Foreign Bank Presence and Credit Growth in Asia: A Disaggregated Credit flows Analysis

(1)

Foreign Bank Presence and Credit

Growth in Asia: A Disaggregated

Credit flows Analysis

August 2010

University of Groningen

Faculty of Economics

Author Supervisor

(2)

1

Abstract

The aim of this paper is to explore the link between foreign bank presence and credit growth, particularly during a crisis, in Asia with the disaggregated measure for credit flows; credit flows to the real- and the financial-sectors. Using a panel dataset covering 6 Asian countries from 1996Q1 to 2008Q4, this paper finds hat foreign bank presence is positively associated with credit growth in both sectors. However, during a crisis, the positive influence of foreign bank presence on credit growth persists only in the real sector, not in the financial sector. The different findings per sector also suggest that the disaggregated measure for credit flows better perform than the traditional aggregate one in uncovering the relationship between foreign bank presence and credit growth.

(3)

2

1. Introduction

Banking markets have increasingly been international on account of financial liberalization and overall economic and financial integration. Between 1995 and 2002, the average share of banking sector assets held by foreign banks in 104 developing countries rose from 18 percent to 33 percent (Micco, Panizza and Yañez, 2006). With the rise in importance of foreign banks, the impacts of foreign bank presence1 on the host country economy have been the subject of a lively debate. Given that the most essential task of banks is financial intermediation2, the role that foreign banks play in credit markets, especially during a crisis, is of great relevance when assessing any benefits or costs that foreign banks can bring about in the host countries. The literature concerned with this issue is broadly divided into two contrasting views. The proponents of foreign bank presence stress that foreign banks can provide a new source of funds with credit markets, and diversify credit risk thereby reducing the severity of a crisis when it does occurs. Contrary to this positive role of foreign banks, critics point to the risk that foreign banks may provoke credit expansion, particularly for speculative investments. Moreover, during a crisis foreign banks might further undermine an already ailing banking sector by rapidly contracting credits. As stated above, the arguments are unequivocal, which suggests that more works need to be done in this research field.

Shifting focus, when measuring credit flows, most of the previous studies have regarded ‘credit’ as a homogeneous resource flow. However, the bank credit supply is directed to the different parts of the economy; to the real-sector and to the financial-sector3 (Werner, 1992, 1997). Whereas credit flows to the real-sector is profoundly linked to output growth, credit flows to the financial-sector is necessarily not (although they indirectly influence the growth in support of the real sector via financing and risk smoothing). Rather, credit flows to the

1_{As an indicator of foreign bank presence for my study, I use the share of banking sector assets held by foreign}

banks in a host country.

2

In simple terms, it means taking deposits from the public and allocating them as loans (credits) to different users.

3

In this paper, credit flows to sectors that are associated with financial transactions including pension funds, insurance, financial intermediation and real estate are defined as credit flows to the financial sector and credit flows to non-financial business such as agriculture, manufacturing, construction and other services as credit

(5)

4 financial-sector, especially to asset markets, have the potential risk to cause asset booms and bubbles. Given this, the increase of aggregate credit does not always guarantee a positive impact of foreign bank presence on the host country economy. What if a large portion of increased credit is directed to asset markets for speculative investments, not to investments in the real-sector? Since credit growth could be inefficient if misdirected to unfit sectors of the economy, the implications of foreign bank presence for host country economy can be different depending on where credit flows go. Hence, it is more useful to take disaggregated credit flows to the real- and financial-sectors into account to precisely interpret the influence of foreign bank presence on a host country economy, not just looking at whether aggregate credit increase or decrease. In addition, Bezemer and Werner (2009) studying the link between credit and economic growth with this disaggregated measure for credit flows find that the disaggregated measure is better predictor of nominal growth than the traditional aggregate one. Therefore, it is worthwhile to take a new approach based on the disaggregation of credit flows.

The aim of this paper is to investigate the impact of foreign bank presence on credit growth, particularly during a crisis. The research questions to be addressed are as follows: (1) Does foreign bank presence promote credit growth in the real- and the financial-sectors, respectively?, (2) In which sector is foreign bank presence more influential on credit growth?, and (3) during a crisis, does foreign bank presence alleviate the adverse effect of the crisis on credit growth in the real- and the financial-sectors, respectively?

(6)

5 The remainder of this paper is organized as follows: Section 2 gives an overview of the existing literature that deals with the relation between foreign bank presence and credit growth, Section 3 presents the data and econometric methodology, Section 4 discusses the empirical results, and finally, my conclusion, limitation of this research and implication for future research are given in the Section 5.

2. Literature Review

As briefly stated in the introduction section, the existing literature that explores the relation between foreign bank presence and credit growth provides the two opposing views and findings. Here, the previous theoretical and empirical studies are reviewed in depth from those two perspectives.

2.1. Foreign bank presence and credit growth

The main argument brought forth by proponents of foreign bank presence is that higher foreign bank presence would result in more loanable funds. Since foreign-owned banks are usually backed by their parent companies, they have the ability to provide fresh money to the host financial market (Eller, Haiss, and Steiner, 2005). Also, the increase in competition caused by foreign bank entry can widen access of borrowers to credit, which may increase aggregate lending (Moreno and Villar, 2005). In the aspect of deposits, greater financial competition would increase interest rate paid on deposits, which would attract more deposits, thus allowing banks to lend more money (Jeon, Miller, and Natke, 2006). Lastly, through technological improvements, e.g. through standardized loan applications, computerized payments processing, and better trained personnel, banks would be able to raise the credit supply (Weller, 2007).

(7)

6 1994-1999. De Haas and Van Lelyveld (2002) show that local credit by foreign-owned banks compared to domestic credit rose in the 1990s in CEE countries, except for Slovenia. Using the data for over 100 developing countries during 1995-2002, Cull and Martínez Pería (2007) find that in countries where the level of foreign bank presence was relatively high and stable, private credit levels were significantly higher than in other countries. However, there are also some studies that provide evidence to the contrary. Using data for 89 low income and lower middle income countries, Detragiache, Gupta and Tressel (2006) find that a larger foreign bank presence is associated with shallower credit markets and slower credit growth. With data for 60 countries from 1985 to 1998, Weller (2007) shows that the credit supply declines in response to increased competition from foreign banks. Yet, the adverse effect of foreign bank presence on the credit supply is less pronounced when the presence of foreign banks is larger.

