The Causality Direction of the Stock Market–Growth Nexus: Application of GMM Dynamic Panel Data and the Panel Ganger Non-causality Tests

Abstract

This study is an attempt to model the linkages between stock market development and economic growth in developed and emerging markets. The causality direction between stock market development and economic growth has also been examined in order to provide solid policy implications. Mainly, the relationship and direction of causality between stock market development and economic growth were tested using dynamic panel data analysis based on the Generalised Methods of Moment (GMM) and panel Granger non-causality, respectively. Additionally, panel unit root tests were also applied to check the stationary of the selected variables to avoid misleading results. The study focuses on 20 countries, both developed and emerging markets, and data were collected over the period 1990–2014 mainly from World Bank data series. Dynamic panel data analysis confirms that there is a statistically significant relationship between stock market development and economic growth in both developed and emerging markets. Further, the study emphasises that only the finance-led growth hypothesis is valid for emerging markets while developed markets support bidirectional causality between stock market development and economic growth, reflecting the existence of both the finance-led growth hypothesis and the growth-led finance hypothesis. Hence, it is crucial to formulate appropriate policies to shift unproductively allocated funds towards stock markets to meet long-term capital requirements to encourage higher economic growth.

JEL Classification: C230, C220, G100, E010

Keywords

Dynamic Panel Data Analysis Stock Market Development Finance-led Growth Hypothesis Granger Non-causality Test Economic Growth

1. Introduction

1.1 Background of the Study

Investigation into the finance–growth relationship has a long history which started with empirical studies by Bagehot (1873) and Schumpeter (1911). However, there has not been a commonly recognised or unique relationship between financial development and economic growth in history. More specifically, there is no consensus about the link between stock market development and economic growth. A great deal of the literature on stock markets and economic growth also illustrates conflicting empirical findings. Four main types of contradictory views on stock markets and economic growth can be identified from reviewing the literature. The first view is the ‘finance-led growth hypothesis’, also known as the ‘supply-leading hypothesis’. According to this hypothesis, stock markets and other intermediate financial institutions promote economic growth. Furthermore, this financial system leads efficient resource allocation and increased savings. Firms and investors can easily acquire the capital requirements which stimulate long-run economic growth. Scholars such as Goldsmith (1969), Hicks (1969), McKinnon (1973), Show (1973), Bencivenga and Bruce (1991), Christopoulos and Efthymios (2004) and Beck and Levine (2004) have supported this view.

A counterargument to the first view, the ‘Growth-Led Finance Hypothesis’ or ‘Demand-Following Hypothesis’ emerged. This states that higher economic growth necessarily increases the demand for financial institutions; therefore, expansion of the financial system is a consequence of higher economic growth. Empirical studies by Robinson (1952), Patrick (1966), Ireland (1994), Dritsaki and Bargiota (2005) and Capasso (2006) have confirmed this hypothesis. Demetriades and Khaled (1996), Luintel and Mosahid (1999) and Arestis and Kul (2001) have developed a third view. According to their conclusions, there is a bidirectional causality between financial development and economic growth. Their argument is that financial sector development raises economic growth and higher economic growth creates new markets by increasing the demand for financial facilities. Finally, Lucas (1988) emphasised there is no causal relationship between financial sector development and economic growth. Furthermore, he stressed that the role of the financial market has been ‘over-stressed’ by the financial literature. These conflicting notions in the empirical literature have motivated rational scholars to re-examine finance–growth linkages. This current study focuses on the stock market and economic growth linkages to provide a rigorous contribution to the existing literature.

The bulk of empirical studies have applied cross-sectional analyses, which do not allow for the observation of potential causality between the stock market and economic growth. In addition, cross-country analysis hinders the country-specific effect, leading to a significantly high error term. Further, many scholars such as Beck and Levine (2004) and Demetriades and Khaled (1996) have criticised cross-country analysis since it treats countries with different characteristics as a homogenous group. Apart from that, time-series analysis specifically focuses on one country, and thus results cannot be generalised. Similarly, time-series analyses do not address the issue of simultaneity. Beck and Levine (2004) tried to improve the existing model specification by introducing dynamic panel data analysis based on the Generalised Method of Moment (GMM) method which supported the supply-leading hypothesis. However, Christopoulos and Efthymios (2004) have criticised the study of Beck and Levine (2004) since the latter did not consider the stationary and the co-integration of the data series. Besides, Christopoulos and Efthymios (2004) have applied dynamic panel analysis based on GMM to overcome the weaknesses of Beck and Levine (2004), but studied only 10 developing countries, so the results cannot be generalised for developed markets. Hence, there still exists a gap that needs to be filled by a concrete and updated study which covers both emerging and developed markets.

The current study attempts to overcome the gaps in the literature. This study especially has applied dynamic panel data analysis based on the GMM estimation technique with proper instruments after carefully examining the stationary of the data using panel unit root tests. This estimation technique overcomes the problems attached to the cross-country and time series analyses. Moreover, the panel Granger causality test has been applied to capture the potential causal direction, particularly between economic growth and stock market development. Following Levine and Zervos (1998), a composite index has been created to measure stock market development. Furthermore, the results of the study can be confidently generalised for both emerging and developed markets, as the study is based on 20 emerging and developed markets over the period 1990–2015.

