, C. M. (2003), ‘An Explanation on the relationship
between Stock Market and Inflation: 1992-2007’, Financial Research, 4, 21-
36 (in Chinese).
Romer, C. D. (1990), ‘The Great Crash and the Onset of the Great Depression’,
Quarterly Journal of Economics, 105, 597-624.
Shirvani, H. and Wilbratte, B (2000), ‘Does Consumption Respond More Strongly to
Stock Market Declines than to Increases?’, International Economic Journal,
14, 41-49.
Shirvani, H. and Wilbratte, B. (2002), ‘The Wealth Effect of the Stock Market
Revisited’, Journal of Applied Business Research, 18, 218-232.
Siklos, P. and Kwok, B. (1999), ‘Stock Returns and Inflation: A New Test of
Competing Hypotheses’, Applied Financial Economics, 9, 567-581.
Singh, R. A. (1993), ‘Response of Stock Prices to Money Supply Announcements:
Australian Evidence’, Accounting and Finance, 33, 43-59.
Sousa, R. M. (2008), ‘Financial Wealth, Housing Wealth and Consumption’,
International Research Journal of Finance and Economics, 19, 167-191.
Starr-McCluer, M. (2002), ‘Stock Market Wealth and Consumer Spending’,
Economic Inquiry, 40, 69-79.
Thorbecke, W. (1997), ‘On Stock Market Returns and Monetary Policy’, Journal of
Finance, 52, 635-654.
Tobin, J. (1969), ‘A General Equilibrium Approach to Monetary Theory’, Journal of
Money, Credit and Banking, 1, 15-29.
Toda, H. Y. and Yamamoto, T. (1995), ‘Statistical Inference in Vector
Autoregressions with Possibly Integrated Processes’, Journal of Econometrics,
66, 225-250.
Tsuji, C. (2007), ‘What Macro-Innovation Risks Really Are Priced in Japan?’,
Applied Financial Economics, 17, 1085-1099.
Wei, Z. Y. and Yong, Z. Z. (2007), ‘An Research on the Relationship between
Economic Growth and Fluctuation of Stock Price: Evidence from HK
(China)’, Financial Research, 3, 112-124.
Wong, W.-K., Khan, H. and Du. J. (2006), ‘Do Money and Interest Rates Matter for
Stock Prices? An Econometric Study of Singapore and USA’, Singapore
Economic Review, 15, 31-51.
Zhao, X.-Q (1999), ‘Stock Prices, Inflation and output: Evidence from China’,
Applied Economics Letters, 6, 509-511.

33
 Table 1. Listed companies on the Chinese stock exchange
1996 1998 2003 2008

Companies which list A shares and B shares 530 851 1287 1625

Companies which only list B shares 85 106 111 109

Companies listing H shares 25 43 93 153
Data source: for 2008: http://www.csrc.gov.cn/; for other years: “China Finance Almanac”
(1996,1998,2003)

Table 2. The size of the Chinese stock market (billions of RMB)

1993 1998 2003 2008
Financing quantum 8.47 82.58 135.72 365.67
Turnover 369.8 2255.19 3211.53 26711.26
Total market value 347.43 1950.57 4245.77 12136.64
Circulating shares market
NA 574.56 1317.85 4521.39
value
Data source: http://www.csrc.gov.cn/

Table 3 Financial ratios for China

Year 1994 1996 1998 2001 2004 2008
Ratio of the value of all
7.65 13.83 23.11 44.72 23.18 40.37
shares to GDP (%)
Ratio of the value of
tradable shares to 1.98 4.03 6.81 14.86 7.31 15.04
GDP(%)
Ratio of loans from
financial institutions to 84.36 85.92 102.51 115.41 117.94 106.45
GDP (%)
Data source: the value of all shares and the value of tradable shares: http://www.csrc.gov.cn/; the loans
from financial institutions: http://www.pbc.gov.cn/ ; GDP: Wind data base for the years 1992 to 2002
China State Statistical Bureau’s “Financial Almanac”, various years, for the period since 2002.

34
Table 4. Tests for stationarity
Variable Lags
SIC 0 1 2 3 4
Intercept, -1.5920 -1.4860 -1.2797 -1.8476 -1.1232
1
no trend (0.4811) (0.5345) (0.6341) (0.3547) (0.7016)
ly
Intercept -2.7164 -2.2985 -2.0838 -3.0793 -3.040
0
and trend (0.2336) (0.4287) (0.5448) (0.1200) (0.1299)
Intercept, -2.6385 -1.6355 -1.4627 -3.3254 -1.7089
1
no trend (0.0905) (0.4590) (0.5461) (0.0177) (0.4219)
ls
Intercept -2.9503 -2.5767 -2.3994 -4.9866 -3.8416
3
and trend (0.1540) (0.2921) (0.3765) (0.0007) (0.0206)
-6.8824 -3.8912 -2.2446 -1.5227 -1.4534
None 3
(0.0000) (0.0002) (0.0250) (0.1189) (0.1352)
∆ly
-10.1346 -6.4643 -4.3612 -2.5787 -2.3246
Intercept 0
(0.0000) (0.0000) (0.0008) (0.1027) (0.1676)
-8.2204 -6.1124 -2.1024 -3.7296 -4.0899
None 0
(0.0000) (0.0000) (0.0350) (0.0003) (0.0001)
∆ls
-8.1501 -6.0768 -2.0065 -3.6064 -4.1625
Intercept 0
(0.0000) (0.0000) (0.2834) (0.0083) (0.0016)
Notes: ly and ls denote the logs of GDP and the stock price index respectively; ∆ denotes the first-
difference operator; the figures in the “SIC” column represent the number of lags chosen on the basis
of the SIC criterion; “Intercept and trend”, “Intercept, no trend”, “Intercept” and “None” denote the
specification of the deterministic components in the ADF equation; p-values for the null hypothesis of
non-stationarity are in parentheses.

