Global financial integration – The effect of foreign capital on financial system development and efficiency in developing and emerging countries

(1)

Global financial integration – The effect of foreign

capital on financial system development and efficiency

in developing and emerging countries

M.Sc. / M.A. Thesis

Supervisors

Dr. D.J. Bezemer (University of Groningen)

Prof. Dr. Tino Berger (University of Göttingen)

Author

Johannes Kinzinger

S2802813 – 21362012

j.p.kinzinger@student.rug.nl

(2)

Abstract

The vast literature regarding the impact of financial globalization on economic growth in developing countries has not been decisive in finding a robust positive relationship. However, it is claimed that a higher integration into global financial markets may bring indirect “collateral benefits” such as financial sector development. The main goal of this thesis is to examine whether or not foreign capital flows into developing countries act as a catalyst for the development of their respective domestic financial systems and lead to a better financial intermediation between savers and borrowers. Using longitudinal data from 86 developing and emerging countries for the time frame between 1980 and 2011 reveals that foreign capital leads to advances in the financial sector development of developing countries. However, it appears to be important to distinguish between different kinds of capital flows, as their impact on the measures of financial development is diversified. In particular, portfolio equity inflows may bring the largest benefits whereas debt inflows might have a detrimental effect in the long run. Moreover, it is shown that developing countries are able to reap the benefits at low levels of initial financial depth.

(3)

List of Illustrations

Figure 1: Quantity and Quality Effects of financial liberalization ... 9

Figure 2: Possible Quality Effects of Financial liberalization ... 10

Figure 3: Possible Quantity Effects of Financial liberalization ... 11

Figure 4: Possible uses of savings ... 18

Figure 5: Leverage vs. squared residuals in the dataset ... 20

Figure 6: Mean of capital inflows per type + Gross Foreign Assets and Liabilities ... 26

Figure 7: Measures of finanancial system efficiency ... 28

Figure 8: Marginal effect of PEI Inflows on ln(bank credit/bank deposits) depending on the level of financial depth (measured by ln(private credit to GDP)) ... 40

Figure 9: Marginal effect of FDI Inflows on ln(bank credit/bank deposits) depending on the level of financial depth (measured by ln(private credit to GDP)) ... 40

Figure 10: Marginal effect of FDI Inflows on ln(bank credit/bank deposits) depending on the level of financial depth (measured by ln(private credit to GDP)) – left: below median; right above median ... 41

Figure 11: Marginal effect of FDI Inflows (left) and PEI Inflows (right) on ln(savings/private credit) depending on the level of financial depth (measured by ln(private credit to GDP)) ... 42

Figure 12: Marginal effect of Debt Inflows on ln(liquid assets/total deposits and short-term funding) depending on the level of financial depth (measured by ln(private credit to GDP)) ... 43

Figure 13: Marginal effect of Debt Inflows on ln(deposit money bank assets (% GDP)) depending on the level of financial depth (measured by ln(private credit to GDP)) ... 44

(5)

List of Tables

Table 1: Literature Review – Financial Liberalization and domestic credit

expansion/access to credit ... 8

Table 2: Descriptive Statistics ... 29

Table 3: Regression results – Dependent Variable: log (Bank Credit/Bank Deposits) ... 30

Table 4: Regression results – Dependent Variable: log (Savings/Private Credit) ... 32

Table 5: Regression results – Dependent Variable: log (Liquid assets to deposits and short term funding (%)) ... 33

Table 6: Regression results – Dependent Variable: log (Deposit Money Bank Assets (% GDP)) ... 35

Table 7: Regression results – Dependent Variable: log (Interest Rate Spread) ... 36

Table 8: Summary of Findings ... 47

Table 9: Results Interaction Model 1: Dependent variable: log(bank credit/bank deposits) ... 67

Table 10: Results Interaction Model 2: Dependent variable: log(savings/private credit) ... 68

Table 11: Results Interaction Model 3: Dependent variable: log(liquid assets/total deposits &s.t. funding) ... 69

Table 12: Results Interaction Model 4: Dependent variable: log(deposit money bank assets (% GDP)) ... 70

(6)

Acronyms

2SLS Two Stage Least Squares

BOP Balance Of Payments

CA Current account

DF Debt Flows

FDI Foreign Direct Investment

FPI Foreign Portfolio Investment

ICF Incoming Capital Flows

IIP International Investment Position

IMF International Monetary Fund

Iqual Institutional Quality

IV Instrumental Variable

GDP Gross Domestic Product

GFAL Gross Foreign Assets and Liabilities (% GDP)

GFDD Global Financial Development Database (World Bank)

GMM Generalized Method of Moments

GNS Gross National Savings (% GDP)

FDSD Financial Development and Structure Database

NAFA Net Acquistion of Foreign Assets

OLS Ordinary Least Squares

PEI Portfolio Equity Investment

PPP Purchasing Power Parity

TI Total Investment (% GDP)

(7)

1 Introduction

In the past decades since the 1980s, financial liberalization has led to a rapid integration of developing economies into international financial markets. These widespread financial liberalization efforts in developing countries were supposed to lead to more efficient financial markets, increased access to credit for domestic firms, more access to international capital and – as a result - lower costs of capital as predicted by the standard economic textbook theory (Kose et al., 2009). The benefits of liberalizing the financial sector and capital account were expected to be large, especially for scarcely capital endowed developing countries. The neo-classical Solow-Model identifies a lack of savings as the main obstacle for economic growth and predicts that more capital flows to developing countries leads to more investments and higher GDP growth. In light of this assumption, higher integration of developing countries into the international financial markets would predict more supply of capital for banks and increasing firm profits. Hence, more firms would gain access to credit as lending activity is expected to increase, which is supposed to lead to higher investment rates and long-run economic growth.

Has this forecast proved to be valid in the past? A meta-analysis of the empirical literature on the relationship between financial liberalization and economic growth covering 60 empirical studies by Bumann et al. (2013) concluded that, on average, there is a positive effect of financial liberalization on growth, but the significance of this effect is only weak. Hence, it appears to be questionable whether or not financial liberalization fosters investment rates in developing countries and the traditional textbook savings-investment nexus has to be reviewed critically. Moreover, a series of studies (Freedman & Click, 2006; Prasad, 2007; Rodrik & Subramanian, 2009) have shown that developing countries seem to be more investment-constrained than savings-constrained as proposed by neoclassical theory. According to Prasad (2007, p.16), “investment does not seem to be highly correlated with net capital inflows, suggesting that it is not constrained by lack of resources”. The underlying problem, why developing countries do not directly benefit in terms of economic growth rates from financial liberalization, may lie in the difficulty to transform savings into investment or the ability to enable investments via credit expansion.

Confirming Bumann et al. (2013), a study by Kose et al. (2009) points out that the majority of empirical studies are unable to find robust evidence supporting the growth benefits of financial liberalization. However, the authors state that the benefits of financial liberalization may be realized through a set of “collateral benefits” which include financial market development, better institutions and governance and macroeconomic discipline.

