Financial Development and Economic Growth

15 downloads 0 Views 1MB Size Report
financial development (FD) and the economic growth (EG). Previous ..... two financial development measurements (broad money & credit) are tested separately,.
Financial Development and Economic Growth -- Toda-Yamamoto Approach based Causality Test on 16 HighPerformance OECD Countries

- Empirical Research Article Assignment of the course “Advanced Econometrical Analysis”

Author: Pengcheng Luo

JÖNKÖPING OCT. 2016

Table of Contents Introduction ............................................................................................................ 1 Econometrical Model .......................................................................................... 4 1.

Variables and Data .............................................................................................. 4

2.

Methodology ....................................................................................................... 8 Empirical Findings .............................................................................................. 9 Conclusion ........................................................................................................ 14 Reference .......................................................................................................... 15 Appendix ........................................................................................................... 17

Table 1 Country List ...................................................................................................... 4 Table 2 VAR lag order ................................................................................................... 9 Table 3 unit root test statistics ..................................................................................... 10 Table 4 unit root test results ......................................................................................... 10 Table 5 Toda-Yamamoto Causality test ...................................................................... 11 Table 6 Main findings .................................................................................................. 11 Table 7 normal Granger causality test results .............................................................. 13 Table 8 Data source ..................................................................................................... 17 Table 9 Toda-Yamamoto Approach result .................................................................. 17 Table 10 Panel cointegration test ................................................................................. 20 Table 11 ols fixed effect model ................................................................................... 21 Table 12 standard granger causality ............................................................................ 23 Figure 1 FD and EG ....................................................................................................... 2 Figure 2 GDP and GDP Growth Rate ............................................................................ 5 Figure 3 Broad Money, as % of GDP ............................................................................ 6 Figure 4 Credit to private sector, as % of GDP ............................................................. 7

Abstract This paper applied the Toda-Yamamoto causality test approach to find the causality link between financial development and economic growth (FD and EG). The paper invested the most advanced 16 OECD economies (GDP per capita above OECD average), from 1980 to 2015, and found out there are bi-directional causality relationship between FD (measured as private sector credit and broad money) and the GDP growth. It is also found that capital accumulation and trade granger-cause the economic growth. Lastly, all the variables (capital, trade, government spending, FD & EG) has a one-directional causality to the capital accumulation. These results are cross-checked with the traditional VAR Granger Causality and Fixed Effects Panel OLS.

Introduction This paper aims to provide further empirical evidence on the causality links between financial development (FD) and the economic growth (EG). Previous studies have used different econometric approach to investigate the developing counties, or the developed economies covering more distant periods. This paper’s objective narrows to the 16 most advanced economies which have an above-OECD-average GDP per capita, and focusing the periods after financial liberalization till most recent time. The existence and direction of causality between FD and EG have been disputed throughout the history. As early as Schumpeter (1911), and later Shaw (1973) had admittedly recognized the interactions between financial development and the economic growth and financing plays a critical role in a country’s economy. Levine (1997) reasoned that financial sector not only promotes the exchange of capital, but also various transactions, resources and information. Thus, these interactions have a significant contribution the long-run growth. The important interactions between FD and EG, and the resulting contribution to the economy had been recognized by a majority of the researchers (World Bank, 1989). Nevertheless, the question of whether there is any causal relationship between a country’s financial development and its economic growth and the direction of such potential causality have long puzzled scholars and policymakers. More importantly, different causality links may have implications of quite the opposite policy direction. The World Bank itself, for example, has changed its financial policy recommendations to the world, from the 1989’s ‘cornerstone’ of economic growth, to 1993, in favor of a financial sector with higher regulations and government intervention. Economists, too, has argued for both sides. McKinnon (1973), Neusser & Kugler (1998), Levin et al. (2000), for instance, has been advocator for the financial sector’s role in promoting growth, he argues that the financial sector can provide the economy with enough capital, and more importantly, the financial market allocates capital efficiently, fulfills the demand and capital goes where it is needed most. On the other hand, economists leading by Lucas (1988) and Goldsmith (1969), found the causal relationship is from economic growth to financial development, that is, as the thriving economy surges capital demand and related financial service, the sector grows. Some (Jung 1986) even argues that the financial sector’s role has been overstated, the under-development of the financial sector in some countries is just a result of lack of demand, and its development is led by the growth of the real economy. Patrick (1966) summarized these two opposing camps as the supply-leading hypothesis and demand-following hypothesis, respectively. Apart from these two, he proposed a middle path- a two-way causality, depend upon which stage the economy is at, i.e. the stage of development hypothesis, which contended that in the early economic development stage, the financial sector provides the economy with abundant and efficient capital allocation, after which, it transforms itself from supply-lending into 1

demand-following as the economy matures. The figure below summarized the various causality channel between the two. Figure 1 FD and EG

Source: Shan & Morris (2010)

