Financial Development Convergence: Does The World Move Towards a Market-based System?

(1)

Financial Development Convergence: Does The World

Move Towards a Market-based System?

MSc International Economics and Business

ABSTRACT

This thesis addresses two questions on financial development. I examine stock market convergence and explore how financial structures evolve during economic development. These questions are important, because financial development can contribute to economic growth and increase the standard of living. Using a broad sample of countries, I show that stock market capitalization relative to GDP converges over time, controlling for macroeconomic variables, institutional quality and financial system efficiency. Stock market growth tends to accelerate at medium levels of economic development, whereas bank credit grows more strongly at lower levels of development. Hence, the results suggest an evolving role for banks and stock markets during the process of economic development, in which stock markets become relatively important.

Name: Koen Seebus Student number: 1924028

Study program: MSc International Economics and Business Supervisor: dr. R.K.J. Maseland

(2)

(3)

I. INTRODUCTION

This thesis addresses two questions on financial development. I examine stock market convergence and explore how financial structures evolve during economic development. Understanding financial development is important, because financial development helps mobilize savings, facilitates efficient allocation of resources, (Greenwood, Sanchez, & Wang, 2010), enhances innovation activities (Aghion, Howitt, & Mayer-Foulkes, 2005), contributes to risk sharing (Saint-Paul, 1992) and is found to boost economic growth (Demirgüç-Kunt & Levine, 2001; Beck & Levine 2002; Levine, 2002). Therefore, the convergence of financial development can help overcome poverty and reduce inequality and could be a first step for developing economies to reach higher standard of living. Moreover, Beck, Demirgüç-Kunt and Levine (2007) find that financial development disproportionately boosts incomes of the poorest quintile and hereby reduces income inequality. According to them, financial development is associated with a decrease of people living on less than $1 a day, after controlling for average growth. These findings emphasize the importance of financial system development, especially for the poor. While developed economies showed strong financial development growth over the last decades, there is very little evidence whether the poor take advantage of this increased international finance know-how. In other words, it is unknown whether a process of financial development convergence exists. Bahadir and Valev (2015) try to answer this in terms of bank credit and find convincing results. They find developing countries grow faster in their distribution of bank credit than developed countries do, possibly by leveraging on knowledge from more developed economies. However, empirical evidence lacks on the topic of stock market convergence.

Many believe stock markets can also contribute to economic growth and help achieve well-being, or that at least a reciprocal relationship exists between stock market development and economic growth (among others: Garcia & Liu, 1999; Beck & Levine, 2004; Caporale, Howells & Soliman, 2004; Lin, Sun & Jiang, 2009). Here, the question for developing countries is not only whether citizens gain more access to financial services, but also whether these countries made full use of available international stock market know-how in order to speed up their process of its development. Contrary to the growth of equity markets within one country over time, I look at differences in stock market development between countries over time. Whether this occurred is an important question, because there are many advantages to having a well-developed capital market. If convergence is not present, policy could take the form of actively providing developing countries with this finance knowledge.

(4)

and Valev (2015) find evidence that bank credit grows faster especially in lower developed economies. I argue that stock markets especially grow in medium or higher developed countries. As mentioned, some authors, such as Garcia and Liu (1999), argue equity markets and economic growth exhibit a mutual or reciprocal relationship. Benefits from stock markets may thus not be straightforward, because they can be particularly advantageous in distinct stages of economic development (Lin et al., 2009). For example, Rioja and Valev (2011) show that stock markets have positive effects on productivity and capital growth, only in high-income countries. They find insignificant results for low-income countries. Furthermore, Demirgüç-Kunt and Levine (2001) show that both banks and capital markets grow alongside GDP, but that markets develop faster than banks, suggesting that the world becomes more market-based. Both researches may reflect reverse causality. The fact that financial systems become more market-based as economies prosper does not mean that equity markets have a larger impact on economic activity in the more advanced economies. It could be that economic growth has a relatively large effect on the development of stock markets.

Subsequently, I examine a related, second question, covering a changing role of banks and markets during economic development. An essential question when comparing financial systems concerns the most efficient way to organize capital transfer from savers to investors. Although financial development can cause economic growth, there is hardly consensus on where this effect comes from: banks, capital markets or a combination (see e.g. Demirgüç-Kunt & Levine, 2001; Demirgüç-Kunt & Maksimovic, 2002; Beck & Levine, 2002). Among others, Lin et al. (2009) theoretically claim that the relevance of banks and capital markets changes when economies develop, because both provide different services. Given these possible differences, it is not surprising that the literature is ambiguous on what type most benefits economic growth. I therefore look into a possible evolving role of banks and markets during the process of economic development. This is important because of its potential policy implications, like possible institutional disadvantages at certain stages of development. Moreover, there are theoretical motivations since empirical research has been unsuccessful at explaining a possible evolving role of banks and markets along the path of economic growth (Song & Thakor, 2010). Furthermore, literature reports mixed findings on whether banks and capital markets are complements or substitutes (e.g. Merton, 1995; Allen & Gale, 1997; Rajan & Zingales, 2003). Addressing the question on financial system evolution can shed more light on this matter.

(5)

institutions and financial system efficiency. While this process could be driven by a relatively large increase in listed companies or in an increase in the valuation of these companies, I do not differentiate between the two possibilities in this research. Similar evidence for bank credit convergence is found.

Using similar methods, I also present evidence that bank credit growth is associated with lower levels of GDP where stock market growth is associated with higher levels of GDP. The combined results suggest that the importance of banks and markets changes during the process of economic development. This is in line with a body of literature that predicts that capital markets become more important and banks become relatively less important along the path of economic development. Additionally, the ratio of stock market capitalization to bank credit also converges. It measures the importance of capital markets relative to that of banks. The predominating existing view distinguishes between bank-based and capital-market-based economies and assumes that banks and capital markets are substitutes. Accordingly, countries should show some form of specialization in either two. My findings contradict this view and suggest that: (1) both markets and banks show strong signs of convergence, after controlling for country-level institutions or financial system efficiency; (2) stock markets grow faster at moderate levels of economic development, however decelerate, as GDP increases further; (3) while banks grow faster at lower levels of economic development; (4) and the ratio of stock markets to banks converges. Hence, stock markets and banks are more likely to be complements than substitutes, because I identify no signs of specialization. Therefore, it seems irrelevant to distinguish between bank-based and market-based economies. Rather, these results indicate that the financial system adapts to different stages of economic development.

These findings are policy relevant and imply estimated elasticities from previous researches can present a misleading message, because the relationship shows signs of non-linearity. Therefore, estimations far from “world average” may be imprecise. The existence and the extent of financial development convergence is an important issue, because policy could take the form of actively providing developing countries with finance knowledge to help them achieve higher standards of living. Even though the results point out that convergence and the evolving role of banks and markets should be taken into account, I do not consider direct policy instruments. Therefore, this thesis does not provide guidance to fitting development of financial systems.