(8)

Calderon & Kubota, 2009; Baltagi et al., 2009; Ahmed, 2012). In the majority of the studies, financial system development is typically measured by the share of private credit to total GDP. However, this variable only partly captures the full extent of financial development, as it does not actually take into account the financial endowments. Based on these findings, the question to be asked is, whether financial liberalization can contribute to higher financial sector efficiency in terms of financial intermediation. Does financial liberalization commit to more supply of loans with respect to the financial endowments?

This paper aims to analyze the impact of financial liberalization efforts on financial sector development in developing countries, focusing on the efficiency of their financial systems. Does global financial integration help to loosen financial constraints in developing countries? Do increasing volumes of capital flows contribute to growth in domestic credit relative to deposits? Or do they end up in liquid assets on banks’ balance sheets such as short-term government bonds or central bank debt, thereby confirming the presumption of an existing investment-constraint? Furthermore, this reseach aims to shed light on the question whether or not there are certain thresholds of financial depth that have to be met before financial liberalization efforts in developing countries have a positive effect on domestic financial sector development.

The contribution of this thesis is particularly based on the use of a large dataset incorporating 86 developing economies from all continents covering a wide time span from 1980 to 2011. Moreover, a variety of measures of financial system development are examined in order to account for the multidimensional construct of financial sectors. Applying a fixed effects panel data analysis, plus a two-stage instrumental variable generalized method of moments (IV-GMM) approach allows to draw conclusions that are robust to potential endogeneity.

The empirical results of this paper show that developing countries may benefit from foreign capital in terms of development of their financial systems. Moreover, the benefits from foreign equity inflows might be reaped already at low levels of initial financial depth. In contrast, debt inflows appear to have adverse effects at low levels of financial depth. In general, equity-like inflows suchs as portfolio equity investments seem to have a more positive effect than debt flows. However, developing countries are most likely not able to fully abstain from debt inflows since it might be the only source of foreign capital.

(9)

2 Literature Review

Proponents of financial liberalization argue that a greater integration of international financial markets – especially a liberalization of the capital account – fosters financial market development and ultimately economic growth. However, in the recent decade a large body of literature opposed the positive views on the financial liberalization efforts in developing countries. Hence, this chapter summarizes the current literature on the effects of financial globalization with respect to overall economic growth and financial system development in particular.

2.1 Financial liberalization and economic growth

The relation between financial liberalization and economic growth has been researched extensively, showing inconclusive results. A meta-analysis of the existing literature regarding this relationship by Bumann et al. (2013) concludes that, on average, there is a positive effect of financial liberalization on growth but the significance of that effect is only weak. Bailliu (2000) shows that capital inflows foster higher economic growth, but only in economies having reached a certain threshold of financial development. Tornell & Westermann (2004) note that financial liberalization leads to higher long-term GDP growth in developing countries with the byproduct of financial fragility.

By applying an empirical analysis, Prasad et al. (2007) conclude that non-industrial countries that have relied more on foreign finance have not grown faster in the long term than countries not relying on foreign capital. They opine that developing countries lack the ability to absorb foreign capital and allocate it to productive uses while the exchange rate tends to appreciate. In a theoretical paper, Rodrik & Subramanian (2009) argue that financial globalization has not generated higher growth rates in emerging markets since foreign capital inflows have adverse effects on the exchange rate. Furthermore, countries that are relying less on foreign capital flows have shown the best growth performances.

Kose et al. (2009) emphasize that the composition of capital flows has a significant influence on growth. Agbloyor et al. (2014) focus on the relation between private capital flows and economic growth in Africa between 1990 and 2007, concluding that private debt flows have an overall negative impact on growth. Only countries with strong initial domestic financial markets observe positive effects on growth. According to Kose et al. (2009), the empirical literature has reached consensus on the fact that debt flows generate the greatest risks from financial openness. In contrast, both foreign direct investment and portfolio equity flows have shown to be more growth enhancing, although the empirical results are still divergent.

(10)

financial systems are underdeveloped. Thus, the relation between financial liberalization and financial system development has to be examined in detail.

2.2 Financial liberalization and financial market development

The large body of literature on the relation between financial liberalization and economic growth highlights the importance of financial sector development in realizing the possible economic benefits. Kose et al. (2009) state that financial integration serves as a “catalyst” for a number of indirect benefits such as the development of the financial sector in developing countries. According to Schmukler (2004), financial globalization can improve the functioning of the financial system in two ways. First, financial globalization can increase the availability of funds. Second, it can improve the infrastructure of the financial system and enhance its efficiency. However, as several studies have shown, the majority of developing countries appear to be more investment-constrained than savings-constrained (Rodrik & Subramanian, 2009; Prasad, 2007). Hence, the relation between financial globalization and the efficiency of financial systems should be emphasized. Following Levine (2005), a financial system has five basic functions:

1) Produce information ex-ante about possible investments and allocate capital 2) Monitor investments & exert corporate governance after providing finance 3) Facilitate the trading, diversification and management of risks

4) Mobilize and pool savings

5) Ease the exchange of goods and services

Financial globalization has an impact on each of these basic functions. First, financial globalization may distort private and privileged information in financial markets as a result of confrontation and demands of new external economic agents (Tovar García, 2012). In particular, the competition between existing and new economic agents may produce better and more consistent information about possible investment opportunities.

Second, financial globalization may improve the monitoring of investments through the introduction of new technologies. Moreover, best practices in financial supervision may spread to developing countries. Schmukler (2004) states that the adoption of international accounting standards can help to improve domestic regulatory and supervisory frameworks.

Third, the integration of developing countries’ financial markets into global markets favors risk diversification. According to Mishkin (2001), foreign banks usually have a well-diversified portfolio, which means that they are less exposed to risk and hence less affected by negative shocks. Domestic economic agents can share risks with foreign agents and build diversified portfolios with significantly lower risk.

(11)

substitute for domestic savings in developing countries. However, Rodrik & Subramanian (2009) argue in the opposite direction stating that external savings will most likely substitute for domestic savings in investment-constrained countries. This phenomenon is confirmed by a study of Rashid (2011). Fifth, financial liberalization is supposed to reduce transaction costs and increase the exchange of goods and services (Tovar García, 2012).

In general, financial liberalization is expected to correct financial market imperfections through the market determination of all institutional interest rates (Ahmed, 2012). This will enhance interaction among savers and borrowers, increase incentives for bank savings and promote the introduction of new financial instruments and risk-sharing possibilities. Moreover, increasing competition in the market is supposed to enforce financial sector efficiency. Rajan and Zingales (2003) argue that powerful interest groups are inhibiting financial development especially in developing countries since better-developed financial markets create opportunities for new firms to enter the market, which might erode profits from the established firms. Thus, the authors opine that financial openness will limit the ability of “incumbents” to block the development of financial sectors.