1.2 Objectives of the Study

The overall objective of the current study is to model the relationship between stock market development and economic growth and examine the causality direction between economic growth and stock market development in both developed and emerging markets. Furthermore, the study attempts to solve the issues related to previous studies; particularly, issues involving modelling techniques and formulating proxy variables were addressed throughout the study. More specifically, this study has two main objectives

Analysing the relationship between stock market development and economic growth of emerging and developed economies; and

Examining the direction of causality between stock market development and economic growth. The second objective examines the existence of unidirectional, bidirectional or non-causality between stock market development and economic growth of both emerging and developed markets during the sample time period.

2. Literature Review

A cross-country study by Atje and Jovanovic (1993) examined the relationship between development of the stock market and the economic growth of 40 countries over the period 1980–88. This study found a direct and statistically significant correlation between stock market development and economic growth. Further, the theoretical emphasis on stock markets, banks and economic growth led them to empirically test the relationships and they specifically found that stock markets influenced economic growth more than they did the banking sector. However, a re-examination of the above relationship by Harris (1997) using the two-stage least squares method was unable to confirm a positive link between expanded stock market activity and economic growth. Another empirical study by Demirgüç and Levine (1996) analysed the relationship between stock market development and output growth. Initially, they found that countries with a developed stock market have well-developed financial intermediaries, while countries with less developed stock markets have less developed financial intermediaries. They concluded that development of stock markets was also a function of the development of banks, pension funds and the insurance sector. Their results justified the positive relationship between stock market development and output growth. A study by Levine and Zervos (1998) evaluated the impact of stock markets and banks on economic growth. This cross-country study was based on 47 countries over the period 1976–93 and it investigated the relationships among stock market liquidity, market size, volatility, integration and economic growth. The findings of the study stressed a direct association between stock market liquidity and economic growth. Further, the results remained the same when other economic and political conditions were included as explanatory variables. Conversely, the study indicated that the proxy for market size, the ratio of stock market capitalisation to GDP, was not a good predictor of economic growth. This study is quite important due to its expansion in the number of observations, as a result of an increase in the sample period and the inclusion of appropriate stock market indicators. However, Levine and Zervos (1998) applied Ordinary Least Squares (OLS) regression which does not capture the simultaneity bias or country-specific fixed effects. Consequently, Zhu, Ash and Pollin (2002) criticised the empirical work by Levine and Zervos (1998) in terms of the way they controlled the outliers. Hence, Zhu et al. (2002) proved that under the appropriate controlling procedure for outliers, stock market liquidity was not a significance predictor of further economic growth.

Beck and Levine (2004) tried to overcome the statistical weaknesses in the existing literature. They applied dynamic panel data analysis based on the GMM technique. Their study covered data from 1976 to 1998, to examine the role of stock exchanges and the banking sector in growth. The results of the study indicated that both stock markets and banks positively affect economic growth. However, Christopoulos and Efthymios (2004) criticised this study highlighting that Beck and Levine (2004) had not considered stationary of the data series and their co-integration process. The study by Rousseau and Wachtel (2000) used the same method as Beck and Levine (2004); however, they considered co-integration of the variables and applied the Vector Error Correction Model (VECM) under the GMM technique. An application of the endogenous growth model by Carporale, Howello and Soliman (2005) used quarterly data from 1971Q1 to 1998Q4 for four countries. Their initial effort was to model the relationship among the stock market, investment and growth in countries such as Chile, South Korea, Malaysia and the Philippines. They also applied the Vector Auto Regressive (VAR) model for the analysis and their findings indicate that the stock market positively influenced the economic growth rate. They indicated that the causality ran from stock markets to economic growth by confirming the supply-leading hypothesis, in alignment with many scholars such as Filer, Hanousek and Campos (1999), Rousseau and Wachtel (2000) and Beck and Levine (2004).

Tang and Kwok (1997) studied the link between stock markets and economic growth of selected Asian countries. They applied both the Johansen co- integration test and the Granger causality test to capture both the long- and short-run relationship and direction of causality. Their research was based on quarterly data over the period 1980–2004. They had mixed results with countries following the finance-led growth hypothesis while others supported the growth-led finance hypothesis. Particularly, Tang and Kwok (1997) confirmed that the stock market variable and growth co-integrated for most of the selected Asian countries, with a one-way causality from stock market to economic growth. Japan and Korea supported the finance-led growth hypothesis, while India followed the growth-led finance hypothesis, and Sri Lanka showed no causal effect. Athapattu and Jayasinghe (2010) found a long-run relationship between the development of the stock market and economic growth for Sri Lanka during the period 1997–2008. They applied the Johansen co-integration test followed by the Granger causality test, and results from the latter confirmed the finance-led growth hypothesis. The Central Bank of Sri Lanka started to produce quarterly data starting from 1997; however, Tang and Kwok (1997) used quarterly data from 1980 even for Sri Lanka. Therefore, there is an issue with the accuracy of the data used by them.