35
Table 5. Tests for co-integration
Deterministic Specification
Intercept and
Intercept in CE Intercept in CE
Lags Statistic trend in CE
no intercept in and in the
and intercept
the VAR VAR
in the VAR
36.6371 9.1596 14.3759
Trace statistic
(0.0001) (0.3508) (0.6265)
1
29.9114 6.9593 8.4693
Max-EV
(0.0002) (0.4940) (0.7776)
29.6691 8.8946 13.2218
Trace statistic
(0.0019) (0.3752) (0.7208)
2
22.8117 7.0616 7.2475
Max-EV
(0.0035) (0.4819) (0.8832)
39.3640 27.2697 42.1642
Trace statistic
(0.0000) (0.0006) (0.0002)
3
26.7391 23.6776 32.6145
Max-EV
(0.0007) (0.0012) (0.0004)
22.9220 18.6790 31.2505
Trace statistic
(0.0210) (0.0160) (0.0097)
4
17.6191 17.2630 17.8338
Max-EV
(0.0266) (0.0163) (0.0829)
16.2087 11.7551 24.9061
Trace statistic
(0.1649) (0.1690) (0.0656)
5
11.92037 11.0488 14.2602
Max-EV
(0.1906) (0.1518) (0.2372)
Note: “CE” denotes the cointegrating equation and “EV” denotes the eigenvalue; the null hypothesis
for each test is that there are zero cointegrating vectors; p-values in parentheses

36
Table 6: Estimated VECM: 4 lags, intercept in the co-integrating equation

error-correction term: ect = lyt − 1.6171lst + 1.7519

[-6.58] [0.96]

Variable ∆ly equation ∆ls equation

-0.0084 0.1544
ect −1
[0.55] [ 4.14]
-0.0956 -0.4736
∆lyt −1
[0.82] [1.68]
0.1698 -0.3198
∆lyt − 2
[ 1.49] [1.16]
0.2785 -0.4816
∆lyt −3
[ 2.41] [1.73]
0.4982 -0.0739
∆lyt − 4
[ 4.10] [0.25]
0.0388 0.3825
∆lst −1
[0.90] [ 3.68]
0.0258 0.1220
∆lst − 2
[ 0.61] [ 1.20]
0.0178 0.3529
∆lst −3
[ 0.43] [ 3.55]
-0.0322 -0.0007
∆lst − 4
[-0.72] [0.01]
R-squared 0.1468 0.4311

Adjusted R-squared 0.0204 0.3468

Note: variables as defined in Table 4; absolute values of t-ratios in brackets

37
Table 7: Tests of short-run Granger causality

Deterministic specification
Alternative
Intercept in CE no
hypothesis No intercept in CE Intercept in CE and
intercept in the
or VAR in the VAR
VAR
2.1987 2.1929 1.4349
ly causes ls
(0.6993) (0.7003) (0.8382)
7.1973 10.0788 5.4419
ls causes ly
(0.1258) (0.0.0391) (0.2449)
Note: variables as defined in Table 4; the tests are based on a VECM with 4 lags; the statistics follow
from a Wald test of the restriction that the lags of the (first-differences of the) first variable are jointly
insignificant in the VECM equation for the (first-differences of the) second. They are χ2-distributed
with 4 degrees of freedom under the null of no causation. Numbers in parentheses are p-values.

Table 8: Tests for long-run Granger causality

Deterministic specification
Alternative
Intercept in CE no
hypothesis No intercept in CE Intercept in CE and
intercept in the
or VAR in the VAR
VAR
0.1668 0.1544 0.1549
ly causes ls
[ 3.9544] [ 4.1414] [ 4.0002]
-0.0149 -0.0084 -0.0138
ls causes ly
[0.8674] [0.5460] [0.8860]
Note: variables are as defined in Table 4; the tests are based on a VECM with 4 lags; the statistics
follow from a t-test of the restriction that the coefficient of the error-correction term in the equation for
the first-difference of the second variable is zero. The numbers are the estimated adjustment coefficient
and the corresponding (absolute) t-ratio in brackets.

38
 Response to Nonfactorized One Unit Innovations
Response of LY to LY Response of LY to LS
6 6

5 5

4 4

3 3

2 2

1 1

0 0
25 50 75 100 25 50 75 100

Response of LS to LY Response of LS to LS
4 4

3 3

2 2

1 1

0 0

-1 -1
25 50 75 100 25 50 75 100

Figure 1 Impulse response functions from a VECM with four lags and an intercept in
the cointegrating equation but not in the VECM. ly and ls are, respectively, the logs
of output and share prices.

39
 Response to Nonfactorized One Unit Innovations
Response of LYTS to LYTS Response of LYTS to LS
6 6

5 5

4 4

3 3

2 2

1 1

0 0
25 50 75 100 25 50 75 100

Response of LS to LYTS Response of LS to LS

4 4

3 3

2 2

1 1

0 0

-1 -1

-2 -2
25 50 75 100 25 50 75 100

Figure 2. Impulse response functions from a VECM with four lags and an intercept in
the cointegrating equation but not in the VECM; output seasonally adjusted using
“Tramo/seats”. lyts and ls are the logs of output (adjusted using Tramo/seats) and
share prices, respectively.

40

View publication stats