Then, does credit grow both in the real- and the financial-sectors? and if it does, in which sector does credit expand more rapidly? Some critics voice the concern, which is that the presence of foreign banks promotes fast credit growth, which is often associated with a rise of speculative investment. There might be three possible explanations for those credit expansions. Firstly, the increase in competition caused by foreign bank entry raises the risk appetite of banks in the fight for market share (Eller et al., 2006). Also, more competition tends to lower interest rate spreads and bank earnings. To compensate, lenders should expand credit to new market segments. But the presence of foreign banks effectively cuts off pathway to less risky investments, so that domestic banks can only expand in riskier segments (Weller, 2007). Secondly, excessive risk-taking by banks can also be caused or exacerbated by deregulation euphoria following financial liberalization4(Grabel, 1995, 1998). She explains that because financial liberalization often props up speculative investments, all lenders may become more optimistic about the future, thus extending more credit to risky investments. Lastly, in a similar context, banks may have insufficient capital to properly account for default risks and thereby overextend lending to risky projects because they have little or nothing to lose with fewer credit market regulations after financial liberalization (Stiglitz, 1994). Based on these theoretical arguments, it is conjectured that credit growth might be

4

(8)

7 faster in the financial sector than in the real sector. As yet, there is to my knowledge no empirical research that does an analysis with the disaggregated credit flows in this research area. The clear-cut answers to above two questions are thus empirical questions that this paper firstly attempts to address.

Overall, the evidence available up to now suggests that bank credit supply may either increase or decrease with foreign bank presence. However, given that the existing literature provides much more evidence on the positive link between foreign bank presence and (aggregate) credit growth, I propose the following hypotheses.

Hypothesis 1a: A higher level of foreign bank presence is positively associated with credit growth in the real sector.

Hypothesis 1b: A higher level of foreign bank presence is positively associated with credit growth in the financia sector.

Hypothesis 1c: The positive effect of foreign bank presence on credit growth is greater in the financial sector than in the real sector.

2.2. Foreign bank presence and credit growth during a crisis

When a crisis occurs, banks in particular face unparalleled liquidity stress hurting their ability to lend. The damage of the 2008 financial crisis on global bank balance sheets in advanced countries – with losses reaching over US $ 4 trillion in the period 2009-2010 (IMF, 2009)5 – provoked a strong credit slowdown. According to Aisen and Franken (2010) studying the evolution of bank credit throughout the 2008 financial crisis, bank credit contracted in 95 percent of all 82 countries covered. Then, does foreign bank presence alleviates or exacerbate this adverse effect of a crisis on credit growth? To get an answer to this question, I review the theoretical and empirical literature that explores how foreign banks behave during the crisis.

(9)

8 In theory, there exist two conflicting possibilities that foreign bank presence can either alleviate or exacerbate the negative effect of crisis on credit growth. On the positive side, the parent bank with (generally) abundant funding sources may act as a “buck-up facility” or lender of last resort during crisis periods (De Haas and Van Lelyveld, 2006). Moreover, during periods of banking stress, foreign bank presence could diversify against country-specific (systemic) risks that can severely impair the capital of the banking system, because foreign banks are diversified across different countries and thus less prone to the adverse effects of a host country bank capital shock (Moreno and Villar, 2005). In these regards, foreign banks may be able to recover relatively fast and keep up their credit supply well (when compared to domestic banks) during and after a crisis. Contrary to this potentially positive role of foreign banks, foreign banks can reallocate their portfolio over different geographical regions on the basis of expected risks and returns (Morgan and Strahan, 2004). That is, foreign banks have alternative investment opportunities to shift their capital to elsewhere when economic environment in a particular host country get bad often making economic downturn even worse. In this line of reasoning, foreign banks may negatively affect credit supply when a crisis occurs.

(10)

9 In summary, although there are two conflicting possibilities that foreign bank presence can either alleviate or exacerbate the negative effect of crisis on credit growth in theory, most of empirical research provides the support for the positive link between foreign bank presence and credit growth during a crisis. Given this, I propose the following hypotheses.

Hypothesis 2a: The higher the level of foreign bank presence, the negative effect of a crisis on credit growth in the real sector will be less pronounced.

Hypothesis 2b: The higher the level of foreign bank presence, the negative effect of a crisis on credit growth in the financial sector will be less pronounced.

3. Empirical Analysis

3.1. Data

Sample

For my analysis, I construct a (unbalanced) panel dataset. To determine which countries to be included, I begin with the countries where were to a certain extent under the influence of the Asian crisis of 1997-1998. However, the sample eventually obtained comprises 6 countries6 covering the period from 1996Q1 to 2008Q4 due to lack of data.

Disaggregated credit flows7 (Dependent variable)

A key element in my study is the construction of disaggregated credit flows; credit flows to the real sector and to the financial sector. To do it, I primarily follow the work of Werner (1992, 1997). The data is collected from the central bank on bank loans denominated in domestic currency to firms, households, and the government. The central banks in countries covered report the financial statements on bank loans classified by economic activity

6

Countries included in the sample are Japan, Korea, Malaysia, Philippines, Singapore, Thailand.

7

(11)

10 (industrial sectors)8. In order to disaggregate credit flows into the real- and the financial-sectors by using this bank loans, the following definitions and assumptions9 are made. Firstly, with respect to business loans, credit flows to sectors that are associated with financial transactions including pension funds, insurance, financial intermediation and real estate are defined as credit flows to the financial sector and credit flows to non-financial business such as agriculture, manufacturing, construction and other services as credit flows to the real

sector. Secondly, in case of household loans, if they are split out into lending for real estate investment (most of which are mortgages) and consumption lending, the former is defined as

credit flows to the financial sector and the latter as credit flows to the real sector. However, if they are not, all of them are counted into credit flows to the financial sector assuming that the former generally accounts for a substantial portion of household loans and the latter is negligible on a basis of the fact that this is the case in two countries (e.g. Thailand, Singapore), where the split data is available. Lastly, government loans are all regarded as

credit flows to the real sector assuming that the government rarely engages in real-estate investments. In line with Cull and Martínez Pería (2007), Detragiache et al. (2006), and Weller (2007), I scale credit by each country’s GDP to control for the size of the real economy, thus reducing the heteroskedasticity. The percentage growth in credit relative to GDP (in the real- and the financial-sectors, respectively) in each country is provided in Appendix E.

