Reason for Reserve? Reserve Requirements and Credit Nada Mora January 21, 2013

Abstract Changes in bank credit can be traced to regulatory shocks such as to capital and reserve requirements. This paper considers the impact of one such policy action on bank credit and its incidence across banks. I make use of a reserve requirement increase in Lebanon that was considerably greater on foreign currency deposits than on domestic currency deposits. All banks cut lending as they scrambled to adjust asset portfolios. But the policy shock disproportionately a¤ected banks with a greater reliance on dollar funding and with low bu¤ers of dollar liquid assets. Exposed domestic-owned banks also adjusted more slowly to the shock than similar foreign-owned banks that obtained outside funding. Matching …rms to their main banks provides descriptive evidence that small …rms were the most a¤ected by the drop in lending. Keywords: Reserve Requirements; Monetary Policy; Bank Credit; Dollarization. JEL Classi…cation: E5, F3, G21, G28.

Economist, Federal Reserve Bank of Kansas City, 1 Memorial Drive, Kansas City, MO, 64198 (E-mail: [email protected]). The views expressed herein are those of the author and do not necessarily re‡ect the positions of the Federal Reserve Bank of Kansas City or the Federal Reserve System. This research was conducted while the author was Assistant Professor at the American University of Beirut (AUB). I wish to thank a number of former students at AUB for their excellent research assistance and dedication; in particular: Roy Doumit, Rana El Helou, Amin Kalawoun, Alice Klat, and Fares Nejm. I also thank Bob DeYoung (the Editor), two anonymous referees, Kristopher Gerardi, participants at the Emerging Market Group Conference in London, the ASSA-MEEA meetings, and seminar participants at the Federal Reserve Bank of Kansas City for helpful comments. I also acknowledge the Lebanese National Council for Scienti…c Research (LNCSR) for a grant that funded this research in 2006. All errors are my own.

1

1

Introduction

Regulatory changes to the banking system can have a nontrivial impact on the balance sheets of banks. In turn, this can translate to changes in bank lending a¤ecting borrowers dependent on banks such as small businesses with relationship loans. For example, China’s central bank frequently raises reserve requirements to directly curb lending growth. Regulatory policy changes also have a distributional e¤ect within the banking system, so that comparing the average bank’s response with that of the non-bank …nancial system is incomplete. Banks di¤er in size, in reliance on retail deposits relative to wholesale funding, in liquid asset holdings, in ownership structure, among other characteristics. As one instance, a recent reserve requirement tightening in China was applied to large banks only.1 Therefore, policy shocks often in‡uence the distribution –not only the availability –of loanable funds and credit to the productive sectors of the economy. The particular case that I take up in this paper is a reserve requirement increase that was signi…cantly greater on deposits denominated in foreign currency than on deposits denominated in domestic currency, and the primary data set that I use to address this question is a panel of over 60 commercial banks operating in Lebanon. The policy change to foreign currency (dollar) accounts had a disproportionate adverse e¤ect on the lending growth of banks with a greater reliance on dollar funding and with smaller bu¤ers of dollar liquid assets available. This paper, therefore, contributes to understanding the real e¤ects of currency-sensitive bank regulation and also speaks to the wider topic of regulatory and monetary policy in emerging market economies (EMEs). The research design has parallels to a paper by Khwaja and Mian (2008) exploiting an aggregate liquidity shock in Pakistan brought about by nuclear tests. They hypothesize that the impact of the aggregate shock varied across banks to the extent that some banks were more reliant on dollar deposits and deposit out‡ows were primarily from dollar accounts because of a convertibility freeze. Bank regulation in Lebanon requires that a larger share of overall local currency (lira) deposits than of foreign currency (dollar) deposits be held as reserves in vault cash and at the central bank. This di¤erence was particularly large prior to June 2001 when the reserve requirement on foreign currency deposits was zero. In contrast, the reserve requirement was 13% on all lira-denominated deposits. On June 2, 2001, Lebanon’s central bank issued circular number 84, which increased the reserve requirement on lira-denominated deposits to 25% on demand deposits but to only 15% on term deposits, which make up the large part of lira deposits (roughly 94% based on central bank data). Concurrently, the requirement on dollar deposits increased from zero to 15%. Therefore, the 1 See

The New York Times, “China moves to curb in‡ation by tightening credit”, February 13, 2010.

2

change to reserve requirements was e¤ectively greater on foreign currency than on domestic currency deposit accounts. This study exploits the variation in di¤erent banks’initial sensitivity to the shock to identify if credit was forced to absorb the policy shock and how banks adjusted their balance sheets. The advantages of using data from Lebanon are the following: 1) Lebanon’s central bank (Banque du Liban, hereafter referred to as BDL) radically changed the law governing commercial bank reserves in 2001 and in so doing a¤ords a natural experiment with clear bank-level implications; 2) the banking system in Lebanon is largely domestically-owned; 3) banks in Lebanon are the most important source of external credit for …rms; and 4) a large part of this credit is in dollars (about 84% in recent years based on statistics published by the BDL).2 These facts, combined, make the Lebanese case suitable to isolate the transmission of reserve requirement policy to bank balance sheets. Reserve requirement policy is actively used in many EMEs for both monetary policy and regulatory …nancial policy purposes. Central bankers in EMEs often confront several competing objectives with limited policy instruments. For example, policymakers may want to target a stable exchange rate and in‡ation rate as well as curb large capital ‡ows and credit cycles. By raising the policy interest rate to tighten monetary policy, the price stability objective may be met but the interest rate increase may attract additional capital in‡ows and further appreciate the domestic currency. As a result, reserve requirement policy has been used in EMEs as a policy tool to e¤ect countercyclical monetary policy in order to limit or cut back money supply and credit growth (through the money multiplier) (see the discussion in Reinhart and Reinhart, 1999; Terrier et al, 2011; Montoro and Moreno, 2011). For example, the People’s Bank of China has raised reserve requirements several times since January 2007 to drain excess liquidity in the banking system. However, in the aftermath of the global downturn in 2008, it lowered reserve requirements to boost domestic credit, as did Brazil, India, and Turkey.3 Even when central banks neutralize the e¤ect of reserve requirement changes on money supply through o¤setting movements in the monetary base (for example, through open market operations), reserve requirement changes serve as a tax on bank intermediation. Such a use is receiving renewed attention not just in EMEs but also in the U.S., in that Kashyap and Stein (2012) show that the central bank can exploit a non-zero and time-varying scarcity value of reserves to tax the negative systemic externality from credit booms.4 Existing evidence of the real 2 The importance of bank intermediation in Lebanon can be seen in a comparative study of the Middle East and North Africa region by Grais and Kantur (2003): Lebanon has the highest share of banking assets to GDP (272%) and the lowest stock market capitalization (10%) in the region. Moreover, domestic-owned banks dominate the banking system, accounting for 80% of system assets. Foreign banks may escape domestic regulatory constraints more easily via cross-border ‡ow of funds within the banking organization, making a study in a country with a foreign-dominated banking system not ideal. 3 See, for example, The Economist, “Hot and bothered”, June 26, 2008; “A monetary malaise”, October 9, 2008; “Not just straw men”, June 18, 2009. 4 Kashyap and Stein show that this can be implemented with a system of reserve requirements on short-term debt, a market for reserves and scarcity of reserves. Thus, reserve requirements can serve as an important policy tool in

3

e¤ects of reserve requirements is taken up in the next section, but is very scarce for EMEs (as also noted by Terrier et al, 2011). There is even less evidence on the e¤ects of di¤erent currency reserve requirements and other currency-sensitive policy in EMEs. Nonetheless, many EMEs confront signi…cant …nancial dollarization that can also blunt the e¤ectiveness of the domestic policy interest rate, leading to the use of alternative policy instruments. For example, the share of deposits denominated in foreign currency exceeds 10% in more than 70 countries (De Nicoló et al, 2005). Financial dollarization persists despite declining and stable in‡ation in EMEs (e.g., the share of foreign currency deposits increased in 1996-2001 from 37% to 48% in Transition countries, from 46% to 56% in South America, and from 53% to 69% in Lebanon; see Gulde et al, 2004). Not only are many EMEs characterized by sizeable foreign currency deposits but credit is typically intermediated in a foreign currency too (Eichengreen and Hausmann, 2005). The application of di¤erent currency reserve requirements in many EMEs can, therefore, impact the cost of funding domestic relative to foreign currency denominated loans. For example, in a separate analysis of a wide range of countries, I …nd that 28 countries apply di¤erent currency reserve requirements; of which, 10 (18) have a higher requirement on domestic (foreign) currency deposits.5 The rest of this paper is organized as follows. Section 2 places this paper within the existing literature and motivates the reserve requirement policy change in Lebanon. Section 3 articulates the hypotheses, sets out the method to trace the e¤ects of the policy change, and describes the data. Section 4 discusses the results, beginning with the aggregate evidence and followed by the bank-level evidence. Finally, Section 5 concludes.

