DO FOREIGN CURRENCY DEPOSITS PROMOTE OR DETER FINANCIAL INTERMEDIARY DEVELOPMENT IN LOW-INCOME COUNTRIES? AN EMPIRICAL ANALYSIS OF CROSS-COUNTRY DATA

Authors


  • A preliminary version of this paper was circulated as Discussion Paper No. 87, Institute of Developing Economies (IDE), January 2007. The author is grateful for the valuable comments and suggestions made by Tsunehiro Otsuki, anonymous referees of the journal, and colleagues at the IDE. Any remaining errors are the responsibility of the author.

Abstract

Foreign currency deposits (FCD) are prevalent in many low-income developing countries, but their impact on bank lending has rarely been examined. An examination of cross-country data indicates that a higher proportion of FCD in total deposits is associated with more private credit only in inflationary circumstances. FCD can lead to a decline in private credit below a certain threshold level of inflation. Given that FCD exhibit persistence, deregulating them in low-income countries could cause more harm than good to financial intermediary development in the long term.

I. INTRODUCTION

The purpose of this paper is to assess the impact of schemes involving foreign currency deposits (FCD) on financial intermediary development in low-income countries. The prevalence of FCD differs substantially among such countries. FCD are prohibited in several countries, but they account for more than half of total deposits in others. As a feature, FCD may be effective in mobilizing savings, especially under circumstances of high inflation. However, mobilized savings in FCD accounts are not always directed toward domestic lending. For example, the balance sheet of the Bank for Foreign Trade of Vietnam (Vietcombank), one of the largest commercial banks in Vietnam, may be observed. In 2001, its foreign assets reached 55% of total assets, which was over 2.5 times that of its domestic lending. The bulk of FCD were channeled into foreign assets in the form of US Treasury bonds. This raises concerns that FCD may provide a channel for capital flight, and this in turn may arrest financial intermediary development.

The relationship between FCD and financial intermediary development is examined in the present paper. As a consensus grows that financial development contributes to economic growth (Beck, Levine, and Loayza 2000; Levine 1997), the question of what accounts for differences in financial development among developing countries arises as an important policy issue as well as a subject of research. In considering factors that may account for the differences in financial development among developing countries, studies have focused on the market share of state-owned commercial banks (La Porta, Lopez-de-Silanes, and Shleifer 2002) and foreign banks (Claessons, Demirgüç-Kunt, and Huizinga 2001) from the viewpoint of corporate governance. Studies have also focused on inflation (Boyd, Levine, and Smith 2001), legal creditor rights (Djankov, McLiesh, and Shleifer 2007), and regulation of the banking industry (Barth, Caprio, and Levine 2004). Apart from De Nicoló, Honohan, and Ize (2003), the impact of FCD on financial intermediary development has seldom been explored in the framework of cross-country data analysis.1

Foreign currency deposits have often been discussed in the context of “dollarization,” the phenomenon in which a foreign currency such as the US dollar is used as a medium of exchange and store of value. In the published literature, focus has been placed on the impact of dollarization on monetary policy (i.e., Baliño, Bennett, and Borensztein 1999) and on the stability of the financial sector (i.e., Burnside, Eichenbaum, and Rebelo 2001). Recently, analyses of the determinants of FCD have been made both empirically (Ize and Levy Yeyati 2003) and theoretically (Broda and Levy Yeyati 2006). A few studies, such as De Nicoló, Honohan, and Ize (2003) and Honohan and Shi (2003), have examined the impact of FCD on bank lending.

In the present paper, focus is placed on financial intermediary development in low-income developing countries. This is based on the assumption that the adverse impact of FCD will be stronger in such countries. Lee (1996) illustrates the process of financial development with particular reference to developing countries. In this process, banks improve credit examinations by learning about their customers through practicing lending. As banks obtain higher profits with improved credit examinations, they return a part of their profits to depositors with a higher interest rate. This in turn leads to further deposits and lending. An argument may be made that in such a process of financial development, depositor access to foreign assets in the early stage of financial development can lead to capital flight and thus deter learning. This can then result in stagnation of financial development. This analysis can be applied to FCD. It is not households but rather risk-aversive banks that allocate FCD to foreign assets in order to reduce risk in their portfolios. The impact of FCD on the learning process may be more significant when banks have inadequate capacity in credit examinations, and this is especially the case for many low-income developing countries.2

The structure of this paper is as follows. In Section II, a simple model of a bank's portfolio based on a mean-variance portfolio model is presented, and the impact of schemes involving FCD on bank lending is analyzed. Section III concerns the prevalence of FCD in low-income countries on the one hand, and the relationship of the real interest rate on local currency deposits with inflation on the other. This section also sheds light on the phenomenon that FCD persist once their proportion in total deposits reaches a certain level. An econometric analysis is performed in Section IV to examine the impact of FCD on the banking sector. In particular, the relationship of the proportion of FCD in total deposits with the deposit-loan ratio, the proportion of foreign assets, and the degree of financial intermediary development are examined. Based on estimations, Section V includes a discussion of the policy issue of whether or not a low-income country should relax regulation of FCD. A summary of analyses and conclusions are presented in Section VI.