Foreign bank presence (Independent variable)

As an indicator for foreign bank presence, foreign bank asset share is used. Many previous studies (e.g Claessens et al, 2001; Micco et al, 2005, 2006) have traditionally used foreign bank asset share as a proxy for the levels of foreign bank presence. In those papers, a bank is defined as foreign if at least 50 percent of its capital is in the hands of non-residents.

Foreign bank asset share is measured by foreign bank asset share in total assets in the banking sector. The asset share of foreign banks in each country is presented in Appendix B. Broadly speaking, it shows that Japan had the lowest level of foreign bank asset share, which was almost negligible (0.1-0.2%). By contrast, although foreign bank asset share in Singapore was very low (on average 1.4%) during the second half of 1990s, it substantially increased up

8

Although the classification categories are to some degree different across countries, they generally comprise agriculture, manufacturing, construction, real estate, financial intermediation and other non-financial services.

9

(12)

11 to about 47% in 2000 and had kept the level of over 40% for the following two years. Malaysia comes next showing around 25% foreign bank asset share for the entire period. Finally, three other countries, namely Korea, Philippines, and Thailand had on average 5% to 9% foreign bank asset share. In those three counties, there were some ups and downs in foreign bank asset share, but foreign bank asset share had moderately changed over time.

Crisis (Independent variable)

I identify and date a crisis during the period of 1996-2008 using two recent studies: Laeven and Valencia (2008) and Inklaar and Yang (2010). According to Laeven and Valencia (2008), there are three types of financial crisis, namely a (systemic) banking crisis, a currency crisis and a (sovereign) debt crisis10. Taking it into count the possibility that the other two types of crises as well as a banking crisis can hurt directly or indirectly the bank ability to lend, I consider all three types of financial crisis for my study. An overview of financial crisis is provided in Appendix C.

My choice of several control variables are discussed in below subsection and a complete list of variables, definitions and data sources are described in Appendix A.

3.2. Regression specification

I use two different sets of regression specifications. Firstly, the regression equations to test whether foreign bank presence has the positive impact on credit growth (hypothesis 1a, 1b, and 1c) are as follows11.

CR(i, t) = α + β₁FBAS(i, t) + β₂CONTR(i, t) + β₃C(i) + β₄T(t) + ε(i, t) (1a)

CF(i, t) = α + β₁FBAS(i, t) + β₂CONTR(i, t) + β₃C(i) + β₄T(t) + ε(i, t) (1b)

10

According to Inklaar and Yang (2010) building on the work of Laeven and Valencia (2008), a banking crisis is identified as a situation where systemically important financial institutions are in distress, a currency crisis as a year with a rapidly depreciating exchange rate and a debt crisis as the year of a sovereign default on privately held debt.

11

To assess whether disaggregated measures for credit flows (CR and CF) perform better in predicting the relationship between foreign bank presence and credit growth than traditional, aggregate measures, I also do the test for the total credit flows (CT). For CT, regression equation is CT(i, t) = α + β1FBAS(i, t) + β2CONTR(i, t) +

(13)

12 Where i is the country index, t is the time index; CR(i, t) is the percentage growth of credit

flows to the real-sector relative to GDP; CF(i, t) is the percentage growth of credit flows to the

financial-sector relative to GDP; FBAS(i, t) is foreign bank asset share in each country as a

proxy for foreign bank presence; CONTR(i, t) is a vector of control variables; C(i) and T(t) are

country and time dummies, respectively and ε(i, t) is a error term.

To capture the effects of other potential factors on credit growth in my regressions, I include five control variables. In line with Ari and Franken (2010) and Law (2008), I use

GDP per capita (% growth) and financial openness12 reflecting characteristics of the countries that may affect the development of local credit. Higher GDP per capita exhibits stronger economic activities and thereby leads to higher credit growth. Countries with more open financial markets are more likely to attract foreign capital providing ‘deepening’ for credit markets. I also include interest rate, banking sector concentration, deposit that account for the characteristics of the countries’ banking systems that could affect the evolution of domestic bank credit. According to Jeon, Miller and Natke (2006), lending by banks depends on the expected returns and risks on investments. Higher interest rate increases the expected return, and thus banks lend more with higher interest rate. Banking sector concentration might improve the supply of credit if it allows banks to reap economies of scale (Clarke et al., 2006). Banking lending also depends on a bank’s deposits (Weller, 2007).

Secondly, the regression equations to test whether foreign bank presence alleviate or exacerbate the negative effect of crisis on credit growth (hypothesis 2a and 2b) are as follows13.

CR(i, t) = α + β₁CRISIS(i, t) + β₂FBAS( i, t) + β₃[CRISIS( i, t) × FBAS( i, t)] + β₄CONTR(i, t) +

β5C(i) + β6T(t) + ε(i, t) (2a)

CF( i, t) = α + β₁CRISIS(i, t) + β₂FBAS( i, t) + β₃[CRISIS( i, t) × FBAS( i, t)] + β₄CONTR(i, t) +

β5C(i) + β6T(t) + ε(i, t) (2b)

12

Financial openness is measured by the sum of foreign assets and liabilities (% of GDP)

13_{I also do the test for the total credit flows (CT). For CT, regression equation is CT}(i, t) = α + β1CRISIS(i, t) +

(14)

13 Where the same notation as above is used. Additionally, CRISIS(i, t) is crisis dummy equal

to one when there is a crisis according to my definition, and otherwise zero.

Most importantly, I include interaction term between FBAS and CRISIS. The negative effect of a crisis on credit growth may vary at the different levels of foreign bank presence (FBAS). The five control variables same as those for Eqs. (1a) and (1b) are also used.

3.3. Estimation methodology

14

Given the nature of a panel data that consists of a group of cross-sectional units (e.g. countries) observed over time, the regression models presented above pose some challenges for estimation. The first is the presence of unobserved country- and time-specific effects. To examine whether it is appropriate to include country- and time-specific fixed effects in the models, I do the Hausman test. The test results show that fixed effects model is better than random effects model15. Therefore, I include the country- and time-specific dummy variables that can account for those fixed characteristics.