Empirical findings on the relation between financial liberalization and financial sector development

Calderon & Kubota (2009) examine the relationship between financial openness and financial development in terms of private sector credit-to-GDP for a sample of 145 countries in the time period 1974 – 2007. Their main finding is that rising financial openness leads to deeper local capital markets in terms of expansion in private credit. The effects are highest in countries with high initial levels of institutional quality. Using data on 21 Sub-Saharan Countries between 1981 and 2009, Ahmed (2012) finds that financial liberalization has a positive impact on financial deepening and resource mobilization in this region while controlling for variables such as institutional quality and inflation. Baltagi et al. (2009) investigate the impact of financial openness, measured by gross foreign assets and liabilities as a share of GDP, on banking sector development (private credit to GDP) for a set of 42 developing countries. The authors find that financial openness is a significant determinant of banking sector development in developing countries. Furthermore, the least open economies may benefit the most in terms of financial sector development by opening both their capital and trade accounts. However, there is a decreasing marginal effect of openness with regard to banking sector development.

(12)

financial globalization has no significant relationship with growth and development of the financial system. Hence, the evidence regarding the relationship between financial liberalization and improvements in financial sector development appears to be inconsistent.

Another important issue concerning this relation is the question to what extent financial systems in developing countries have to be developed before being integrated into international financial markets? According to Kose et al. (2009), there is evidence that opening the capital account without having established well-developed financial sectors can have detrimental effects by making the structure of capital inflows unfavorable. Furthermore, this might increase the vulnerability to sudden stops and reversals of capital flows.

The argument that financial liberalization can significantly contribute to financial system development appears to be logically inconsistent, as many studies emphasize the importance of functional financial systems as a prerequisite for opening the domestic financial markets. Rodrik & Subramanian (2009) doubt the theoretical consistency and practical feasibility of the preceding reforms needed to support financial globalization in developing countries. Kose et al. (2009) suggest that there are ways to improve the benefit-risk calculus, but it is unlikely that there is a uniform approach for liberalizing financial markets in developing countries. Thus, the issue of threshold conditions for financial market development has to be further examined in this research.

2.3 Financial liberalization and domestic credit expansion

The relation between financial liberalization and financial sector development has to be investigated from another point of view: domestic credit growth. Does financial liberalization lead to higher financial intermediation between savers and borrowers and hence to more credit expansion in the economy?

Several studies emphasize access to credit as a major growth constraint for firms in developing counties, especially for small and medium sized enterprises (Beck & Demirguc-Kunt, 2006; Beck & de la Torre, 2006). However, it is important to mention that measuring the private sector’s access to credit is extremely difficult and often only possible with firm-level surveys, which are rarely available for developing countries. In the available body of literature, credit provided to the private sector with respect to GDP is often the main dependent variable of interest. However, this variable does not distinguish between small, medium and large enterprises. Therefore, it is difficult to make conclusions about increasing financing opportunities for firms that have been excluded from the financial sector before liberalizing financial markets based on this variable.

(13)

to credit in developing countries concluding that foreign bank penetration improves firms’ access to credit, including small- and medium sized firms (SMEs). However, Rashid (2011) shows that foreign banks tend to allocate less of their assets and deposits to lending by using bank-level data from 81 developing countries. Moreover, foreign bank entrance may decrease the deposit base of domestic banks thereby reducing their lending activities. Knill and Lee (2014) assess the impact of foreign portfolio investment (FPI) volatility on access to finance of small listed firms. They find that volatility of FPI is significantly related with a decreased access to finance only when nations are closer to crises. However, even in times of crises the positive impact of the level of FPI on small firms’ access to finance surpasses the negative effects of volatility of FPI. By using firm-level data of 57 developing and transition countries, O’Toole (2012) finds that financial liberalization reduces the probability of being credit constrained by 5 to 20 percent depending on the firm’s initial position.

Several newly published studies examine the effect of international capital flows on domestic credit growth. Furceri et al. (2012) analyze the effect of capital inflow surges on the evolution of domestic credit growth in 112 developing and emerging economies between 1970 and 2007. They find that in the two years following the capital inflow surge, the credit-to-GDP ratio increases by about two percentage points. However, after seven years the effect is reversed and the credit-to-GDP ratio decreased by almost seven percentage points. Arslan & Taskin (2014) show a significant positive relationship between capital flows and domestic credit expansion with the effect being strongest in upper-middle income countries. Similarly, Lane & McQuade (2013) conclude that domestic credit growth in European countries is strongly related to debt inflows but not equity inflows. Moreover, they state that there is both a direct and an indirect relation between international debt flows and domestic credit growth. The direct relation works, for instance, through the international funding activities of banks while the indirect relation means that debt flows have an impact on domestic macro and financial variables. In turn, they can affect both supply and demand factors influencing domestic credit expansion. Arena et al. (2015) find that surges in capital inflows are associated with credit booms under less flexible exchange rate regimes using a database on bank credit, covering 135 developing countries in the period 1960-2011. In line with these findings, Frost and van Tilburg (2014) emphasize the significant effect of capital flows on domestic credit excesses and banking crises. Magud et al. (2014) analyze the impact of exchange rate flexibility on credit markets during periods of large capital inflows while focusing on emerging markets. The authors find that relatively inflexible exchange rate regimes are positively associated with growth in private credit-to-GDP.

(14)

potential side effects on macroeconomic variables influencing both supply and demand factors of domestic credit. These effects have to be taken into account by developing countries when considering the liberalization of financial markets.

Table 1: Literature Review – Financial Liberalization and domestic credit expansion/access to credit Study Number of Countrie s/Time Period Dependent Variable/Regression Methodology Financial Liberalization Measure Type of data Main Findings Clarke et al. (2001)

38 / 1999 Enterprises perceptions about interest rates and long-term loans

Tobit/ Ordered Probit Models

Foreign Bank Assets

Firm- Level-Data

POSITIVE EFFECT of foreign bank penetration on firms’ access to credit Kalderon & Cubota (2009) 145 / 1974-2007

Private credit by deposit money banks/GDP

Panel Data (OLS, IV) Models

Foreign Assets + Liabilities/GDP

Macro-data

POSITIVE EFFECT of higher integration in world capital markets on the development of domestic financial markets Rashid

(2011)

81 / 1995

– 2009

Loans to Total Assets; Growth in Deposit Share; Volatility in Loans to Total Assets; Interest Rate Spread; Credit to Private Sector/GDP OLS/GLS/System GMM Panel Ratio of foreign Banks, deposit share of foreign banks Bank- Level-Data

NEGATIVE EFFECT of foreign bank presence on loans to total Assets; Substitution of domestic for foreign savings

O’Toole (2012)

57 /

2005-2009

Probability of being credit constrained

Standard Probit Model (Cross-Section) Financial Liberalization index Firm-level survey data

POSITIVE EFFECT of financial liberalization on access to finance for SMEs Furceri et al. (2012) 112 / 1970 – 2009