In summarising, several important strands can be identified in the existing literature. A great deal of empirical works have been based on either time-series or cross-country analysis rather than on panel data analysis. However in the cross-country analyses, the studies have assumed that the selected countries are homogenous, and do not account for country-specific institutional conditions such as the legal, political and financial systems. Apart from this, the cross-country studies do not provide the direction of causality and hence the causality effect among the variables may be ambiguous. Unlike the cross-sectional analyses, the time-series analyses permit observation of country-specific conditions and potential causality directions (Rousseau & Wachtel, 1998). However, most of the time-series analyses have specifically focused on a single country so their results may not be generalised. Similarly, time-series analyses do not address the issue of simultaneity (Beck & Levine, 2004), so it is crucial to introduce a dynamic panel speciation. In addition, stock market development indicators used by the scholars also need to be revised. An appropriate composite index should be recommended to represent overall stock market behaviour rather than an individual indicator. Numerous empirical works have also skipped the growth explanatory variables which leads to an exaggeration of the impact of stock market development. The current study attempts to fill these gaps and overcome the weakness in the existing literature.

3. Data and Methodology

3.1 Data

The current paper uses the annual dataset of the World Bank data series which consists of 20 countries, both developed and emerging markets, relating to the period 1990–2015 (see Table 1).

Table 1
List of Sample Countries

Developed Markets Emerging Markets

Japan
USA
UK
Australia
New Zealand
Switzerland
Canada
France
Germany
Norway Sri Lanka
India
Saudi Arabia
Thailand
Hong Kong
China
Brazil
South Africa
Bangladesh
Malaysia

Developed Markets	Emerging Markets
Japan USA UK Australia New Zealand Switzerland Canada France Germany Norway	Sri Lanka India Saudi Arabia Thailand Hong Kong China Brazil South Africa Bangladesh Malaysia

Source: Created by the author.

3.2 Econometric Methodology

The current study has applied concrete econometric techniques to overcome modelling issues highlighted in the literature. First, panel unit root tests were employed to check the stationary of the data panel. After that, GMM panel data analysis was carried out followed by a panel Granger causality test to capture the nature of the relationship and causality direction between economic growth and stock market development, respectively.

3.2.1 Panel Unit Root Tests

The panel unit root tests are crucial to identify the stationary of data series (Christopoulos & Efthymios, 2004). Keeping in mind that the study by Beck and Levine (2004) was criticised by Christopoulos and Efthymios (2004) due to the absence of panel unit root tests, the current study has employed the Im, Pesaran and Shin (IPS) and the Augmented Dicky–Fuller (ADF) panel unit root tests to check stationary of the panel.

The Im, Pesaran and Shin and Augmented Dickey-Fuller tests: Im, Pesaran and Yongcheol (2003) expressed the first-order autoregressive stochastic process y_i,t considering a sample of N cross-section observation and T time periods.

y_{i, t} = (1 - θ_{i}) γ_{i} + θ_{i} y_{i, t - 1} + e_{i t}

(1)

Taking the difference of (1)

Δ y_{i t} = δ_{i} + μ_{i} y_{i, t - 1} + e_{i t}

(2)

where

δ_{i} = (1 - θ_{i}) μ_{i}, μ_{i} = - ​ (​ 1 - θ_{i}) a n d Δ y_{i t} = y_{i t} - y_{i, t - 1}

such that, the null hypothesis of unit root becomes

H_{0} : β_{i} = 0 f o r a l l i

The same null hypothesis is also valid for the Fisher Chi-square ADF test as well. If the null hypothesis can be rejected, the series are stationary at the levels.

3.2.2 GMM Dynamic Panel Data Analysis

The GMM dynamic panel approach was used under the 2SLS method to estimate the empirical model in Equation (3). Equation (3) was construed for both emerging and developed markets

\ln R G D P_{i, t} = β_{3} S T O C K_{i, t} + γ X_{i, t} + δ_{1, i} + ε_{1 i, t}

(3)

In Equation (3), lnRGDP indicates the log of real GDP while STOCK stands for the stock market development variable. STOCK was created by employing the method used by Levine and Zervos (1998). X is a vector of other control variables which includes capital formation, the labour force, per capita income, inflation, openness, external debt and a dummy variable for an economic crisis. Further, δ and ϵ are the unobserved country-specific effect and the error term, respectively.

Incorporating the dynamic nature into Equation (3)

\begin{matrix} \ln R G D P_{i, t} - \ln R G D P_{i, t - 1} = (α_{1} 1) \ln R G D P_{i, t - 1} + γ ​_{1} S T O C K_{i, t} + \\ β_{1} X_{i, t} + η_{1 i} + e_{1 i, t} \end{matrix}

(4)

The above equation can be rearranged as

y_{i, t} = α_{1} \ln R G D P_{i, t - 1} + γ_{1} S T O C K_{i, t} + β_{1} X_{i, t} + η_{1 i} + e_{1 i, t}

(5)

The first difference of Equation (5) was constructed to eliminate unobserved country-specific effects.

\begin{matrix} y_{i, t} - y_{i, t - 1} = α_{1} (\ln R G D P_{i, t - 1} - \ln R G D P_{i, t - 2}) + \end{matrix} \begin{matrix} γ_{1} (S T O C K_{i, t} - S T O C K_{i, t - 1}) + \end{matrix} \begin{matrix} β_{1} (X_{i, t} - X_{i, t - 1}) + (e_{1 i, t} - e_{1 i, t - 1}) \end{matrix}

(6)

The endogeneity problem can be overcome by introducing lag values of the regressors as instruments based on the following moment conditions

E [\ln R G D P_{i, t - s} (e_{1 i, t} - e_{1 i, t - 1})] = 0 f o r s \geq 2; t = 3, .... T

(7)

E [S T O C K_{i, t - s} (e_{2 i, t} - e_{2 i, t - 1})] = 0 f o r s \geq 2; t = 3, .... T

(8)

E [X_{i, t - s} (e_{2 i, t} - e_{2 i, t - 1})] = 0 f o r s \geq 2; t = 3, .... T

(9)

Further, the Sargan test was used to test the overall validity of the moment condition and the instruments, while the serial correlation test checks whether the error terms are serially correlated.