The second challenge is whether my time-series data is stationary or nonstationary16? The main reason why it is important to know this is that there is a danger of obtaining apparently significant regression results from unrelated data when nonstationary time series are used in a regression model. Such regression is said to be spurious. It results from the fact that one series with a stochastic trend is related to another series with another stochastic trend. Thus, I perform the Phillips-Perron test17 to check for the stationarity of my data. With the test results reporting that while some variables are stationary, the others are nonstationary, it is

14_{Most of econometric theories that I refer to in this paper is based on the “Principles of econometrics ” (Cater}

Hill, Griffiths, and Lim, 2008)

15

The Hausman test shows that while fixed effects model is better than random effects model in the regression models for CF, this is not the case in the models for CR. Given that the fixed effects model is definitely more suitable in some regression models, I use the country- and time-sepecific dummies.

16

If the time-series data is nonstationary, it displays trending behavior. In contrast, stationary time-series display behavior that can be described as irregular ups and downs, or more like fluctuations. In econometric context, E(yt) = µ (constant mean); var(yt) = σ2 (constant variance); cov (yt, yt+s)= cov (yt, yt-s) = γs

(covariance depends on s, not t)

17

I do the Phillips-Perron test per country. The result shows that FBAS, GDP per capita, and banking sector

(15)

14 necessary to make nonstationary variables stationary. However, this step should be taken only when there is no cointegration between the nonstationary variables, which are “integrated of order 1” or I(1)18. Time series are cointegrated if they tend to move together through time and the regression with nonstationary variables - I(1) variables - is acceptable providing those variables are cointegrated. The augmented Dickey-Fuller test is used to see if my time-series data is cointegrated. Since the test does not find cointegration between FBAS and CR (or CF) in every country covered19, I decide to convert the nonstationary variable to its stationary counterpart by taking the first difference, that is, ∆

y

t =

y

t –

y

t-420. Here, it is noteworthy that

differencing tends to increase the standard deviation to (potentially) high level, making it much harder to find the result. Having a higher variance means that there is a lower probability of obtaining a coefficient estimate close to its true value. Given this disadvantage of differencing and the fact that my primary interest lies in exploring the relationship between foreign bank presence (FBAS) and credit growth (CR and CF), only CR and CF variables are differenced instead of differencing all the variables including control variables. FBAS variable is already stationary, thus differencing is not necessary21.

The third challenge is whether there are autocorrelation in the regression models22. With time-series data, it is (highly) likely that the observations will be correlated over time. Ignoring autocorrelation if it is present does not affect the coefficient estimates. They are still unbiased, but inefficient even in large sample sizes. Moreover, the usual least squares standard errors are no longer correct and hence there exists the possibility that we could make the misleading inferences. To see whether the errors in my regression models exhibit

18

Time-series, which can be made stationary by taking the first difference, are said to be ‘integrated of order 1’, and denoted as I(1).

19

According to the results of the augmented Dickey-Fuller test, CR and FBAS variables are not cointegrated in Janpan, Malysia, and Singapore and CF and FBAS variables are not cointegrated in Japan, Korea, and Philippines.

20_{Given that the dependent variable (credit data) is quarterly data, ∆}_y_{t is the change in the value of dependent}

variable y from time t-4 to time t.

21

Now, my regression models presented above take the form like ∆CR(i, t) = α + β1FBAS(i, t) + β2CONTR(i, t) +

β3C(i) + β4T(t) + ε(i, t); ∆CF(i, t) = α + β1FBAS(i, t) + β2CONTR(i, t) + β3C(i) + β4T(t) + ε(i, t). The same goes for Eqs.

(2a) and (2b).

22_{When there exists autocorrelation, the error term takes the form of e}

t = ρet-1 + vt . While E(et) = 0; Var(et) =

(16)

15 autocorrelation, Durbin-Watson test23 is used and its results discover the existence of autocorrelation. To capture autocorrelation in the errors, I assume that the errors in my regression models follow AR(1) error model (First-Order Autoregressive Error model)24.

All in all, my baseline model employs stationary variables (CR, CF, and FBAS) and AR(1) error model. Given the dynamic relationship between the stationary variables with an AR(1) error model, I use the Generalized Least Squares (GLS) to estimate my regression models. This estimation procedure first transforms the original models in Eqs. (1) and (2) with the autocorrelated error term

e

t into a new model that has an error term

v

t that is

uncorrelated over time, and then applies least squares to these transformed models. Also, to verify whether my estimation strategy demonstrated above is appropriate, I estimate some variants of the baseline model.

4. Results

In my narrow (N=6) and long (T=52) panel dataset consisting of quarterly data, there are on average 300 observations for each variable. The detailed descriptive statistics and bivariate correlations are reported in Appendix D.

4.1. Foreign bank presence and credit growth

Table 1 shows the results for the impact of foreign bank presence on credit growth regarding hypotheses 1a, 1b and 1c. The estimates of the baseline model (Ⅳ) are reported in column (7) - (9). The results of other three specifications (Ⅰ) - (Ⅲ), which are estimated in order to check whether my baseline model presents the robust results, are reported in column (1) - (6).

23

I perform Durbin-Watson test per country. The results shows that there exists autocorrelation in the regression models for Japan, Korea, Malaysia, and Thailand in case of CR and in Korea, Malaysia, Singapore, Thailand in case of CF.

24

(17)

16 I start with the results of the baseline model. From the table, it follows that the FBAS, an indicator of foreign bank presence, has a statistically significant and positive effect on the credit growth in the real- and the financial-sectors (CR and CF, respectively) at the 1% and 5% level, respectively. That is, an increasing share of foreign banks is positively associated with the credit growth in both sectors strongly supporting the hypotheses 1a and 1b.