(6)

II. LITERATURE REVIEW

Literature on the topic of financial system convergence is rare and reports mixed results. On the one hand, Djankov, McLiesh and Shleifer (2007) find very little convergence in creditor right scores, which they find to be a determinant of private credit to GDP. Moreover, Antzoulatos et al. (2011) find evidence for non-convergence of financial institution characteristics in a sample of thirty-eight industrial and developing countries during 1990–2005. On the other hand, Veysov and Stolbov (2011) report financial development convergence for 1980-2010. Bruno et al. (2012) also show strong convergence in the OECD in terms share holdings and insurance products, but they find mixed evidence for debt securities and deposits. Bahadir and Valev (2015) present evidence for global convergence in private credit levels to GDP and other measures of financial development.

Next to mixed findings, the term capital market “convergence” is often used interchangeably with market integration, for instance by calculating common stochastic trends, examining valuation congruence or looking at news and volatility spillovers (e.g. Baele, 2005; Shackman, 2006). Mylonidis and Kollias (2010) find evidence of stock market integration in four major markets in the Eurozone and show that it is very much an ongoing process. However, Bianco, Gerali and Massaro (1997) and Schmidt, Hackethal and Tyrell (2001) conclude that there is limited financial institution convergence for EU countries, despite the initiation of the European Single Market.

(7)

Should we expect stock market convergence?

Reduced barriers, the global move towards market-based economies1_{and the integration of financial}

systems can cause capital markets to converge in terms of size. Moreover, advantages of backwardness in combination with knowledge transfers, increased economic development, and diminishing returns of capital can also contribute to stock market convergence. Finally, stylized facts give an indication that such a process is indeed present.

First of all, a global trend of economic integration can cause capital markets to converge, i.e. integration can cause inequality to decrease over time. Financial system integration opens up investment opportunities, reduces barriers to trade international financial products and eases access to capital that was first unavailable. This can lead to increased trading, positively affecting liquidity in the market. In turn, higher liquidity positively affects stock market capitalization (Garcia & Liu, 1999). The effects of capital market integration may be biggest for lower developed countries, because the marginal difference in the size of the opening playing field is bigger for these countries. Especially the European Union has seen strong financial system integration. According to Giannetti (2002), European countries with lower relative sizes of financial development grow faster than their developed counterparts. They assume that financial integration in Europe increases the supply of finance in the less financially developed countries. This process can be reflected in an expansion in the size of national financial systems – relative to domestic GDP.

Related, the global general move towards free-market policies (where markets rather than central organs set prices) has opened up the playing field for capital investments in financial services (Goldberg, 2004; Bahadir & Valev, 2015). Free-market policies and liberalization policies can stimulate firm activity (Baldwin, 1997). Opening up the economy yields an opportunity to open up businesses, including financial institutions. For example, new banks can create numerous possibilities for companies to thrive and enter the stock These positive effects may be most noticeable in so-called transition economies, because these countries are often relatively closed economies. Therefore, marginal effects are possibly bigger in these countries.

Second, relatively low financial development can benefit from advantages of backwardness. Less developed countries can borrow and use technologies, business models and knowledge from more advanced economies. In this sense, imitation may be easier than innovation. Economic liberalization enables the spread of sector know-how to developing countries. International banks, rating agencies, brokerage houses or other international financial institutions bring expertise on how to successfully

1_{Here, free market policies are defined in the sense of demand and supply: open markets, which set prices. Not to be confused}

with capital-market-based systems, by which I mean a system with a relative high presence of stock and bond markets.

2_{I do not add all control variables at once, because this leads to data loss.}

(8)

organize capital markets. For example, in countries such as Estonia, 75% of the banks are foreign owned. This knowledge transfer especially strengthens financial development in low-development areas, simply because knowledge from the frontier is transferred ‘backwards’. Such a backwards knowledge transfer is particularly encouraging for developing economies given that financial development can boost economic growth. Consequently, financially low-developed countries have an incentive to catch up with the frontier as a means to enhance GDP. Arguably, the advantages of backwardness are greater at moderate levels of backwardness. Specifically, at moderate levels, human capital permits actual exploitation of backwardness and its opportunities, whereas very low levels of economic and financial development often go hand in hand with poor education, institutions and infrastructure. This prevents these countries from taking full advantage of backwardness. Later on, I will discuss the effect of stock market convergence at medium levels of economic development.

Third, economic prosperity may have spillover effects into the financial system, because demands for most financial services are determined by the demands of the real economy (Lin et al., 2009). While financial systems can cause economic growth by activities such as fund pooling, risk diversification and liquidity management, the opposite may also be true. In reality, the relationship between financial and economic development is most likely reinforcing and bidirectional. The financial and real sectors interact during all stages of development (Garcia & Liu, 1999). Boyd and Smith (1998) theoretically demonstrate that stock markets grow as GDP increases. Their model allows for financial structure changes as countries go through different stages of economic development. Thus, economic growth could in turn lead to growth of the financial system. Also, efficiency of the financial sector increases with its size, because economies of scale and learning-by-doing effects (Garcia & Liu, 1999). Moreover, economic growth can affect stock market capitalization through increased corporate earnings, more investment opportunities and decreased default probabilities. Hence, a growing real economy can wield positive externalities on a country’s financial sector. These externalities may be of particular importance for relatively low developed nations, since these are the ones who have seen high economic growth. Furthermore, general income growth across countries may have pushed firms and investors above a certain threshold level of income to start using more sophisticated financial services such as bond markets and stock markets. Parallel to the increase in stock market capitalization, GDP per capita has strongly increased over the past 20 years. As I will explain later, this could be of special importance for moderately developed nations in terms of economical development.

(9)

(Demirgüç-Kunt & Levine, 1996). Nonetheless, on average, even these listings are smaller or at best equal to average listings in developed markets.

If the factors discussed above have played a more important role in lower- and medium-developed countries as opposed to medium-developed countries, then I should observe capital market convergence. On the other hand it may be that globalization and knowledge transfers did not affect low developed countries in particular, but equally affected the world economy as a whole or even led to divergence.

Last, stylized facts in the Beck and Demirgüç-Kunt (2009) dataset give an impression that convergence is present in stock markets. Over the past 20 years, global stock markets have seen a strong increase and almost doubled in relative size. In 1990, the average stock market capitalization for all countries records around 30% of GDP, where in 2010 it rose to more than 55% of GDP for all countries (data allowing). Hence, stock markets have strongly gained in importance over the years. These figures also suggest that the world is adapting a more capital-market orientated system, as opposed to a bank-centered system. Table 1 shows the average stock market capitalization relative to GDP for 48 countries. The countries are sorted along the 33rd_{and 66}th_{percentile of stock market capitalization in}

1990, forming three groups with low, medium and high levels op stock market development. Stock market capitalization strongly increased for the lowest developed countries, by almost 720%, whereas the average rise for all 48 countries was 152%. The numbers suggest that a process of financial convergence is at work for capital markets.

Table 1.