II. MODEL OF A BANK PORTFOLIO WITH FOREIGN CURRENCY DEPOSITS

One of the distinctive features of FCD is that households can deposit their foreign currency-denominated assets in local banks without altering the denomination of assets. When, as a result of high inflation, households are not confident in the local currency and local currency–denominated deposits, they may hold a large part of their foreign assets (in the form of foreign currency) outside the formal banking system. In such cases, introducing FCD might be effective in mobilizing foreign currency savings that were otherwise held outside the banking system. However, there is still a question relating to how large a portion of such mobilized FCD may be intermediated to borrowers in the domestic market.

To address this question, a simple model of a bank portfolio in a static partial equilibrium framework is presented below. Here, the economy includes households and a bank. The number of households is normalized to unity.

A. Households

Households may get foreign currency (the US dollar for example) from export transactions or overseas remittances from relatives. Foreign currency may be allocated into three types of assets: (1) local currency–denominated deposits (Lf), (2) foreign currency deposits (F), and (3) foreign currency held in hand (D). Here, Lf is measured in terms of the foreign currency. A regulatory parameter that indicates the extent of regulations on FCD is denoted by φ(0 ≤ φ ≤ 1). Here, a higher φ stands for less restriction and more convenience for FCD. Household demand for F may then be defined as follows:

F ≡ F(φ), F′ > 0. (1)

The above definitions indicate that households tend to prefer FCD when restrictions are deregulated.

Apart from their income in foreign currency, households have local currency income, and they allocate a part of it to local currency deposits, Ll. This is also measured in terms of the foreign currency. Thus, their local currency deposits sum to L ≡ Lf + Ll. The local currency deposit can be assumed to be an imperfect substitute for FCD, and it is a decreasing function of φ:

L ≡ L(φ), L′ < 0.(2)

B. Bank

The portfolio of a bank may be considered in the framework of a mean-variance approach. The bank accepts both local and foreign currency deposits. For tractability of analysis, it may be assumed that there is no reserve requirement for both types of deposits. Furthermore, both lending and deposit interest rates are exogenous.3 Focusing analysis on FCD, it may be further assumed that all local currency deposits are directed to local currency–denominated loans in the domestic market. However, FCD are allocated between foreign currency–denominated loans in the domestic market and US Treasury bonds, with the proportions of z1 and z2, where zi ∈ [0, 1] (i = 1, 2) and z1 + z2 = 1. Therefore, total lending to the domestic market is

C ≡ L(φ) + F(φ) · z1.(3)

The first term on the right-hand side of the equation matches local currency loans (by assumption), and the second term refers to foreign currency loans. Note that all components are measured in terms of the foreign currency.

For given interest rates, the stochastic rates of return of assets may be denoted as follows:

image(4)
image(5)

where RL and RF are the expected returns of local and foreign currency loans, respectively. These expected returns take into account the bank's credit examination capacity and, therefore, the rates of default on loans. As in Lee (1996), the screening capacity is considered to depend on the bank's practice of lending, so that RL and RF are in proportion to the accumulated amount of loans the bank has provided. When the country is still in the early stages of financial development, the bank has less experience, and the expected rates of return tend to become lower for given interest rates. ɛL and ɛF are disturbance terms. For US Treasury bonds, variance may be assumed to be zero, and the expected rate of return is RTB. Hence, the mean and variance of the bank's portfolio, , are inline image

For the utility of the bank, U(), the expected utility function may be defined as

image(6)

where ρ(ρ > 0) refers to the risk aversion coefficient. For a given set of L, F, and the rates of return of assets, the bank maximizes the expected utility with respect to z1, subject to the constraint 0 ≤ z1 ≤ 1.

C. Bank Lending

Substituting the mean and variance of the portfolio into equation (6), solving the maximization problem yields

image(7)

Therefore, for low-income countries where RF is low in relation to RTB because of the bank's inadequate screening capacity, the bank allocates less to loans. The bank holds more US Treasury bonds, and this in turn leads to a slower learning process and slower financial intermediary development.

