CC BY
74
ISSN 2311-8768 (Online) ISSN 2073-4484 (Print)
Economic and Statistical Research
УДК 336.7
STOCK MARKET DEVELOPMENT AND ECONOMIC GROWTH: EMPIRICAL EVIDENCE FROM KAZAKHSTAN
Evgeny V. ILYUKHIN
PhD in Biology, Associate Professor, Department of Economics, Management and Informatics, Institute of Aviation Technology and Management (Ulyanovsk State Technical University), Ulyanovsk, Russian Federation evgeny.ilyukhin@gmail.com
The study examines a causal relationship between the stock market and economic growth variables for Kazakhstan. The stock market is found to promote economic growth in the country but the causation between the variables is not always significant. Consistent efforts are required on the part of the Kazakhstan authorities in order to ensure well-organized and competent operation of the stock market because the more efficient market attracts investors. For example, it can be achieved through stimulation of trading activities in the Kazakhstan Stock Exchange.
Keywords: stock market development, economic growth, causality relationship, policy implications
Introduction
The limited capital availability or its ineffective usage is the point of problems of Kazakhstan's economic development. Stock markets can play an important role by providing capital and stimulating its effective use in order to promote economic growth. Some countries are heavily dependent on the stock market development level while regulating economic development. Despite evidence from different countries where the relationship between stock market development and economic growth has been almost unequivocally established, it is not yet discovered for Kazakhstan. Both academicians and practitioners clearly feel a lack of researches and materials in the Kazakh context that would be based on reliable statistics. It provides the reason for conducting this study. As a result, the subject is still important for studying and discussing at the present time. In the context of the study, it aims to assess the contribution of the
Kazakhstan stock market to economic growth. The studies on similar cases of countries are presented below. Caporale et al investigated a sample of four countries (Chile, Malaysia, Philippines and South Korea) over the period from 1971:Q1 to 1998:Q4 using a vector autoregressive (VAR) framework based on an endogenous growth model. They find the significance of causality between the stock market components, investment and economic growth in line with the model [3]. Examining a relationship between stock market performance and economic growth in Iran with causality tests, Oskooe discovers the short-run causality between stock prices and economic growth. It would mean that stock market is the main economic indicator of the Iranian economic growth in the short run [6]. Studying the stock market development and economic growth measures from Nigeria over the period from 1990:Q1 to 2009:Q4 and employing cointegration and vector error correction (VEC) model, Adenuga finds that the indicators used to describe the country's stock market development are significant and positively related to economic growth. In addition, the author notes that the simplified trading promotes investment, facilitates efficient capital allocation and stimulates the long-run economic growth [1]. Regmi provides evidence of a causal relationship between stock market development and economic growth for Nepal over the period from 1994 to 2011. Using a unit root test, cointegration, VEC model and NEPSE composite index as a stock market development indicator, the author discovers that stock market development significantly contributes to economic growth of the country [7].
Methods
In order to study the impact of stock market development on economic growth in Kazakhstan, quarterly data with a sample period from 1997:Q1 to 2012:Q4 are adopted. The matrix proposed by Boot, Feibes and Lisman is used in order to translate annual data into quarterly data [2]. The point is to assure that enough data for analysis are obtained. The measures are calculated on the basis of available data of the World Bank Indicator. The estimates are conducted using econometric computer software package GRETL. In this study, a quantitative analysis of the short-run impact is based on VAR and considered as 'Granger causality'. This kind of causality does not necessarily mean causality relationship but the fact that a change in one factor precedes a change in another one. The tests are carried out on the basis of equations 1 and 2:
Yt =a0 +]T aY +£ P, Xt - + 8,
1=1 1=1
X t = Yo +1 yY-1 +15, Xt-1 , i =1 i =1
where X is a stock market development indicator, Y is economic growth and the subscripts t and t-i denote the current and lagged values. The lag length is determined based on Akaike (AIC), Schwarz (SIC), Hannan-Quinn (HQC) criteria.
Based on the previous studies, a multivariate model is adopted with some changes in order to take into consideration the specifics of Kazakhstan's economy and start testing the long-run relationships between the model variables [1, 4]. The stock market variables [(tv), (lc) and (mc)] in equation 3 are entered into the model. EG = P0 + P1n + P2ir + P3sr + P4(tv) + +P5 (lc) + P6 (mc) + P7cf + Pscpr + 8t.