Average stock market cap to GDP

Country groups, according to their levels of stock market development

Low Medium High All

1990 5.1 19.7 72.1 32.3

2000 21.7 66.2 150.3 79.4

2010 41.7 53.4 147.6 81.5

Increase 1990-2010 (%) 718% 171% 105% 152%

Number of countries 16 16 16 48

Bahadir and Valev (2015) show related figures: private sector credit as percent of GDP had doubled in the three decades before the financial crisis of 2008. In addition, they show actual convergence for credit levels, also when controlled for the quality of institutions in these countries. An increase in credit levels to the private sector can be a sign of increased investment opportunities. In time, this can lead to firm growth, hereby possibly increasing market capitalization of listed firms. Thus, the convergence in terms of credit might have spillovers, which can cause capital markets to converge as well.

(10)

Stock market convergence: The medium development hypothesis

Contrary to bank credit, stock markets may be relatively unimportant to developing economies. Bahadir and Valev (2015) find that credit growth is especially high in the financially least developed countries. Small local banks and microfinance institutions are probably more important for small enterprises and households, whereas stock markets may only be relevant for firms from a certain size onwards. I thus need to examine the role of economic development simultaneously to answer the question of stock market convergence, As mentioned, some authors, for example Garcia and Liu (1999), argue equity markets and economic growth exhibit a two-way relationship. Therefore, benefits from stock markets are not straightforward. Arguably, stock markets are particularly advantageous in later (i.e. non-developing) stages of economic development, whereas banks are important for standardized agreements in earlier stages of the economy (Lin et al., 2009). Combining this logic with that of convergence, stock market growth is presumably highest at medium levels of economic development. I call this the medium development hypothesis.

It is hard to imagine stock markets developing overnight. Rather, countries experience economic development over a long period of time and possibly adapt their financial structure accordingly. Initially, nations in Southeast Asia relied on smaller banks and only in later stages stock markets became more important. The United States went through a similar process. In the early phases of its economic development, the United States relied mostly on banks and the New York Stock exchange became important after the creation of large, capital intensive industrial firms in the late 19th

century (Lin et al., 2009). Increases in GDP per capita have pushed households and firms within developing countries above a certain threshold of income (Aghion et al., 2005). This allows them to start using more sophisticated financial services such as the capital markets. In general, global poverty has strongly declined over the last decades, even showing sings of income convergence for middle-income countries. Southeast Asia accounts for a large fraction of the success (Sala-i-Martin, 2006; Lakner & Milanovic, 2013). Moreover, there are also signs of real convergence on a global scale (Rodrik, 2006; Shleifer, 2009) and low-income countries like Sub-Saharan African countries (Jones & Olken, 2008). Rodrik (2012) shows that manufacturing industries exhibit strong unconditional convergence in GDP per worker. Stock markets can be especially advantageous for these large, sometimes high-tech or capital-intensive firms, including industries as manufacturing (Lin et al., 2009). Therefore, stock market growth and GDP per capita are likely to grow together, perhaps as soon as industries with higher technological innovation risk start to play an important role in an economy. For these reasons, I expect capital markets to develop relatively faster in countries, which have medium levels of economic development.

(11)

Financial structure evolution

Both banks and capital markets are needed to have a fully functioning financial system. If the financial market is composed of banks only, the market will fail to achieve efficient allocation of capital because of asymmetric information problems. Thus, the development of stock markets is also necessary to achieve full efficiency of capital allocation (Caporale et al., 2004). The medium development hypothesis states that stock markets grow particularly at moderate levels of GDP while banks grow at lower levels of GDP. It follows that stock markets and banks could have changing or evolving roles as economies grow. I call this financial structure evolution.

(12)

arrangements. More recently, Lin et al. (2009) propose that as countries advance through different stages of economic development and at each stage of development the relevance of banks and capital markets changes. Both provide different financial services with its demand often being determined by demands in the real economy.

Empirical research supports this logic, but is inconclusive. Demirgüç-Kunt and Levine (1996) find as countries reach middle income, stock markets and non-bank financial intermediaries grow relatively faster than before, suggesting a catalytic effect once a certain development level is reached. In addition, Demirgüç-Kunt and Levine (2001) find that stock markets tend to develop faster than banks when economies grow, suggesting the world moves towards a more market-based system as it develops economically. Rioja and Valev (2011) show that in high-income countries stock markets are found to have sizable positive effects on both productivity and capital growth, but they find insignificant results for low-income countries even though stock markets are present in these countries. According to the authors, these nations have probably not reached a minimum size and liquidity to effectively supply funding to firms. However, if the medium development hypothesis is correct, the process described by Rioja and Valev (2011) may suffer from reverse causality. It may be that they find such results, because stock markets only (strongly) develop when GDP increases.

If financial structures do evolve, there are several implications: (1) stock markets (bank credit) should grow faster at medium (low) levels of economic development; (2) and stock markets and banks are more likely to be complements than substitutes. Therefore, countries do not necessarily specialize in neither stock markets nor banks, but mainly shift their emphasis as they develop; (3) Finally, stock markets should thus ‘outgrow’ banks once countries reach a stage of medium economic development, i.e. on average, the relative importance of stock markets should increase as countries become richer. This leads to the following additional hypotheses:

Hypothesis 3: Bank credit growth is higher at low levels of economic development.

Hypothesis 4: Countries converge in terms of their financial structure, i.e. the ratio of stock markets to bank credit will converge.

(13)

III. DATA & METHOD Stock market convergence

I will explore capital market convergence, using several measures from the Financial Structure Database (Beck, Demirgüç-Kunt & Levine, 2009), the World Bank World Development Indicators and World Governance Indicators. The methodology is partly based on the one used by Bahadir and Valev (2015), who measure financial development convergence in terms of deposit money bank credit. My primary focus, however, is on stock market capitalization as percent of GDP (ST MARKET CAP) since this measure has been most extensively used in the literature to measure capital market development (e.g. Demirgüç-Kunt & Levine, 1996; Garcia & Liu, 1999; La Porta, Lopez-De-Silanes & Schleifer, 2006). First, and most importantly, it can be seen as a measure for quality. Stock markets with high market capitalizations, relative to GDP, have been found to be more liquid, more integrated in the world market and show less mispricing (Demirgüç-Kunt & Levine, 1996). Second, it can be seen as a measure for capital-market-importance to the economy, because it shows the share of stock market activity within the entire economy. This is relevant for examining financial structures and contrasting between capital-based and bank-capital-based economies. I also use the stock market total value traded as percent of GDP (VALUE TRADED) and stock market turnover ratio (TURNOVER), which is the ratio (%) of the value of total shares traded to average real market capitalization. Variables and data sources are presented in Appendix A. Summary statistics are presented in Appendix B.