Next, the impact of FCD on bank lending may be evaluated. Differentiating C with respect to φ yields

image(8)

There are three channels through which the scheme of FCD affects the volume of credit:

  • 1Substitution Effect (between FCD and local currency deposits). There is substitutability between two types of deposits, and deregulation on FCD leads to a reduction in local currency deposits. This in turn results in a decline of local currency–denominated lending (∂L/∂φ < 0).
  • 2Savings Mobilization Effect. In addition to the substitution between local currency and FCD, a deregulated FCD scheme may attract household foreign assets that would otherwise be held outside the banking system (∂F/∂φ ) · &#x007a;̂1 > 0.
  • 3Portfolio Adjustment Effect. Banks may react to changes in regulation on FCD and in the proportion of FCD to local currency deposits. This effect depends on the risk-averseness of the bank. Intuitively, an increase in the proportion of FCD enables the bank to reduce credit risk by increasing US Treasury bond holdings, and this results in a decline in lending.

Substituting (7) into (8) yields (for the case 0 < &#x007a;̂1 < 1)

image((8′))

In this particular case, the whole increment in FCD is channeled to US Treasury bonds. The remaining effects are a decline in local currency loans resulting from the decline in local currency deposits and the associated change in foreign currency loans. Unless the disturbances of foreign and local currency loans are perfectly correlated, a unit of decline in local currency loans leads to less than a unit of increase in foreign currency loans. Therefore, the sign of the term in brackets is positive, and ∂C/∂φ < 0. That is, deregulation on FCD results in less domestic credit, and this in turn implies a slower financial intermediary development in the long term.

It may be appropriate to consider that the substitution effect between FCD and local currency deposits is relatively lower under higher inflation (2L/∂π∂φ < 0). As will be shown in the subsequent section, the real interest rate on local currency deposits tends to be negative under high inflation, so that households will be less likely to change their foreign currency income into local currency deposits under such a circumstance. Accordingly, it may be argued that the adverse effect of FCD on credit is less pronounced under higher inflation.

Based on these results, the following sections include an examination of the effects of FCD on financial intermediary development in low-income countries.

III. FOREIGN CURRENCY DEPOSITS AND LOCAL CURRENCY DEPOSITS

A. Foreign Currency Deposits

Figure 1 shows a histogram of the prevalence of FCD in 76 developing countries.4 These are countries where GDP per capita as of 1995 was less than US$3,000. For the value of the proportion of FCD in total deposits, an average for 2002 through 2004 is used.5 For example, this diagram shows that the proportion of FCD is between 31% and 40% in 13 countries. Notable variation in the prevalence of FCD among developing countries is observed from this diagram.6

Figure 1.

Histogram of the Proportion of Foreign Currency Deposits in Low-Income Developing Countries (Average for 2002–4)

Source: Compiled from the IMF Country Report (various issues) and statistics of the central bank for each country.
Note: The sample consists of countries with a GDP per capita of less than US$3,000 in 1995 and in which data on foreign currency deposits (FCD) is available. The sample includes the following countries:
Transition Economies in Eastern Europe and former Soviet Union: Albania (A), Armenia (A), Azerbaijan (A), Belarus (A), Bosnia and Herzegovina (C), Bulgaria (A), Georgia (A), Kazakhstan (A), Kyrgyz (A), Latvia (A), Lithuania (A), Macedonia, FYR (A), Moldova (A), Romania (A), Russian Federation (A), Tajikistan (C), Ukraine (A)
Middle East and Northern Africa: Algeria (D), Egypt (A), Iran (A), Jordan (A), Morocco (A), Syria (A), Tunisia (A), Turkey (A), Yemen (A)
Sub-Saharan Africa: Angola (A), Burundi (D), Cape Verde (A), Congo, Dem. Rep. (B), Eritrea (C), Ethiopia (A), Gambia (A), Ghana (A), Guinea (C), Guinea-Bissau (A), Kenya (A), Lesotho (A), Liberia (D), Madagascar (A), Malawi (A), Mozambique (A), Namibia (A), Nigeria (E), Rwanda (A), Sao Tome & Principe (C), Sierra Leone (D), Sudan (B), Swaziland (A), Tanzania (A), Uganda (A), Zambia (A)
Latin America and Caribbean: Belize (A), Bolivia (A), Colombia (A), Costa Rica (A), Dominica (A), Dominican Rep. (A), Guatemala (A), Haiti (A), Honduras (A), Jamaica (A), Nicaragua (E), Paraguay (A), Peru (A), Suriname (D)
Asia and Pacific: Bangladesh (E), Cambodia (A), China (D), Indonesia (C), Laos (A), Mongolia (A), Nepal (A), Pakistan (A), Papua New Guinea (E), Philippines (C), Sri Lanka (A), Thailand (C), Tonga (A), Vietnam (A)
(A) = the countries that consist of the sample for Figure 1 and Tables 1, 4, and 5. (B) = the countries that consist of the sample for Figure 1, and Tables 1 and 4. (C) = the countries that consist of the sample for Figure 1 and Table 1. (D) = the countries that consist of the sample for Figure 1. (E) = the countries that do not appear in Figure 1 or Table 1, but consist of the sample for Tables 4 and 5.