Empirically, there are quite abundant researches supporting the supply-lead view: that the causality is from FD to EG. On this side, researchers such as King and Levine (1993), Levine (1997) and Levine & Zervos (1998), used different econometric methods and covering different cross-sectional data, found evidence that are in line with this hypothesis. Demetriades & Hussein (1996), on the other hand, criticized the above-mentioned line of studies, they argued that cross-sectional studies may have weak statistical inference, as these countries tend to vary a lot in their development stages. These two researchers are in favor of time-series approach on this topic, and argued that contemporaneous correlation test is not enough. In their 16 country studies, they found the results objecting the one directional from FD to EG, supporting bi-directional causality and some evidence on EG to FD. Other studies by this time, for example, Blackburn and Huang (1998), or Khan (2001), both have taken a micro economics approach on the financial intermediaries, had also recognized two-way causality between FD and EG. More recently, Shan & Morris (2010) compared OECD countries, to an Asian developed economy, South Korea with a developing world sample- China. They used quarterly data up to 1990s, but found no evidence suggesting a causality link from FD to EG, directly, or indirectly through investment and productivity. Their results suggest a demand-follow pattern of the FD to EG.

2

Calderon & Liu (2003)’s study may be more ambitious, by including a large dataset of 109 countries from both developed and developing world. used Geweke decomposition approach to get around the econometric problems with traditional Granger causality test. This approach was ground-breaking at that time in the panel analysis with causality tests. The authors tested 109 countries from 1960’ – 94’, developing and developed. Evidence suggests that causality from both directions coexist; but FD contributes more to the causal relationship in developing countries group; FD also takes time to impact on the real economy. Lastly, the effect from FD to EG are through two sources: capital accumulation and technological change. The causality from FD to the two sources are stronger in developing countries, while in developed world, the reverse (capital & technology to FD) is true. Shan et al. (2001) on the other hand, contended that the traditionally used simple model could produce misleading results, they argued for the use of VAR model for its multi-variable feature and eliminates the bias with single-equation models. Their study investigated 10 countries, found half of them with bi-directional causality, 3 of them has the causality from EG to FD. The debate, both theoretical and empirical, has been conflicting over time, especially after the 2008 Financial Crisis. This paper aims to provide further empirical evidence on the causality links between financial development and the economic growth, making use of the newer Toda-Yamamoto approach (1995), which is built upon VAR model, while providing a way around its limitations. The following sections of this paper are organized as: first a summary of the variables and the Toda-Yamamoto approach will be used in this study. Then the empirical findings are detailed in section III, with cross-validation using traditional VAR model and panel least squares. Lastly, the paper will provide a brief conclusions and policy implication.

3

Econometrical Model

1. Variables and Data As mentioned, this paper focuses on the 16 most developed countries from the Organization for Economic Co-operation and Development (OECD) with annual data from 1980 to the latest data available, 2015. As Shan & Morris (2002) noted, since the 1980s, the industrialized economies have undertaken financial deregulations and reforms. This paper covers this up-to-date period in an effort to make sense of how the developed world’s financial sector and economic growth are related. The OECD, which founded to facilitate trade and economic growth, now has 35 member countries. According to the World Bank data, these countries have an average GDP (PPP) per capita of 39,741 US Dollars. Due to data quality and the objective of this paper, only the countries with GDP per capital number higher than the average line are selected, see next table for the list of these 16 countries (in descending order of GDP per capita). These countries share high homogeneity in their development in various aspect, as Demetriades & Hussein (1996) criticized many of the previous cross-sectional studies, that the results are with less value, with countries vary a lot in development status.

Table 1 Country List

Country Luxembourg Norway Switzerland US Ireland Netherlands Austria Germany Denmark Iceland Sweden Australia Canada Belgium UK Finland

As this paper investigates the causality links between economic growth and the financial development (EG & FD), indicators for these two are the main variable under investigation.

4

More precisely, economic growth is defined as the GDP per capita growth rate (labelled as ‘growth’) from the World Development Indicators database (World Bank 2016). As the figure of GDP and its growth rates shown below, the 16 countries’ economic development level and their growth pattern are quite parallel, hence it eliminated potential problems one may face when doing cross-country study with a very high cross-sectional difference.

Figure 2 GDP and GDP Growth Rate

120,000

Luxembourg Norway Switzerland Sweden United States United Kingdom Iceland Ireland Netherlands Austria Australia Germany Denmark Canada Belgium Finland

100,000 80,000 60,000 40,000 20,000 0 1980

1985

1990

1995

2000

2005

2010

2015

12

Luxembourg Norway Switzerland Sweden United States United Kingdom Iceland Ireland Netherlands Austria Australia Germany Denmark Canada Belgium Finland

8 4 0 -4 -8 -12 1980

1985

1990

1995

2000

2005

2010

2015

For the financial development level, it depends on various progress of the financial intermediaries and institutions, which is difficult to be captured into one single variable. However, there are two widely used and arguably good measurement of the financial sector’s development, the ratio of broad money over GDP, and the share of credit to private sector on GDP. This paper will use both (‘broad money’ and ‘credit’), and they are also obtained from the World Development Indicators database. A detailed data source of the variables used is included in the appendix table 8.