Similar to Bahadir and Valev (2015), I measure convergence by calculating the growth rate of capital market development in a period of time and relating this to the level of capital market development in the first year of that specific period. I transform the annual data into non-overlapping, four-year periods: 1991-1995, 1996-1999, 2000-2003, 2004-2007, and 2008-2011. The periods are to reduce endogeneity issues and to smooth out annual variation although retain time variation on the country level. Beck, Levine and Loayza (2000) and Bahadir and Valev (2015) use the same method in their research with five-year periods. Instead, I use four-year periods in order to obtain an additional time observation. For robustness, I also use the annual data panel in unreported estimations and obtain qualitatively similar results. I start with a simple empirical equation of unconditional ! convergence and will gradually add control variables for robustness, which leads to data loss.

!"!" = ! + !!"#!"+ !!+ !!+ !!" (1)

where gsit is the average annualized growth of stock market development in country i, during a four-year

period t. SMDit is the initial level of stock market development (i.e. the level in the first year of that

(14)

traded or stock market turnover ratio, all divided by GDP. !! captures the country-specific effects,

where !! captures the time-specific effects. I estimate my regressions with fixed effects in combination

with the non-overlapping four-year periods to capture country and time-specific effects. All variables are transformed into logs and the independent variables are timed at the start of the period.

The coefficient of interest is !. !<0 implies capital market development convergence such that capital markets grow faster in countries or periods with a lower initial levels of capital market development. !>0 means divergence, because capital market development grows faster in countries that already have relatively high levels of capital market development. !=0 implies that capital market development growth is not related to initial levels of capital market development and therefore that differences across countries remain equal over time.

Next, I add a vector of macroeconomic control variables:

!"!" = ! + !!"#!"+ !!!"+ !!+ !!+ !!" (2)

where X is a vector of macroeconomic control variables. As before, all variables are transformed into logs and times at the beginning of the four-year period. The macroeconomic controls are GDP per capita (GDP PER CAPITA), trade openness (TRADE), the savings rate (SAVINGS) and inflation (INFLATION).

(15)

For robustness, I add control variables for institutional quality (I) and financial system efficiency (Eff) in equation 3.2_{A more efficient financial system, treated here exogenously, can have}

important impact on the speed of financial development growth.

!"!" = ! + !!"#!"+ !!!"+ !!!"+ !!""!"+ !!+ !!+ !!" (3)

where I and Eff are vectors of institutional control variables. As before, all variables are transformed into logs and timed at the beginning of the four-year period. I use institutional controls regulatory quality, control of corruption and rule of law, which for reasons of multicollinearity are combined into governance (GOVERNANCE), and Voice and Accountability (ACCOUNTABILITY). I use the Net Interest Margin (NET INTEREST MARGIN), Cost (COST) and Return on Assets (RETURN ON ASSETS) as controls for financial system efficiency. Net Interest Margin is defined as the accounting value of bank’s net interest revenue as a share of its interest-bearing (total earning) assets; Cost is the total cost divided by total income of all commercial banks in a country; and Return on Assets is the average return on assets (Net Income / Total Assets) in a country.

Especially for capital markets, institutional development can be important for the functioning of the system. Demirgüç-Kunt and Levine, (1996) argue that rule of law and regulatory quality are important predictors of stock market development. For example, property rights or investor protection laws should be properly enforced in order to encourage investors to contribute to the market who otherwise fear expropriation. In turn, high levels of corruption would discourage investors to participate in the market. La Porta et al. (2006) find that laws mandating disclosure and liability rules benefit stock markets, however, do not find that general public enforcement has benefit. For these reasons, I add institutional control variables from the World Bank’s World Governance Indicators (WGI) and World Development Indicators (WDI). The inclusion of the data comes at a cost, because the data is only sporadically available from 1996 onwards.

To test the medium development hypothesis I enter GDP per capita as an explanatory variable into the model. I am interested in whether capital markets grow faster in low-developed countries or in countries with medium levels of economic development. Developing countries arguably do not need sophisticated financial institutions like stock markets as much as they need banks. In equation 4b, GDP PER CAPITA (SQUARED) is used to test for a non-linear relationship between stock market growth and GDP per capita.

(16)

!"!" = ! + !!!"#!"+ !!!"#!"+ !!!"#!"! + !!!"+ !!+ !!+ !!" (4a)

!"!" = ! + !!!"#!"+ !!!"#!"+ !!!"#!"! + !!!"+ !!!"+ !!""!"+ !!+ !!+ !!" (4b)

Despite using non-overlapping time periods, reverse causality problems may still be present. Growth rates in the last year of a period explain the starting point for the next period. This can result in covariance between the independent variables and the error term. Moreover, I calculate growth rates using the initial level in that same period. Reversely, starting levels of that specific period may thus be explained by the growth rates. The set of previously described control variables can also present reverse causality issues in a similar way as the growth rates. Prior research also reports difficulties in establishing causality between financial development and economic growth. Financial development growth in period 1 may seem to explain the GDP per capita in period 2, but GDP per capita explains financial development growth instead.

To deal with the reverse causality issues, I follow Bahadir and Valev (2015) and estimate equations 1 – 3 using Generalized Method of Moments (GMM) dynamic panel techniques, which have been widely used in finance and growth literature (Blundell & Bond, 1998; Beck et al., 2000). The method was designed to estimate panels with a small number of time-series (here 5 periods), but many cross sectional units (here over 100 countries). The GMM panel estimator exploits the time-series variation in the data, accounts for unobserved country-specific effects and controls for endogeneity for all of the independent variables, including the stock market development variables. To realize this, the panel estimator uses internal instruments, i.e. instrumental variables based on previous observations of independent variables. More specifically, the Arellano-Bover (1995), Blundell-Bond (1998) GMM systems estimator combines: (1) a GMM difference estimator where lagged levels of the independent variables are used as instruments under the conditions: (1a) the error term is not serially correlated; (1b) the lagged levels of the independent variables are weakly exogenous, i.e., they can be affected by current and past growth rates, but are uncorrelated with future realizations of the error term and (2) a GMM levels estimator that uses the lagged differences of the independent variables as instruments under the conditions: (2a) the error term is not serially correlated; (2b) there is no correlation between the difference in the independent variables and the country specific effect, but there may be correlation between the levels of the independent variables and the country-specific error term.

(17)

In this section I test for a possibly changing role for banks and stock markets during economic development. Therefore, I look at bank credit convergence, the medium development hypothesis and the interplay of bank credit, stock markets and financial structures, respectively. Equation 5a tests for bank credit convergence, next to stock market convergence.Banking development is operationalized as credit by deposit money banks to the enterprise sector, as a percent of GDP. Equation 5b examines the medium development hypothesis for bank credit by including GDP PER CAPITA SQUARED.

!"!" = ! + !!!"!"+ !!!"#!" + !!!"+ !!+ !!+ !!" (5a)

!"!" = ! + !!!"!"+ !!!"#!"+ !!!"#!"! + !!!"+ !!+ !!+ !!" (5b)

where gbit is the average annualized growth of bank credit in country i, during a four-year period t.

BCit is the initial level of bank credit (CREDIT) (i.e. the level in the first year of that same four-year

period t). X is the vector of macroeconomic control variables: trade openness (TRADE), the savings rate (SAVINGS) and inflation (INFLATION).