The proportion of FCD for each country is summarized in Appendix Figures A.1–A.5. Some regularity may be found in the prevalence of FCD according to region. The proportion of FCD is generally high in transition economies in Eastern Europe and the former Soviet Union. The proportion of FCD is higher in Latin America and the Caribbean region. It is also relatively lower in Asia, the Middle East, and Northern African countries. Among sub-Saharan African countries, the proportion of FCD varies considerably.

From these observations, it is possible to relate the proportion of FCD in total deposits to some economic and social factors.7 First, those countries that have experienced high inflation tend to have a high proportion of FCD. Examples include transition economies and Latin American countries such as Bolivia, Nicaragua, and Peru. Turkey, the Democratic Republic of Congo, and Lao PDR also fall into this category. Second, the high proportion of FCD is observed in countries where there is an unstable political situation, such as civil war. Prime examples are in sub-Saharan Africa and include Angola, Liberia, Mozambique, and Sudan. Other examples are Cambodia and Bosnia and Herzegovina. Third, the proportion of FCD is often high in countries that have a large number of emigrant workers and emigrants who send money back to their home countries. It is possible that these countries deregulate the FCD scheme to encourage inbound remittances for the accumulation of foreign reserves. Examples include Pakistan, the Philippines, Vietnam, and some Central American countries.8

Characteristics of the background of FCD vary from temporary ones, such as high inflation and political instability, to relatively permanent and structural ones, such as foreign remittances from emigrant workers. However, a high proportion of FCD tends to demonstrate inertia regardless of background. Table 1 shows the transition matrix summarizing the change in the proportion of FCD between 1994–96 and 2002–4. Data was collected for 70 countries for both periods. For example, it can be seen from this table that the proportion of FCD was in the range of 10% to 25% in 15 countries in 1994–96. Among these countries, nine moved to the range of 25% to 50% in 2002–4 and one country to over 50%. The proportion of FCD has, in general, an upward trend and inertia in the higher end of the sampled countries. Except for transition economies, once the proportion of FCD reached 25%, it rarely declined among sampled countries. This implies that once FCD spread and their convenience is established, it is difficult to lower their proportion.

Table 1. 
Transition Matrix of the Proportion of Foreign Currency Deposits (FCD) in Low-Income Developing Countries, 1994–96 by 2002–4
 (Number of countries)
FCD Ratio (%), 2002–4 AverageSubtotal
0–1010–2525–5050–100
  1. Source: Same as Figure 1.

FCD Ratio (%), 1994–96 Average
 0–10164 0 020
 10–25 23 9 115
 25–50 0016 925
 50–100 10 0 910
Subtotal197251970

B. Local Currency Deposits and Interest Rates

From the perspective of depositors, one of the drawbacks of local currency deposits relative to FCD is that their real interest rate can be negative under high inflation. In fact, the proportion of FCD in total deposits is high in several countries that experienced high inflation. Regardless of high inflation, however, a counteracting higher nominal interest rate on local currency deposits may eradicate such a disadvantage of local currency deposits. Therefore, it is worthwhile looking into the actual relationship of the real interest rate on local currency deposits with inflation.

Figure 2 plots the inflation rate and the ex post real interest rate on local currency deposits in low-income developing countries for the period of 1994 through 2001. A negative correlation between inflation rate and the real deposit rate is observed, especially under high inflation circumstances. It might be the case that the volatility of price levels becomes larger as inflation goes up, with the result that a change in the nominal interest rate does not always catch up with inflation. This might often bring about a negative real interest rate. Therefore, depositors under inflationary circumstances anticipating a negative real interest rate may prefer FCD to local currency deposits. In other words, the substitutability between FCD and local currency deposits might become lower under higher inflation.

Figure 2.

Inflation and the Real Interest Rate on Local Currency Deposits in Low-Income Developing Countries, 1997–2001

Source: International Monetary Fund, International Financial Statistics CD-ROM.
Note: The real interest rate is obtained by deducting the annual inflation rate from the deposit interest rate on local currency deposits. The sample does not include those for which data are not available.