2

Context

As this paper straddles the literature on reserve requirements and that on dollarization, I review the elements of each that pertain to this paper. The speci…c context in Lebanon is described in the last part of this section.

2.1

Context in the Literature: Reserve Requirements

As indicated in the introduction, reserve requirement policy is well and active in EMEs. The large part of the literature has, however, focused on the past experience of the U.S. and the consensus conjunction with the federal funds rate. 5 Speci…cally, among the top 50 dollarized economies (with a foreign currency deposit share of at least 25%), the following have a higher reserve requirement on domestic currency deposits: Uruguay, Lebanon, Macedonia FYR, Slovenia, Philippines, Ghana, Tanzania, and Egypt. While those with a higher requirement on foreign currency deposits are Peru, Paraguay, Turkey, Sao Tome and Principe, Romania, Russia, Honduras, and Pakistan. A detailed table is available upon request, which is based on my interpretation of reserve requirement policy using a database of over 150 countries drawing on Barth et al (2001, updated 2003) and the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions.

4

…nding is that reserve requirement policy is a distortionary tax on depository organizations and intermediary activity.6 It is useful to review this literature because the policy reasons provided in today’s EMEs overlap with those in the historical U.S. For example, one rationale for reserve requirements is to ensure adequate liquidity insurance in the event of funding out‡ows. Robitaille (2011) suggests that reserve requirements –which can be used to meet the Basel III liquidity coverage ratio requirement –may be a preferred option in EMEs because of …nancial market underdevelopment that leaves an undersupply of e¤ectively liquid assets. A second rationale for reserve requirements is to ensure a predictable demand for reserves, which is needed if monetary policy targets direct control over reserves (as is the case in some EMEs and was the case in the U.S. prior to interest rate targeting). For example, an increase in the reserve requirement leads to a lower level of deposits and loans that can be supported for a given monetary base. As discussed in the introduction, policymakers in EMEs often use reserve requirements as part of a countercyclical toolkit to mitigate credit ‡uctuations caused by foreign capital ‡ows. Reinhart and Reinhart (1999) show that central banks in EMEs experiencing capital in‡ows intervene to buy foreign exchange as a way of preventing exchange rate appreciation. To avoid increasing the money supply, central banks try to maintain price stability by concurrent open market operation sales of treasuries and by increases in reserve requirements to keep money supply constant (or to decrease it if credit expansion is deemed excessive).7 Similarly, the adverse e¤ects of capital out‡ows on credit supply can be o¤set by reducing reserve requirements. In a review, Montoro and Moreno (2011) show that such a countercyclical reserve requirement policy helped smooth credit in Brazil and other Latin American countries in the recent global …nancial crisis. But the cost of holding non-interest bearing reserves means foregone interest from potential loans and imposes a tax burden on depository institutions, borne by bank borrowers, depositors and shareholders. The tax impact is present even when central banks move to neutralize the contractionary monetary impact of a requirement-induced shift in the reserve demand schedule (by increasing the supply of reserves). In this way, reserve requirements are often thought of as regulatory policy rather than as part of countercyclical monetary policy (Loungani and Rush, 1995). The tax on bank intermediation e¤ectively drives a wedge between the rate that a bank pays its depositors and its cost of funding. This serves to lower …nancial intermediation and credit by raising the lending to 6 With the exception of the novel argument put forth by Kashyap and Stein (2012) that the reserve requirement “tax” – instead of being distortionary – can indeed serve as a Pigouvian tax to tax the negative externality stemming from excessive short-term debt. 7 Schnabl and Schobert (2009) also discuss why central banks may prefer to apply reserve requirements and not rely exclusively on “market-oriented sterilization” such as open market operations or remunerating (at market rate) required reserves. Open market sales are costly because of the higher domestic market rate than the return on foreign assets. Therefore, central banks try to reduce sterilization costs by increasing reserve requirements (although this comes at the cost of increasing the implicit tax on …nancial intermediation).

5

deposit rate spread, for example.8 Thus, reserve requirements can be used to in‡uence the …nancial condition and credit of depository institutions relative to that of other …nancial institutions. The maintained assumption is that banks cannot easily substitute reservable liabilities with other funding sources due to information frictions, as discussed by Kashyap and Stein (2000) in the context of the bank lending channel. One strand of the literature has focused on the reaction of interest rates and bank equity returns to changes in reserve requirements in order to identify how the tax burden is distributed. If bank deposits o¤er special transaction and liquidity services, then the cost of higher reserve requirements can be passed on to bank depositors in the form of lower deposit rates (Black, 1975). Fama (1985) challenges this view, arguing that banks were also required to hold reserves on CDs, but CDs are easily substitutable with other money market instruments like commercial paper. Fama conjectures that the reserve tax is borne instead by bank-dependent borrowers (see also, James, 1987). That is, bank loans – being inside and not outside debt – must therefore be special. Subsequent studies examined bank abnormal equity returns around periods of reserve requirement changes. For example, Slovin et al (1990) …nd that announcements of increases in requirements lowered bank stock returns and had an opposite e¤ect on nonbank …nancial …rms (the latter gained at the expense of bank shareholders). As Slovin et al point out, this …nding is consistent with the hypothesis that by raising the cost of bank loans, lending is reduced and banking activities are scaled back. The higher reserve requirement leads to a lower franchise value, which is re‡ected in a fall in shareholder wealth upon announcement. My paper’s focus on the variation across banks also has parallels to Cosimano and McDonald (1998) who …nd that banks with more CD …nancing bene…ted more from the reduction in the reserve requirement on CDs in 1990, re‡ected in a greater positive abnormal equity return. In more recent research on EMEs, Gelos (2009) investigates the drivers of the comparatively high intermediation spreads in Latin American countries, …nding that a higher reserve requirement is a signi…cant determinant. Robitaille (2011) discusses how reserve requirement policy in Brazil taxes large banks to subsidize small banks that are exempt, but that over time banks have shifted from demand deposits with a high reserve requirement to other funding sources such as CDs. Using an aggregate VAR framework applied to Brazil, Glocker and Towbin (2012) show that changes in reserve requirements are associated with changes in lending spreads in the hypothesized direction. Indeed, a reserve requirement shock is found to have a greater impact on lending spreads and credit supply than an interest rate shock.9 8 For example, if a bank pay depositors a rate r D and the reserve requirement ratio is denoted by T , then the bank’s cost of funds will be rD =(1 T ) (e.g., Hubbard, 2002). 9 Because their analysis is on aggregate data, they identify the separate interest rate and reserve requirement shocks via the restriction that positive reserve requirement shocks increase reserves (through accommodative adjustment of the monetary base by the central bank) while positive interest rate shocks lower reserves.

6

This last paper closely complements the second strand of the literature, which has focused on the use of reserve requirement policy to in‡uence bank credit and real investment. Romer and Romer (1993) argue that direct credit actions taken by the Federal Reserve in the postwar period –including reserve requirement policy – had a greater impact on bank lending than the conventional credit channel interpretation of monetary policy.10 Similarly, Loungani and Rush (1995) …nd that reserve requirement changes led to signi…cantly lower aggregate C&I lending and investment, controlling for other measures of monetary policy and real equity returns. And in the prewar period, the Federal Reserve doubled reserve requirements in 1936-37 to “prevent an uncontrollable expansion of credit in the future”11 , arguably leading to a severe downturn in 1937-38 because of the resulting monetary contraction (Friedman and Schwartz, 1963). When the Federal Reserve locked up reserves, banks moved to restore their desired liquidity cushion by reducing lending. The putative excess reserves were not excess in an economic sense. Cargill and Mayer (2006) support the Friedman and Schwartz hypothesis by exploiting the variation between Federal Reserve member banks and nonmember banks, both facing similar weak credit demand but with nonmember banks not subject to the change in the reserve requirement. Member banks cut lending relative to nonmember banks and increased their cash ratios. This view has, however, been challenged most recently in a paper by Calomiris et al (2011) that argues that member and nonmember banks di¤ered in the intrinsic determinants of their demand for reserves.12