I want to know whether the level of bank credit is associated with the capital market growth. Demirgüç-Kunt and Levine (1996) argue that capital markets and banks and non-bank financial intermediaries are complements rather than substitutes. Hence, it may be possible that convergence is accelerated by the presence of high levels of bank credit, while it is also possible that countries with high levels of credit market development already have a high level of financial development in general, reducing the scope for further improvements.

Equations 6a and 6b explore the interplay between stock markets and bank credit:

!"

= ! + !

!

!"#

!"

+ !

!

!"

(6a)

before, X is a vector of macroeconomic control variables.

(18)

equation 7. Divergence for the structure ratios indicates specialization in either banks or markets.3 I examine whether convergence or divergence is present for financial structures as a whole in equation 7, which is similar to earlier convergence equations:

!(!"#/!")!" = ! + !!(!"#/!")!" + !!!"+ !!!"+ !!+ !!+ !!" (7)

where !(!"#/!")!" is the growth in either the NARROW or BROAD structure measure during a four

year non-overlapping period, whereas the right hand variables are timed at the beginning of this period. To explore whether the world adopts a more capital-market-based system. I estimate time trends in equation 8a – 8c.

(!"#/!")!" = ! + !" + !!+ !!" (8a)

!"#!" = ! + !" + !!+ !!" (8b)

!"!" = ! + !" + !!+ !!" (8c)

More importantly, I am interested whether the ratio will increase as economies grow, by testing equation 9.

(!"#/!")!" = ! + !!!"#!"!!+ !!!"#!"!!! + !!!"!!+ !!+ !!+ !!" (9)

where (!"#/!")!" is either the NARROW structure measure in which SMDit is the level of stock

market capitalization to GDP; BCit is the level of bank credit to GDP or the BROAD structure measure,

in which SMDit is the stock market capitalization and bond market capitalization combined. X is a

vector of macroeconomic control variables. Contrary to prior estimations, I estimate equation 9 without non-overlapping time periods, but use annual data instead.

3_{Divergence (negative coefficient) of the structure ratio implies countries with relatively high initial levels of stock markets}

(19)

IV. RESULTS Stock market convergence

Table 2 shows results for estimating equation 1 with several measures for stock market development. I do not include any control variables here, but use time- and country-specific fixed effects. All three columns show highly significant results for unconditional convergence of stock market development across different countries and time periods, confirming hypothesis 1. The coefficients vary between -0.225 and -0.240 for stock market turnover and value traded respectively. The coefficient for my most important measure, stock market capitalization to GDP, equals -0.236. This means an increase of 20% in initial stock market capitalization is associated with a decline in annual stock market growth of around 4.72 percentage points. This is a moderate effect, since the average annual growth is high, around 26%.

Table 2.

Unconditional stock market convergence

The table reports fixed effects regressions for panel data in the period 1991-2011. Unreported time dummies are used to capture time specific effects. The sample covers the Financial Structure Database (Beck et al., 2009), the World Bank World Development Indicators and World Governance Indicators. The dependent variables are growth variables during a four-year non-overlapping period. Specifications 1 – 3 concern stock market capitalization, stock market value traded and stock market turnover ratio, respectively. The explanatory variables are lagged levels of stock market capitalization, stock market value traded and stock market turnover ratio at the start of the four-year period. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 (1) (2) (3) VARIABLES ST MARKET CAP GROWTH VALUE TRADED GROWTH TURNOVER GROWTH ST MARKET CAP -0.236*** (0.0148) VALUE TRADED -0.240*** (0.0159) TURNOVER -0.225*** (0.0210) Constant 0.770*** 0.328*** 0.612*** (0.0449) (0.0342) (0.0587) Observations 425 418 413 R-squared 0.641 0.604 0.408 Number of countries 104 105 103

(20)

and value traded. GDP PER CAPITA SQUARED is also highly significant. The relation thus shows clear signs of concavity and therefore provides evidence for the medium development hypothesis, i.e. stock markets grow faster at moderate levels of GDP per capita.

For robustness I estimate the same equation using two-step Generalized method of moments (GMM) dynamic panel techniques (from Arellano & Bover, 1995; Blundell & Bond, 1998) in Table 3b. I do this to address endogeneity in the initial level of stock market development and other explanatory variables. The Sargan tests for the validity of instruments show no signs of over identification of the restrictions, except in column 4 where the p-value equals 0.0902. In column 4– 6 of Table 3b I use three lags instead of one lag. Table 3b confirms the convergence results from Table 3a. GDP per capita also shows a positive and significant association with stock market capitalization growth, providing further evidence that stock markets and GDP grow together.

Table 3a.

Stock market convergence: macroeconomic factors and the influence of GDP per capita

(21)

Table 3b.

Stock market convergence: robustness using GMM

The table reports results from Generalized Method of Moments (Arellano-Bover, 1995; Blundell-Bond, 1998) for panel data in the period 1991-2011. The sample covers the Financial Structure Database (Beck et al., 2009) the World Bank World Development Indicators and World Governance Indicators. The dependent variables are growth variables during a four-year non-overlapping period. Specifications 1 – 3 concern stock market capitalization, stock market value traded and stock market turnover ratio, respectively and use one lag. Specifications 4 – 6 use three lags. The explanatory variables are lagged levels of stock market capitalization, stock market value traded and stock market turnover ratio at the start of the four-year period. Macroeconomic control variables are real GDP per capita, trade balance, the gross domestic savings rate and the inflation rate. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

(1) (2) (3) (4) (5) (6) VARIABLES ST MARKET CAP GROWTH (GMM) VALUE TRADED GROWTH (GMM) TURNOVER GROWTH (GMM) ST MARKET CAP GROWTH (GMM) VALUE TRADED GROWTH (GMM) TURNOVER GROWTH (GMM) ST MARKET CAP -0.303*** -0.198*** (0.0347) (0.0452) VALUE TRADED -0.365*** -0.255*** (0.0295) (0.0506) TURNOVER -0.318*** -0.270*** (0.0294) (0.0450) GDP PER CAPITA 0.215*** 0.200 0.212 0.104** 0.0735 0.118 (0.0443) (0.132) (0.136) (0.0479) (0.135) (0.142) TRADE 0.0937 0.0953 -0.166 0.0550 0.0904 -0.222* (0.0838) (0.197) (0.192) (0.0703) (0.193) (0.124) SAVINGS 0.193 1.068 0.837 0.510* 0.582 0.851 (0.191) (1.045) (0.713) (0.278) (0.962) (0.782) INFLATION -0.109** -0.0714 -0.123 -0.0799* -0.00276 -0.0342 (0.0470) (0.0868) (0.105) (0.0423) (0.146) (0.0909) Constant -1.900** -6.731 -4.113 -2.824** -3.451 -3.523 (0.864) (5.173) (3.529) (1.308) (4.798) (3.952) Sargan test 0.3182 0.1328 0.1828 0.0902 0.1930 0.3074 Observations 386 376 373 355 342 337 Number of countries 98 99 97 93 93 90

(22)

Table 4.