What is the implication of a negative real interest rate on the relationship of FCD with the bank credit given to the private sector? When the real interest rate is negative, FCD may exert a savings mobilization effect with less of a substitution effect between FCD and local currency deposits. Therefore, when the real interest rate on local currency deposits is negative, FCD may add more to financial intermediation development. The next section will examine this argument quantitatively.

IV. EMPIRICAL ANALYSIS

A. Methodology and Data

Through an analysis of cross-country data from low-income countries, the impact of FCD on bank lending and, therefore, on financial intermediary development in these countries may be examined.

With regard to the behavior of banks, Section II shows that the deregulation of FCD has three channels for exerting influence. These include: (1) a substitution effect between FCD and local currency deposits, (2) a savings mobilization effect, and (3) a portfolio adjustment effect. The aggregate effects of these three channels may include less credit being given to the private sector in the long term, which implies slower financial intermediary development.

What kinds of features might be observed in a bank's portfolio if FCD lead to less private credit? One might be a higher proportion of foreign assets in total assets, and another might be a lower loan-deposit ratio. In addition to the relationship of FCD with financial intermediary development, possible impacts of FCD on the balance sheet of the aggregate banking sector are examined.

To quantify the regulatory conditions of FCD, the FCD ratio is used as a proxy variable. The FCD ratio refers to the proportion of FCD as a percentage of total deposits (demand and savings/fixed deposits) in deposit money banks. This is based on the assumption that more deregulation on FCD leads to a higher FCD ratio. An index of foreign assets of banks ([foreign assets – foreign liabilities]/[total deposits + foreign liabilities]) is calculated in percentages. Foreign liabilities refer to foreign borrowings of banks and do not include FCD. In general, the proportion of foreign assets to total assets is correlated with the proportion of foreign liabilities to total liabilities. In addition, the proportion of foreign liabilities to total liabilities differs substantially among sampled countries. Therefore, an index such as (foreign assets/total deposits) does not clearly reflect the extent of capital flight as a result of FCD. Therefore, the abovementioned index is used instead.

Two variables are used to measure financial intermediary development: (1) the value of deposit money banks’ claims on the private sector divided by GDP, and (2) the value of M2 (the sum of currency in circulation and the demand deposits and the fixed/savings/foreign currency deposits of deposit money banks) divided by GDP. Both “private credit” and M2 to GDP are commonly used in the financial development literature (e.g., King and Levine 1993; Levine, Loayza, and Beck 2000). Since the present analysis focuses on financial intermediary development in terms of bank lending, the former index better corresponds to the analytical framework of Section II than the latter. Taking into account that FCD may divert financial resources to foreign countries while deposits increase, it is worthwhile examining differences, if any, in the impacts of FCD on two measures of financial intermediary development: (1) private credit, and (2) M2/GDP.

Apart from the above variables, regressions include inflation rates, real interest rates on local currency deposits, and per capita income as control variables. Inflation per se is considered to exert adverse effects on financial intermediation (Boyd, Levine, and Smith 2001). Per capita income is often used as a proxy variable for the level of economic development, and it is considered to be positively correlated with the level of financial development (Beck, Levine, and Loayza 2000).

All data, except for the amount of FCD and per capita income, are collected from the International Monetary Fund (IMF), International Financial Statistics CD-ROM. The data on FCD are compiled from various issues of the IMF Country Report and various statistical bulletins of the central bank for each country. For the data on FCD, the earliest period available for a large number of countries is 1994. The data on per capita income is GDP per capita found in the World Bank, World Development Indicator CD-ROM.

Table 2 includes a summary of the descriptive statistics of the variables and their correlation coefficients. As expected, negative correlations are observed for private credit with indices of the foreign assets ratio, the FCD ratio, and inflation. Positive correlations are observed with the loan-deposit ratio and the real interest rate on local currency deposits.

Table 2. 
Descriptive Statistics
 Private Credit (2002–4 Average)M2/GDP (2002–4 Average)Net Foreign Assets Ratio (2002–4 Average)Loan-Deposit Ratio (2002–4 Average)FCD Ratio (1994–96 Average)Inflation (1997–2001 Average)Real Interest Rate on LCD (1997–2001 Average)Ln (GDP p.c.) (1994–96 Average)
  1. Source: Author's calculation.

  2. Notes: 1. Significance level is in brackets.

  3.      2. FCD = foreign currency deposits; LCD = local currency deposits.