The model includes the following variables: Economic Growth (£'G)/GDP growth (annual %): it is measured by the rate of change in real GDP. According to the demand-driven hypothesis, economic expansion will create new demand for financial services. An increase will stimulate setting up the larger and more sophisticated financial institutions to satisfy new demand for their services.
Macroeconomic Stability (n, CPZ)/Inflation, consumer prices (% per annum): it is measured as the consumer price level, CPI (inflation). It is known that a low inflation can stimulate investors to pay more attention to economy.
Investment Ratio (/r)/Gross fixed capital formation (% of GDP): it is calculated as gross fixed capital formation divided by nominal GDP. According to the endogenous economic theory, investments are positively related to economic growth.
Savings Ratio (sr)/Gross savings (% of GDP): it is calculated as gross domestic savings as a percentage of GDP. Gross savings are calculated as gross national income less total consumption, plus net transfers. Usually, larger savings lead to higher availability of capital that could flow through stock market. Turnover velocity (tv): it is the ratio of turnover to market capitalization. It is necessary to find the market development indicators that are independent of stock prices. Given that the market role is to reallocate capital among its most productive users, such an indicator would be appropriate.
Change in the number of listed companies (%) (lc): it is calculated as annual percentage of increase/decrease in the number of listed domestic companies. It would be an indication of financial deepening for country. Market capitalization of listed companies (% of GDP) (mc): it is the stock price times the number of outstanding shares of the companies listed in the country's stock exchanges except for investment companies, mutual funds or other collective investment. Capital flows/Foreign direct investment, net inflows (% of GDP) (cf): it is measured using foreign direct investment as a percentage of GDP. It is associated with institutional and regulatory reforms, adequate disclosure, listing requirements and fair trading procedures. Banking sector development/Domestic credit provided by the financial sector (% of GDP) (cpr): it is the value of domestic credit provided by the banking system to the private sector in relation to GDP. It measures the banking system activity in one of its main function: directing savings.
The expectations for the variables are presented as:
Pi <0;P2;P3,...,P8 >0.
Augmented Dickey-Fuller (ADF) specification for unit root involves the estimation of one of the following equations (4, 5 and 6) respectively [8]:
AXt = PXt-1 +£ 5 ,AXt-, +8t,
j=i
p
AXt =ao +pXt-1 +Z5, AXt-1 +8t,
j=1
p
AXt = ao + a/ + pXt-1 + X5, AXt-1 + 8t.
,=1
In order to ensure that the errors are uncorrected, the additional lagged terms are included. The allowed maximum lag length is four and proceeds to the appropriate lag by checking the mentioned information criteria. The null hypothesis is that the variable Xt is a non-stationary series (H0: p = 0) and is rejected when p is significantly negative (Ha: p<0). If the calculated ADF statistics is higher than critical values, then the null hypothesis (H0) is rejected and the series is stationary or integrated of order zero I(0). Alternatively, non-rejection of the null hypothesis implies non-stationarity leading to the test for the difference of the series until stationarity is reached and the null hypothesis is rejected. The cointegration tests are conducted using the maximum likelihood framework [5]. The point is to find whether the long-run relationship exists between the variables. The appropriate optimal lag-length is determined in order to get standard normal error terms. Whereas quarterly data are used, six lags are initially allowed. In the test, SIC is preferred due to its more strict theoretical support.
If the variables of economic growth equation are cointegrated, it is necessary to estimate the short-run dynamics within VEC model for capturing the adjustment speed to equilibrium in the case of any shock to any of the independent variables. The generalized specification framework of VEC model is expressed and extended for the three models in equation:
k-1
AEG = ß0 + £ß, Aeg,-, + £a, An,-, +
¡=1 i=0 k-1 k-1 k-1
A,r-, +& Asrt-, +& A(tv)<-,
+
i=0
k-1
A(lc)t-i +& A(mc),-, + i=0 i =0
Table 1
Sinmmary Statistics of Variables
Zn Acf,- + ZKi AcPrt- + Qecmt- + s< > i=0 i=0
A is the first difference of the series. Pff Pf a, 5., y, \ k. and Q is the model estimated parameters.