Stock market convergence: institutions and financial system efficiency

The table reports fixed effects regressions for panel data in the period 1991-2011 and GMM techniques respectively. The sample covers the Financial Structure Database (Beck et al., 2009), the World Bank World Development Indicators and World Governance Indicators. Specifications 1 – 3 concern stock market capitalization, stock market value traded and stock market turnover ratio, respectively. Unreported time dummies are used to capture time specific effects. The dependent variables are growth variables during a four-year non-overlapping period. The explanatory variables are lagged levels of stock market capitalization, stock market value traded and stock market turnover ratio at the start of the four-year period. Macroeconomic control variables are real GDP per capita, trade balance, the gross domestic savings rate and the inflation rate. Institutional control variables include world governance indicators and financial system efficiency variables. For robustness, specifications 4 – 6 use GMM techniques with three lags, because one lag did not satisfy the Sargan test. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 (1) (2) (3) (4) (5) (6) VARIABLES ST MARKET CAP GROWTH VALUE TRADED GROWTH TURNOVER

(23)

One implication of financial structure evolution is that banks should grow faster at low levels of GDP per capita and stock markets at moderate or higher levels of GDP per capita. In Table 5a, I compare stock market capitalization growth and bank credit growth, which are my most important measures for stock market development and bank development respectively. The results are in line with hypothesis 3 and the theory of financial structure evolution (and also reconfirm hypotheses 1 and 2). Although using a different time period than Bahadir and Valev (2015), columns 3 and 4 confirm their earlier findings on bank credit convergence. Interestingly, stock market growth and bank credit growth show an opposite relationship with GDP per capita. Stock markets seem to grow faster at moderate levels of GDP per capita. However, bank credit growth and GDP per capita show a negative and convex relationship in column 4, because GDP PER CAPITA (β = -0.350) and GDP PER CAPITA SQUARED (β = 0.0299) are significant. Hence, bank credit growth seems to be higher at low levels of GDP and slows down as GDP per capita increases.

Table 5a.

Stock market capitalization and bank credit convergence compared

The table reports fixed effects regressions for panel data in the period 1991-2011. Unreported time dummies are used to capture time specific effects. The sample covers the Financial Structure Database (Beck et al., 2009), the World Bank World Development Indicators and World Governance Indicators. The dependent variables are growth variables during a four-year non-overlapping period. Specifications 1 and 2 concern stock market capitalization. Specifications 3 and 4 concern bank credit. The explanatory variables are lagged levels of stock market capitalization and bank credit at the start of the four-year period. Macroeconomic control variables are real GDP per capita, trade balance, the gross domestic savings rate and the inflation rate. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

(1) (2) (3) (4)

VARIABLES ST MARKET CAP

GROWTH

ST MARKET CAP GROWTH

CREDIT GROWTH CREDIT GROWTH

(24)

Table 5b.

Stock market capitalization and bank credit convergence compared

The table reports fixed effects regressions for panel data in the period 1991-2011. Unreported time dummies are used to capture time specific effects. The sample covers the Financial Structure Database (Beck et al., 2009), the World Bank World Development Indicators and World Governance Indicators. The dependent variables are growth variables during a four-year non-overlapping period. Specifications 1 and 2 concern stock market capitalization. Specifications 3 and 4 concern bank credit. The explanatory variables are lagged levels of stock market capitalization and bank credit at the start of the four-year period. Macroeconomic control variables are real GDP per capita, trade balance, the gross domestic savings rate and the inflation rate. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

(1) (2) (3) (4)

VARIABLES ST MARKET CAP

GROWTH

ST MARKET CAP GROWTH

CREDIT GROWTH CREDIT GROWTH

ST MARKET CAP -0.249*** -0.254*** 0.0385*** 0.0407*** (0.0163) (0.0163) (0.0105) (0.0104) CREDIT -0.0511 -0.0477 -0.159*** -0.160*** (0.0322) (0.0315) (0.0205) (0.0192) GDP PER CAPITA 0.175 1.363*** 0.0497 -0.472*** (0.109) (0.307) (0.0529) (0.180) GDP PER CAPITA SQUARED -0.0762*** (0.0212) 0.0335*** (0.0107) TRADE -0.0940 -0.0892 -0.00874 -0.0108 (0.0679) (0.0615) (0.0388) (0.0407) SAVINGS 0.781** 0.924** -0.00907 -0.0717 (0.348) (0.352) (0.204) (0.197) INFLATION -0.0475* -0.0392 -0.0301 -0.0338 (0.0258) (0.0253) (0.0391) (0.0395) Constant -3.978** -9.195*** 0.259 2.551** (1.820) (2.325) (1.034) (1.278) Observations 385 385 385 385 R-squared 0.666 0.689 0.386 0.418 Number of countries 99 99 99 99

Table 5b adds financial development to examine whether banking levels affect stock market convergence and stock markets levels affect bank credit convergence.4 Stock market convergence is robust to the inclusion of CREDIT. Specifically, in columns 1 and 2, bank credit does not significantly affect the growth in stock markets. Bank credit convergence is also robust to ST MARKET CAP in columns 3 and 4. Stock markets even affect bank credit growth in column 3 and 4, although the impact is small (β = 0.0385 and β = 0.0407). This provides support for theory that states banks credit and stock markets could be complements. Moreover, GDP PER CAPITA (SQUARED) remains highly significant in column 2 with a negative coefficient. Bank credit seems to grow faster at low levels of GDP per capita. Column 4 affirms the convex relation between bank credit growth and GDP per capita, since GDP PER CAPITA SQUARED is positive and significant (β = 0.0335). Bank credit growth slows down if GDP per capita grows. Therefore, results in Table 5b are in line with the medium development hypothesis and financial structure evolution. These results are also in line with Lin et al, (2009) who claim developing countries benefit by starting out using banks.

4

_{In unreported regressions I also included ST MARKET CAP SQUARED and CREDIT SQUARED. Both showed no}

(25)

Table 6.