Then, in which sector is the influence of foreign bank presence on credit growth larger? I firstly compare the marginal effects (dCR/dFBAS, dCF/dFBAS) in the regression models for

CR – in column (7) and for CF – in column (8). They are 0.00315 and 0.00355, respectively. The marginal credit growth in the financial sector obtained as a result of 1% increase in FBAS is slightly larger than that in the real sector. However, the difference between them is small (0.00040). Also, the standard deviation of coefficient of FBAS in the regression model for CF is larger, which implies that coefficient of FBAS in the model for CF has lower probability of being close to its true value. Taken together, it is difficult to say that foreign bank presence is more influential in the financial sector than in the real sector. Rather, the impact of foreign bank presence might be statistically similar in both sectors. In addition, I compare coefficients of FBAS in the regression models for CR and for CF in the economic context. In this case, it should be taken into account that CR and CF, my dependent variables, are % growth and the average growth of credit growth in each sector is different. That is, we need to know how large these coefficients are compared to the average growth of credit growth in their corresponding sectors. While the average growth of credit growth in the real sector is 0.00976%, the one in the financial sector is 0.0052%. Hence, the coefficient of FBAS in the regression model for CR, 0.00315 % additional growth obtained as a result of 1% increase in

FBAS, implies about 32% additional growth given the average growth of CR. By contrast, the coefficient of FBAS in the regression model for CF, 00355 % additional growth is about 68% additional growth given the average growth of CF. In other words, the influence of foreign bank presence on credit growth is approximately twice larger in the financial sector than in the real sector from the economic perspective. As shown above, the statistical and economic implications are not consistent. That is, it is not clear in which sector the influence of foreign bank presence on credit growth is larger and thereby hypothesis 1c is inconclusive.

(18)

17 effects of foreign bank presence on credit growth, only GDP per capita (% growth) variable enters negatively and significantly in the regression model for CR – in column (7) and GDP

per capita and financial openness variables in the regression model for CF- in column (8). Higher GDP per capita leads to higher credit growth. Yet, this positive relationship between them is not linear. In reality, as GDP per capita increases, credit growth rises rapidly at first, and then begins to increase at a decreasing rate indicating a nonlinear relationship between them. Also, given that my dependent variables (CR and CF) are the growth of the growth of credit (e.g. ∆CRt = CRt –

CR

t-4, where CR is % growth in credit flows to the real sector

relative to GDP), they can be obtained by the slope of the curve between GDP per capita and

credit growth. In economic words, the countries with higher GDP per capita, which are closer to the maximum of credit growth, have relatively slower credit growth. In case of

financial openness, it is measured by the sum of foreign assets and liabilities. While foreign assets are positively related to credit growth, liabilities are negatively. The possible explanation for the negative sign of financial openness is that whereas the positives effect of foreign assets on credit growth is already captured by FBAS, the negative effect of liabilities on credit growth still remains25.

4.2. Foreign bank presence and credit growth during a crisis

Next, table 2 shows the results for the effect of foreign bank presence on credit growth especially during a crisis regarding hypotheses 2a and 2b. FBAS still has a statistically positive and significant effect on credit growth in the real- and the financial-sectors at the 1% and 5% levels, respectively. Yet, Crisis dummy variable comes out with the unexpectedly positive signs, but is insignificant. Since there were only one or two episodes of crises per country during 1996-2008, Crisis dummy variable has only around 10% variation. Possibly,

Crisis dummy is not significantly different from zero due to this data limitation.

Importantly, I include the interaction term between FBAS and Crisis dummy to test whether foreign bank presence alleviates the (supposed) negative effect of the crisis26 on

25

When estimating the regression equations with excluding FBAS variable, the sign of coefficient for financial

openness becomes positive in the regression models for CR.

26

(19)

18 credit growth in the real- and the financial-sectors. This interaction term is positive and significant at the 5% level in the regression model for CR - in column (7) - supporting the hypothesis 2a. That is, the higher level of foreign bank presence, the (supposed) adverse effect of a crisis on credit growth is less pronounced in the real sector. Here, it should be kept in mind that the result of this interaction term may be only an approximation for the potential impact of foreign bank presence on credit growth in the real sector during a crisis because of data limitation such as Crisis dummy variable having only around 10% variation. Even though there is this shortcoming, this is an important result providing more solid evidence in favor of foreign bank presence. Often voiced is the concern that foreign banks will not have an attachment to domestic borrowers (Peek and Rosengren, 2001). Thus, foreign banks faced with the adverse economic shock such as a crisis can shift their capital to elsewhere. Yet, the significant and positive effect of interaction term between FBAS and Crisis suggests that foreign bank presence helps sustain credit growth even during a crisis. On the other hand, the interaction term is insignificant in the model for CF – in column (8) and thus hypothesis 2b is not supported. The possible explanation for insignificant effect might be that credit flows in the financial sector is (potentially) more volatile during a crisis compared to those in the real sector, thus the role of foreign bank presence in moderating the negative effect of crisis may be less effective. Conclusively, the significant interaction term for CR suggests that the positive relationship between foreign bank presence and credit growth persists only in the real sector when there is a crisis.

Finally, I estimate the effect of FBAS on the total credit (CT) to compare the predictive power of disaggregated credit flow measure (CR and CF) to aggregate one – estimates are in column (9) in both table 1 and 2. With the aggregate measure, I also find the statistically significant and positive effect of FBAS on the total credit. Yet, interaction term is insignificant. That is, aggregate measure fails to find evidence on whether foreign bank presence alleviates the (supposed) negative effect of crisis. Hence, disaggregated measure better performs when uncovering the relationship between foreign bank presence and credit growth.

(20)

19

Other three specifications

Turning to estimations of other three specifications (Ⅰ) - (Ⅲ) of the baseline model in the table 1 and 2, model specification (Ⅰ) presents a LSDV (Least Squares Dummy Variable)27 model with nonstationary variables and not correcting for autocorrelation problem. Since this model uses nonstationary variables, I expect that the results of it may spuriously indicate a significant relationship between foreign bank presence and credit growth. Unlike my expectation, whereas FBAS has the positive and significant effect on CR - in column (1), it does not on CF - in column (2). Yet, the results clearly show that 4 out of 5 control variables are significant, especially in the model for CR. In terms of control variables, this model is

said to be spurious and less suitable. Model specification (Ⅱ) uses stationary variables - CR,

CF, and FBAS - and addresses autocorrelation with ARDL (Auto-Regressive Distributed Lags) model28. The estimates of this model are similar with those of the baseline model. This model finds the positive and significant effects of FBAS on both CR and CF. Moreover, it does find the significant effect of interaction term between FBAS and Crisis dummy in the model for CF- in column (4), whereas the baseline model does in the model for CR- in column (7). Overall, this model is appropriate same as the baseline model, but the baseline model is slightly better than the model (Ⅱ) in a sense that it finds significant results at the higher level and more control variables are also significant in the baseline model. Lastly, with respect to stationarity strategy, all the variables are differenced to make them stationary29 in model specification (Ⅲ). This model finds the significant and positive effect of FBAS on CR - in column (5), but it does not find it on CF- in column (6). As stated in estimation methodology subsection, this might be due to the disadvantage of differencing, which increases the standard deviation of variables to high levels and thus makes it difficult to find the results. Hence, this model is also less suitable.