2.2

Context in the Literature: Dollarization

To brie‡y summarize the elements of the literature on dollarization that are relevant for this paper, Baliño et al (1999) and Gulde et al (2004) discuss how a dollarized banking system helps remonetize an economy in which con…dence in the local currency has been undermined. But dollarization results in a di¢ cult choice for monetary authorities when deciding on appropriate policy targets. Dollarization also increases the fragility of the banking sector and the central bank has less ‡exibility to act as a lender of last resort. As a result, these authors propose that both commercial and central banks hold larger international reserves and arrange external lines of credit. They also make an 1 0 That is, it was not open market operations, by altering the supply of reserves, that greatly a¤ected the opportunity cost of funds available to banks. But credit actions like reserve requirements directly shifted the demand schedule for reserve balances. 1 1 Cited in Goodfriend and Hargraves (1987) as Board of Governors of the Federal Reserve System, Annual Report, 1937, page 2. 1 2 The question of what to do with excess reserves has received renewed attention in the U.S. where excess reserves increased from $1.5 billion to more than $1 trillion by 2009 due to Federal Reserve asset purchases (Keister and McAndrews, 2009). Current proposals for an “exit strategy” are contingent on economic recovery and involve less explicit methods than locking up reserves. In addition to a natural wind-down of liquidity facilities and assets, measures include paying interest on excess reserves, which allows the central bank to target a path for short-term interest rates independently of the level of reserves. That said, a number of economists are critical and fear that excess reserves will be multiplied into more lending and deposits, fuelling money growth and in‡ation (e.g., Feldstein, Financial Times, April 19, 2009; indeed, Calomiris, Wall Street Journal, March 12, 2012, calls for reserve requirements to be raised as an insurance policy against in‡ation).

7

interesting prudential case for imposing higher reserve requirements on foreign currency deposits in order to mitigate the higher systemic liquidity risk not internalized by domestic banks and depositors. Dollar deposits are only partially covered by liquid dollar assets and, unlike domestic currency deposits, no adjustment is possible by having the central bank print money to expand reserves. A number of papers have studied the e¤ect of monetary policy and depreciations on the competitiveness and balance sheet channels of borrowers in dollarized economies (e.g., Chang and Velasco, 2001; Bleakley and Cowan, 2008). In contrast, very little research exists on the e¤ect of currencysensitive bank regulatory policies. Broda and Levy-Yeyati (2006) look at one such aspect, relating to deposit insurance. They show that banking systems in countries like Argentina are excessively dollarized because of the equal treatment of peso and dollar deposit accounts in the event of bank liquidation, meaning that dollar claims experience a valuation gain when the exchange rate is devalued. But abstracting from deposit insurance motives, risk-averse consumers still may prefer to lend in foreign currency to risk-neutral domestic …rms in order to insure against real shocks (Rappoport, 2009). The idea is that depreciations occur at the same time as recessions. In practice, consumers intermediate dollar savings through the banking system. Banks with dollar deposits then are induced to lend in dollars in order to match the currency composition of their assets to their liabilities because of regulatory constraints or internal limits on currency mismatch (Calvo, 2002). The empirical evidence supports currency matching by banks (e.g., Luca and Petrova, 2008; Section 3 in this paper). This additional matching constraint can limit the extent of adjustment to a currency-sensitive policy shock, as discussed in the next section.

2.3

Context in Lebanon

As mentioned in the introduction, the BDL announced an increase in reserve requirements in mid2001. This section illustrates that the regulatory change to reserve requirements was introduced to tighten monetary policy in the face of capital out‡ows, declining foreign assets at the BDL, and pressure on the exchange rate peg. I show how given the monetary environment faced by the BDL, the use of open market operations was not a viable policy tool, and thus the BDL employed its reserve requirement tool. A primary driver of capital pressure was the rapidly worsening …scal imbalance, which led to a government debt to GDP ratio of 164% by 2001 (Schimmelpfennig and Gardner, 2008). Table 1 provides supportive aggregate evidence, where Panel A describes annual data from 1996 to 2004 and Panel B zooms into the period surrounding the reserve requirement change. The exchange rate peg to the dollar survived the pressure (also seen in the rising Eurobond spreads), but as the BDL intervened to sell foreign exchange, its foreign assets declined beginning in 2000 and dropped

8

more in 2001 reaching 0.34 as a fraction of end-of-year BDL assets from 0.66 in 1999. Tight monetary policy also is seen in the increase in the interbank rate from an average 7.6% in 2000 to 9.7% in 2001, as well as the contraction in money supply growth (in particular, domestic monetary aggregates like M1 and M2, but also M3 that includes foreign currency monetary aggregates). Then, the conclusion of the “Paris II”donor conference in November 2002 and its agreement of …nancing support produced a positive signal (e.g., the Eurobond spread decreased from over 1000 bps in October 2002 to 860 by end-November 2002; and the interbank rate dropped from 8.3% to 5.5% over the same period). In response, capital in‡ows surged and net foreign assets at the BDL reversed course as shown in the 2002Q4 …gure in Panel B (also see Schnabl and Schobert, 2009; Schimmelpfennig and Gardner, 2008; Poddar et al, 2006). To better understand the central bank’s policy stance in 2001, Poddar et al (2006) discuss the BDL’s two main targets: 1) the spread between foreign currency deposit rates o¤ered domestically and those in international markets and 2) the spread between local currency and foreign currency deposit rates domestically. The …rst target is to attract su¢ cient capital in‡ows into the country and the second target is to encourage local currency deposits with adequate compensation for currency risk. Beginning with the …rst target, Panel B shows the deposit rates on Eurodollar deposits compared with domestic dollar deposits.13 The spread between dollar deposit rates and the Eurodollar rate steadily increased beginning in early 2001, as domestic dollar rates did not fall as much as international rates, revealing capital out‡ow pressure. Therefore, the tight policy re‡ects an e¤ort by the BDL to maintain in‡ows. To achieve a monetary contraction, reserve requirements on foreign currency accounts can be tightened to lower the multiplier, especially because central banks in dollarized economies have less control over the supply of reserves (i.e., they cannot print foreign money for open market operations). Moreover, the BDL relied on the reserve requirement policy tool to contract money supply because it was forced to keep buying government bonds, as shown by the increase in the BDL’s securities portfolio over 2000-02 in Panel B. This purchase more than o¤set the decline in foreign assets so that the BDL’s balance sheet actually trended up in 2000-02 more so than in prior years. This meant that it could not rely on open market sales to e¤ect a monetary contraction. The second target helps explain why the reserve requirement increase was greater on dollar than on lira deposit accounts. Deposit out‡ows were higher from local currency deposits, resulting in a shift in the overall fraction of deposits denominated in foreign currency from 0.56 at the beginning of 2000 to 0.70 in 2001-02. And as shown in Panel B, the spread between local currency deposits and dollar deposits steadily increased (until the conclusion of Paris II). Applying a higher reserve 1 3 The 1-month maturity Eurodollar rate is the most comparable because the majority of deposits in Lebanon have a maturity of one month or less (Poddar et al, 2006).

9

requirement tax on dollar than on lira denominated deposits supports this spread and promotes lira deposits (or at least worked to mitigate the out‡ow). Putting the pieces together, therefore, the increase in reserve requirements was applied as part of an overall contractionary monetary policy designed to impose a greater tax on foreign currency deposits. Unlike the recent use of reserve requirements in EMEs to sterilize capital in‡ows as discussed previously, the 2001 episode shared a greater similarity with reserve requirement changes in Brazil over the 1994-2001 period. For example, Takeda et al (2005) show that the monetary stance in Brazil was captured by reserve requirement policy in addition to short-term interest rates, and that increases in reserve requirements were associated with periods of …nancial market turbulence (albeit the objective of reducing money supply is common to either case of sterilizing in‡ows or reacting to capital account pressure). Takeda et al also …nd that the lending of small banks was less a¤ected because more exemptions were given to small banks than to large banks. The increase in the reserve requirement was re‡ected in higher bank reserves as shown in Table 1, where bank reserves scaled by total BDL assets increased from an average 0.41 in 2001Q3 to 0.53 by 2001Q4 (the requirement was applied in October).14 Reserves then declined as the multiplier process repeated itself and monetary aggregates contracted. This process was curtailed, however, with the Paris II agreement that resulted in capital in‡ows and an associated increase in reserves – this time excess reserves –beginning in 2002Q4. Section 4 assesses the ‡uctuations in reserves and their di¤erent consequences in detail.