Domestic Credit to private Sector/GDP Impulse-Response-Functions Capital Inflows (aggregate, FDI, PEI, Debt) Macro-Data

MIXED EFFECT; short term: Positive relation between Capital Inflows and Credit/GDP; medium-term: negative effect on Credit/GDP Lane & McQuade (2013) 29 / 1993 – 2008

Credit to private sector/GDP

OLS/IV Panel Regression

Int. Capital Flows (equity and debt flows)

Macro-data

MIXED EFFECT – Strong effect of debt flows on domestic credit growth/no effect of equity flows Knill &

Lee (2014)

39 / 1996

– 2005

Probability of Capital Issuance

Probit Model

FPI Volume and Volatility

Firm-level data

POSITIVE EFFECT of foreign portfolio flows on small firms’ access to credit Frost & van Tilburg (2014) 43 / 1975 - 2011 Change in Credit to GDP; Deviation of Credit to GDP from trend; Banking Crisis

Multivariate Regression Models

Current account balance, Gross Capital Inflows

Macro-Data

MIXED EFFECT of foreign capital flows; contributing to both domestic credit excesses and banking crises Arslan & Taskin (2014) 101 / 1970 – 2009

Real private credit growth; Private Credit to GDP

Fixed Effects/GMM estimations

Net Capital Flows; Capital Openness Index

Macro-Data

POSITIVE EFFECT of capital flows on domestic credit growth, especially in upper-middle income countries Magud et al. (2014) 25 / 1993 – 2008 Domestic Credit/GDP; Domestic Credit in Foreign Currency/Total Domestic Credit

OLS/GLS/IV Estimations

Capital Inflows Macro-Data

POSITIVE EFFECT of capital inflows on bank credit growth; countries with less flexible exchange regimes attract more inflows Arena et al. (2015) 135 / 1960-2011

Dummy variable indicating credit boom periods

Panel Probit Methods

Capital Inflows Macro-Data

(15)

2.4 Theoretical framework

As acknowledged by Abiad et al. (2008), financial liberalization is typically expected to bring two particular benefits: quality effects, manifested in a more efficient allocation of capital; and quantity effects, reflected by higher volumes of savings and investment. Figure 1 illustrates the framework for both channels.

Figure 1: Quantity and Quality Effects of financial liberalization

2.4.1 The quality effects of financial liberalization

The quality effect implies that financial liberalization depicted by foreign capital inflows may contribute to the development of financial systems in developing countries through various channels. The three main transmission channels are presented in Figure 2. First, it is expected that financial liberalization leads to a decreasing spread between lending and deposit rates as interest rates become determined by the international markets. Moreover, financial liberalization contributes to a better diversification of risk, which reduces cost of offering loans.

(16)

countries.1_{O’Toole (2012) finds that financial liberalization may increase the financing constraints}

of small firms and firms whose creditworthiness is difficult to assess. Moreover, financial liberalization may lead to excessive risk-taking by banks and a higher fragility of financial intermediaries. These effects can be viewed as negative quality effects of financial liberalization.

Figure 2: Possible Quality Effects of Financial Liberalization

2.4.2 The quantity effects of financial liberalization

Besides the quality effects of financial liberalization it is expected that financial liberalization leads to more savings and investments due to higher deposit interest rates through the removal of interest rate ceilings (McKinnon, 1973; Shaw, 1973). This is the typical economic textbook assumption, which suggests that developing countries lack sufficient capital resources in order to finance investments.

Figure 3 outlines the possible impact of foreign capital on both savings and investment rates in developing countries. Channel 1 shows the typical assumption of neocleassical theory that foreign capital inflows are supposed to increase the availability of external financing opportunities. Thereby, loan interest rates are reduced and more bank lending activities are enabled, which is assumed to result in higher domestic investment rates.

Channel 2 states that foreign capital flows bring in physical capital but also know-how and technological enhancements. This will most likely lead to higher firm incomes and increasing profits. Hence, firms are demanding more credits in order to expand their businesses. Moreover, the firms gain in collateral and banks are willing to provide credit as the problems of moral hazard and adverse selection are mitigated (Alhassan et al., 2013).

However, studies have shown that banks in developing countries are highly liquid (Freedman & Click, 2006) and hence, a lack of capital is not a major constraint for increased bank lending. This is in line with the argumentation of Rodrik & Subramanian (2009) who claim that most developing

1_{Relationship lending refers to banks that engage in multiple interactions with a borrower and thereby invest in}

(17)

countries are rather investment-constrained than savings-constrained. This may be manfiested through a lack of access to credit based on asymmetric information or low perceived returns of investment projects. In this case, foreign savings may substitute for domestic savings. This is outlined by channel 3 in Figure 3. The supply of credit by banks to private sector companies is not enhanced. Typically, this results in boosts of consumption but with no effect on real investment rates.

Furthermore, in investment-constrained economies it is most likely that the additional resources are used to buy financial assets or real estate (Channel 4). This is reflected by large liquidity ratios of banks in developing countries (Freedman & Click, 2006). However, the demand pressure of foreign capital flows may not be offset by a higher supply of goods and services, which will ultimately end in asset price booms and is confirmed by Mendoza & Terrones (2013).

Figure 3: Possible Quantity Effects of Financial Liberalization

Another macroeconomic variable affected by foreign capital inflows is the exchange rate, which is shown by channel 5. Rodrik & Subramanian (2009) underline this effect. Improved domestic financial intermediation tends to depreciate the exchange rate by closing the ex ante gap between desired investments and savings. However, foreign capital has the opposite effect on the exchange rate – it appreciates it. In turn, this leads to lower investment demand by the tradables industry and slows down the growth rate. Several studies have delivered evidence for foreign-capital induced exchange rate appreciation (Mendoza & Terrones, 2013; Magud et al., 2014).

(18)

liberalization affects domestic saving and total investments, which confirms the assumption that foreign savings substitute for domestic savings in developing countries. However, Hermes & Lensink find a positive relation between financial liberalization and private investment and a negative relation regarding public investments. This may indicate that financial liberalization induces a substitution from public to private investments.

Moreover, based on the works of Keynes & Schumpeter, several economists argue that causation does not run from savings to investments but in reverse direction (Dullien, 2009). According to this view, autonomous increases in investment may create savings, as banks are able to create credits ex nihilo (“out of nothing”)2. This may provide a feasible pathway for developing countries to increased domestic investment and economic growth rates without prior savings or foreign capital. Dullien (2009) argues that foreign borrowing by banks increases the vulnerability of the financial sector to exchange rate fluctuations, which limits the ability of central banks to support a domestic credit-financed investment expansion. Furthermore, “dollarization” of economies in developing countries should be prevented, as central banks are no longer able to supply base money in order to accommodate credit processes3. Both, foreign borrowing and increasing dollarization are typically results from financial liberalization reforms in developing countries, which may have adverse effects on investment rates.

However, according to Dullien (2009), a functioning financial system has to be in place in order for the Schumpeterian-Keynesian investment-savings nexus to work. This brings us back to the initial question of this research: How does the development and efficiency of financial systems in developing countries change due to financial liberalization?