5

King and Levine (1993) for instance used the previous one, M2/GDP. Demetriades and Hussein (1996) noted that for developing countries, this measure may not be the best, because currency consists the larger part of the money stock. This however, is not the case of this paper, which focuses on the most developed countries. This paper used World Bank-defined broad money, as M2 definition varies across countries, as it captured the abundance of capital (including credits, loans which are correlated with FD). The other most common variable for FD is the credit to the private sector, as a share to the GDP figure. This variable directly measure the capital (such as loans) from the financial sector. It reflects not only how free or repressed the financial sector is, but also high or low reserve requirement, the development level of the private sector (AbuBader & Abu-Qarn 2008, Levine & Zervos 1993). The following figures plot these two variables against time. Apart from a peak of credit issued in Iceland 2006 and Canada’s increasing money stock in 2001, no systematical structural break was witnessed. We have again witnessed similar pattern- that is both indicators suggest a steady but slow financial development, even though these economies are at very mature stage.

Figure 3 Broad Money, as % of GDP

200

Luxembourg Norway Switzerland Sweden United States United Kingdom Iceland Ireland Netherlands Austria Australia Germany Denmark Canada Belgium Finland

160

120

80

40

0 1980

1985

1990

1995

2000

2005

6

2010

2015

Figure 4 Credit to private sector, as % of GDP

350

Luxembourg Norway Switzerland Sweden United States United Kingdom Iceland Ireland Netherlands Austria Australia Germany Denmark Canada Belgium Finland

300 250 200 150 100 50 0 1980

1985

1990

1995

2000

2005

2010

2015

Conventionally, in addition to the main variable under investigation, some variables are included into the regressions, too, used as control variables, as they are proved to have correlations with FD and EG (Levine & Renelt 1992). They are: investment capital formation (‘capital’), value of the international trade (‘trade’) and the government spending (‘expense’). Again, they are obtained from the same mentioned database.

7

2. Methodology The results from traditional correlation tests and differences-indifferences approach are largely invalid because of endogeneity problems. To solve this, the commonly used approach is the Granger causality test and its variants, which used mostly in long timeseries data, however, with potential non-stationarity and cointegration in a panel dataset, the traditional causality test power is still limited due to the Wald statistic’s distribution. Toda-Yamamoto (1995) proved that by their approach, the Wald statistic follows X2 distribution regardless the process’s order of integration (stationary or I(1), I(2), or whether it is cointegrated). This paper used this Toda-Yamamoto VAR approach (1995) to examine the causal direction between financial development and economic growth. This approach eliminates the limits of previous methods, makes it possible to estimate VAR in levels, the results are more intuitive to interpret, compared to variables taken first differences. The Toda-Yamamoto approach is to augment the k order of the VAR with the including of extra d lags. Here d is the maximal order of integration. We regress the new augmented VAR model, and using Wald tests for standard test of noncausality. For instance, we have a simplified k-order VAR (variables F and Y, for financial development and economic growth), then we include the additional #d terms of both Y and F variables in both function, now we have a p = k+ d max order VAR, Then a Wald test can be applied to see if the coefficients of variables of the original (k-order variables) jointly = 0 (null hypothesis: β1i=0, ϕ1i=0, respectively no causality from Y to F, no causality from F to Y), rejection leads to causality conclusion.

Based upon this model, the paper included the three control variables, moreover, the two financial development measurements (broad money & credit) are tested separately, named as ‘credit model’ & ‘broad money model’.

8

Empirical Findings Following the steps of Toda & Yamamoto (1995) we firstly determine the VAR order. In both model, shaded areas in the following output table, suggests that the optimal lag order is at lag 2 or 3. Moreover, tests on the residual (LM serial correlation test) are carried out, and the statistics not reject the null of no serial correlation on the lag order of 2. Thus, for both models, we conclude the VAR lag length of 2 (k=2) is selected.

Table 2 VAR lag order Credit Model: LM stat 2 27.57029 p-value 0.3280 VAR Lag Order Selection Criteria Endogenous variables: GROWTH EXPENSE CAPITAL TRADE CREDIT Sample: 1980 2015 Included observations: 247 Lag

LogL

LR

FPE

AIC

SC

HQ

0 1 2 3 4 5 6 7 8 9 10

-4661.511 -3171.017 -3098.932 -3073.719 -3057.626 -3040.601 -3016.357 -2989.617 -2975.614 -2952.658 -2919.510

NA 2908.577 137.7481 47.15982 29.44944 30.46631 42.40170 45.68661 23.35698 37.36030 52.60768*

1.77e+10 124221.9 84865.45 84771.11* 91215.59 97480.73 98348.63 97341.43 106945.5 109438.6 103287.1

37.78552 25.91916 25.53791 25.53619* 25.60831 25.67288 25.67901 25.66491 25.75396 25.77051 25.70454

37.85656 26.34540 26.31936* 26.67283 27.10015 27.51993 27.88125 28.22236 28.66661 29.03837 29.32759

37.81412 26.09077 25.85253* 25.99381 26.20894 26.41652 26.56565 26.69456 26.92661 27.08618 27.16321

* indicates lag order selected by the criterion LR: sequential modified LR test statistic (each test at 5% level) FPE: Final prediction error AIC: Akaike information criterion SC: Schwarz information criterion HQ: Hannan-Quinn information criterion Broad Money Model: : LM stat 27.46544 p-value 0.3330 VAR Lag Order Selection Criteria Endogenous variables: GROWTH EXPENSE CAPITAL TRADE BROADMONEY Exogenous variables: C Included observations: 173 Lag