3 3.1

Empirical Design Testable Hypotheses

This section motivates the mechanism by which banks, varying in their initial share of local currency deposits, respond di¤erently to the same reserve requirement shock. This transmission links bank lending with the currency composition of a key component of bank liabilities, namely deposits. To the extent that bank loans are not perfect substitutes with bonds and other sources of …nance, the investment and real activity of bank borrowers will be adversely a¤ected by declines in loan supply (Bernanke and Blinder, 1988). Disruptions also occur because of bank-speci…c relationships so that borrowers cannot costlessly switch between banks. Following Loungani and Rush (1995), loan supply is given by Ls = f (rL ; rB )D(1

T ), where rL is the rate on bank loans, rB is the rate on bonds, D

is the quantity of bank deposits, and T is the required reserve ratio.15 In equilibrium, loan demand 1 4 Speci…cally, these …gures are based on BDL monthly data that show that reserves shot up from 7,424 LBP billion in September to 9,474 by October. Note the data are not split according to the currency denomination (the next section analyzes bank-level data that supports a greater increase in foreign currency reserves). 1 5 As noted by Loungani and Rush, the focus in Bernanke and Blinder’s model was on credit-market shocks that a¤ect f ( ; ) as well as on monetary policy movements in the supply of reserves that a¤ect the loan rate, and by

10

equals loan supply so that equilibrium rates, rL and rB are determined. An increase in T , therefore, reduces loan supply and pushes up lending spreads. Adapting this model to a two-currency banking system, the required reserve ratio is now denoted with a bank subscript i; so that Ti ( ) depends on the initial fraction of D that is denominated in local currency. Comparing two otherwise identical banks, if the requirement increases more on foreign- than on local-currency deposits, the bank with a greater fraction of local currency deposits will experience a lower e¤ective increase in Ti and so its lending supply will not fall as much. As discussed previously, even when the central bank o¤sets the shock by expanding the monetary base, the reserve requirement will nonetheless have real e¤ects by driving a wedge between the rate paid on deposits and the bank’s cost of funding, rD =(1

T ). Again, the bank reliant on local currency

deposits will experience a lower e¤ective increase in its tax Ti and in its cost of funds. Moreover, because banks match the currency composition of their assets to their liabilities and prefer to lend dollar-denominated credit16 , dollar-denominated declines in credit are not o¤set by increases in lira-denominated credit. In this simple model, a reserve requirement increase automatically translates to a cutback in lending because reserve requirement constraints are binding. Thus, bank lending is hypothesized to absorb the shock. In practice, banks may have su¢ ciently large excess reserves to absorb the shock. But as shown in Table 1, average excess reserves scaled by total assets declined substantially from 7.5% in 2001Q3 to 3.1% by 2001Q4. This also masks cross-sectional di¤erences discussed in the next section as well as the possibility that these smaller bu¤ers may no longer be excess in an economic sense (similar to the argument made for the U.S. during 1936-37).17 A second available option to banks is access to alternative funding sources in order to substitute away from reservable deposits. This is di¢ cult in EMEs such as Lebanon in which deposits comprise the bulk of funding (Table A1). Even in the U.S., Kashyap and Stein (2000) show that only large banks can access securities markets and Cetorelli and Goldberg (2012) show that only large banks with global operations are able to lessen the e¤ect of a domestic liquidity shock through cross-border internal ‡ows within the organization. Nonetheless, this possibility is tested in the next section (focusing on possible di¤erences in adjustment by exposed banks across domestic and foreign ownership classes). Finally, banks with higher foreign currency liquid assets should be able to draw down these bu¤ers to absorb extension, investment demand by …rms. These shocks shift the so-called commodities and credit curve in their model. Note also that banks are assumed unable to perfectly substitute reservable deposits with other sources of funds due to frictions. 1 6 In a separate bank survey (results available upon request), a greater number of banks responded that they prefer to lend in dollars compared with liras. The main reasons provided were currency matching considerations and perceived exchange rate risk. 1 7 Indeed, average excess reserve bu¤ers were a relatively stable 7.5% in the several years preceding the reserve requirement change. Banks also moved to restore them by 2003, which supports a deliberate portfolio allocation decision.

11

the shock, similar to the reasoning in Kashyap and Stein (2000).18 Yearly panel regressions are estimated on the roughly 60 banks operating in Lebanon over the 1996-2004 period to test the hypotheses outlined above. This degree of cross-sectional and time variation is needed because analyzing total bank credit in the banking system cannot convincingly answer the research question at hand. For example, there may be a decline in credit because of a decline in the demand for loans and not because of a fall in bank lending supply induced by the regulatory shock. The following speci…cation, therefore, employs a di¤erence-in-di¤erences approach in that it compares the lending growth of banks after and before the reserve requirement change in 2001 as a function of a bank’s deposit currency composition and its liquidity bu¤er, controlling for other characteristics:

ln(Loans)it

=

+ I01;t + 2 F CLAit 1

1 LCDit 1

+

+

2 (I01;t

1 (I01;t

F CLAit

LCDit 1)

+ Xit

1) 1

+ + Yt

(1) 1

+ ui + "it ,

where the dependent variable is the growth of a bank’s lending, the dummy I01 takes on a value of 1 in 2001 and 0 otherwise, LCD is the bank’s share of deposits in local currency, and F CLA is the share of its assets that are foreign currency liquid assets (securities holdings). X and Y capture various controls and their interaction terms, both bank-speci…c Xi (such as capital and size), and macroeconomic Y (such as the Lebanese interbank rate and real economic activity), and ui is the unobserved bank …xed e¤ect. First lags of the explanatory variables, such as the bank’s lira deposit share and liquidity, are used to avoid biasing the results with a possible contemporaneous adjustment by banks.19 This regression provides a simple test of the hypotheses outlined earlier in this section –namely, whether banks with a larger share of foreign currency deposits were more constrained in their lending in 2001 than in other years relative to banks with a smaller dollarized deposit share (Hypothesis 1). That is, loan growth is expected to have been overall negative in 2001 due to the tightening in the reserve requirement but the greater the proportion of liras in a bank’s deposits, the less adversely a¤ected was this bank than an otherwise identical but more dollarized bank (predict 1 8 In Kashyap and Stein (2000), monetary policy (the federal funds rate) is the state variable. Cross-sectional di¤erences in the correlation between a bank’s loans and its liquid bu¤er are then related to the state variable. Analogously, the state variable in this study is the 2001 reserve requirement change and cross-sectional di¤erences in the correlation between a bank’s loans and its local currency deposit share (or its foreign liquid assets) are evaluated. 1 9 Note that unlike simple di¤erence-in-di¤erences (DD) in which an indicator is used to distinguish between the treated and control groups, this regression DD facilitates the study of policies that have di¤ering treatment intensity across groups or individual banks. The intensity of the e¤ect here relates to a bank’s year-before …nancial position. Note also that an advantage of the regression generalization is that one can relax the restrictive assumption that changes in the outcome variable over time must be equal across treatment and control banks in the absence of intervention.

12

1

> 0). Similarly, banks with a higher bu¤er of liquid foreign currency assets should be able to

draw on these to cushion any negative e¤ect on their lending growth (predict

2

> 0) (Hypothesis 2).

The identifying assumption of this empirical strategy is that variation in banks’…nancial positions should be associated with the intensity of the impact, and therefore, the key coe¢ cients of interest center on the interaction of the respective …nancial positions with the policy shift. While the shift in policy represents an economy-wide shock, the coe¢ cient

(or even disaggregated group-year

e¤ects), captures not just the regulatory shock but also other confounding changes such as demand shocks that may impact all banks without regard to the currency composition of the bank’s deposit base.20 The results of equation (1) are discussed in Section 4.