27

The LSDV model (also called the Fixed Effects Model) is used to estimate the panel data because the countries in my sample are not selected at random from a pool of worldwide countries. The counties in South Asian region, which were to a certain degree under the influence of 1997-1998 Asian crisis are chosen

28 In contrast to an AR(1) error model, which explicitly including lags through an autocorrelated error, ARDL model can capture the same dynamic effects by simply adding lagged values of Yt-1 and X t-1 to the original linear equation. Whereas the AR(1) error mode is estimated by the GLS, ARDL model by OLS.

29 In model specification (Ⅲ), I convert nonstationary control variables as well as the credit (CR and CF) data and FBAS variables to its stationary counterparts by taking the difference. GDP per capita and Banking sector

(21)

20

Table 1. Foreign bank presence and credit growth & allocation

Note: Standard errors are within parentheses. ***1%, **5%, * 10% significance

The estimates of country- and time-dummy (in case of LSDV, time dummy) is not reported in the table because they are almost insignificant in each model. (Ⅰ)

Nonstationary

not correcting for autocorrelation (LSDV)

(Ⅱ)

Stationary, ARDL (LSDV)

(Ⅲ)

Stationary (all var.), AR(1) (GLS)

(Ⅳ)*Baseline model

(22)

21

Table 2. During a crisis foreign bank presence and credit growth & allocation

Note: Standard errors are within parentheses. ***1%, **5%, * 10% significance.

The estimates of country- and time-dummy (in case of LSDV, time dummy) is not reported in the table because they are almost insignificant in each model. (_Ⅰ)

Nonstationary

not correcting for autocorrelation (LSDV)

(_Ⅱ)

Stationary, ARDL (LSDV)

(Ⅲ)

Stationary (all var.), AR(1) (GLS)

(_{Ⅳ)*Baseline model}

(23)

22

5. Conclusions

A sizeable body of research has found that foreign bank presence helped to sustain credit growth, although some studies have provided evidence to the contrary. However, this positive evidence is relatively scarce for Asian countries. Therefore, this paper aims to investigate whether the foreign bank presence has a positive effect on credit growth in Asian credit markets, particularly during a crisis. For this study, I disaggregate credit flows to the real- and the financial sectors to see whether the influence of foreign bank presence is different per sector.

The results clearly show that a higher level of foreign bank presence is positively associated with credit growth in both sectors. Yet, it is not clear in which sector this positive influence of foreign bank presence on credit growth is larger. Although foreign bank presence is more influential in the financial sector than in the real sector in an economic context, there is not statistically obvious difference between them. Next, when there is a crisis, the outcomes are different depending on sectors. The results indicate that when the presence of foreign banks is larger, the (supposed) adverse effect of the crisis on credit growth in the real sector is less pronounced, but fail to find this evidence for the financial sector. Hence, during a crisis, the positive influence of foreign bank presence on credit growth persists only in the real sector.

The overall picture that emerged from the findings suggests that foreign banks also play a positive role in the Asian credit markets by promoting the credit growth. However, this evidence is more solid for the real sector, since foreign bank presence is firmly positively associated with credit growth regardless of whether there is a crisis. In addition, the different findings per sector indicate that the disaggregated measure for credit flows better perform than the traditional aggregate one in uncovering the link between foreign bank presence and credit growth.

Limitations

(24)

23 provided by Micco et al. (2005). Because of this, I use the average of foreign bank asset share over 1996-2002 for the period from 2003 to 2008 (except Singapore) assuming the moderate change in foreign bank asset share. In Singapore, foreign bank asset share substantially increased in 2000 to almost 50 %, and had kept the level of over 40% for the two consecutive years. In case of using the average value, foreign bank asset share sharply drops to 20.3%, which is against a moderate trend assumption. Given this, I thus use the value of last year for the period after 2002. Thirdly, crisis data has low variation (approximately 10%) in it. Lastly, the disaggregated measure for credit flows is not perfect. Yet, with the present reporting conventions of central banks on bank loans, this method following the work of Werner (1992, 1997) could be the second-best.

Future research

(25)

24

References

Aisen, A., and Franken, M. 2010. Bank credit during the 2008 financial crisis: A cross-country comparison. IMF working paper/10/47

Bezemer, D. and Werner, R. 2009. Disaggregated credit flows and growth in Central Europe. Munich Personal PePEc Archive.

Cho, D., and Hong, K. 2001. Currency crisis of Korea: Internal weakness or external interdependence? In T. Ito & A. O. Krueger (Eds.), Regional and global capital flows:

Macroeconomic causes and consequences, East Asia seminar on economics: Vol. 10. Chicago: University of Chicago Press.

Claessens, S., Demirgüç-Kunt, A., and Huizinga, H. 2001. How does foreign entry affect the domestic banking market?. Journal of Banking and Finance, 25: 891–911.

Clarke, G., Cull, R. and Martínez Pería, M.S. 2006. Foreign bank participation and access to credit across firms in developing countries. Journal of Comparative Economics, 34: 774-795 Crystal, J.S., Dages, B.G., and Goldberg, L.S. 2001. Does foreign ownership contribute to sounder banks in emerging markets? The Latin American experience. Federal Reserve Bank of New York, May.

Crystal, J.S., Dages, B.G., and Goldberg, L.S., 2002. Has foreign bank entry led to sounder banks in Latin America? Current Issues in Economics and Finance 8: 1–6.

Cull, R., and Martínez Pería, M.S. 2007. Foreign bank participation and crises in developing countries. World Bank Policy Research Working Paper 4128, February

Dages, G., Goldberg, L., and Kinney, D. 2000. Foreign and domestic bank participation in emerging markets: Lessons from Mexico and Argentina. Economic Policy Review (Federal

Reserve Bank of New York), 6: 17–36.