LogL

LR

FPE

AIC

SC

HQ

0 1 2 3 4 5 6 7 8 9 10

-3057.917 -2079.539 -2022.573 -2010.242 -1996.512 -1978.329 -1968.345 -1939.045 -1925.891 -1902.956 -1875.021

NA 1888.892 106.6871 22.38170 24.12734 30.89957 16.39012 46.40605 20.07317 33.67360 39.39904*

1.64e+09 26862.52 18576.67* 21545.29 24626.10 26791.34 32128.99 30917.75 35994.45 37591.24 37245.45

35.40944 24.38773 24.01819* 24.16465 24.29493 24.37375 24.54734 24.49763 24.63458 24.65845 24.62452

35.50058 24.93455* 25.02068 25.62282 26.20878 26.74327 27.37255 27.77851 28.37114 28.85069 29.27244

35.44642 24.60957 24.42489* 24.75622 25.07137 25.33505 25.69351 25.82866 26.15048 26.35921 26.51015

* indicates lag order selected by the criterion

9

Secondly, in order to obtain the maximal order of integration (d max), we need to apply the unit root tests. For panel unit root tests, the results are included in table 3, and summarize in table 4. The null of these test statistics is unit root, hence the rejection (starred item) indicates stationarity. Growth and Capital variables are tested to be stationary, while the other three have unit root in level, stationary at first difference, so that they are I(1) process. These results are summarized in table 4, in which, it is evident that the maximal order of integration (d max) is 1. So our augmented VAR models have the order of k+d = 3, which means that each of the variable will be regressed on the lagged terms of its own and lagged terms of all other variables (with lag length =3).

Table 3 unit root test statistics Series

Level LLC

Growth

Breitung

IPS

First Difference ADF

PP

LLC

Breitung

IPS

ADF

PP

-11.682*

-9.588*

-10.1565*

152.383*

174.976*

--

--

--

--

--

2.171

0.768

2.345

20.729

13.453

-11.163*

-7.105*

-9.673*

157.288*

160.626*

-0.131

0.530

0.149

17.172

9.631

-12.625*

-10.525*

-11.230*

132.269*

151.406*

Capital

-3.053*

-4.641*

-3.941*

64.930*

42.988***

--

--

--

--

--

Trade

-2.571*

-1.765**

-1.453***

43.261** *

33.701

-18.729*

-15.677*

-16.688*

260.101*

664.359*

-1.021

-3.212*

-0.221

32.612

24.702

-15.241*

-13.126*

-14.635*

226.231*

316.138*

Credit Broadmoney

Expense

Nb. significance level: * 0.01, **0.05, ***0.10

Table 4 unit root test results Series

Unit Root Result

Growth

I(0)

Credit

I(1)

Broadmoney

I(1)

Capital

I(0)

Trade

I(1)

Expense

I(1)

10

Lastly, we can test the zero restriction on the original k parameters with Wald tests, rejection indicating causality from independent variable to dependent. The test statistics and associated p-values are in the following table 5, panel (i) are results using ‘broad money’ as financial development measure, panel (ii) used credit as FD measurement. The shaded coefficients are statistically significant from 0, suggesting a causality from that variable to the dependent variable. (appendix table 9 for detailed output)

Table 5 Toda-Yamamoto Causality test (i)

Broadmoney as financial development measure

(ii)

Dependent variable: GROWTH

CREDIT as financial development measure

Dependent variable: GROWTH

Chi-sq

Prob.

Chi-sq

BROADMONEY

4.659522

0.0973***

CAPITAL

1.881673

EXPENSE

0.369891

TRADE

2.586226

Prob.

CREDIT

5.751273

0.0564***

0.3903

CAPITAL

11.80067

0.0027*

0.8311

EXPENSE

0.049632

0.9755

0.2744

TRADE

5.968784

0.0506***

Dependent variable: BROADMONEY

Dependent variable: CREDIT

Chi-sq

Prob.

Chi-sq

Prob.

GROWTH

6.880875

0.0321**

GROWTH

11.21957

0.0037*

CAPITAL

0.604424

0.7392

CAPITAL

0.089204

0.9564

EXPENSE

3.242917

0.1976

EXPENSE

0.658223

0.7196

TRADE

1.480964

0.4769

TRADE

1.81108

0.4043

Nb. starred items indicating rejection of null of non-causality (significance level: * 0.01, **0.05, ***0.10)

Table 6 Main findings Broad money model EG & FD Other Variables

Bi-directional (EGFD and FD EG) All variablescapital

Credit model Bi-directional EG & FD (EGFD and FD EG) Capital EG; Other Variables Trade EG All variablescapital

In both model with different FD measurement, the test results suggesting a bidirectional causality link of financial development (FD) and the economic growth (EG), regardless if the FD is measured as share of broad money or the share of private sector credit. We also found that capital available for investment and country’s trade volume are significant related to the economic growth. Lastly, all the variables, has a one-directional causality to the capital accumulation. One can also use the cointegration test to cross-check the robustness of the causality test. For the ADF and Pedroni statistics, the null is no cointegration. We used Kao and 11

Pedroni test statistics (tested under no trend and with trend for the latter). The results are all rejecting the null at 0.01 significance (output see appendix table 10), which suggesting there are cointegration in our panel. Cointegration among series indicating there must be causality existed between them, so our results are to some degree, vilified by cointegration test.