2.5 Financial liberalization measures and their impact on financial system

development

In order to estimate the effects of financial globalization on financial development in developing countries, this thesis is relying on ‘de facto’ measures of financial globalization.

Several studies looking at the effects of financial liberalization have relied on the use of ‘de jure’ measures of capital account openness, which reflect legal regulations rather than actual visible effects. Kose et al. (2009) emphasize that a distinction between ‘de jure’ and ‘de facto’ measures is important, as many countries have strict capital controls on paper but still experience high ‘de facto’ levels of integration – in form of capital inflows or stocks of foreign assets and liabilities. Hence, this thesis is relying on ‘de facto’ measures such as aggregated and disaggregated capital inflows and overall stocks of foreign assets and liabilities with respect to GDP, as recommended

2_{The process of ex nihilo credit creation is explained in detail in Dullien (2009).}

(19)

by Kose et al. (2009). The existing literature recommends looking at disaggregated capital flows since foreign direct investment (FDI), portfolio equity investment (PEI) and debt flows may have a different impact on financial development.

In order to measure financial system development, it is typically necessary to take various dimensions into account. The main variables of interest used in this research are savings/credit and bank credit/bank deposits, which indicate the efficiency of intermediating savings or deposits into loans. Moreover, the liquidity ratio (liquid assets/total deposits and short-term funding), lending-deposit rate spread and lending-deposit money bank assets (% GDP) will be used. The rationale of each variable is explained in chapter 3.2.1.

2.5.1 Foreign Direct Investments (FDI)

Kose et al. (2009) note that capital flows with “equity-like” features i.e. FDI and portfolio equity flows are more stable, less prone to reversals and also provide transfers of managerial and technological advancements. Hence, it is expected to increase firms’ income and profitability, which should ultimately lead to higher demand for credits and a higher willingness of banks to extend credits. Using a data set for the Czech Republic between 1994 and 2003, a study by Manole & Spatareanu (2014) examines the relationship between technological spillovers from FDI and firm’s access to external finance. The authors find that firms that have access to external finance benefit more from FDI in their own industry through increased productivity. Moreover, their results suggest that well-developed financial markets may be needed to fully benefit from FDI inflows. Tornell & Westermann (2004) note that the greater the share of inflows, in form of FDI, and the lower the share of debt flows, the lower is financial fragility. Moreover, it is shown that the highest share of FDI is directed towards financial institutions and the tradables sector in developing countries. This leads to the following hypothesis:

H1: FDI inflows have a positive influence on financial sector development and efficiency of financial intermediation

2.5.2 Portfolio Equity Investments (PEI)

(20)

foreign equity investors require higher standards regarding financial system development in order to invest in developing countries.

Nonetheless, Lane & McQuade (2013) do not find a significant relation between equity investments and domestic credit growth. This finding is supported by Furceri et al. (2012) who state that portfolio equity inflow episodes have no significant effect on domestic credit. Hence, the following hypothesis is proposed:

H2: PEI has no significant effect on financial system development and efficiency in developing countries

2.5.3 Debt flows

According to Kose et al. (2009) there is a general consensus in the literature on financial globalization that debt flows include the greatest risks among capital flows. Debt flows, which include portfolio debt flows and banks loans, are associated with exchange rate volatility and financial fragility. This change in macroeconomic variables may lead to lower investment rates and hence lower credit demand by domestic firms (Lane & McQuade, 2013).

However, debt flows increase the savings level among banks in developing countries, which may be allocated towards credits for the private sector. Several studies have shown that domestic credit growth is strongly associated with debt flows (Lane & McQuade, 2013; Furceri, 2012). Moreover, Tornell & Westermann (2004) and Kose et al. (2009) argue that a reduction of risky bank flows may not necessarily be beneficial as capital-poor countries with no access to FDI and PEI may still be able to benefit from debt flows to finance investments. This leads to the following hypothesis: H3: Debt flows are positively related to financial system development and efficiency in developing countries

2.5.4 Gross Foreign Assets + Liabilities (% GDP)

Both Rodrik & Subramanian (2009) and Baltagi et al. (2009) propose using gross foreign assets and liabilities with respect to GDP as a measure of financial globalization. As mentioned before, Baltagi et al. (2009) find a positive effect of this variable on banking development measured by private credit-to-GDP.

The existing literature shows an ambiguous picture regarding the effect of financial globalization on both overall financial development and access to finance in developing countries. However, the majority of studies conclude that a higher integration of developing countries into the global financial markets has positive impact on financial system development and increased credit expansion to the private sector. This leads to the following hypothesis:

(21)

3 Methodology & Data

As seen in Table 1, the relationship between financial liberalization and domestic credit growth is extensively investigated in the literature. However, this paper is relying on different measures of financial system development apart from domestic credit growth. Moreover, the existing body of literature uses a large variety of measures of financial liberalization. As mentioned before, this research is relying on ‘de facto’ measures as recommended by Kose et al. (2009).

3.1 Dataset description

The dataset used in this research is compiled using a variety of statistical sources on a country level scale. It contains data of 86 developing and transition countries covering different time periods. Countries included in the database are classified by the World Bank as either low-income, lower-middle income or upper-middle income economies4_{. The complete list of countries can be}

found in Appendix A.5

The overall time span of the dataset reaches from 1980 up to 2011. Countries with a population of less than one million inhabitants and countries with substantial missing financial data are excluded. Data on financial system development is mainly taken from the World Bank’s Global Financial Development Database (GFDD) and Financial Development and Structure Dataset (FDSD). The GFDD is an extensive dataset of financial system characteristics for 203 economies. It contains annual data, starting from 1960. It has been last updated in November 2013 and contains data up to 2011 for 105 indicators capturing various aspects of financial institutions and markets. The FDSD contains indicators of financial development and structure across countries and over time. It includes a range of indicators (31 indicators in total), starting from 1960 that measure the size, activity, and efficiency of financial intermediaries and markets.

Financial data regarding capital flows and accumulation of foreign assets and liabilities is taken from a dataset constructed by Lane and Milesi-Ferretti (2007)6_{. They have assembled a}

comprehensive dataset on the foreign assets and liabilities of advanced, emerging and developing countries for the period 1970-2011. Their estimates are largely based on data of countries’ balance of payments (BOP) and international investment positions (IIP), which are published in the International Financial Statistics by the International Monetary Fund (IMF). The authors mention that the estimates rely heavily on information published by individual countries and international organizations, but they should be considered as their own estimates and not official data.