3.2.3 Panel Test of Granger Non-causality

The Granger non-causality hypothesis developed by Hurlin (2005) was applied to capture the exact causality direction between stock market development and economic growth of the selected countries. Assume that x and y are stationary variables, observed in T periods and for N countries. The following heterogeneous autoregressive model [given in Equation (10)] can be considered, in which i = 1, …, N for each country and t = 1, …, T for each time period.

y_{i, t} = α_{i} + \sum^{​} ​ γ_{i}^{}_{}^{(k)} y_{i, t - k} + \sum^{​} ​ β_{i}^{}_{}^{(k)} x_{i, t - k} + ε_{i, t}

(10)

Both $γ_{i}^{}_{}^{(k)}$ and $β_{i}^{}_{}^{(k)}$ vary across countries. For each country i = 1, …, N, the innovation $ε_{i, t}, \forall t = 1, ...., T$ are i.i.d. $(0, σ_{i}^{}_{}^{2})$ and are independently distributed across groups. The attached null and alternative hypotheses to Equation (10) are as follows:

H_{0} = β_{i} = 0 \forall_{i} = 1, ... N

(11)

H_{1} = β_{i} \neq 0 \forall_{i} = N_{1} + 1, N_{2} + 2, ..., N

(12)

If the null hypothesis is accepted, x does not Granger cause y for all countries in the panel. In contrast, rejection of the null hypothesis confirms that x Granger causes y for all countries in the panel.

The test statistics that can be used to check the hypotheses is defined as the average of the individual Wald statistics, as given below

W_{N, T}^{H n c} = \frac{1}{N} \sum^{​} ​ W_{i, T}

(13)

Based on given Wald statistics, both standardised and approximated standardised statistics can be calculated as indicated in Equations (14) and (15), respectively.

Z_{N, T}^{H n c} = \sqrt{\frac{N}{2 K}} (W_{N, T}^{H n c} - K) \frac{\underset{\to}{d}}{T, N \to \infty} N (0, 1)

{\tilde{Z}}_{N, T}^{H n c} = \frac{\sqrt{N [W_{N, T}^{H n c} - N^{- 1} \sum^{​} ​ E (W_{i, T})]}}{\sqrt{N^{- 1} \sum^{​} V a r (W_{i, T})}}

(14)

If the standardised and approximated standardised statistics are higher than that of the respective critical values, the homogenous non-causality (HNC) hypothesis is rejected and in turn x Granger causes y.

4. Results and Discussion

4.1 Results of the Unit Root Tests

Tables 2 and 3 summarise the results of the unit root test for emerging and developed markets, respectively. According to the IPS and ADF tests, most of the variables are non-stationary at level forms. However, all the series of the emerging and developed markets are stationary at the first difference.

4.2 Results of the GMM Dynamic Panel Data Analysis for Emerging Markets

The results of the GMM dynamic panel estimations for emerging markets are given in Table 4. LNRGDP (the economic growth variable) is the dependent variable along with a set of independent variables, including the composite index for stock market development (STOCK). According to the results, stock market development positively affects economic growth in emerging markets. Further, this relationship is statistically significant at the 1 per cent level confirming that stock market development stimulates economic growth of emerging markets.

Table 2
Unit Root Results for Countries with Emerging Markets

Series Level
First Difference

IPS ADF IPS ADF

CRISIS 0.0000* 0.0000* 0.0000* 0.0000*

LNCF 0.1762 0.1421 0.0000* 0.0000*

LNIMEX 0.8748 0.9439 0.0003* 0.0001*

LNINFLATION 0.0291 0.0310 0.0000* 0.0000*

LNLF 0.7960 0.5807 0.0293 0.0132

LNPERCAP 1.0000 0.9998 0.0038* 0.0134

LNRGDP 1.0000 1.0000 0.0000* 0.0000*

LNDEBT 0.5623 0.1452 0.0002* 0.0013*

STOCK 0.0000* 0.0000* 0.0000* 0.0000*

Series	Level	First Difference
CRISIS	0.0000***	0.0000***	0.0000***	0.0000***
LNCF	0.1762	0.1421	0.0000***	0.0000***
LNIMEX	0.8748	0.9439	0.0003***	0.0001***
LNINFLATION	0.0291**	0.0310**	0.0000***	0.0000***
LNLF	0.7960	0.5807	0.0293**	0.0132**
LNPERCAP	1.0000	0.9998	0.0038***	0.0134**
LNRGDP	1.0000	1.0000	0.0000***	0.0000***
LNDEBT	0.5623	0.1452	0.0002***	0.0013***
STOCK	0.0000***	0.0000***	0.0000***	0.0000***

Source: Author’s calculation based on World Bank data.