Furthermore, for comparison of different econometric methods, we also applied simple panel OLS, and the original VAR model-based Granger Causality tests (the results are detailed in appendix table 11, summarized here). Firstly, for the panel least squares, Fixed Effect model is applied, because of the homogeneous feature of the 16 countries investigated, thus it’s assumed that the crosssectional effect is fixed, and we're more interested in the impact of variables that vary over time. We estimated two regressions under each FD measurement model (growth and FD measurements are regressed upon lagged terms and control variable): GROWTH C GROWTH(-1) GROWTH(-2) CREDIT CREDIT(-1) CREDIT(-2) CAPITAL TRADE EXPENSE CREDIT C GROWTH GROWTH(-1) GROWTH(-2) CREDIT(-1) CREDIT(-2) CAPITAL TRADE EXPENSE

GROWTH C GROWTH(-1) GROWTH(-2) broadmoney broadmoney (-1) broadmoney (-2) CAPITAL TRADE EXPENSE broadmoney C GROWTH GROWTH(-1) GROWTH(-2) broadmoney (-1) broadmoney (-2) CAPITAL TRADE EXPENSE

While we found out the FD’s lagged term and growth rate are more significantly correlated (at least at 0.05 level), than the FD contemporary term and growth; the reverse is also true: that the lagged growth variables are correlated with FD variables. Secondly, the FD and EG variables are of course found out to be autocorrelated with their own lagged term. Moreover, trade and capital accumulation has a positive effect over credit and broad money, while expense tend to positively related to the economic growth.

Additionally, the normal VAR based Granger Causality tests are also applied, table below summarized the results, the output is included in the appendix (table 12). Although the bi-directional causality is found to be true between FD & EG, however, it is evident from the output, that there is a greater inclination to the rejection of the noncausality compared to the Toda-Yamamoto approach, and even to the fixed panel OLS.

12

Table 7 normal Granger causality test results (i)

Broadmoney as financial development measure Dependent variable: GROWTH

(ii)

CREDIT as financial development measure Dependent variable: GROWTH

BROADMONEY

CREDIT

CAPITAL

CAPITAL

EXPENSE

EXPENSE

TRADE

TRADE

Dependent variable: BROADMONEY

Dependent variable: CREDIT

GROWTH

GROWTH

CAPITAL

CAPITAL

EXPENSE

EXPENSE

TRADE

TRADE

Nb. shaded items indicating rejection of null of non-causality (significance level 0.10)

13

Conclusion In summary, relationship between financial sector development measured by private sector credit and broad money, and the GDP growth, more precisely, there are causality links found on both direction. It is also found that capital accumulation and trade granger-cause the economic growth. Lastly, all the variables, has a onedirectional causality to the capital accumulation. This result indicates that for the most-developed economy, the financial development still has a significant contribution to the economic growth, in the meantime, the growth of the economy also facilitates the development of the financial sector. Contribution of this paper, combined with the previous findings on the developing countries, and the studies on periods before 1980s, provided a policy reference which supports Patrick (1966)’s stage of development hypothesis, that in the early stage, financial development cause economic growth, but as it stepping into the developed world, the evidence of the reverse is found, the supplement of this paper’s finding would be that after the economy matures, the causality is bi-directional. The limitation of this paper is that it only focused on the homogenous group of the developed countries, for the period of time that financial deregulation has already happened. Further studies could make use data available on the developing world whose financial liberalization is undergoing.

14

Reference Abu-Bader, S. & Abu-Qarn, A. (2008). Financial Development and Economic Growth: Empirical Evidence from Six MENA Countries. Review Of Development Economics, 0(0), 080414152145312-???. http://dx.doi.org/10.1111/j.1467-9361.2008.00427.x Blackburn, Keith & Victor TY Huang (1998). A theory of growth, financial development and trade, Economica 65 107-124 Calderon, Cesar & Lin Liu (2003). The Direction of Causality between financial development and economic growth, Journal of Development Economics 72, p321-334 Demetriades, P. & Hussein, K. (1996). Does financial development cause economic growth? Time-series evidence from 16 countries. Journal Of Development Economics, 51(2), 387-411. http://dx.doi.org/10.1016/s0304-3878(96)00421-x Goldsmith, R.W. (1969), “Financial Structure and Development,” New Haven, Yale University Press Jung, W. S. (1986). Financial development and economic growth: International evidence. Economic Development and Cultural Change, 34, 336–346 Khan, A. (2001). Financial development and economic growth. Macroeconomics Dynamics, 5, 413–433. King, R.G., Levine, R., 1993a. Finance and growth: Schumpeter might be right. Quarterly Journal of Economics 108, 717–738. Levine, R., 1997. Financial development and economic growth: views and agenda. Journal of Economic Literature 35, 688–726. Levine, R., Loayza, N., Beck, T., 2000. Financial intermediation and growth: causality and causes. Journal of Monetary Economics 46, 31–77 Levine, R., & Zervos, S. (1998). Stock Markets, Banks, and Economic Growth. The American Economic Review, 88(3), 537-558. Retrieved from http://www.jstor.org/stable/116848 Levine, R. & Renelt, D. (1992). A Sensitivity Analysis of Cross-Country Growth Regressions. American Economics Review 82 (4): 942-63, 1992..