3.2

Bank Data

Annual data on individual bank balance sheets and income statements for 1996 to 2004 were collected from various issues of the annual publication, Bilanbanques, produced since the early 1980s by Freddie Baz of Banque Audi (one of the top three banks).21 These were not available in electronic format so the relevant series were carefully entered by hand by several research assistants and crosschecked. We also collected information on bank mergers and other bank characteristics such as majority foreign ownership (based on shareholder information) from Bilanbanques as well as from direct correspondence with the BDL. There were a number of bank exits and name changes over the sample period that were accounted for in the empirical analysis by creating a unique bank code. Another issue was the treatment of mergers and acquisitions, of which there were 14 occurring in the period. The acquirer’s balance sheet variables will spike upwards on completion of the purchase. To ensure this does not a¤ect the results of the regressions, banks involved in merger activity were pre-merged so that their structure throughout the sample re‡ects the group’s composition at the end of the sample (merger information is available upon request). Descriptive statistics on the Bilanbanques data are shown in Table 2 (key regression variables) and Table A1 (snapshots of average bank balance sheets over time).22 First, among the main stylized 2 0 The inference will be valid as long as the variation in banks’deposit currency composition is itself not endogenous to variation in lending opportunities such as to bank-speci…c demand that may have occurred with the policy shock. To control for this possibility, I control for other bank characteristics and their interaction with the policy shift, in addition to macroeconomic controls. I also focus on di¤erences within various groups (e.g., within the group of domestic- and foreign-owned banks, as well as within group partitions based on banks’ pre-announcement reserve positions). 2 1 The 1996-2004 period is an appropriate sample for the analysis, characterized by relative macroeconomic stability. The earlier period followed the end of the civil war in 1991 and was characterized by high depreciation and in‡ation. Moreover, Bilanbanques had a change in format and expanded the set of variables from 1996 (reported beginning with the 1998 edition). The period after 2005 was also characterized by an increase in instability due to several political assassinations and a war during the summer of 2006. 2 2 Note that most of the data in Bilanbanques were reported in USD millions and disaggregates for some series such as deposits were reported according to domestic- and foreign-currency denominations, both reported in USD millions (the exchange rate was …xed during most of the sample period (Table 1)). There are some discrepancies in the statistics shown in Table A1 in that the sum of the domestic- and foreign-currency items for some of the balance sheet items does not quite add up to the total because of a typically larger set of banks reporting totals than currency

13

patterns, the share of total banking system assets held by domestic commercial banks increased from 74% in 1996 to 85% by 2004 (Table A1). Therefore, domestic banks dominate the banking system, and their role has grown over time. Second, there was a general decline in loans and in securities (mostly government bonds) over the period (e.g., domestic banks’loans decreased from about 30% of total assets to 20% and holdings of securities from about 39% to 30%). Mirroring this decline was an increase in cash and balances held at the central bank in 2001 and an even more marked rise in 2003 (following Paris II …nancing in‡ows) (reserves and excess reserves patterns are discussed in detail in the next section).23 Third, the main source of funding was deposit funding (over 70% for both types of banks as shown in Table A1), of which on average 37.4% was denominated in local currency but with signi…cant variation across banks as shown in Table 2 (standard deviation 21.2%). In addition, foreign banks tend to rely more on interbank loans and funds borrowed from head o¢ ce and a¢ liates in the organization. Fourth, banks matched the overall currency composition of their assets to their liabilities (e.g., the average local currency asset share was 39.2% as shown in Table 2; 45% (30%) for domestic (foreign) banks). And this currency matching holds at the individual bank level in which, as shown in Table 2, the average asset-liability currency mismatch was only 2.5% (standard deviation 4%). Turning to earnings patterns, pro…tability, measured by accounting return on assets, averaged 1% over the sample (consistent with Peters et al ; 2004) but declined over the period as provisions and non-performing loans increased (Tables 2 and A1). Finally, the banking system’s overall interest margin on assets decreased (e.g., from 4.5% to 2.3% for domestic banks), re‡ecting the general shift from lending to safe assets over the period, even though this shift cannot be attributed to a decrease in the net interest margin on loans & advances. For example, the di¤erence between the interest rate received on loans & advances and that paid on deposits remained roughly constant over the period at about 3% (5%) for domestic (foreign) banks.24

4 4.1

Results: The Response to the Reserve Requirement Change Aggregate Evidence

This section focuses on the trends in lending behavior and reserves around the 2001 reserve requirement change using monthly consolidated commercial bank data from the BDL (available from end-1997). The monthly frequency allows for an evaluation of the policy change both at the time disaggregates. 2 3 In addition to bank reserves, securities, interbank assets, and total loans, “other assets” comprise the residual component of assets (consisting of items such as …xed assets, goodwill, and investments in related parties) and is a stable 10% (20%) of average domestic (foreign) bank assets. 2 4 The more favorable spread for foreign banks re‡ects their lower deposit rates (owing to their relatively larger dependence on foreign currency denominated deposits in addition to possibly greater customer con…dence).

14

it was announced in June 2001 and at the time it took e¤ect in October 2001. As a cross-check on the Bilanbanques data, the aggregated individual bank data is found to closely compare with the BDL aggregate data series. That the data match con…rms that the information reported by banks in Bilanbanques is consistent with the (presumably more reliable) information they report to the BDL and therefore using the individual bank data –only publicly available from Bilanbanques –is valid for the bank-level analysis. Figure 1 shows year-on-year lending growth (and by currency denomination). Lending growth denominated in liras increased at the time of the announcement through the end of 2002, whereas foreign currency credit growth decreased and this rate actually turned negative. The decline in dollar credit growth was also re‡ected in a decline in overall credit growth because most credit is extended in dollars (83.4% from Table 2). While the decline in credit growth at the time of the regulatory announcement may have coincided with a general decline in credit over the preceding period (possibly due to demand conditions), the fact that loan growth in liras went up and that foreign currency loan growth slid more sharply in 2001 hints at a regulatory constraint on loan supply at the time. This evidence supports the hypothesis that banks were overall more inclined to lend in dollars before the announcement date in June 2001. As a result, the share of total bank credit in liras rose from about 13% in the year before June 2001, reaching 15% by December 2001, and stabilizing at 18% by 2002. The aggregated annual Bilanbanques data are similar with a more pronounced jump in the lira share: rising from a ‡at 11% in 1996-2001 to 15% by 2002. Figure 2 traces the corresponding adjustment in reserves. While the share of bank credit in liras increased immediately at the time of the announcement in June 2001 (as expected when banks react to news), the ratio of reserves to total assets only increased in October 2001, which is when the law took e¤ect (and indicates that the regulatory change was binding). Reserves, as a share of bank assets, rose from roughly 10% before October 2001 to 13% in October, and 15% by December. The BDL does not report bank reserves according to currency, but Bilanbanques provides the split of reserves by domestic and foreign currency. The latter annual data support the di¤erential impact of the increase in reserve requirements. Reserves in foreign currency increased, on average, from 7% to 11% of assets while those in lira remained largely stable at about 4% of assets (see also Table A1).25 Table 2 also shows the bank-level distribution of excess reserves by currency type for 2000-03, which declined signi…cantly from an average 8.8% (standard deviation 10.1%) in 2000 to 4.3% (standard deviation 6.3%) in 2001 for foreign currency excess reserves, persisting through 2002. While the average bank may have been able to absorb the shock without cutting back loans, the question remains an empirical one because these are the outcomes of deliberate portfolio reallocation 2 5 Note that reserves include voluntary (excess) reserves in addition to required reserves. Therefore, reserve holdings (cash in vault and balances at the BDL) denominated in foreign currency were not zero before 2001.

15

decisions and banks may have withheld new lending to restore bu¤ers. Figure 2 also shows the sharp increase in reserves beginning in December 2002, reaching over 30% of assets and remaining quite high in recent years. As discussed previously, the increase was a result of the positive conclusion of the “Paris II” donor conference, which led to a large capital in‡ow (Poddar et al, 2006; Schnabl and Schobert, 2009). Therefore, the increase was the outcome of a surge in excess reserve balances, not an increase in required reserves as there was no regulatory change during this time. In robustness checks in the next section, I …nd that the increase in reserves in 2003 had no impact on bank lending, which is consistent with it not being a regulatory-driven increase in the demand for required reserves.