De Haas, R.T.A., and Van Lelyveld, I.P.P. 2002. Foreign bank penetration and private sector credit in Central and Eastern Europe. De Nederlandsche bank, Staff Reports

De Haas, R.T.A., and Van Lelyveld, I.P.P. 2006. Foreign banks and credit stability in Central and Eastern Europe. A panel data analysis. Journal of Banking & Finance, 30: 1927-1952. Detragiache, E., and Gupta, P. 2006. Foreign banks in emerging market crisis: Evidence from Malaysia. Journal of Financial Stability 2(3): 217-242

Detragiache, E., and Gupta, P., and Tressel, T. 2006. Foreign banks in poor countries: Theory and evidence, IMF working paper No. 06/18, Washington D.C., January

(26)

25 The engine of growth for Central and Eastern Europe?, Europainstitut, Working paper No. 69, December

Eller, M., Haiss, P., and Steiner, K. 2006. Foreign direct investment in the financial sector and economic growth in Central and Eastern Europe: The crucial role of the efficiency channel. Emerging Markets Review, 7, pp. 300-319

Grabel, I. 1995. Speculation-led economic development: a post-Keynesian interpretation of financial liberalization programmes in the Third World, International Review of Applied

Economics, 9(2): 127–149.

Grabel, I. 1998. Rejecting exceptionalism: reinterpreting the Asian financial crises, in J. Michie and J.Grieve Smith (Eds), Global Instability and World Economic Governance. London: Routledge.

Haber, S., and Musacchio, A. 2005. Contract rights and risk aversion: Foreign banks and the Mexican economy, 1997-2000. Stanford University.

Inklaar, R., and Yang, J. 2010. The severity of financial crises and tolerance for uncertainty. This research is supported by European Commission as part of the 7th Framework Programme, Theme 8, “Socio-Economic Sciences and Humanities”

Jeon, Y., and Miller, S.M. 2001. Domestic and foreign bank performance in Korea during the Asian financial crisis. Paper presented at the Allied Social Science Association/Korea–

American Economic Association meetings.

Jeon, Y., and Miller, S.M., and Natke, P.A. 2006. Do foreign bank operations provide a stabilizing influence in Korea?. The Quarterly Review of Economics and Finance, 46: 82-109. Kaminsky, G. L., and Reinhart, C. M. 2001. Bank lending and contagion: Evidence from the Asian crisis. In T. Ito & A. O. Krueger (Eds.), Regional and global capital flows:

Macroeconomic causes and consequences, East Asia seminar on economics: Vol. 10. Chicago: University of Chicago Press.

Laeven, L., and Valencia, F. 2008. System banking crises: A new database. IMF Working

Paper, WP/08/224.

Law, S.H. 2008. Does a country’s openness to trade and capital accounts lead to financial development? Evidence from Malaysia. Asian Economic Journal, 22(2): 161-177

Micco, A., U. Panizza, and M. Yañez, 2005, Bank ownership and performance. Research Department Working Papers; No. 518 (Inter-American Development Bank).

(27)

26 matter?. Journal of Banking and Finance, 31(1): 219-241.

Moreno, R., and Villar, A. 2005. The increased role of foreign bank entry in emerging markets. Bank for International Settlements (BIS), Paper No. 23

Morgan, D., and Strahan, P.E. 2004. Foreign bank entry and business volatility: Evidence from U.S. states and other countries. In: Ahumada, L.A., Fuentes, J.R. (Eds.), Banking Market Structure and Monetary Policy. Central Bank of Chile, Santiago, pp. 241–269.

Peek, J., and Rosengren, E.S., 2000. Implications of the globalization of the banking sector: The Latin American experience. New England Economic Review (September/October), 45– 63.

Schularik, M., and Taylor, A. 2009. Credit Booms Gone Bust: Monetary Policy, Leverage Cycles and Financial Crises, 1870–2008. National Bureau of Economic Research, NBER Working Paper, No. 15512 (Cambridge, Massachusetts: MIT Press).

Stiglitz, J. 1994. The role of the state in financial markets, in Proceedings of the World Bank

Annual Conference on Development Economics, 1993: Supplement to The World Bank

Economic Review and The WorldBank Research Observer. Washington, D.C.: World Bank,

pp. 19–52.

Tornell, A. 2001. Lending booms and currency crises: Empirical link. In T. Ito & A. O. Krueger (Eds.), Regional and global capital flows: Macroeconomic causes and consequences.

East Asia seminar on economics: Vol. 10. Chicago: University of Chicago Press.

Tornell, A., and Westermann, F. 2002. Boom-Bust Cycles in Middle Income Countries: Facts and Explanation. IMF Staff Papers, Vol. 49, IMF Annual ResearchConference (2002), pp. 111–155 (International Monetary Fund: Washington).

Weller, C.E. 2007. The presence of multinational banks and the supply and quality of credit in emerging economies. International Review of Applied Economics, 21(2): 273-292

Werner, R. 1992. Towards a Quantity Theorem of Disaggregated Credit and International Capital Flows. Paper presented at the Royal Economic Society annual conference, York, April 1993

(28)

27

Appendix A. Data definitions and sources

Variables Definition Source

Dependent/Independent variables

Credit growth CR

CF

CT

Foreign bank asset share

Crisis Control variables GDP per capita Interest rate Bank concentration Deposit Financial openness

% growth in credit flows to a real sector relative to GDP (GDP is at 1990 constant prices)

(1996-2008, quarterly )

% growth in credit flows to a financial sector relative to GDP (GDP is at 1990 constant prices) (1996-2008, quarterly )

Sum of CR and CF (1996-2008, quarterly )

Share of banking sector assets held by foreign banks in a country (1996-2002)

Bank, currency or debt crisis (0/1)

% growth in GDP per capita (1996-2008)

Lending interest rate (1996-2005)

Asset share of the three largest banks (1996-2007)

The ratio of deposits to GDP

Foreign assets + liabilities (% of GDP)

The central bank of each country

Micco et al. (2005)

Laeven and Valencia (2008)

UN National Accounts

World Development Indicator (2007)

Financial Structure (2007)

(29)

28

Appendix B. Foreign Bank Asset Share (%)

Source : Micco et al. (2005)

Note: Foreign bank asset share data provided by Micco et al (2005) covers only from 1996 to 2002. The authors have never updated the data. Only Thailand central bank reports foreign bank asset share, but this data is not consistent with the data of Micco et al (2005). Due to this data limitation, I use the average of foreign bank asset share over 1996-2002 for the period from 2003 to 2009 (except Singapore) assuming the moderate change in foreign bank asset share. In case of using the average value for Singapore, foreign bank asset share sharply drops to 20.3%, which is against a moderate trend assumption. Given this, I thus use the value of last year for the period after 2002.