15

Levine, R. & Zervos, S. (1993). What We Have Learned about Policy and Growth from Cross-Country Regressions? The American Economic Review, Vol. 83, No. 2, Papers and Proceedings of the Hundred and Fifth Annual Meeting of the American Economic Association (May, 1993), pp. 426-430. Lucas, Robert E. Jr., “On the Mechanics of Economic Development,” Journal of Monetary Economics 22 (1988):3–42 McKinnon, R. I. (1973). Money and capital in economic development. Washington, DC: Brookings Institution, DC. Neusser, K., Kugler, M., 1998. Manufacturing growth and financial development: evidence from OECD countries. Review of Economics and Statistics 80, 638–646 Patrick, H.T., 1966. Financial development and economic growth in underdeveloped countries. Economic Development and Cultural Change 14, 174–189 Schumpeter, J.A., 1911. The Theory of Economic Development. Harvard Univ. Press, Cambridge, MA. Shaw, E.S., 1973. Financial Deepening in Economic Development. Oxford Univ. Press, London. Shan, J. & Morris, A. (2010). Does Financial Development 'Lead' Economic Growth?. International Review Of Applied Economics, 16(2), 153-168. Shan, J., Morris, A., & Sun, F. (2001). Financial Development and Economic Growth: An Egg-and-Chicken Problem?. Review Of International Economics, 9(3), 443-454. http://dx.doi.org/10.1111/1467-9396.00291 Toda, H. & Yamamoto, T. (1995). Statistical inference in vector autoregressions with possibly integrated processes. Journal Of Econometrics, 66(1-2), 225-250. World Bank, 1989. World Development Report. Oxford Univ. Press, New York. World Bank, 1993. “the Role of the State in Financial Markets,”, world bank conferences on development economics 1993, Washington DC, by Stiglitz, Joseph E. World Bank. 2016. Database of World Development Indicators. Available at: http://data.worldbank.org/data-catalog/world-development-indicators

16

Appendix Table 8 Data source

Label in WDI- World Bank database

Variable Definition

NY.GDP.PCAP.KD.ZG GDP per capita growth (annual %) NE.TRD.GNFS.ZS Trade (% of GDP) NE.GDI.TOTL.ZS Gross capital formation (% of GDP) GC.XPN.TOTL.GD.ZS Expense (% of GDP) FS.AST.PRVT.GD.ZS Domestic credit to private sector (% of GDP) FM.LBL.BMNY.GD.ZS Broad money (% of GDP) Data Source: http://data.worldbank.org/data-catalog/world-development-indicators

Table 9 Toda-Yamamoto Approach result Credit model

VAR Granger Causality/Block Exogeneity Wald Tests

Dependent variable: CREDIT Excluded

Chi-sq

df

Prob.

EXPENSE GROWTH TRADE CAPITAL

0.658223 11.21957 1.811080 0.089204

2 2 2 2

0.7196 0.0037 0.4043 0.9564

All

19.66656

8

0.0117

Dependent variable: EXPENSE Excluded

Chi-sq

df

Prob.

CREDIT GROWTH TRADE CAPITAL

12.20101 7.366832 0.021594 0.242520

2 2 2 2

0.0022 0.0251 0.9893 0.8858

All

21.67898

8

0.0055

Dependent variable: GROWTH Excluded

Chi-sq

df

Prob.

CREDIT EXPENSE TRADE CAPITAL

5.751273 0.049632 5.968784 11.80067

2 2 2 2

0.0564 0.9755 0.0506 0.0027

All

19.49833

8

0.0124

17

Dependent variable: TRADE Excluded

Chi-sq

df

Prob.

CREDIT EXPENSE GROWTH CAPITAL

3.059321 4.441566 12.00192 5.854050

2 2 2 2

0.2166 0.1085 0.0025 0.0536

All

36.64669

8

0.0000

Dependent variable: CAPITAL Excluded

Chi-sq

df

Prob.

CREDIT EXPENSE GROWTH TRADE

15.94490 9.509106 30.42220 16.18410

2 2 2 2

0.0003 0.0086 0.0000 0.0003

All

73.38908

8

0.0000

Broad money model VAR Granger Causality/Block Exogeneity Wald Tests Included observations: 244

Dependent variable: CAPITAL Excluded

Chi-sq

df

Prob.

BROADMONE Y EXPENSE GROWTH TRADE

0.019153 15.09810 7.342391 5.120418

2 2 2 2

0.9905 0.0005 0.0254 0.0773

All

39.67655

8

0.0000

Dependent variable: BROADMONEY Excluded

Chi-sq

df

Prob.