Descriptive statistics
 No. of observations6666656566666466
 Mean21.736.013.989.427.123.5−5.2 6.5
 Standard deviation16.421.119.026.723.551.237.4 0.9
 Minimum1.04.5−18.028.00.0−0.7−207.7 4.5
 Maximum71.5118.698.0153.092.8275.316.6 8.2
Correlations
 Private credit1       
 M2/GDP0.7271      
[0.0000]       
 Net foreign assets ratio−0.4470.011     
[0.0002][0.9372]      
 Loan-deposit ratio0.52080.0292−0.65141    
[0.0000][0.8172][0.0000]     
 FCD ratio−0.3193−0.30570.1006−0.17341   
[0.0090][0.0126][0.4254][0.1671]    
 Inflation−0.2411−0.26890.2146−0.24030.27421  
[0.0512][0.0290][0.0861][0.0538][0.0259]   
 Real interest rate on LCD0.1070.131−0.16750.1237−0.1377−0.95821 
[0.4001][0.3023][0.1894][0.3340][0.2780][0.0000]  
 Ln (GDP per capita)0.50690.3548−0.43270.4216−0.1231−0.0852−0.0656 1
[0.0000][0.0035][0.0003][0.0005][0.3249][0.4964][0.6067] 

B. Proportion of Net Foreign Assets and Loan-Deposit Ratios

Regressions were performed to examine the relationship of net foreign assets and loan-deposit ratios with FCD. To deal with endogeneity between dependent and explanatory variables, an average for 2002–4 for the dependent variable of each regression and predetermined variables for explanatory variables (an average of the FCD ratios in 1994–96) were used. For control variables, an average of inflation in 1997–2001 and an average of per capita income in 1994–96 were used.

Results of ordinary least squares regressions are summarized in Table 3. Standard errors of the regressions in this table are White heteroskedasticity robust standard errors. For the regression of the net foreign assets ratio, the net foreign assets ratio appears to be influenced by the level of per capita income but not by FCD. One possible interpretation is that the level of per capita income stands for the proportion of creditworthy borrowers in the economy such that the banks in the countries of the low level of per capita income divert funds from domestic loans to foreign assets. Another interpretation concerns the corresponding accounts for international settlements that the banks maintain in foreign countries. Given that the size of the banking sector is in proportion to the level of per capita income (Levine 1997), the proportion of such assets may tend to get higher in the countries with small banking sectors.

Table 3. 
Estimation Results: Net Foreign Assets Ratio and Loan-Deposit Ratio Thumbnail image of

For regressions involving loan-deposit ratios, coefficients of the FCD ratio have the expected sign but are not statistically significant. Analogous to the net foreign assets ratio, the level of per capita income appears to exert more significant impact on the loan-deposit ratio. The intercept dummy for transition economies is highly significant. The high loan-deposit ratio in transition economies may be related to stagnated savings mobilization. These economies made a transition from a planned economy to a market economy around the late 1980s and the early 1990s. Therefore, the banking sector is relatively new, and savings mobilization has not progressed in comparison to the level of economic development. Furthermore, the banking system has often experienced crises in these economies. In the aftermath of such crises, intensified regulations have required banks to increase paid-up capital. As a result, commercial banks in transition economies tend to have high capital-to-deposit ratios, and these in turn often appear as high loan-deposit ratios.

On the whole, no firm relationship between FCD and the balance sheet structure of the aggregated banking sector is confirmed. Rather, the balance sheet structure appears to be more influenced by macroeconomic conditions and socioeconomic characteristics.

C. Financial Intermediary Development, FCD, and Inflation

The relationship of FCD to financial intermediary development (FID) was examined using the following specification:

image(9)

The dependent variable, FID, refers to two measures of financial intermediary development; namely, the ratios of private credit and M2 to GDP. Two specifications are used as dependent variables: (1) the level of an average percentage point of FID for 2002–4, and (2) the difference in such averages for 2002–4 and for 1994–96. If FCD exerts an adverse effect on FID, it should have a negative correlation with changes in FID. Taking the difference also alleviates the problem of omitted variables. It eliminates fixed effects that would affect the level of FID.

For explanatory variables, as in the previous subsection, predetermined variables are used to handle endogeneity between explanatory variables and dependent variables. For example, the FCD ratio refers to values in 1994–96.

This specification also includes an interaction term of the FCD ratio and inflation. As suggested in Section III, under high inflationary circumstances, the adverse effect of FCD on financial intermediary development may be weak, and FCD can add to it through a savings mobilization effect. β2 is expected to capture such an effect. However, as also suggested in Section II, FCD per se are expected to have a negative effect on financial intermediary development; as is inflation. Therefore, the expected signs of coefficients are β1 < 0, β2 > 0, and β3 < 0.