Financial structure convergence

The table reports fixed effects regressions for panel data in the period 1991-2011 in specifications 1 - 4. Unreported time dummies are used to capture time specific effects. The sample covers the Financial Structure Database (Beck et al., 2009), the World Bank World Development Indicators and World Governance Indicators. The dependent variables are growth variables during a four-year non-overlapping period. Specifications 1 and 2 concern the narrow measure (the log ratio of stock market capitalization to bank credit). Specifications 3 and 4 concern the broad measure (the log ratio of stock market capitalization plus bond market capitalization to bank credit). The explanatory variables are lagged levels of these ratios at the start of the four-year period. Macroeconomic control variables are real GDP per capita, trade balance, the gross domestic savings rate and the inflation rate. Other control variables include governance indicators and accountability from the World Bank. Specification 5 is estimated using GMM with one lag. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 (1) (2) (3) (4) (5) VARIABLES STRUCTURE GROWTH (NARROW) STRUCTURE GROWTH (NARROW) STRUCTURE GROWTH (BROAD) STRUCTURE GROWTH (BROAD) GMM (NARROW) (1 lag) STRUCTURE (NARROW) -0.275*** -0.289*** -0.269*** (0.0174) (0.0163) (0.0397) STRUCTURE (BROAD) -0.188*** (0.0416) -0.232*** (0.0489) GDP PER CAPITA -0.127 2.688*** -0.174 2.416*** 0.291*** (0.135) (0.391) (0.215) (0.693) (0.103) GDP PER CAPITA SQUARED -0.182*** (0.0260) -0.169*** (0.0497) TRADE -0.0132 -0.0140 0.0232 -0.0805 0.0689 (0.0936) (0.0789) (0.144) (0.126) (0.135) SAVINGS 0.627 0.887** 0.563 0.713 0.264 (0.459) (0.432) (1.353) (1.021) (0.300) INFLATION 0.0906** 0.125*** -0.108 0.0444 -0.0575 (0.0432) (0.0400) (0.133) (0.108) (0.0541) GOVERNANCE -0.0271 0.0154 -0.0539 0.0220 -0.0599*** (0.0225) (0.0223) (0.0443) (0.0436) (0.0227) ACCOUNTABILITY -0.0357 -0.0500 -0.0397 -0.0551 -0.101 (0.0553) (0.0473) (0.0965) (0.0678) (0.113) Constant -2.449 -14.37*** -0.768 -10.99* -4.030** (2.514) (2.812) (5.938) (6.198) (1.799) Sargan test 0.6468 Observations 330 330 142 142 314 R-squared 0.637 0.723 0.490 0.616 Number of countries 99 99 42 42 97

(26)

Low initial banking activity (relative to stock markets) is associated with faster bank credit growth whereas low initial levels of capital market activity (relative to banking) show faster growth herein. Furthermore, the results show a positive relation between GDP per capita and the relative growth in stock markets, providing further evidence for the medium development hypothesis. For robustness, I include the estimation by GMM in column 5, which yields similar results.5

A third implication of financial structure evolution involves a growing importance for stock markets as economies develop beyond moderate levels of GDP per capita. Table 7a depicts time trends in financial structures and shows positive and significant coefficients in all specifications. The NARROW measure (ratio of stock market capitalization to bank credit) grows over time (β = 0.0228). Stock markets capitalization (β = 0.0523) and bank credit (β = 0.0295) also grow significantly over time. These results suggest a significant increasing role for stock markets over time, relative to banks, despite that bank credit to GDP also grows over time. These results are in line with hypothesis 5.

Table 7.

Time trend of financial structures

The table reports time trends for panel data in the period 1991-2011, using country fixed effects. The sample covers the Financial Structure Database (Beck et al., 2009), the World Bank World Development Indicators and World Governance Indicators. The dependent variables are growth variables during a four-year period. Specifications 1 concerns the narrow measure (the log ratio of stock market capitalization to bank credit) and specification 2 concerns the broad measure (the log ratio of stock market capitalization plus bond market capitalization to bank credit). Specifications 3 and 4 concern stock market capitalization and bank credit respectively. The explanatory variable is time, measured in years. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

(1) (2) (3) (4)

VARIABLES NARROW BROAD ST MARKET CAP BANK CREDIT

YEAR 0.0228*** 0.0166*** 0.0523*** 0.0295*** (0.00555) (0.00591) (0.00495) (0.00430) Constant -45.99*** -33.10*** -101.3*** -55.30*** (11.12) (11.84) (9.909) (8.610) Observations 1,543 727 1,543 1,543 R-squared 0.055 0.057 0.265 0.237 Number of countries 107 45 107 107

5_{Unfortunately, for the broad measure, there were not many observations when using non-overlapping periods in combination}

(27)

Table 8.

Financial structures and economic development

The table reports fixed effects regressions for panel data in the period 1991-2011. Unreported time dummies are used to capture time specific effects. The sample covers the Financial Structure Database (Beck et al., 2009), the World Bank World Development Indicators and World Governance Indicators. The dependent variables are growth variables during a four-year period. Specifications 1 and 2 concern the narrow measure (the log ratio of stock market capitalization to bank credit) with a sample split according to economic development. Specifications 3 and 4 concern the narrow measure with the full sample. Specifications 5 and 6 concern the narrow measure. For robustness, specifications 7 and 8 concern the broad measure (the log ratio of stock market capitalization plus bond market capitalization to bank credit). The explanatory variables are lagged levels of stock market capitalization and bank credit at the start of the four-year period. Macroeconomic control variables are real GDP per capita, trade balance, the gross domestic savings rate and the inflation rate. Institutional control variables include world governance indicators and financial system efficiency variables. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Sample split6 Full Sample

Full Sample (including governance)

(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES NARROW

LOW-MID

NARROW MID-HIGH

NARROW NARROW NARROW NARROW BROAD BROAD GDP PER CAPITA -1.098** 1.015*** 0.546 3.288*** 1.404*** 7.064*** 1.300*** 5.141*** (0.468) (0.337) (0.415) (1.217) (0.351) (1.811) (0.340) (1.239) GDP PER CAPITA SQUARED -0.178** (0.0758) -0.370*** (0.112) -0.252*** (0.0795) TRADE 0.388* 0.470** 0.394** 0.416** 0.587** 0.656*** -0.0462 -0.0166 (0.228) (0.223) (0.165) (0.174) (0.235) (0.239) (0.245) (0.228) SAVINGS 0.377 1.097 0.585 0.739 0.743 0.905 5.253* 4.911* (1.568) (1.224) (1.060) (1.053) (0.988) (1.074) (2.704) (2.524) INFLATION 0.0937 -0.534*** -0.435*** -0.413*** 0.129 0.0585 -0.468** -0.389* (0.248) (0.168) (0.148) (0.151) (0.208) (0.184) (0.232) (0.222) GOVERNANCE -0.0869 -0.0449 0.136* 0.170** (0.0765) (0.0799) (0.0768) (0.0714) ACCOUNTABILITY 0.441*** 0.334** 0.0131 0.0306 (0.163) (0.153) (0.191) (0.187) INTEREST MARGIN 0.0645 0.0631 0.0760 0.0800 (0.0859) (0.0838) (0.0505) (0.0514) COST -0.0289 -0.0209 -0.0423 -0.0148 (0.0938) (0.104) (0.0513) (0.0538) RETURN ON ASSETS -0.00299 (0.0421) (0.0437) -0.0395 -0.00642 (0.0187) (0.0184) -0.0303 Constant 1.580 -16.03** -8.934 -20.01*** -19.33*** -40.57*** -38.00*** -50.09*** (7.696) (7.302) (5.659) (7.416) (6.093) (9.957) (11.70) (13.14) Observations 470 1,331 1,801 1,801 987 987 429 429 R-squared 0.338 0.355 0.276 0.291 0.234 0.281 0.496 0.532 Number of countries 34 77 111 111 105 105 45 45