CAPITAL EXPENSE GROWTH TRADE

0.604424 3.242917 6.880875 1.480964

2 2 2 2

0.7392 0.1976 0.0321 0.4769

All

17.01239

8

0.0300

Dependent variable: EXPENSE Excluded

Chi-sq

df

Prob.

CAPITAL

10.85793

2

0.0044

18

BROADMONE Y GROWTH TRADE

4.099460 1.849272 10.47034

2 2 2

0.1288 0.3967 0.0053

All

28.14691

8

0.0004

Dependent variable: GROWTH Excluded

Chi-sq

df

Prob.

CAPITAL BROADMONE Y EXPENSE TRADE

1.881673

2

0.3903

4.659522 0.369891 2.586226

2 2 2

0.0973 0.8311 0.2744

All

8.861858

8

0.3541

Dependent variable: TRADE Excluded

Chi-sq

df

Prob.

CAPITAL BROADMONE Y EXPENSE GROWTH

1.396004

2

0.4976

0.623536 1.071368 4.554430

2 2 2

0.7322 0.5853 0.1026

All

12.86163

8

0.1167

19

Table 10 Panel cointegration test Broad money model

Credit model

No trend – Kao

ADF -8.899

No trend – Kao

ADF -6.038

Panel v-Statistic

Panel vStatistic Panel rho

-0.132916

Panel PP

-5.412462

Panel ADF

-5.105012

1.154950

Group Rho

0.574380

-8.020939

Group PP

-6.334658

-7.321563

Group ADF

-4.842318

-1.378624 Panel rho -0.011214 Panel PP -6.444557 No trend Pedroni

Panel ADF -6.870421 Group Rho

No trend - Pedroni

-0.999923

Group PP Group ADF

Weighted Statistic Prob. Statistic Prob. 1.071891 0.8581 1.399001 0.9191

Statistic Panel vStatistic -0.553295 Panel rhoStatistic 0.852780 Panel PPStatistic -6.501652 Panel ADFStatistic -6.714708

Panel vStatistic Panel rhoStatistic 0.893305 0.8142 0.846625 0.8014 Panel PPStatistic 4.667294 0.0000 5.762386 0.0000 Panel ADFStatistic 4.293636 0.0000 5.185114 0.0000 With trend Pedroni

Alternative hypothesis: individual AR coefs. (between-dimension)

With trendPedroni

Statistic Prob. Group rhoStatistic Group PPStatistic Group ADFStatistic

6.145735 0.0000 5.095362 0.0000

20

Statistic

Prob.

0.7100 -3.303191

0.9995

0.8031

1.808305

0.9647

0.0000 -7.288630

0.0000

0.0000 -6.464930

0.0000

Alternative hypothesis: individual AR coefs. (between-dimension)

Group rhoStatistic Group PPStatistic Group ADFStatistic

2.156165 0.9845

Prob.

Statistic

Prob.

2.882288

0.9980

-8.538738

0.0000

-6.515674

0.0000

Table 11 ols fixed effect model Dependent Variable: GROWTH Total panel (unbalanced) observations: 444 Variable

Coefficient

Std. Error

t-Statistic

Prob.

C GROWTH(-1) GROWTH(-2) CREDIT CREDIT(-1) CREDIT(-2) CAPITAL TRADE EXPENSE

6.377146 0.327828 -0.207842 -0.020430 0.036638 -0.028354 0.041802 -0.007327 -0.120542

1.666091 0.049942 0.049626 0.009777 0.014786 0.010064 0.046720 0.004952 0.030455

3.827609 6.564113 -4.188157 -2.089560 2.477887 -2.817469 0.894740 -1.479746 -3.958067

0.0001 0.0000 0.0000 0.0373 0.0136 0.0051 0.3714 0.1397 0.0001

Effects Specification Cross-section fixed (dummy variables) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.282826 0.243552 2.157913 1955.767 -959.1710 7.201372 0.000000

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

1.768300 2.481100 4.428698 4.650094 4.516007 1.973459

Dependent Variable: CREDIT Method: Panel Least Squares Total panel (unbalanced) observations: 444 Variable

Coefficient

Std. Error

t-Statistic

Prob.

C GROWTH GROWTH(-1) GROWTH(-2) CREDIT(-1) CREDIT(-2) CAPITAL TRADE EXPENSE

-0.539879 -0.503625 0.703544 0.234354 1.130313 -0.196498 0.625973 0.037606 -0.289276

8.415213 0.241020 0.258102 0.251229 0.049258 0.049516 0.230171 0.024582 0.153356

-0.064155 -2.089560 2.725835 0.932833 22.94701 -3.968355 2.719603 1.529840 -1.886309

0.9489 0.0373 0.0067 0.3514 0.0000 0.0001 0.0068 0.1268 0.0599

Effects Specification Cross-section fixed (dummy variables) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.946859 0.943949 10.71413 48212.86 -1670.646 325.3678 0.000000

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

21

96.01384 45.25469 7.633542 7.854938 7.720850 2.019615

Dependent Variable: GROWTH Method: Panel Least Squares Total panel (unbalanced) observations: 261 Variable

Coefficient

Std. Error

t-Statistic

Prob.