Results
Table 1 shows the summary statistics for the variables. The means range from 0.021505 for cfto 6.7500 for cpr. It indicates that the variables exhibit significant variation in runs of magnitude, suggesting that estimation at the levels may introduce some bias in the results. The Jarque — Ber statistics is significant except for ir EG, sr and cf. The refore, the null hypothesis that the series are normally distributed can be rejected. The numbers in Table 2 demonstrate that there is an inverse relation between lc and the rest of variables. They also indicate that mc and tv variables are positively related to other variables except for lc. Causality tests in context of the standard VAR procedure with the null hypothesis of that stock market does not cause EG, find some causations in the proposed models. The estimates are presented in Table 3. Model C with mc included indicates notable results: there is rather strong positive causality between mc, cpi, ir, sr and cpr especially. Another causality for cpr is related to Model A with tv included. It is worth noting that mc and tv can influence economic growth through other variables. The results of the ADF test presented in Table 4 show that ir, sr, lc and cpr are stationary at the levels while the rest of variables are non-stationary at the levels (I(0) series). However, the second group of variables becomes stationary after the first difference. It indicates they are I(1) series.
Based on the figures from Table 5 representing unrestricted rank of cointegration test results, it can
,=0
,=0
Indicator CF CPI CPR EG IR LC MC SR TV
Mean 0.021505 2.2737 6.7500 1.6953 5.7812 0.11344 0.050573 6.0938 2.0199
Median 0.020982 1.9271 5.0902 1.7945 5.7494 0.039810 0.028387 6.0023 1.4072
Minimum 0.005634 0.64638 0.82161 -0.51431 2.0224 -0.22623 0.0068799 1.7696 0.0000
Maximum 0.047523 6.5636 20.587 4.0344 11.311 1.3534 0.20336 12.066 6.5211
Std. Dev. 0.009806 1.2230 4.8676 1.1894 2.2003 0.31163 0.045594 2.4416 1.7875
Skewness 0.057367 1.7105 0.96733 0.038724 0.39501 3.1166 1.4449 0.49117 1.0197
Kurtosis -0.28831 3.2417 0.24416 -0.85142 -0.28031 9.0483 1.6371 -0.26265 0.078599
Jarque-Ber 3.73206 59.2323 10.1402 1.94909 1.87388 321.929 29.4151 2.75724 11.1069
5% percentile 0.0086477 0.84123 1.3926 -0.43569 2.2752 -0.11163 0.0094415 2.6168 0.0000
95% percentile 0.038889 5.0406 15.839 3.6856 10.463 1.1466 0.15389 10.934 5.8303
Interquartile range 0.014465 1.0684 7.7413 1.9368 2.9542 0.11568 0.064539 3.4921 2.2102
Observations 64
Table 2
Correlation Matrix
Indicator CF CPI CPR EG IR LC MC SR TV
CF 1.0000 0.4901 0.5219 0.3734 0.7499 -0.2151 0.4067 0.6484 0.4682
CPI 0.4901 1.0000 0.2812 0.0902 0.4342 -0.0047 0.1734 0.4553 0.1240
CPR 0.5219 0.2812 1.0000 0.0996 0.6835 -0.3123 0.7651 0.7628 0.3049
EG 0.3734 0.0902 0.0996 1.0000 0.5751 -0.3918 0.1245 0.5689 0.5281
IR 0.7499 0.4342 0.6835 0.5751 1.0000 -0.1884 0.6404 0.9452 0.6013
LC -0.2151 -0.0047 -0.3123 -0.3918 -0.1884 1.0000 -0.2404 -0.2384 -0.1950
MC 0.4067 0.1734 0.7651 0.1245 0.6404 -0.2404 1.0000 0.6050 0.2326
SR 0.6484 0.4553 0.7628 0.5689 0.9452 -0.2384 0.6050 1.0000 0.5397
TV 0.4682 0.1240 0.3049 0.5281 0.6013 -0.1950 0.2326 0.5397 1.0000
Table 3
Causality Test in the Context of VAR
Direction F-Value | Causality Lag Obs
Model A
TV Causes EG 0.00053231 N 1 63
TV Causes CPI 1.4616 N 1 63
TV Causes IR 2.0226 N 2 62
TV Causes SR 2.1482 N 2 62
TV Causes CF 1.2574 N 2 62
TV Causes CPR 3.5652* Y 2 62
Model B
LC Causes EG 1.4686 N 2 62
LC Causes CPI 0.57362 N 2 62
LC Causes IR 0.92161 N 2 62
LC Causes SR 0.65551 N 2 62
LC Causes CF 0.96829 N 2 62
LC Causes CPR 0.0098679 N 2 62
Model C
MC Causes EG 0.00026627 N 1 63
MC Causes CPI 4.3359* Y 1 63
MC Causes IR 3.3160* Y 2 62
MC Causes SR 3.8832* Y 2 62
MC Causes CF 0.042388 N 1 63
MC Causes CPR 5.0434** Y 2 62
Note: **, * represents 1% and 5% level of significance.