According to the Balance of Payments Manual by the IMF, capital inflows and outflows are claims

4_{The country classification can be accessed via http://data.worldbank.org/about/country-and-lending-groups.}

5_{Latvia, Lithuania, Russian Federation and Uruguay have been classified as “High Income Countries” for the year}

2015. Nonetheless, they are included in the sample as they only got their high-income status in 2012.

(22)

and transactions between a country’s residents and nonresidents (Lane & Milesi-Ferretti, 2007). External assets and liabilities are divided into the following categories:

• Portfolio investment, subdivided into equity securities and debt securities; • Foreign direct investment, which refers to equity participations above 10%;

• Other investment (which includes debt instruments such as loans, deposits, and trade credits); • Financial derivatives; and

• Reserve assets.

Balance of payments data measures net capital inflows and outflows during a specific time period (typically one year) while the IIP data measures the stocks of external assets and liabilities at the end of the period. The dataset by Lane and Milesi-Ferretti contains stock measures for both assets and liabilities of portfolio equity investments, foreign direct investments and debt investments. Specifically, debt assets and liabilities include the sum of portfolio debt securities and other debt instruments. Since the dataset contains a discrete time period, the change in stock measures from the end of the previous year to the end of the following year is equal to the corresponding flow variable per year. Hence, by taking first differences of the stock measures in the dataset it is possible to obtain the corresponding capital inflows and outflows per year per country. Furthermore, the authors calculated the stock measures of foreign assets and liabilities incorporating valuation changes over time.

Data about GDP per capita (ppp-adjusted), investments and savings-rates, inflation rates and current account positions are taken from the IMF world economic outlook database. The institutional quality of a country is measured by using the Institutional Quality database set up by Aljaz Kuncic (2012), which captures the quality of legal, political and economic institutions of countries. Since annual capital flows are highly volatile, three-year non-overlapping periods of the dataset are constructed. Moreover, this makes it possible to study effects in the medium run.

3.2 Model specification

The baseline panel regression model looks as follows:

!"_!"! =!∝ ! +!!_!!"#_!"!!+ !_!!"#$_!"!!+ ß_!!_!"!+ !_!+ !_!+ ! !_!" (1)

(23)

Both ICFit-1 and GFALit-1 are the main explanatory variables of interest. α is the constant, !i are the unobserved country-specific effects while ωt depicts period dummies to control for time fixed effects and εit is the regular error term.

A more specific model is presented in equation (2), which includes disaggregated capital flows.

!!"!" =!∝ +!!!!"#!"!!+ !!!"#!"!!+ !!!"!"!!+ !!!"#$!"!!+ ß!!!"!+ ! !!+ !!+ ! !!" (2) FDIit-1, PEIit-1 and DFit-1 depict the one-period lagged inflows of foreign direct investments (FDI), portfolio equity inflows (PEI) and debt inflows (DF).

In the empirical literature it is emphasized that having well-developed financial markets strongly increases the likelihood of countries to benefit from financial globalization. In order to examine possible “thresholds” of financial depth of developing countries, an interaction between private credit-to-GDP and the capital inflows is introduced to the model. The model specification and the results are shown in chapter 4.5.

3.2.1 Dependent variable

The dependent variable in the model used in this research aims to capture domestic financial system development and reveal potential quality and quantity effects of financial liberalization. However, the concept of financial system development is multidimensional and hence a variety of aspects have to be considered. The indicators of financial system development focus on both depth and efficiency of the domestic financial sector.

Bank credit/bank deposit (%): This variable captures the efficiency of domestic banks in

channeling their available deposits into credits, which is the typical core task of banks and other financial institutions. Bank deposits represent the main funding base for financing credits. Thus, this variable depicts the possible ‘quality effect’ of financial liberalization. A low ratio may imply that banks are not earning optimal returns on the loans issued, whereas a high ratio (above 100 percent) shows that the bank is borrowing from external sources, which it allocates towards higher yielding loans. However, a large ratio may reflect high risk-taking by banks and/or an insufficient assessment of the creditworthiness of borrowers, resulting in increasing liquidity risk. The World Bank Global Financial Database offers data for this variable.

Savings/Private Credit by deposit money banks and other financial institutions: This variable is

(24)

countries might be revealed. However, it has to be noticed that savings do not represent the availability of financing to fund investments. Savings represent income not consumed, whereas finance is access to purchasing power. Thus, investments require financing rather than savings. Savings depict the “real endowments” of financial intermediaries that are transferred into investments (Borio & Disyatat, 2011). Figure 4 outlines this issue:

Figure 4: Possible uses of saving – Source: Howells and Bain (1998)

Only when the allocation of this residual income is supported by the issuance of financial claims (e.g. credit to the private sector), new financial assets and liabilities are created. In Figure 4, ‘NAFA’ stands for the net acquisition of financial assets, which can either result in the accumulation of ‘hoards’ (i.e. acquisition of liquid assets) or lending activities by banks. Nonetheless, savings is used as proxy for finance in this research in order to estimate the efficiency of the financial system. Eventually, both Bank credit/Bank deposit and Savings/Private Credit do not reveal whether capital is allocated efficiently in terms of investment projects with the highest return.

Liquidity ratio (Liquid assets to deposits and short-term funding (%)): This variable captures

how much of the available deposits and short-term funding is held in liquid assets such as central bank debt or government bonds and is not used as a loan to either the public or private sector of the economy. A high liquidity ratio may point towards an existing investment constraint. The liquidity ratio represents the counterpart to the ratio of bank credit to bank deposits. Thus, this research aims to obtain a more comprehensive picture of the effects of foreign capital on bank behavior in developing countries by employing the liquidity ratio as a dependent variable.

Deposit money bank assets to GDP (%) is a typical measure of depth or size of the financial

(25)

Interest rate spread (lending rate minus deposit rate): Financial liberalization is expected to

lower the spread between the interest rate charged by banks on loans to private sector customers and the interest rate paid by commercial or similar banks for demand-, time-, or savings deposits as defined by the World Bank. The interest rate spread is comparable to the profit margin of banks. As financial liberalization is expected to remove market distortions, the market should determine all interest rates and the spread between lending rate and deposit rate is supposed to decline due to higher competition in the banking market. However, the terms and conditions attached to these rates differ by country, and thus limit their comparability.

3.2.2 Explanatory variables

As shown in equations (2) and (3), the main explanatory variables are aggregated capital inflows (ICF), disaggregated capital inflows (FDI, PEI, DF) and gross foreign assets and liabilities (GFAL). Both capital inflows and gross foreign assets and liabilities are taken as percentage of GDP. Regarding capital inflows this research incorporates three main monetary flows in international trade: Foreign Direct Investment (FDI); Foreign Debt inflows (DF) and Portfolio equity investment (PEI). By using net inflows instead of net flows it is aimed to mitigate the problem that net flows only depict the current account balance, which does not reveal the degree to which investments are financed from abroad. Moreover, the impact of cross-border capital flows on domestic financial conditions remains unclear by using the current account balance as a proxy for financial integration. According to Borio and Disyatat (2011, p.9), “a balanced current account only implies that domestic production equals domestic spending, not that domestic saving finances domestic investment”.