C GROWTH(-1) GROWTH(-2) BROADMONEY BROADMONEY(-1) BROADMONEY(-2) CAPITAL TRADE EXPENSE

2.511305 0.344733 -0.218981 -0.037440 0.057511 -0.035554 0.095995 0.012258 -0.095962

2.583660 0.065099 0.063032 0.015457 0.020935 0.015582 0.055248 0.014628 0.049515

0.971995 5.295557 -3.474117 -2.422162 2.747149 -2.281815 1.737535 0.838005 -1.938030

0.3320 0.0000 0.0006 0.0162 0.0065 0.0234 0.0836 0.4028 0.0538

Effects Specification Cross-section fixed (dummy variables) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.258058 0.209406 1.928121 907.1069 -532.9120 5.304156 0.000000

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

1.658968 2.168490 4.213885 4.446057 4.307211 1.891858

Dependent Variable: BROADMONEY Method: Panel Least Squares Total panel (unbalanced) observations: 261 Variable

Coefficient

Std. Error

t-Statistic

Prob.

C GROWTH GROWTH(-1) GROWTH(-2) BROADMONEY(-1) BROADMONEY(-2) CAPITAL TRADE EXPENSE

-15.02440 -0.627129 0.588839 0.308035 0.920135 -0.012682 0.305379 0.159463 0.197019

10.55080 0.258913 0.278785 0.263537 0.064016 0.064442 0.226664 0.059078 0.203814

-1.424006 -2.422162 2.112164 1.168846 14.37346 -0.196791 1.347276 2.699184 0.966660

0.1557 0.0162 0.0357 0.2436 0.0000 0.8442 0.1791 0.0074 0.3347

Effects Specification Cross-section fixed (dummy variables) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic)

0.949997 0.946718 7.891187 15194.08 -900.7133 289.7287 0.000000

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

22

73.91075 34.18617 7.032286 7.264458 7.125612 2.083360

Table 12 standard granger causality Credit model

VAR Granger Causality/Block Exogeneity Wald Tests Date: 11/01/16 Time: 12:40 Sample: 1980 2015 Included observations: 436

Dependent variable: CREDIT Excluded

Chi-sq

df

Prob.

CAPITAL GROWTH TRADE EXPENSE

0.908850 12.91240 0.090789 3.757804

2 2 2 2

0.6348 0.0016 0.9556 0.1528

All

29.24242

8

0.0003

Dependent variable: CAPITAL Excluded

Chi-sq

df

Prob.

CREDIT GROWTH TRADE EXPENSE

23.87796 40.76835 13.17540 11.81744

2 2 2 2

0.0000 0.0000 0.0014 0.0027

All

88.54420

8

0.0000

Dependent variable: GROWTH Excluded

Chi-sq

df

Prob.

CREDIT CAPITAL TRADE EXPENSE

33.87755 38.79328 7.230911 1.746727

2 2 2 2

0.0000 0.0000 0.0269 0.4175

All

70.85516

8

0.0000

Dependent variable: TRADE Excluded

Chi-sq

df

Prob.

CREDIT CAPITAL GROWTH EXPENSE

1.735876 4.324735 30.55786 0.998587

2 2 2 2

0.4198 0.1151 0.0000 0.6070

All

44.96902

8

0.0000

df

Prob.

Dependent variable: EXPENSE Excluded

Chi-sq

23

CREDIT CAPITAL GROWTH TRADE

1.739900 25.43988 4.171766 1.446308

2 2 2 2

0.4190 0.0000 0.1242 0.4852

All

42.61084

8

0.0000

Money model VAR Granger Causality/Block Exogeneity Wald Tests Included observations: 255

Dependent variable: BROADMONEY Excluded

Chi-sq

df

Prob.

GROWTH TRADE EXPENSE CAPITAL

8.871642 3.514302 5.143078 0.177092

2 2 2 2

0.0118 0.1725 0.0764 0.9153

All

17.99112

8

0.0213

Dependent variable: GROWTH Excluded

Chi-sq

df

Prob.

BROADMONEY TRADE EXPENSE CAPITAL

9.441260 2.609531 0.539817 8.424185

2 2 2 2

0.0089 0.2712 0.7634 0.0148

All

21.89806

8

0.0051

Dependent variable: TRADE Excluded

Chi-sq

df

Prob.

BROADMONEY GROWTH EXPENSE CAPITAL

1.306628 12.76329 0.214986 2.371997

2 2 2 2

0.5203 0.0017 0.8981 0.3054

All

20.44030

8

0.0088

Dependent variable: EXPENSE Excluded

Chi-sq

df

Prob.

BROADMONEY GROWTH TRADE CAPITAL

0.202899 1.687345 6.891683 21.14612

2 2 2 2

0.9035 0.4301 0.0319 0.0000

All

30.78653

8

0.0002

24

Dependent variable: CAPITAL Excluded

Chi-sq

df

Prob.

BROADMONEY GROWTH TRADE EXPENSE

1.301344 10.27744 2.837084 16.22902

2 2 2 2

0.5217 0.0059 0.2421 0.0003

All

39.82489

8

0.0000

25