Table 4
ADF Unit Root Test
Variable Level 1st Difference Remarks
EG -2.8406 -3.89341** I(1)
CPI -3.32439 -4.24127** I(1)
IR -4.47917** I(0)
SR -4.21533** I(0)
TV -2.91025 -4.56586** I(1)
LC -3.80284** I(0)
MC -2.97444 -4.2563** I(1)
CF -3.39101 -4.2563** I(1)
CPR -3.18065* I(0)
Note: **, * represents 1% and 5% level of significance.
be concluded that there is the long-run relationship between the variables. It is indicated with at least one existing cointegrating vector in the models. The long-
run in economic growth equitation can be interpreted by normalizing the estimates of the unconstrained cointegrating vector to economic growth. The parameters of the cointegrating vectors for the long-run economic growth are presented in the ecm equations for each model.
Table 6 presents the ECMs for the models with tv, lc and mc included as measures of stock market development demonstrating that the previous quarter's disequilibrium gets adjusted up to the long-run equilibrium from 5.81 percent (Model A) to 367.09 percent (Model C) within a quarter. In other words, the coefficient of the error correction which measures the speed of adjustment back to equilibrium whenever the system is out of equilibrium indicates that an adjustment can be both relatively slow and notably fast. The R2 measures indicate that the variation in economic growth is explained by the final variables that entered into the model and ranges between 64-71 percent. The F-test statistics shows that the overall model fit is significant at around 1.0 percent. The error correction run in the economic growth equation is statistically significant with a correct negative sign.
Conclusions and Policy Implications
The stock market development and economic growth of Kazakhstan over the period from 1997:Q1 to 2012:Q4 have been examined. The three stock market variables in the short- and long-run are not always significantly related to the economic growth variables. Whereas the stock market is a tool for economic growth for purposes of the study, I would recommend integrating the stock market into the entire economic system of the country while designing economic policies. The key policy implication is that the country needs a well-organized stock market so to accelerate and maintain economic growth. Therefore, Kazakhstan's authorities should make consistent efforts to create such a stock market because a more efficient market attracts investors. It
Экономико-статистические исследования Economic and Statistical Research -49-
Table 5
Cointegration Tests
Rank (H0) Trace Statistics p-value Rank (H0) Max Eigen Stat p-value
Model A (1 lag)
r = 0 94.071 0.5980 r = 0* 47.546 0.0103
r < 1 46.525 0.9812 r < 1 21.122 0.8277
r < 2 25.403 0.9965 r < 2 8.6205 0.9989
r < 3 16.783 0.9696 r < 3 6.6446 0.9928
r < 4 10.138 0.8402 r < 4 5.6619 0.9090
r < 5 4.4763 0.6368 r < 5 2.8422 0.2355
ECM1 = EG - 1.3102tv - 15.339cpi + 7.4008ir — 7.4735sr + 738.51cf+ 1.1648cpr
Model B (1 lag)
r = 0 96.333 0.3178 r = 0* 49.920 0.0045
r < 1 46.413 0.9818 r < 1 15.078 0.9903
r < 2 31.336 0.9595 r < 2 12.746 0.9638
r < 3 18.589 0.9345 r < 3 7.9067 0.9763
r < 4 10.683 0.8043 r < 4 2.8263 0.8067
r < 5 4.3639 0.6525 r < 5 1.5376 0.2513
ECM2 = EG — 2.3903lc — 3.5009cpi + 1.3705ir — 1.1219sr + 162.06cf + 0.057920cpr
Model C (1 lag)
r = 0* 120.45 0.0117 r = 0* 57.228 0.0003
r < 1 63.218 0.5909 r < 1 28.904 0.3113
r < 2 34.314 0.9054 r < 2 14.746 0.8998
r < 3 19.568 0.9072 r < 3 8.1368 0.9716
r < 4 11.431 0.7501 r < 4 6.9071 0.8134
r < 5 4.5238 0.6302 r < 5 2.8335 0.8057
ECM3 = EG — 128.25mc — 7.5849cpi + 9.7202ir — 7.4483sr — 1.9577cf+1.5539cpr
Note: r — number of cointegrating vectors. Tests find either 1 cointegating equation in trace test at the 0.05 level or 1 cointegrating equation in max-eigen test. 'denotes rejection of the hypothesis at the 0.05 level.