Notes: ***significant at 1 per cent; **significant at 5 per cent. IPS, Im, Pesaran and Shin; ADF, Augmented Dicky–Fuller.

Table 3

Unit Root Results for Countries with Developed Markets

Series	Level		First Difference
Series	IPS	ADF	IPS	ADF
CRISIS	0.0000***	0.0000***	0.0000***	0.0000***
LNCF	0.0165**	0.0142**	0.0000***	0.0000***
LNIMEX	0.2761	0.2118	0.0000***	0.0000***
LNINFLATION	0.0000***	0.0000***	0.0000***	0.0000***
LNLF	1.0000	0.9692	0.0012***	0.0005***
LNPERCAP	0.9937	0.9755	0.0000***	0.0000***
LNRGDP	0.5809	0.5795	0.0000***	0.0000***
LNDEBT	0.8251	0.7562	0.0032***	0.0056***
STOCK	0.0003***	0.0021***	0.0000***	0.0000***

Source: Author’s calculation based on World Bank data.

Notes: ***significant at 1 per cent; **significant at 5 per cent. IPS, Im, Pesaran and Shin; ADF, Augmented Dicky–Fuller.

Table 4

GMM Dynamic Panel Analysis—Impact of Stock Market Development on Economic Growth in Emerging Markets

Dependent Variable: LNRGDP Estimation Technique: Panel GMM - 2SLS method (first difference) Observations: 118
Variable	Coefficient	Std. Error	t-statistic	Prob.
LNRGDP (−1)	0.4795	0.0678	7.0721	0.0000***
STOCK	0.0209	0.0054	3.8504	0.0002***
LNCF	0.0212	0.0433	0.4892	0.6256
LNLF	0.3794	0.1478	2.5662	0.0116**
LNOPEN	0.1954	0.0424	4.6105	0.0000***
LNPERCAP	0.2343	0.0367	6.3811	0.0000***
LNINFLATION	0.0010	0.0063	0.1660	0.8684
LNDEBT	−0.0264	0.0109	−2.4220	0.0359**
CRISIS	−0.0172	0.0077	−2.2108	0.0291**
Instrument rank J-statistic	32.0000
Instrument rank J-statistic	25.0809
Sargan test ( p -value)¹	0.4013
Serial correlation ( p -value)²	0.6031

Source: Author’s calculation based on World Bank data.

Notes: ¹ Sargan test has the null hypothesis that the over-identifying restrictions are valid.

² The serial correlation test has the null hypothesis that the error terms are not serially correlated.

***significant at 1 per cent; **significant at 5 per cent.

Apart from that, the lag of real GDP, log of the labour force, log of openness and log of per capita income also positively affect economic growth and are highly statistically significant, while the log of inflation and of capital formation are statistically insignificant. In fact, labour force is one of the key drivers of economic growth in emerging markets. South Asian, Latin American and African countries particularly still rely on their labour force, since the contribution of technical progress is considerably low. Openness of the economy has also been considered a source of growth by many scholars such as Banda (2005), Levine and Zervos (1998) and Beck and Levine (2004). They have verified that there is a positive relationship between economic openness and economic growth in both developed and developing countries. According to Banda (2005) free trade and economic liberalisation stimulate the allocation efficiency of an economy which, in turn, stimulates higher economic growth. In addition, Krueger (1997), Helpman and Krugman (1969) highlighted the inefficiencies of import-substitution policies, and mentioned that free trade shifts resources from inefficient import-substitution activities to efficient comparative advantage ones. In addition, per capita income is also a common factor in growth regressions and the current study observed a positive relationship between the log of per capita income and economic growth which is consistent with the findings of Banda (2005), Beck and Levine (2004), Levine (1993), Levine and Zervos (1998) and Osinubi (2002).

Economic crisis and the log of external debt variables have indicated a negative relationship with the economic growth as expected, and is significant at the 5 per cent level. Yuncu (2007) also found the same results but did not include the recent global financial crisis in his crisis variables. Conversely, inflation has become an insignificant factor in explaining economic growth. Ireland (1994) also obtained the same results, and stressed that the impact of inflation on economic growth is considerably low, and may die out entirely in the long run. Further, he stated that capital formation is insignificant for emerging markets. In fact, the marginal physical product of capital in South Asian and Latin American countries was remarkably lower, due to technical inefficiencies. Hence, the contribution of existing capital formation has been not sufficient to promote economic growth. In the context of model specification, the null hypothesis of the Sargan test suggests that over-identifying restrictions are valid. According to the p-value (0.4013) of the Sargan test, the null hypothesis cannot be rejected; therefore, the included instruments in the model are valid. Similarly, the null hypothesis of the serial correlation test indicates that the error terms are not serially correlated. Further, the insignificant p-value (0.6031) of the serial correlation test confirms that the error terms are not serially correlated. Therefore, the estimated GMM dynamic panel data model aligns with the econometric theory.

4.3 Results of the GMM Dynamic Panel Data Analysis for Developed Markets

This section discusses the relationship between stock market development and economic growth of developed markets. Table 5 summarises the stock market–growth relationship by assigning economic growth (LNRGDP) as the dependent variable. According to the estimated results, all the variables are positively affect by economic growth except the economic crisis variable.