4.2 4.2.1

Individual Bank Evidence The Baseline Results

Table 3 presents the estimation results of equation (1), accounting for the panel nature of the sample by including bank …xed e¤ects to capture any bank-speci…c variation in lending not accounted for by the explanatory variables. The standard errors are robust and clustered at the bank level. The dependent variable is the annual growth of loans (excluding substandard and doubtful loans). The results also are robust to using total loans or C&I loans. On average, loans are close to 25% of the average bank’s assets (Table 2; or 27% if the excluded non-current loans are added back). In addition to the key variables of interest (I01

LCD and I01

F CLA), the regressions include a

set of common bank characteristics (capital ratio and size) and a set of macroeconomic factors as controls. Including bank controls helps address the alternative view that the local currency deposit share may be simply picking up the correlation of other bank attributes with loan growth in 2001. And including macroeconomic controls helps address another alternative view that credit demand, not supply, was the driver in 2001 and the a¤ected borrowers happened to be concentrated at banks sensitive to the policy change. Robustness checks include di¤erent speci…cations for these and other factors and neither alternative view is supported by the results. Before focusing on the key cross-sectional di¤erences, the …rst column of Table 3 shows the estimate of the mean impact on loan growth in 2001. Lending growth signi…cantly declined by an average 8.8% at the time of the policy shift. As discussed earlier in Section 3.1, however, this decline may not be fully attributable to the policy shock as there were other economy-wide shocks. The second column, therefore, examines di¤erential changes based on the treatment intensity. The results of the baseline regression in Table 3(2) support the two hypotheses outlined in Section 3.1. The lending growth of banks with an initially higher composition of local currency deposits was less adversely a¤ected by the requirement change, as seen in the statistically signi…cant

16

positive coe¢ cient (0.423) on the interaction term I01

LCD. And banks with higher foreign

currency liquid assets were at least partly insulated from the shock as also seen in the coe¢ cient on the interaction term I01

F CLA (0.522, signi…cant at the 10% level). In line with the adverse

e¤ect on average lending growth in 2001, the coe¢ cient on the 2001 indicator is also negative.26 Note also that the coe¢ cients on the terms LCD and F CLA are insigni…cant and not interesting economically as they re‡ect general correlations such as di¤erences in intermediation technologies and risk preferences. The other controls are generally insigni…cant. Larger banks tend to have lower loan growth on average but there was no signi…cant di¤erence in 2001. Finally, the lending growth of banks more exposed to the local economy (as proxied by their local currency deposit share) generally increases when interbank rates fall or when real economic activity picks up, but these coe¢ cients are statistically not di¤erent from zero at standard con…dence levels.27 The estimated di¤erence-in-di¤erences (DD) e¤ects also are economically signi…cant. One can compare the response of two otherwise identical banks with di¤erent local currency deposit shares, one at the 75th percentile and the other at the 25th percentile (see Table 2). The di¤erence implies that lending growth at the 25th percentile bank declines 10.9% more than that at the 75th percentile bank in response to the policy shock.28 Or one can quantify the insulating e¤ect of moving a bank at the mean value of local currency deposit share to a less exposed value. For example, had the mean bank’s exposure been in the upper quartile, its lending growth would have been 4.8% higher. Similarly, comparing two banks equally exposed to the reserve requirement tax (identical LCD and other attributes), the bank with a greater bu¤er of liquid foreign currency securities is able to absorb more of the shock. Therefore, the bank holding the 75th percentile foreign currency liquid assets experienced a 6% smaller decline in loan growth than the 25th percentile bank, partly o¤setting the adverse impact. In the working paper version available on the author’s website, I present additional speci…cations without bank …xed e¤ects as well as with year e¤ects to sweep out all aggregate shocks. The results are robust to these changes (Table A2). The DD test can be further partitioned according to a bank’s pre-announcement foreign currency reserves. The hypothesis tested in columns (3) and (4) of Table 3 is that lending activity should 2 6 Note however, that the estimated coe¢ cient on the indicator (-0.376) is not meaningful in and of itself to describe the e¤ect on average lending growth in 2001. This is because the indicator shows up in other macroeconomic terms and interaction terms in the regression model. Evaluating the e¤ect at, for example the mean values of …nancial positions, yields a 9% average decline, comparable to the average e¤ect in the …rst column. The common approach in bank-level studies is to focus on the interaction terms and not report the e¤ect of the monetary policy shock (e.g., Kashyap and Stein, 2000; Cetorelli and Goldberg, 2012), likely because of unobserved loan demand di¤erences as discussed by Kashyap and Stein. 2 7 In unreported regressions including the macroeconomic factors but without their respective interaction with LCD, economic activity enters with a positive e¤ect and the interbank rate enters with a negative e¤ect statistically signi…cant at the 10% level. 2 8 Speci…cally, the 75th percentile is 0.487 and the 25th percentile is 0.230. The di¤erence, 0.257 is also roughly equal to the standard deviation of the local currency deposit share (0.212). The di¤erence-in-di¤erences e¤ect is equal to 0.423*0.257 = 0.109.

17

absorb a greater amount of the shock for a bank starting out with low foreign currency reserve bu¤ers than for a bank with high bu¤ers (to the extent that the increase in the reserve requirement should be a binding constraint on the former bank). Empirically, the set of banks in column (3) are those whose pre-announcement foreign currency reserve to asset ratio was below median (average 19992000) while column (4) is for the above-median group. As expected, the …rst set is more sensitive to the policy shock (e.g., the coe¢ cient on the interaction term I01

LCD is a statistically signi…cant

1.131 compared with an insigni…cant 0.185 for the a priori less constrained set; similar e¤ects are found for the interaction term I01

F CLA).

Additionally, columns (5)-(7) explore whether there were signi…cant di¤erences in the reaction between foreign (majority-owned) banks and domestic banks. First, observe that the DD e¤ect relating to variation in dollarized deposits remains statistically and economically signi…cant within each of the respective ownership groups, although the DD e¤ect relating to foreign currency liquid assets is not statistically di¤erent from zero in these smaller samples. In order to statistically test whether the currency composition DD e¤ect di¤ers systematically between foreign and domestic banks, I follow the triple interaction approach applied in Bertrand et al (2007). The results in column (7) indicate that the lending response in 2001 was not statistically di¤erent between the two groups. The interaction term I01

LCD

F oreign is statistically insigni…cant so that a foreign

bank with a similarly low share of local currency deposits as a domestic bank was just as adversely impacted by the shock (signi…cant coe¢ cient 0.515 on I01

LCD). However, what is interesting

is that exposed foreign banks were able to adjust more quickly to the policy shock than similarly exposed domestic banks (discussed in the next section). A second observation on the contrast between foreign and domestic banks is that foreign banks, as a whole, appear to have endured a milder adverse shock surrounding the policy change, possibly due to less adverse demand conditions. But the null of no di¤erence from the average domestic bank cannot be rejected (also see Table A2). Finally, the evidence suggests that larger and better capitalized foreign banks actually reduced their lending more than smaller and less capitalized foreign banks (and these e¤ects are systematically di¤erent from domestic banks). Care should be taken not to overinterpret such di¤erences. The results, however, are compatible with elements of the inconclusive literature on foreign banks’ response to local shocks. This is because healthy foreign banks can be quick to pull out from EMEs when they have favorable investment opportunities elsewhere to deploy their capital. At the same time, healthy foreign banks have been shown to play a stabilizing role too. The topic is beyond the scope of this paper.

18

4.2.2

The Adjustment to the Reserve Requirement Change

This section takes up the question of whether and how the initially a¤ected foreign banks adjusted to the policy shock. Table 4 shows the response of lending growth over a wider event window (200102), in which the results for all banks are in column (1) and domestic and foreign partitions follow in columns (2)-(4). The results in column (1) indicate that bank lending did not recover from the shock by 2002. While the coe¢ cient estimate on I01

02

LCD is smaller than its impact in the …rst

year (0.218), it remains, nonetheless, statistically signi…cant. And foreign currency liquidity bu¤ers continued to dull the impact into 2002. Average lending growth in the 2001-02 period also remained negative (as before, the coe¢ cient on the period indicator does not capture the full e¤ect being picked up by the macroeconomic and other terms: Separate regressions show that average lending growth signi…cantly declined 12.2% throughout this period; 14.1% (8.1%) for domestic (foreign) banks, though the di¤erence between foreign and domestic banks is not statistically di¤erent at standard levels). Unlike the immediate response in 2001, signi…cant di¤erences between the sensitivities of foreign and domestic banks as a function of their …nancial positions are apparent over the longer horizon. The contrasting results on the interaction terms in columns (2) and (3) indicate that adjustment remained on the lending margin for domestic banks but that foreign banks resorted to other available margins of adjustment. For example, the coe¢ cient estimate on I01

02

LCD is a statistically

signi…cant 0.336 for domestic banks but an insigni…cant 0.082 for foreign banks. To test these di¤erences, column (4) con…rms that foreign banks exposed through a more dollarized deposit share were able to undo the shock’s lending impact by 2002.29 The hypothesized explanation is that foreign banks have access to funds from their cross-border parent bank and other a¢ liates. This reasoning is analogous to Cetorelli and Goldberg (2012) who …nd that global banks are able to mitigate the e¤ect of domestic liquidity shocks because they manage liquidity through cross-border internal funding markets. The last three columns of Table 4 test whether the policy shock was indeed absorbed with a liability substitution. The regressions presented are for the set of foreign banks (23 of the total 64 banks). In the interest of space, the corresponding regressions for domestic banks are not shown but the results are insigni…cant for that group. The dependent variable in column (5) is the growth of other …nancing, which includes liabilities under …nancial instruments such as CDs and bonds not counting toward bank capital. And the dependent variable in columns (6) and (7) is the growth of deposits (account balances) placed by the bank’s head o¢ ce and other a¢ liates (due to the even smaller sample size available 2 9 For

example, the null that I01 02 LCD + I01 02 LCD F oreign = 0 cannot be rejected. Note also that the reversal for foreign banks was in 2002. For example, in a regression with two separate year dummies, the coe¢ cient on I01 LCD is positive while I02 LCD is negative.