Table 6
Estimates of Error Correction Models
Variable | Coefficient | Std. Error | t-Statistics | p-value
Model A
const 0.615985 0.479098 1.286 0.20871
d(EG(-3)) -0.501082 0.219728 -2.2805 0.03011**
d(CPI(-3)) 0.542735 0.266746 2.0347 0.05112*
ECM1 -0.058143 0.0551256 -1.0547 0.30025
RI = 0.681642; F-test = 12.8891
Model B
const 0.0160957 0.160596 0.1002 0.92086
d(EG(-2)) -0.504508 0.24254 -2.0801 0.04646**
d(EG(-3)) -0.644194 0.236085 -2.7287 0.01069**
d(CPI(-1)) -0.491831 0.276411 -1.7793 0.08567*
d(CPR(-3)) -0.224105 0.128328 -1.7464 0.09134*
ECM1 -0.221882 0.489755 -0.453 0.65389
RI = 0.714166; F-test = 13.94212
Model C
const 5.52631 6.60074 0.8372 0.40909
d(EG(-3)) -0.467279 0.220545 -2.1187 0.0425**
d(CPI(-3)) 0.591244 0.334585 1.7671 0.08739*
d(IR(-4)) -0.656624 0.385112 -1.705 0.09853*
d(CPR(-3)) -0.300717 0.167125 -1.7994 0.08203*
ECM1 -3.67092 4.90158 -0.7489 0.45973
RI = 0.641432; F-test = 12.79586
Note: **, * represents 5% and 10% level of significance.
■23 (257) - 2015
ФИНАНСОВАЯ АНАЛИТИКА: FINANCIAL ANALYTICS:
проблемы и решения science and experience
can be achieved through stimulating trading activities in the local stock exchange.
References
1. Adenuga A.O. Stock Market Development Indicators and Economic Growth in Nigeria (1990 - 2009): Empirical Investigations. Central Bank of Nigeria: Economic and Financial Review, 2010, no. 48-1, pp.33-70.
2. Boot J.C.G., Feibes W., Lisman J.H. Further Methods of Derivation of Quarterly Figures from Annual Data. Journal of the Royal Statistical Society. Series C (Applied Statistics), 1967, no. 16-1, pp. 75-76.
3. Caporale G., Howells P., Soliman A. Endogenous Growth Models and Stock Market Development: Evidence from Four Countries. Review of Development Economics, 2005, no. 9-2, pp. 166-176.
4. Dritsaki C., Dritsaki-Bargiota M. The Causal Relationship between Stock, Credit Market and Economic Development: An Empirical Evidence for Greece.
Economic Change and Restructuring, 2005, vol. 38, iss. 1, pp. 113-127.
5. Johansen S., Juselius K. Maximum Likelihood Estimation and Inference on Cointegration with Applications to the Demand for Money. Oxford Bulletin of Economics and Statistics, 1990, vol. 52, iss. 2, pp.169-210.
6. Oskooe S.A.P. Emerging Stock Market Performance and Economic Growth. American Journal of Applied Sciences, 2010, no. 7-2, pp. 265-269.
7. Regmi U.R. Stock Market Development and Economic Growth: Empirical Evidence from Nepal. Administration and Management Review, 2012, no. 24-1, pp. 1-28.
8. Seddighi H., Lawler K., Katos A. Econometrics: A Practical Approach. London, Routledge, 2000.