Table 5
GMM Dynamic Panel Analysis: Impact of Stock Market Development on Economic Growth in Developed Markets

Dependent Variable: LNRGDP
Estimation Technique: Panel GMM-2SLS method (first difference)
Observations: 106

Variable Coefficient Std. Error t-Statistic Prob.

LNRGDP (−1) 0.5309 0.1440 3.6861 0.0003***

STOCK 0.0269 0.0128 2.0944 0.0384**

LNCF 0.0032 0.0563 0.0570 0.9546

LNLF 0.4017 0.3050 1.3172 0.1903

LNOPEN 0.0917 0.0431 2.1278 0.0355**

LNPERCAP 0.0608 0.0251 2.4210 0.0170**

LNINFLATION 0.0007 0.0033 0.2242 0.8230

LNDEBT −0.0242 0.0094 −2.5744 0.0277**

CRISIS −0.0198 0.0051 −3.8772 0.0002***

Instrument rank
J-statistic 31.0000

15.9511

Sargan Test ( p -value)¹ 0.8573

Serial Correlation
( p -value)² 0.0991

Dependent Variable: LNRGDP Estimation Technique: Panel GMM-2SLS method (first difference) Observations: 106
LNRGDP (−1)	0.5309	0.1440	3.6861	0.0003***
STOCK	0.0269	0.0128	2.0944	0.0384**
LNCF	0.0032	0.0563	0.0570	0.9546
LNLF	0.4017	0.3050	1.3172	0.1903
LNOPEN	0.0917	0.0431	2.1278	0.0355**
LNPERCAP	0.0608	0.0251	2.4210	0.0170**
LNINFLATION	0.0007	0.0033	0.2242	0.8230
LNDEBT	−0.0242	0.0094	−2.5744	0.0277**
CRISIS	−0.0198	0.0051	−3.8772	0.0002***
Instrument rank J-statistic	31.0000
15.9511
Sargan Test ( p -value)¹	0.8573
Serial Correlation ( p -value)²	0.0991

Source: Author’s calculation based on World Bank data.

Notes: ¹ The Sargan test has the null hypothesis that the over-identifying restrictions are valid.

² The serial correlation test has the null hypothesis that the error terms are not serially correlated.

***significant at 1 per cent; **significant at 5 per cent.

The coefficient of the stock market development variable is positive and statistically significant at the 5 per cent significance level, reflecting that liquidity, the efficient allocation of resources and capital generated by stock markets expedite the pace of economic growth in developed countries. These results for developed markets are also consistent with cross-country and panel data analyses by Levine and Zervos (1998) and Beck and Levine (2004). However, Yuncu (2007) found an insignificant relationship between development of the stock market and economic growth for developed markets. He considered both stock markets and the banking sector to represent the financial sector, which may underestimate the contribution of the stock market. At the same time, he has used the value-traded ratio as a proxy for stock market development rather than including a composite index which could cover all aspects of the stock market.

Another finding is that both economic openness and per capita income raise the economic growth rate significantly in developed economies. In fact, these developed markets are highly liberalised and follow outward-oriented economic policies while gaining from international trade by applying comparative advantages. Despite the positive labour force and capital formation link with economic growth positively, neither the labour force nor capital formation is significant. The main reason is that major developed economies are currently focusing on technology and total factor productivity as the engines of economic growth rather than on factor accumulation. Simultaneously, some developed countries have already employed most of their capital stock and are only maintaining capital stock instead of creating new capital. Though developed countries have a stable economic set-up, the global economic crisis and accumulated external debt have adversely affected their economic growth rate. Thus, the economic crisis variable and log of external debt negatively link with economic growth and are significant at the 1 per cent and 5 per cent significance levels, respectively. The goodness-of-fit of the model was evaluated using the Sargan test and the serial correlation test. These tests verify the appropriateness of the instruments and serial correlation of the errors, respectively. The p-values attached to these tests indicate that the instruments are valid and the error terms are not serially correlated.

4.4 Results of the Panel Granger Non-causality Test

The panel Granger non-causality test was employed to identify the exact causality direction between stock market development and economic growth based on the HNC hypothesis. The hypothesis was tested based on standardised statistics (Z_HNC) and approximated standardised statistics $(\tilde{Z})$ . The results of the test for developed and emerging markets are given in Tables 6 and 7, respectively.

Further, the tests were repeated for three lags in order to check the robustness of the causality direction. According to Table 6, all the test statistics for ‘STOCK to RGDP’ are significant at the 1 per cent level and hence the null hypothesis (that the stock market does not Granger-cause economic growth) can be rejected. Therefore, stock market development essentially stimulates economic growth in countries with emerging stock markets. However, a reverse causality, from economic growth to stock market development does not exist in emerging markets as the null hypothesis for ‘RGDP to STOCK’ cannot be rejected at any lag order. Consequently, only a unidirectional causality, which runs from stock market development to economic growth, exists in emerging markets, supporting the finance-led growth hypothesis.