19

for this variable, the speci…cation in column (7) reestimates column (6) dropping the insigni…cant controls). The results support the hypothesis that foreign banks less vulnerable to the policy shock were less active in seeking to raise other funding. For example, the coe¢ cients on the interaction term I01

02

LCD are signi…cantly negative (-0.486 in column (5); -1.867 in column (7)). In other

results, there is evidence that part of the shock was also absorbed on the asset side by the more exposed foreign banks decreasing their investments in a¢ liated companies.30 To quantify the escape route taken by foreign owned banks, one can compare the initial economic impact on a (foreign) bank’s lending growth to the additional funding it obtains. Continuing the example in the previous section, we saw that moving a bank at the mean value of local currency deposit share to a less exposed value such as to the upper quartile would have led to about a 5% higher lending growth. Such an e¤ect translates to about $8 million in new credit for the average foreign bank in 2001. Likewise, moving this bank to the upper quartile would have reduced its other …nancial borrowing and borrowing from a¢ liates in a range between $5.5 and $8 million over the 2001-02 period.31 Therefore, the evidence supports the view that foreign banks were able to bounce back more quickly, largely owing to their access to non-deposit funding from cross-border a¢ liates and other debt issuance. However, it does not follow that the adverse e¤ect on lending in the economy would have been undone by 2002. First, domestic banks are the leading source of credit in Lebanon (as they are in many other EMEs as shown in Barth et al (2001)). As discussed previously, the share of total banking system assets held by domestic banks was on average 80% over the period. Second, lenders are not perfectly substitutable due to informational frictions and other non-negligible switching costs. Therefore, the results of this paper are a cautionary tale for policymakers. To explore the credit distributional e¤ects in more detail, I rely on information about …rm-bank relationships from Kompass, a business directory database (with the caveat that the analysis is solely descriptive based on a randomized 10% sample from the database of about 8,000 companies). The advantage of this alternative data source is that information on companies such as industrial sector, size (number of employees), and age (date of establishment), in addition to the names of their main banks is provided. Its limitation is that no information is reported on the quantity of loans extended by a particular bank to a borrower. Tabulating the …rm-level data shows that the fraction of …rms 3 0 There is no evidence that the more vulnerable foreign banks signi…cantly borrowed more on the interbank market (while Table A1 indicates that interbank borrowing increased on average for foreign banks, there does not appear to be signi…cant cross-sectional di¤erences; similarly, the cross-sectional di¤erences identi…ed for head o¢ ce and other funding are not apparent in Table A1). Finally, a¤ected domestic banks issued more subordinated and long-term debt in 2001-02. 3 1 The mean foreign bank’s lending was $162.1 million in 2000. And using the DD estimates from Table 3(7) and substituting for the mean (upper quartile) foreign bank’s local currency deposit share of 0.250 (0.347) yields $8.1 million in additional credit. A similar methodology based on the DD estimates in columns (5)-(7) of Table 4 yield the $5.5 to $8.0 million range for total other borrowing, depending on the estimate used.

20

whose main bank is a domestic bank is 0.72, which compares well with the fraction of total system loans in Bilanbanques accounted for by domestic banks (0.75) (the results table is available upon request). Breaking down the data further, domestic banks are found to have a higher incidence of loans to the smallest …rms (0.28) than foreign banks have to the smallest …rms (0.21). Similarly, a greater incidence of borrowers from domestic banks were young …rms. Therefore, the matched data a¤ords descriptive evidence that …rms with high agency costs experienced the more persistent credit cutbacks a¤ecting domestic banks in 2001-02. 4.2.3

Robustness Checks

I conducted additional robustness tests with alternative measures of bank controls (e.g., risk-based capital, pro…tability, local currency loan share) and macroeconomic controls (e.g., alternative growth estimates). The results are available in the working paper (Table A3). Hypotheses such as a di¤erential demand for credit or a di¤erential reaction to aggregate capital ‡ows do not add up to convincing alternative explanations. Brie‡y, of note, are two results. First, banks more reliant on wholesale funds (interbank borrowing) experienced pressure in 2001 as re‡ected by a pullback in loan growth, signi…cant at the 10% level. Nonetheless, this e¤ect was not responsible for the robust currency composition DD e¤ect. That is, the possibility that banks reliant on local currency deposit funding were relatively immune to the policy shock in 2001 because of their lower dependence on interbank borrowing (and conditions in the stressed interbank market, not deposit makeup, were largely responsible) does not receive support. A second main test is on the 2003 large increase in excess reserves that followed the Paris II conference and capital in‡ows, which I argue is a “placebo” (nonexistent) event because there was no regulatory reserve requirement change at the time. The results support this prior in that the interaction terms of the 2003 indicator with LCD and F CLA are insigni…cant.

5

Conclusion

The results of this paper have shown that the transmission channel of monetary policy and regulatory shocks can work through the composition of bank deposits and the ownership structure of the banking system, both of which complicate the conduct of policy and lead to signi…cant e¤ects on lending. The results speak to the wider relevance of the transmission of liquidity shocks than, for example, Khwaja and Mian (2008), which focuses on a shock induced by a political event. In many EMEs, reserve requirement policy is employed to complement monetary and macroprudential policy because the policy interest rate is an insu¢ cient instrument for achieving several objectives including a stable currency, capital account, and price level. For example, an increase in reserve requirements

21

often is applied as part of a contractionary monetary policy, as was the case in Lebanon. The additional twist in this case was the application of a higher tax on foreign currency deposits in an e¤ort to attract local currency deposits. The use of currency-sensitive policy is underexplored in the literature but is an available policy instrument in EMEs with signi…cant deposit dollarization. Whether intended or not, such a policy change can lead to a drop in lending as banks scramble to adjust balance sheet portfolios in response. Absent easy substitution with non-deposit funding and absent su¢ ciently large foreign currency liquidity bu¤ers, the shock is absorbed by lending activity. As was the case here, the asset side is often not decoupled from the liability side because of currency matching considerations and exchange rate risk in EMEs, so that the decline in dollar credit is not fully undone by an increase in local currency credit. One implication is that monetary and regulatory policy in EMEs can have signi…cant distributional consequences to the extent that certain banks (and by extension, their bank-dependent borrowers) are ex ante more sensitive to the policy change (e.g., through the currency composition of their balance sheet). A second implication is that some banks are better able to thwart the policy objective ex post (e.g., through parent …nancing of foreign-owned banks). If the implementation of policy changes were gradually phased in, credit distortions may be moderated. While the shock analyzed in this study was one event, in other EMEs such as Brazil, reserve requirement changes are frequently applied. Despite their recurrence, they are still accompanied by signi…cant e¤ects on lending (Takeda et al, 2005; Glocker and Towbin, 2012). To the extent that being insulated to such changes requires holding more liquid assets, this comes at the cost of earnings that could be obtained from less liquid assets such as loans (Kashyap and Stein, 2000). To conclude, reserve requirements have signi…cant e¤ects – whether they result (intentionally or not) in a reserve requirement distortionary tax or whether they are used to apply, in a novel way, an optimal tax on credit booms fueled by short-term debt (Kashyap and Stein, 2012).

References [1] Baliño, Tomás, Adam Bennett, and Eduardo Borensztein. (1999). “Monetary Policy in Dollarized Economies.” Occasional Paper 171, International Monetary Fund. [2] Barth, James R., Gerard Caprio, and Ross Levine. (2001). “The Regulation and Supervision of Banks Around the World: A New Database.” Policy Research Paper 2588, World Bank. [3] Bernanke, Ben S., and Alan S. Blinder. (1988). “Credit, Money, and Aggregate Demand.” American Economic Review 78, 435-439. [4] Bertrand, Marianne, Antoinette Schoar, and David Thesmar. (2007). “Bank Deregulation and Industry Structure: Evidence from the French Banking Reforms of 1985.” Journal of Finance 62, 597-628.