Table 8 indicates what factors are associated with the financial structure measure. A higher measure indicates a higher emphasis on capital markets relative to banks. Table 8 reports supportive results for financial structure evolution, i.e. the relation between GDP per capita and financial structures. In column 1, the sample split6 depicts a significant negative relation between STRUCTURE and GDP PER CAPITA for low to medium developed economies. Contrarily, in column 2 the relation is positive and

6_{The sample is split according to levels of economic development. I do not split the group 50/50, because the dataset contains}

(28)

significant for medium to high-developed economies. Hence, banks are relatively important early in the process of economic development and stock markets become important later on. The full sample shows insignificant results for GDP PER CAPITA in column 3, but when I include GDP PER CAPITA SQUARED it reveals a concave relationship. Higher GDP per capita is highly significantly related to a higher emphasis on stock markets relative to banks, particularly at moderate levels of GDP per capita and thus confirm hypothesis 5. The results are robust to inclusion of institutional control variables in columns 5 – 8 and illustrate similar patterns for both the narrow and broad measures (the broad measure includes bond markets in the definition of capital markets).

(29)

V. CONCLUSION

For a variety of countries, I present evidence that stock market capitalization to GDP, and other measures of stock market development, converge over time. Countries with lower initial levels of stock market development exhibit stronger growth in its subsequent stock market development. These findings are robust to different estimation methods and control variables. As economic development increases, both stock markets and banks develop, but stock markets seem to grow particularly at moderate levels of GDP per capita whereas bank credit grows stronger at lower levels. This suggests that the comparative importance of banks and stock markets changes as countries move through different stages of economic development and choose the appropriate financial structure accordingly. Furthermore, as countries grow richer, stock markets seem to become more relevant. On the other hand, I find no evidence that stock markets and banks are substitutes or that some kind of specialization occurs. Rather, I find that financial structures converge in the sense that countries with relatively low banking or stock market activity start to grow faster in their use of banking or stock market activity respectively, possibly in order to catch up.

These findings are consistent with the view that (1) financial institutions are complements rather than substitutes; (2) and they provide different financial services; (3) with a stronger emphasis on capital markets as economies develop. It is more useful to distinguish countries by overall financial structure than to characterize economies as bank or market-based, because countries with better-developed stock markets also have better-developed banks and non-bank financial intermediaries (Huybens & Smith, 1999; Demirgüç-Kunt & Levine 1996; 2001). My results support this reasoning.

Therefore, this research recommends that policies on the development of financial institutions take the evolution of these structures into account. In other words, it seems unnecessary for countries to specialize in either banks or capital markets. Developing economies are likely better off when they focus on banks and switch the emphasis to stock market development in later stages of their economic development.

VI. DISCUSSION

(30)

development of financial intermediation system profitable where the establishment of an efficient financial system permits faster economic growth. This is in line with my results as both show they grow together. However, if initial GDP per capita is caused by stock market growth through reverse causality, the medium development hypothesis may be untrue in the sense that it is not necessarily higher levels of economic development that make stock markets grow, but some other correlating variable. I believe this to be unlikely, also given my use of GMM techniques.

My sample contains data only from 1990 onwards and is reduced when I introduce control variables. Understandably, especially developing countries have fewer data available. This yields a limited perspective and biases my results towards more developed nations. Perhaps most countries in the dataset have already reached the medium development. I try to address this issue by a sample split. Future datasets containing developing countries over a longer period can hopefully shed more light on how economic development and financial structures interact at earlier stages of development.

Regarding financial structures as a whole, I find convergence in the sense that countries with relatively low stock market (banking) activity start to grow in the use of stock market (banking) activity. I am unable to distinguish whether this effect is due to low levels of banking or low levels of stock market capitalization or that both factors drive this effect. Although the results do not point in the direction of substitutes, I am unable to conclude complementarity. Results from Table 5b give only a small indication, because credit growth is still present at high initial levels of stock market capitalization. Moreover, the time trends in Table 7 indicate that both banks and stock markets grow over time, not just one or the other. However, stock markets simply seem to grow faster. To further assess financial structure evolution, future research could take the form of testing complementarity while taking a evolutionary role of banks and markets into account.

(31)

VII. REFERENCES

Aghion, P., Howitt, P., & Mayer-Foulkes, D. (2005). The effect of financial development on convergence: Theory and evidence. The Quarterly Journal of Economics, 120(1), 173–222. Allen, F., & Gale, D. (1997). Financial markets, intermediaries, and intertemporal smoothing. Journal

of political Economy, 105(3), 523-546.

Antzoulatos, A. A., Panopoulou, E., & Tsoumas, C. (2011). Do financial systems converge?. Review of International Economics, 19(1), 122-136.

Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of error-components models. Journal of econometrics, 68(1), 29-51.

Baele, L. (2005). Volatility spillover effects in European equity markets. Journal of Financial and Quantitative Analysis, 40(02), 373-401.

Bahadir, B., & Valev, N. (2015). Financial development convergence. Journal of Banking & Finance, 56, 61-71.

Baldwin, R. E. (1997). The causes of regionalism. The World Economy, 20(7), 865-888.

Beck, T., & Levine, R. (2002). Industry growth and capital allocation:: does having a market-or bank-based system matter?. Journal of Financial Economics, 64(2), 147-180.

Beck, T., & Levine, R. (2004). Stock markets, banks, and growth: Panel evidence. Journal of Banking & Finance, 28(3), 423-442.

Beck, T., Demirgüç-Kunt, A., & Levine, R. (2007). Finance, inequality and the poor. Journal of economic growth, 12(1), 27-49.

Beck, T., Demirgüç-Kunt, A., & Levine, R. (2009). Financial institutions and markets across countries and over time-data and analysis. World Bank Policy Research Working Paper Series, Vol. Beck, T., Levine, R., & Loayza, N. (2000). Finance and the Sources of Growth. Journal of financial

economics, 58(1), 261-300.

Bianco, M., Gerali, A. and R. Massaro (1997), “Financial Systems Across ‘Developed Economies’: Convergence or Path Dependence?”, Research in Economics, 51, pp. 303-331.

Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of econometrics, 87(1), 115-143.

Blundell, R., and S. Bond. 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87: 115–143.

Boot, A. W., & Thakor, A. V. (2000). Can relationship banking survive competition?. The journal of Finance, 55(2), 679-713.

Boyd, J. H., & Smith, B. D. (1998). The evolution of debt and equity markets in economic development. Economic Theory, 12(3), 519-560.

Boyd, J. H., Levine, R., & Smith, B. D. (2001). The impact of inflation on financial sector performance. Journal of monetary Economics, 47(2), 221-248.

Bruno, G., De Bonis, R., & Silvestrini, A. (2012). Do financial systems converge? New evidence from financial assets in OECD countries. Journal of Comparative Economics, 40(1), 141-155.