Table 6
Results of the Granger Non-causality Test for Emerging Markets

Lag Order (K)

K = 1 K = 2 K = 3

Test statistics STOCK to RGDP

W_HNC 2.3511 3.9463 4.7853

Z_HNC 5.3421* 5.5423* 4.3423*

${\tilde{Z}}_{}$ 4.4323* 4.1232* 2.4345

Test statistics RGDP to STOCK

W_HNC 1.7342 2.3423 2.9843

Z_HNC 2.4312** 0.7786 0.0873

${\tilde{Z}}_{}$ 1.8021 0.3245 −0.4326

Lag Order (K)
Test statistics	STOCK to RGDP
W_HNC	2.3511	3.9463	4.7853
Z_HNC	5.3421***	5.5423***	4.3423***
${\tilde{Z}}_{}$	4.4323***	4.1232***	2.4345**
Test statistics	RGDP to STOCK
W_HNC	1.7342	2.3423	2.9843
Z_HNC	2.4312**	0.7786	0.0873
${\tilde{Z}}_{}$	1.8021	0.3245	−0.4326

Source: Author’s calculation based on World Bank data.

Notes: *** Significant at 1%, ** Significant at 5%.

Table 7

Results of the Granger Non-causality Test for Developed Markets

Lag Order (K)
	K = 1	K = 2	K = 3
Test statistics	STOCK to RGDP
W_HNC	4.7856	5.6773	6.4573
Z_HNC	14.3472***	9.6232***	7.4552***
${\tilde{Z}}_{}$	12.4535***	7.8786***	5.6353***
Test statistics	RGDP to STOCK
W_HNC	4.2311	4.6728	5.5663
Z_HNC	11.9636***	6.8979***	4.7868***
${\tilde{Z}}_{}$	10.3537***	5.6565***	3.5633***

Source: Author’s calculation based on World Bank data.

Notes: *** Significant at 1%.

Developed markets in contrast support both the finance-led growth and growth-led finance hypotheses. As Table 7 indicates, the null hypotheses attached to both ‘STOCK to RGDP’ and ‘RGDP to STOCK’ can be rejected at the 1 per cent significance level and, in turn, the alternative hypothesis, can be accepted which is that stock market development Granger-causes economic growth and economic growth Granger-causes stock market development, respectively.

The robustness of the results holds at a higher level as all the lag orders confirm rejecting the null hypotheses at the 1 per cent level. Thus, bidirectional causality exists in developed markets reflecting their stable economic growth and well-developed financial systems. The results are consistence with Tang and Kwok (1997) and Athapattu and Jayasinghe (2010), although both studies are time-series analyses which have applied only the time-series Granger casualty test.

5. Conclusions

This study attempts to investigate whether stock market development promotes economic growth of emerging and developed economies, and to observe the direction of causality focusing on the finance-led growth hypothesis and the growth-led finance hypothesis. The results of GMM panel data analyses confirm that stock market development significantly stimulates economic growth in both emerging and developed markets. In developed markets, there is a bidirectional causality between stock market development and economic growth, which confirms the existence of both the finance led-growth and the growth led-finance hypotheses. In contrast, there is only a unidirectional causality which runs from stock market development to economic growth in emerging markets, indicating that only the finance led-growth hypothesis is valid for emerging markets. In addition to the main findings, all the variables (lag of real GDP, labour force, capital formation, openness, per capita income and inflation) are positively related with economic growth in both emerging and developed markets, except for the economic crisis and external debt variables. Economic crisis and the log of external debt have a negative and significant impact on economic growth. Capital formation, the labour force and inflation do not significantly affect economic growth. Since development of the stock market stimulates economic growth in emerging economies, it is important to stimulate stock market activities under higher levels of regulation. Further, as the causality runs from stock market development to economic growth, it is clear that the stock market can be used to stimulate economic growth. Hence, it is highly recommended that necessary policies are formulated to shift unproductively allocated funds towards stock markets, to meet long-term capital requirement which is essential for long-term economic growth. Further, it is crucial to promote a favourable economic and political climate to obtain the optimal contribution from stock markets.

Footnotes

Appendix

References

Arestis

P.P.O.D.

, & Kul

B.L.

(2001). Financial development and economic growth: The role of stock markets. Journal of Money, Credit and Banking, 33(1), 16–41.

Athapattu

A.R.

, & Jayasinghe

(2010). Stock market performance and economic growth: The case of Sri Lanka. Proceedings of International Conferences on Information and Business (pp. 83–92). Colombo, Sri Lanka: University of Colombo.

Atje

, & Jovanovic

(1993). Stock markets and development. European Economic Review, 37(2), 632–40.

Bagehot

(1873[1999]). Lombard street. Homewood, IL: Irwin.

Banda

O.G.

(2005). Foreign investment inflows, government institutions, external openness and economic growth in developing countries: A theoretical and empirical investigation (unpublished PhD thesis).

Beck

, & Levine

(2004). Stock markets, banks, and growth: Panel evidence. Journal of Banking and Finance, 28(3), 423–42.

Bencivenga

V.R.

, & Bruce

D.S.

(1991). Financial intermediation and endogenous growth. Review of Economic Studies, 58(2), 195–209.

Capasso

(2006). Endogenous information frictions, stock market development and economic growth. The Manchester School, 76(2), 204–22.

Carporale

G.M.

, Howello

, & Soliman

A.M.