Appendix C. Financial Crisis

Source : Laeven and Valencia (2008), Ranciere, Tornell and Westermann (2006) NA = not informed Country 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Japan 0.2 0.2 0.2 0.2 0.1 0.0 0.0 0.1 0.1 0.1 0.1 0.1 0.1 0.1 Korea, Rep. of 2.1 2.2 5.0 4.7 7.6 4.9 9.2 5.1 5.1 5.1 5.1 5.1 5.1 5.1 Malaysia 24.9 25.0 26.0 23.1 25.4 22.7 22.8 24.3 24.3 24.3 24.3 24.3 24.3 24.3 Philippines 8.2 8.5 8.3 8.6 9.4 9.8 8.3 8.7 8.7 8.7 8.7 8.7 8.7 8.7 Singapore 2.7 1.5 1.5 1.4 46.6 48.3 40.6 40.6 40.6 40.6 40.6 40.6 40.6 40.6 Thailand 6.9 7.1 6.5 10.2 9.0 7.3 6.7 7.6 7.6 7.6 7.6 7.6 7.6 7.6 Country Banking (starting year) Currency crisis (starting year)

(30)

29

Appendix D. Descriptive Statistics and Correlations

Table D1. Descriptive Statistics

Variables Obs. Mean Std.Dev. Min Max

Credit growth CR CF CT ∆CR ∆CF ∆CT

Foreign bank asset share Crisis

GDP per capita Interest rate

Banking sector concentration Deposit Financial Openness 276 276 276 252 252 252 336 336 312 240 288 288 240 0.00013 0.05024 0.01784 0.00976 0.00520 0.00572 12.73095 0.10714 0.39532 7.63123 0.91579 1.03759 2.44737 0.10815 0.15740 0.07816 0.13268 0.19920 0.086292 13.23405 0.30976 0.11939 3.85698 0.87604 0.51812 2.65897 -0.46395 -0.28914 -0.16588 -0.47788 -0.86413 -0.37579 0 0 -0.33642 1.67717 0.24251 0.33883 0.54129 0.35716 0.87513 0.26069 0.51496 0.81616 0.24024 48.3 1 0.27945 16.77742 3.96482 2.30157 9.94563

Table D2. Bivariate Correlations

Variable 1 2 3 4 5 6 7 8 9

1. ∆CR 1

2. ∆CF 0.3764 1

3. Crisis 0.1236 0.0111 1

4.Foreign bank asset share 0.0637 0.0278 -0.0781 1

5.GDP per capita -0.1230 -0.0967 -0.6541 0.1183 1

6. Interest rate -0.0151 0.0432 0.4169 0.1292 -0.3316 1

7. Banking sector concentration -0.0064 0.0472 0.0824 0.0688 -0.2544 0.1050 1

8.Deposit -0.0000 -0.0325 0.0648 -0.2490 -0.1482 -0.7477 0.0078 1

(31)

30

Appendix E. The Bank Credit Growth in 6 Asian countries

Figure E1. The evolution of Bank credit growth for Japan, 1996Q1-2008Q4

-0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 19 96/Q 1 19 96/Q 4 19 97/Q 3 19 98/Q 2 19 99/Q 1 19 99/Q 4 20 00/Q 3 20 01/Q 2 20 02/Q 1 20 02/Q 4 20 03/Q 3 20 04/Q 2 20 05/Q 1 20 05/Q 4 20 06/Q 3 20 07/Q 2 20 08/Q 1 20 08/Q 4 CR CF

Source: Author’s caculations.

Note: CR is the percentage growth in credit flows to the real-sector relative to GDP and CF is the percentage growth in credit flows to the real-sector relative to GDP

Figure E2. The evolution of bank credit growth for Korea, 1999Q1-2008Q4

-0.1 0 0.1 0.2 0.3 0.4 0.5 1996 /Q1 1996/ Q4 1997 /Q3 1998 /Q2 1999 /Q1 1999 /Q4 2000 /Q3 2001 /Q2 2002/ Q1 2002 /Q4 2003 /Q3 2004/ Q2 2005 /Q1 2005 /Q4 2006/ Q3 2007 /Q2 2008 /Q1 2008 /Q4 CR CF

(32)

31 Figure E3. The credit growth for Malaysia, 1997Q1-2008Q4

-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 19 96/Q 1 19 96/Q 4 19 97/Q 3 19 98/Q 2 19 99/Q 1 19 99/Q 4 20 00/Q 3 20 01/Q 2 20 02/Q 1 20 02/Q 4 20 03/Q 3 20 04/Q 2 20 05/Q 1 20 05/Q 4 20 06/Q 3 20 07/Q 2 20 08/Q 1 20 08/Q 4 CR CF

Figure E4. The credit growth for Phillippines, 2000Q1-2008Q4

-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 19 96/Q 1 19 96/Q 4 19 97/Q 3 19 98/Q 2 19 99/Q 1 19 99/Q 4 20 00/Q 3 20 01/Q 2 20 02/Q 1 20 02/Q 4 20 03/Q 3 20 04/Q 2 20 05/Q 1 20 05/Q 4 20 06/Q 3 20 07/Q 2 20 08/Q 1 20 08/Q 4 CR CF

(33)

32 Figure E5. The credit growth for Singapore, 1997Q1-2008Q4

-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 1996 /Q1 1996 /Q4 1997 /Q3 1998 /Q2 1999 /Q1 1999 /Q4 2000 /Q3 2001 /Q2 2002 /Q1 2002 /Q4 2003 /Q3 2004 /Q2 2005 /Q1 2005 /Q4 2006 /Q3 2007 /Q2 2008 /Q1 2008 /Q4 CR CF

Figure E6. The credit growth for Thailand, 1996Q1-2008Q4

-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 1996 /Q1 1996 /Q4 1997 /Q3 1998 /Q2 1999 /Q1 1999 /Q4 2000 /Q3 2001 /Q2 2002 /Q1 2002 /Q4 2003 /Q3 2004 /Q2 2005 /Q1 2005 /Q4 2006 /Q3 2007 /Q2 2008 /Q1 2008 /Q4 CR CF

Foreign Bank Presence and Credit Growth in Asia: A Disaggregated Credit flows Analysis