The model includes a second measure of financial globalization: gross foreign assets and liabilities as a share of GDP (GFAL), as used by Rodrik & Subramanian (2009). In contrast to capital flows, this is a stock measure and is not as volatile as capital inflows. Moreover, it also incorporates foreign capital investments by developing countries and thus, acts as a solid proxy for ‘de facto’ global financial integration.

3.2.3 Control variables

(26)

a country has achieved certain developments. A description of all variables including their sources is reported in Appendix B.

3.3 Data Analysis

The baseline model regresses the total capital inflows on five measures of financial system development, namely bank credit/bank deposits, savings/private credit, liquidity ratio and interest rate spread. Afterwards, the effect of disaggregated capital flows on the five dimensions is measured, both with and without the control variables. In general, each proxy for financial development measures the concept in different ways and hence, possible measurement errors may be controlled for through the multidimensional measurement.

The initial dataset as outlined above was found to be unbalanced, which means that some countries do not offer entries on each variable for each year between 1980 and 2011. This applies especially for Eastern European countries for which only data from the early 1990s on is available. 7

In addition, the dataset is examined for influential observations. A highly influential observation is one that substantially changes the estimates of the regressors. Outliers can be detected by using Cook’s distance, which “measures the aggregate change in the estimated coefficients when each observation is left out of the estimation” (StataCorp, 2013). Values for Cook’s distance that are greater than 4/n (n is the number of observations per variable) may be problematic and hence this “threshold” is applied in this research. The outlier analysis resulted in the finding of 27 influential observations. Figure 5 shows the most influential observations in the dataset plotting leverage (i.e. a measure of how far an independent variable deviates from its mean) versus squared residuals.

Figure 5: Leverage vs. squared residuals in the dataset

7_{Financial data for Eastern European countries only became available after the dissolution of the UDSSR in 1991.} Albania

Albania Albania Albania

Albania

Albania ArgentinaArgentinaArgentinaArgentinaArmeniaArgentinaArmeniaArmeniaArgentinaArgentinaArgentinaArmenia

ArmeniaAzerbaijanAzerbaijan AzerbaijanAzerbaijanAzerbaijan Burundi

Burundi_BurundiBurundi_Burundi_Burundi BurundiBurundi

BeninBurkina FasoBurkina FasoBeninBeninBeninBeninBurkina FasoBurkina FasoBeninBenin Burkina FasoBurkina FasoBurkina FasoBangladeshBurkina FasoBulgaria BulgariaBangladeshBulgariaBangladeshBangladeshBangladeshBangladeshBulgariaBangladesh BulgariaBelarusBulgaria BelarusBelarusBelarus BelarusBelarus Belize_BelizeBelizeBelizeBelizeBelizeBelizeBelize BoliviaBoliviaBoliviaBoliviaBoliviaBoliviaBoliviaBolivia

Brazil BrazilBrazilBrazilBrazilBrazilBrazilBrazil

Botswana BotswanaBotswanaBotswanaBotswana

BotswanaBotswanaBotswana Central African RepublicCentral African RepublicCentral African RepublicCentral African Republic Central African RepublicCentral African RepublicCentral African Republic

Central African Republic ChinaChinaChinaChinaChina China

China

Cote d'Ivoire Cote d'Ivoire Cote d'IvoireCote d'IvoireCote d'IvoireCote d'IvoireCote d'IvoireCameroon Cameroon CameroonCameroonCameroon CameroonCameroonCameroon

Congo, Rep.Costa RicaColombiaColombiaColombiaColombiaColombiaColombiaCosta RicaCongo, Rep. Congo, Rep.ColombiaColombia Congo, Rep.Congo, Rep.Congo, Rep.Congo, Rep. Costa RicaCosta Rica

Costa Rica Dominican Republic Dominican RepublicDominican Republic Dominican RepublicDominican Republic Dominican RepublicEcuadorEcuadorEcuadorEcuadorEcuadorDominican RepublicAlgeria AlgeriaEcuadorDominican RepublicEcuadorAlgeriaAlgeriaAlgeriaAlgeriaEcuadorAlgeriaAlgeria

Egypt, Arab Rep.GeorgiaEgypt, Arab Rep.GeorgiaEthiopiaGhanaEgypt, Arab Rep.EthiopiaEgypt, Arab Rep.EthiopiaGhanaEthiopiaEthiopiaGhanaGhanaGhanaEgypt, Arab Rep.GeorgiaGabon GabonEthiopiaGabonGabonEthiopiaGabonEgypt, Arab Rep.GabonGhanaEgypt, Arab Rep.Egypt, Arab Rep.GhanaGhana GabonGabon Gambia, The_{Gambia, The}

Gambia, The_{Gambia, The} Gambia, The

Gambia, TheGambia, TheGuinea-BissauGuinea-BissauGuinea-BissauGuinea-BissauGuinea-BissauGuinea-Bissau Guinea-Bissau GuatemalaGuatemala

Guatemala Guatemala GuatemalaGuatemalaGuatemala

GuatemalaHondurasGuyanaGuyana GuyanaGuyanaGuyanaGuyana HondurasHondurasHondurasHondurasHondurasHondurasHonduras Haiti HaitiHaiti

HaitiIndonesiaIndonesiaHaitiHaitiIndonesiaIndonesiaIndonesia Indonesia IndonesiaIndonesiaIndiaIndiaIndiaIndiaIndiaIndiaIndiaIndia

JamaicaCambodiaCambodiaJamaicaJamaicaJamaicaCambodiaCambodiaKenyaJordanKenyaJamaicaCambodiaCambodiaJamaicaKenyaJordanJamaicaJordanJamaicaJordanKenyaKenyaKenyaJordanJordanKenyaKenyaJordanJordan Sri LankaSri LankaSri LankaLesothoLesothoLesothoLithuaniaSri LankaSri LankaSri LankaSri LankaLesothoLesothoSri LankaLithuaniaLesothoLithuaniaLesothoLesotho LithuaniaLatvia LatviaLatvia LatviaLatvia

MoroccoMoldovaMoroccoMoldovaMoroccoMoroccoMoroccoMoroccoMoroccoMoroccoMoldova MoldovaMoldova

MoldovaMadagascarMadagascarMadagascarMadagascarMadagascarMadagascarMadagascarMadagascarMexico Mexico

MexicoMali MaliMexicoMaliMexicoMexicoMexicoMexico Mali MaliMali

MaliMali MongoliaMongoliaMongolia Mongolia

MongoliaMongoliaMozambiqueMozambique MozambiqueMozambique Mozambique MozambiqueMozambique Mauritius MauritiusMauritiusMauritiusMauritiusMauritius

Mauritius Mauritius

Malawi MalawiMalawiMalawiMalawiMalawiMalawi

MalawiNigerNigerNigerNigerNigerNigerMalaysiaNigerMalaysiaNigerMalaysia MalaysiaMalaysiaMalaysiaMalaysiaMalaysia Nigeria

Nigeria Nigeria Nigeria

NigeriaNepalNigeriaPakistanNigeriaNepalNepalNigeriaPakistanPakistanPakistanPakistanPakistanPakistanNepalNepalNepalNepalPakistanNepal Panama