The following control variables are included in regressions: (1) an average of FID for 1994–96 for regressions with the change in FID as the dependent variable, (2) an average of per capita income for 1994–96 (in logarithm), and (3) an intercept dummy variable for sub-Saharan African countries. The initial level of FID (an average for 1994–96) is expected to reflect convergence, if any, in financial intermediary development among developing countries.

Table 4 includes a summary of estimation results. Regressions (1) through (4) include the level of private credit as the dependent variable. Regressions (5) through (8) include the difference of private credit. While the results between two types of specification of dependent variables are mostly consistent, the fit of the model is higher for those models with the level dependent variable. In consideration of endogeneity between the dependent and explanatory variables, the models with the differenced dependent variables might be preferred in terms of robustness of estimations.9

Table 4. 
Estimation Results: Financial Intermediary Development, Foreign Currency Deposits (FCD), and Inflation Thumbnail image of

Among the models with the differenced dependent variables, the coefficient on the FCD ratio is not statistically significant for Model (5). This might be due to the fact that the specification of Model (5) does not take it into account the varying effect of FCD in accordance with inflation. In contrast, when the interaction term of inflation and the FCD ratio is included, the coefficient on the FCD ratio becomes highly significant. Among regressions (5) to (8), which are nested with one another, regression (8) is selected with the adjusted R2.

For regression (8), the coefficient of the interaction term of inflation and the FCD ratio is positive. This implies that FCD add to financial development when the annual inflation rate is over 43% (β1 + β2 (Inflation) = 0). This result is more or less consistent with De Nicoló, Honohan, and Ize (2003). Although their sample includes not only low-income countries but also middle-income developing and developed countries, De Nicoló, Honohan, and Ize (2003) point out that the threshold level of yearly inflation is in the 20% to 30% range. Because of the adverse impact of inflation per se, however, private credit is stagnant under such levels of inflation.

Table 4 also includes results of regressions with M2/GDP as the dependent variable (regressions (9) to (16)). In contrast to the regressions with private credit (regressions (5) to (8)), the coefficient on FCD is not always negative or statistically significant (regressions (13) to (16)). One interpretation of these results is that because M2/GDP counts the liabilities side of the banking system, this measure of financial intermediary development reflects the savings mobilization effect of FCD more than the portfolio adjustment effect. It is worthwhile noting that the net effect of FCD appears differently on two measures of financial intermediary development: private credit and M2/GDP.

D. Financial Intermediary Development, Foreign Currency Deposits, and the Real Interest Rate

Considering that FCD may be more effective in savings mobilization under a negative real interest rate on local currency deposits, the relationship of FCD with financial intermediary development is reexamined by replacing Inflation in equation (9) by the real interest rate on local currency deposits as below:

image(10)

where Interest Rate stands for an average of the ex post real interest rate on local currency deposits for 1997–2001. The other variables are the same as with previous regressions. The expected signs are γ1 < 0 but γ3 > 0. For the sign on the coefficient of the interaction term, as a negative real interest rate on local currency deposits would alleviate the substitution effect and give an impetus to the savings mobilization effect of FCD, the expected sign is γ2 < 0.

Table 5 includes a summary of estimation results with equation (10). For regressions with private credit as the dependent variable, because the results are more or less consistent between the models with the level dependent variable ((1) to (4)) and those with the differenced dependent variable ((5) to (8)), analysis proceeds with the latter as in the previous sub-section. Among regressions (5) to (8), regression (8) is selected with the adjusted R2. For this regression, all the coefficients of interest have the expected signs. This indicates that FCD contribute to financial development when the real interest rate on local currency deposits is below –9.0, although with low statistical significance. For regressions with M2/GDP, similar results are obtained as with Table 4.

Table 5. 
Estimation Results: Financial Intermediary Development, Foreign Currency Deposits (FCD), and the Real Interest Rate Thumbnail image of

Comparing the results in Tables 4 and 5, inflation appears to be a more important factor for private credit than the real interest rate. Higher inflation per se may exert an adverse impact on private credit, and a higher nominal interest rate on local currency deposits does not completely offset such an adverse impact. Apart from this, the estimated coefficients of regression (8) in Table 5 are consistent with their counterpart in Table 4. That the impact of FCD on bank lending is positive only with the negative real interest rate on local currency deposits is compatible with the fact that it is positive only in inflationary circumstances.

V. POLICY DISCUSSION

The above results indicate that FCD add to private credit under circumstances of medium and high inflation. This may be interpreted as the savings mobilization effect of FCD surpassing their portfolio adjustment (capital flight) effect under medium and high inflation. At the same time, net effects of FCD on private credit appear to be negative under low inflation as the portfolio adjustment (capital flight) effect of FCD seems to exceed their savings mobilization effects.