(2005). Stock market development and economic growth: The causal linkage. Journal of Economic Development, 29(1), 33–50.

10.

Christopoulos

D.K.

, & Efthymios

G.T.

(2004). Financial development and economic growth: Evidence from panel unit root and cointegration tests. Journal of Development Economics, 73(1), 55–74.

11.

Demetriades

P.O.

, & Khaled

A.H.

(1996). Does financial development cause economic growth? Time series evidence from 16 countries. Journal of Development Economics, 51(2), 387–411.

12.

Demirgüç

K.A.

, & Levine

(1996). Stock markets corporate finance and economic growth: An overview. World Bank Economic Review, 10(2), 223–40.

13.

Dritsaki

, & Bargiota

M.D.

(2005). The causal relationship between stock, credit market and economic development: An empirical evidence for Greece. Economic Change and Restructuring, 38(1), 113–27.

14.

Filer

K.R.

, Hanousek

, & Campos

F.N.

(1999). Do stock markets promote economic growth? (Working Paper Number 267). New York: City University of New York.

15.

Goldsmith

R.W.

(1969). Financial structure and development. New Haven, CT: Yale University Press.

16.

Harris

R.D.F.

(1997). Stock markets and development: A re-assessment. European Economic Review, 41(1), 139–46.

17.

Helpman

, & Krugman

(1969) Market structure and foreign trade. Cambridge, MA: MIT Press.

18.

Hicks

(1969). A theory of economic history. Oxford: Clarendon Press.

19.

Hurlin

(2005). Un Test Simple de lHypothèse de Non Causalité dans un Modèle de Panel Hétérogène. Revue Economique, 56(3), 799–809.

20.

K.S.

, Pesaran

M.H.

, & Yongcheol

(2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53–74.

21.

Ireland

P.N.

(1994). Money and growth: An alternative approach. American Economic Review, 84(1), 47–65.

22.

Krueger

A.O.

(1997). Trade policy and economic development: How we learn. American Economic Review, 87(1), 1–22.

23.

Levine

(1993). Stock markets, growth, and tax policy. The Journal of Finance, 46(4), 1445–65.

24.

Levine

, & Zervos

(1998). Stock markets, banks, and economic growth. The American Economic Review, 88(3), 537–58.

25.

Lucas

R.E.

(1988). On the mechanics of economic development. Journal of Monetary Economics, 22(1), 3–42.

26.

Luintel

K.B.

, & Mosahid

(1999). A quantitative re-assessment of the finance-growth nexus: Evidence from a multivariate VAR. Journal of Development Economics, 60(2), 381–405.

27.

Mckinnon

R.I.

(1973). Money and capital economic development. Washington, DC: Brookings Institution.

28.

Patrick

H.T.

(1966). Financial development and economic growth in underdeveloped countries. Economic Development Cultural Change, 14(2), 174–89.

29.

Pedroni

(2004). Panel cointegration: Asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis. Econometric Theory, 20(3), 597–625.

30.

Robinson

(1952). The generalization of the general theory and other essays. New York: St. Martin’s Press.

31.

Rousseau, & Wachtel (2000). Equity markets and growth: Cross-country evidence on timing and outcomes, 1980–1995. Journal of Banking and Finance, 24(12), 1933–57.

32.

Schumpeter

J.A.

(1911). Theorie der WirtschaftlichenEntwicklung. Leipzig: Dunker & Humblot [The theory of economic development (trans. Opie

Redvers

). Cambridge, MA: Harvard University Press, 1934].

33.

Show

(1973). Financial deepening in economic development. New York: Oxford University Press.

34.

Tang

G.Y.N.

, & Kwok

K.H.

(1997). Day-of-the-week effect in international portfolio diversification: January vs non-January. Japan and the World Economy, 9(3), 335–52.

35.

Yuncu

L.E.

(2007). Essays on financial development and economic growth (PhD Dissertation). Ankara: Department of Management, Bilkent University.

36.

Zhu

, Ash

, & Pollin

(2002). Stock market liquidity and economic growth: A critical appraisal of the Levine/Zervos model (Working Paper No. 47). Political Economy Research Institute (PERI), University of Massachusetts-Amherst.

The Causality Direction of the Stock Market–Growth Nexus: Application of GMM Dynamic Panel Data and the Panel Ganger Non-causality Tests

Abstract

Keywords

1. Introduction

1.1 Background of the Study

1.2 Objectives of the Study

2. Literature Review

3. Data and Methodology

3.1 Data

Table 1 List of Sample Countries Developed Markets Emerging Markets JapanUSAUKAustraliaNew ZealandSwitzerlandCanadaFranceGermanyNorway Sri LankaIndiaSaudi ArabiaThailandHong KongChinaBrazilSouth AfricaBangladeshMalaysia

3.2.1 Panel Unit Root Tests

4.1 Results of the Unit Root Tests

4.2 Results of the GMM Dynamic Panel Data Analysis for Emerging Markets

Footnotes

Appendix

References

Table 1
List of Sample Countries

Developed Markets Emerging Markets

Japan
USA
UK
Australia
New Zealand
Switzerland
Canada
France
Germany
Norway Sri Lanka
India
Saudi Arabia
Thailand
Hong Kong
China
Brazil
South Africa
Bangladesh
Malaysia