Panama_Panama_PanamaPanama_Panama Panama

PanamaPeru Peru Peru

Peru PeruPeruPeruPeru PhilippinesPhilippinesPhilippinesPhilippinesPhilippinesPhilippinesPhilippines PhilippinesPapua New Guinea Papua New Guinea Papua New Guinea Papua New GuineaPapua New GuineaPapua New Guinea Papua New Guinea Papua New Guinea

Paraguay ParaguayParaguayParaguayParaguay

ParaguayParaguayRussian FederationRussian FederationParaguay Russian FederationRussian Federation Russian FederationRussian Federation_Rwanda

RwandaRwandaRwandaRwanda Rwanda Sudan _Sudan Sudan Sudan Sudan Sudan

SudanSenegal SenegalSenegal SenegalSenegalSenegal

SenegalSwazilandSenegalSwaziland El SalvadorEl SalvadorEl SalvadorEl SalvadorEl SalvadorEl SalvadorEl Salvador El Salvador SwazilandSwazilandSwazilandSwaziland

Swaziland

Syrian Arab Republic Syrian Arab Republic Syrian Arab RepublicSyrian Arab RepublicSyrian Arab Republic Syrian Arab RepublicSyrian Arab Republic Syrian Arab Republic

Chad

Chad Chad_TogoTogo_TogoTogoTunisiaTunisiaThailandThailandTurkey_TogoTurkeyTogoTunisiaTunisiaTunisiaTunisiaTunisiaChadTunisiaChadChadThailandThailandThailandThailandThailand Thailand TurkeyTurkeyTurkey

Turkey

TurkeyTurkeyTanzaniaUgandaUkraineTanzaniaTanzaniaUgandaUgandaUgandaUgandaUgandaTanzaniaTanzaniaUgandaTanzaniaUgandaTanzaniaUkraine Tanzania Ukraine UkraineUruguayUruguayUkraineUkraine

UruguayUruguayUruguayZambiaZambia ZambiaUruguayVietnam VietnamUruguayUruguaySouth AfricaSouth AfricaSouth AfricaVietnamSouth AfricaVietnamSouth AfricaSouth AfricaSouth AfricaVietnamSouth Africa Zambia

Zambia_Zambia_Zambia

0 .2 .4 .6 .8 1 L e ve ra g e 0 .005 .01 .015 .02

(27)

Figure 5 reveals that there are some influential outliers (e.g. Mauritius, China and Ukraine), which should be taken into account. In total, the outliers amount for 2.8 percent of all observations in the model. However, the outliers are not deleted from the dataset but a dummy variable is created, which takes on the value 1 if the observation is a potentially influential observation. The outliers are excluded in the baseline regressions, but they are included in the models as a robustness check. The methodology applied in this research benefits from the advantages of panel data analysis. Compared to cross-sectional analyses, panel data analysis allows specifying and estimating more complicated and realistic models. However, since the same units are observed over a specific time period it is no longer appropriate to assume that different observations are different from each other (Verbeek, 2012).

As for conventional panel data analysis, it has to be checked whether or not there are random effects present in the data and a pooled OLS model is suitable for the analysis. In a pooled OLS regression model, it is assumed that there are no cross-country differences and all countries interact with the variables in the same way. The pooled model assumes that the error term has zero mean and variance. In addition, errors are assumed to be uncorrelated over time, across individuals and the independent variables. Thus, no random individual heterogeneity is presumed to be present in the data.

However, in the dataset used in this research, cross-country differences are most likely to be present as data of countries from Latin America, Eastern Europe, Asia and Africa is included. Hence, a pooled OLS regression appears to be inappropriate. In order to verify this assumption, a Breusch-Pagan Lagrange multiplier test is executed to test for random effects. The null hypothesis of this test states that the variance across all countries is zero. The alternative hypothesis states that the variance is greater than zero and random effects across countries are present. The test is applied to each regression model in this research and if the p-value is lower than 0.05 (5% significance level) the null hypothesis can be rejected and random effects are present.

(28)

within individuals (Carter Hill et al., 2012). Moreover, the fixed effects estimator assumes that the individual heterogeneity between countries is correlated with the explanatory variables.

In contrast, the random effects model assumes that the individual intercepts are “treated as drawings from a distribution with mean µ and variance σ2” (Verbeek, 2012, p. 342). Hence, the individual heterogeneity is treated as random. The error term of the random effects estimator consists of two parts: an individual component that does not vary over time and a remainder component that is assumed to be uncorrelated over time. Hence, all correlation of the error terms is due to the individual heterogeneity. Furthermore, it is assumed that the individual heterogeneity and the error terms are “mutually independent and independent of all explanatory variables” (Verbeek, 2012, p. 348).

In order to decide whether a random effects or a fixed estimator is suitable, a Hausman Test is performed for each model. The null hypothesis of the test states that the independent variables and the individual heterogeneity are uncorrelated and a random effects estimator is preferred over the fixed effects estimator, as it is both consistent and efficient. The alternative hypothesis states that there is a correlation between the unobserved individual heterogeneity (!i) and the independent variables (xit) and only the fixed effects estimator is consistent. In general, the fixed effects estimator is consistent irrespective as to whether or not xit and !i are correlated, while the random effects estimator is only consistent and efficient when xit and !i are not correlated. Thus, the Hausman Test tests whether the fixed effects and the random effects estimators are significantly different from each other. If the p-value of the test statistic is smaller than 0.5 (5% significance level) the null hypothesis can be rejected and the fixed effects estimator is preferred.

(29)

Finally, a test for time-fixed effects and the inclusion of period dummies is performed as the fixed effects model only accounts for time-invariant effects across individuals.

The results for all described model specification tests are given in Appendix C. The results from the Hausman-test show that fixed effects are the valid estimation technique for all models. Moreover, clustered standard errors have to be used as both heteroskedasticity and first-order autocorrelation haven been detected. Furthermore, multicollinearity has been detected for the control variables investment rate (TI) and institutional quality (Iqual). Hence, the coefficients and standard errors of both variables in the same model have to be treated with caution. Since FDI inflows and gross foreign assets and liabilities show relatively high VIF values in the model employing the liquidity ratio as dependent variable, both variables are not used in the same regression.

It is decided to introduce the control variables stepwise into the model in order to control for changes in both significance and magnitude of the coefficients of the main explanatory variables. Time fixed effects have been found to be present in all models and hence dummies for each 3-year period are included. Furthermore, time dummies also capture the bias caused by autocorrelation and address a possible omitted variable bias.

3.4 Endogeneity

The models used in this research may be subject to endogeneity, which may lead to biased regression results. Economies with more developed financial systems may attract more capital inflows but capital inflows may act as a catalyst for further financial system development. Thus, the direction of causality might not be clear.

Global financial integration – The effect of foreign capital on financial system development and efficiency in developing and emerging countries