As a policy prescription for low-income countries that suffer from high inflation and have a large informal foreign exchange parallel market, deregulation on FCD could give an immediate remedy for financial intermediary development. However, this prescription must be applied with caution. Experiences in many low-income countries have shown that once FCD become prevalent, they develop inertia. Therefore, when countries succeed in containing inflation, prevalent FCD might do more harm than good under low inflation circumstances. FCD can then be an obstacle to financial intermediary development in the long run.

In terms of the portfolio adjustment (capital flight) effect of FCD, controlling bank holdings of foreign assets by regulation might not always be a solution for financial intermediary development. Forcing banks to provide loans to domestic borrowers that do not have foreign currency revenues might leave banks with exchange rate risks regardless of whether the denomination of loans is in local or foreign currency (Burnside, Eichenbaum, and Rebelo 2001). Furthermore, creating a worse exchange rate risk may destabilize the banking sector, and this in turn may retard financial intermediary development.

VI. CONCLUDING REMARKS

This paper examined the impact of schemes involving FCD on the financial intermediation of banks in low-income countries. There are three channels through which FCD exert influence on financial intermediation: (1) the substitution effect between FCD and local currency deposits, (2) the savings mobilization effect with which banks absorb the foreign currency that is otherwise held outside the banking system, and (3) the portfolio adjustment (capital flight) effect. In low-income developing countries, the credit examination capacity of banks is underdeveloped, and profitability of lending may remain low due to credit risks. FCD encourage such banks to increase holdings of foreign assets to alleviate high credit risk in their portfolios. As a result, the adverse portfolio adjustment effect of FCD on private credit may be greater in low-income countries.

The prevalence of FCD and the background of the diffusion of FCD differ substantially among low-income countries. Nevertheless, once FCD attain a high proportion of deposits, they develop inertia in many countries regardless of the background.

An econometric analysis indicates that FCD add to private credit under circumstances of medium and high inflation. However, results also imply that the opposite will be the case under circumstances of low inflation. Given that FCD develop inertia, even a country with high inflation should be wary of relaxing regulations on FCD because it may exert an adverse effect on private credit and financial intermediary development after the containment of inflation.

Footnotes

  • 1

    Baliño, Bennett, and Borensztein (1999) offer a valuable survey on FCD in developing countries. Although their study is comprehensive, they do not discuss the impact of FCD on financial development. De Nicoló, Honohan, and Ize (2003) claim to be the first to assess the impact of foreign currency deposits on financial deepening.

  • 2

    There are several empirical studies that focus on low-income countries. Detragiache, Tressel, and Gupta (2006) apply a framework of analysis similar to that of this paper to account for the background of financial development in low-income countries from the viewpoint of foreign and state ownership of banks. Rioja and Valev (2004) also differentiate low-income developing countries from other developing and developed countries in their empirical analysis of the financial development and economic growth nexus. They find that the impact of marginal growth in financial development on economic growth differs between low-income developing countries and others.

  • 3

    The assumption that interest rates are exogenous can be interpreted as being regulated in the context of developing countries.

  • 4

    In this paper, FCD are deposits of residents into foreign currency accounts. In general, the foreign currency deposits of non-residents are not counted as “money” in macroeconomic statistics. Such foreign currency deposits of non-residents are not necessarily substitutable with the local currency, so they are disregarded in this paper.

  • 5

    This average is used to partially alleviate the influences of shocks such as an abrupt decline in the exchange rate on the proportion of FCD.

  • 6

    Apart from the countries included in this histogram, there are virtually no FCD in 12 West African countries that constitute the CFA franc monetary union.

  • 7

    For detailed analysis on the determinants of the prevalence of FCD, see Ize and Levy Yeyati (2003).

  • 8

    A large amount of inbound overseas remittances from emigrant workers also exists in India. However, Indian regulation deals with such foreign income in non-resident foreign currency accounts with a preferred deposit interest rate. These deposits are not counted as “money” in macroeconomic statistics, so they are also disregarded in this analysis. For workers’ remittances in Central and Latin America, see Amuedo-Dorantes and Pozo (2004).

  • 9

    Of course, the problem of endogeneity may be eradicated with the use of instrumental variables for endogenous explanatory variables. However, it remains another problem to find appropriate instrumental variables.

Appendix Fig. 1. Changes in the Proportion of Foreign Currency Deposits by Region

inline image
inline image
inline image

Sources: Compiled from the IMF Country Report (various issues) and statistics of the central bank for each country.

Ancillary