22

[5] Black, Fischer. (1975). “Bank Funds Management in an E¢ cient Market.”Journal of Financial Economics 2, 323-339. [6] Bleakley, Hoyt, and Kevin Cowan. (2008). “Corporate Dollar Debt and Depreciations: Much Ado about Nothing?” Review of Economics and Statistics 90, 612-626. [7] Broda, Christian, and Eduardo Levy-Yeyati. (2006). “Endogenous Deposit Dollarization.”Journal of Money, Credit, and Banking 38, 963-988. [8] Calomiris, Charles W., Joseph Mason, and David Wheelock. (2011). “Did Doubling Reserve Requirements Cause the Recession of 1937-1938? A Microeconomic Approach.”Working Paper 16688, NBER. [9] Calvo, Guillermo. (2002). “On Dollarization.” Economics of Transition 10, 393-403. [10] Cargill, Thomas F., and Thomas Mayer. (2006). “The E¤ect of Changes in Reserve Requirements during the 1930s: The Evidence from Nonmember Banks.”Journal of Economic History 66, 417-432. [11] Cetorelli, Nicola, and Linda S. Goldberg. (2012). “Banking Globalization and Monetary Transmission.” Journal of Finance 67, 1811-1843. [12] Chang, Roberto, and Andres Velasco. (2001). “Monetary Policy in a Dollarized Economy Where Balance Sheets Matter.” Journal of Development Economics 66, 445-464. [13] Cosimano, Thomas F., and Bill McDonald. (1998). “What’s Di¤erent among Banks?” Journal of Monetary Economics 41, 57-70. [14] De Nicoló, Gianni, Patrick Honohan, and Alain Ize. (2005). “Dollarization of Bank Deposits: Causes and Consequences.” Journal of Banking and Finance 29, 1697–1727. [15] Eichengreen, Barry, and Ricardo Hausmann. (2005). Other People’s Money: Debt Denomination and Financial Instability in Emerging Market Economies, Chicago: University of Chicago Press. [16] Fama, Eugene F. (1985). “What’s Di¤erent About Banks?” Journal of Monetary Economics 15, 29-39. [17] Friedman, Milton, and Anna J. Schwartz. (1963). A Monetary History of the United States, 1867 - 1960. Princeton, N.J.: Princeton University Press. [18] Gelos, R. Gaston. (2009). “Banking Spreads in Latin America.”Economic Inquiry 47, 796-814. [19] Glocker, Christian, and Pascal Towbin. (2012). “The Macroeconomic E¤ects of Reserve Requirements.” Working Paper 374, Banque de France. [20] Goodfriend, Marvin, and Monica Hargraves. (1987). “A Historical Assessment of the Rationales and Functions of Reserve Requirements.” In Monetary Policy in Practice, edited by Marvin Goodfriend. Federal Reserve Bank of Richmond. [21] Grais, Wa…k, and Zeynep Kantur. (2003). “The Changing Financial Landscape: Opportunities and Challenges for the Middle East and North Africa.” Policy Research Paper 3050, World Bank. [22] Gulde, Anne-Marie, David Hoelscher, Alain Ize, David Marston, and Gianni De Nicoló. (2004). “Financial Stability in Dollarized Economies.” Occasional Paper 230, IMF. [23] Hubbard, R. Glenn. (2002). Money, the Financial System, and the Economy. Boston, MA.: Pearson Education, Inc. 23

[24] James, Christopher. (1987). “Some Evidence on the Uniqueness of Bank Loans.” Journal of Financial Economics 19, 217-235. [25] Kashyap, Anil K., and Jeremy C. Stein. (2000). “What Do a Million Observations on Banks Say about the Transmission of Monetary Policy?” American Economic Review 90, 407-428. [26] Kashyap, Anil K., and Jeremy C. Stein. (2012). “The Optimal Conduct of Monetary Policy with Interest on Reserves.” American Economic Journal: Macroeconomics 4, 266-282. [27] Keister, Todd, and James McAndrews. (2009). “Why are Banks Holding So Many Excess Reserves?” Sta¤ Report 380, Federal Reserve Bank of New York. [28] Khwaja, Asim I., and Atif Mian. (2008). “Tracing the Impact of Bank Liquidity Shocks: Evidence from an Emerging Market.” American Economic Review 98, 1413-1442. [29] Loungani, Prakash, and Mark Rush. (1995). “The E¤ect of Changes in Reserve Requirements on Investment and GNP.” Journal of Money, Credit, and Banking 27, 511-526. [30] Luca, Alina, and Iva Petrova. (2008). “What Drives Credit Dollarization in Transition Economies?” Journal of Banking and Finance 32, 858-869. [31] Montoro, Carlos, and Ramon Moreno. (2011). “The Use of Reserve Requirements as a Policy Instrument in Latin America.” BIS Quarterly Review, March, 53-65. [32] Peters, David, Elias Raad, and Joseph F. Sinkey. (2004). “The Performance of Banks in Postwar Lebanon.” International Journal of Business 9, 259-285. [33] Poddar, Tushar, Mangal Goswami, Juan Solé, and Victor E. Icaza. (2006). “Interest Rate Determination in Lebanon.” Working Paper 06/94, International Monetary Fund. [34] Rappoport, Veronica. (2009). “Persistence of Dollarization after Price Stabilization.” Journal of Monetary Economics 56, 979-989. [35] Reinhart, Carmen M., and Vincent R. Reinhart. (1999). “On the Use of Reserve Requirements in Dealing with Capital Flow Problems.” International Journal of Finance and Economics 4, 27-54. [36] Robitaille, Patrice. (2011). “Liquidity and Reserve Requirements in Brazil.” International Finance Discussion Paper 1021, Board of Governors of the Federal Reserve System. [37] Romer, Christina D., and David H. Romer. (1993). “Credit Channel or Credit Actions? An Interpretation of the Postwar Transmission Mechanism.” In Changing Capital Markets: Implications for Monetary Policy, A Symposium by the Federal Reserve Bank of Kansas City, Jackson Hole, Wyoming, 71-116. [38] Schimmelpfennig, Axel, and Edward H. Gardner. (2008). “Lebanon – Weathering the Perfect Storms.” Working Paper 08/17, International Monetary Fund. [39] Schnabl, Gunther, and Franziska Schobert. (2009). “Global Asymmetries in Monetary Policy Operations: Debtor Central Banks of the MENA Region.” Manchester School 77, 85-107. [40] Slovin, Myron B., Marie E. Sushka, and Yvette M. Bendeck. (1990). “The Market Valuation E¤ects of Reserve Regulation.” Journal of Monetary Economics 25, 3-19. [41] Takeda, Tony, Fabiana Rocha, and Márcio I. Nakane. (2005). “The Reaction of Bank Lending to Monetary Policy in Brazil.” Revista Brasileira de Economia 59, 107-126. [42] Terrier, Gilbert, Rodrigo Valdés, Camilo E. Tovar, Jorge Chan-Lau, Carlos FernándezValdovinos, Mercedes García-Escribano, Carlos Medeiros, Man-Keung Tang, Mercedes Vera Martin, and Chris Walker. (2011). “Policy Instruments to Lean against the Wind in Latin America.” Working Paper 11/159, International Monetary Fund.

24

Total

Foreign Currency

Lira .3

.2

.1

0

1998m6

1999m12

2001m6

2002m12

2004m6

-.1

FIG. 1. Aggregate Lending Growth (All banks, monthly data, year-on-year percent change) NOTES: The vertical line is the announcement date of the reserve requirement change (June 2001). Source: Banque du Liban (www.bdl.gov.lb)

.3

.25

.2

.15

.1 1998m6

1999m12

2001m6

2002m12

2004m6

FIG. 2. The Ratio of Bank Reserves to Total Assets (All banks, monthly data) NOTES: The first vertical line is the announcement date of the reserve requirement change (June 2001). The second vertical line is the implementation date (Oct. 2001). The increase in reserves in 2003 was not the result of changes to required reserves but followed the Paris II Donor conference (November 2002) and associated capital inflows. See section 4. Source: Banque du Liban (www.bdl.gov.lb)

Table 1. Descriptive Statistics of the Aggregate Economy Panel A. Annual time series All years 1996 1997 1998

1999

2000

2001

2002

2003

2004

-0.6

6.0

2.1

6.3

8.3

Coincident indicator of economic activity (% change)

3.2

1.1

2.3

3.7

0.1

Real GDP growth (IMF estimate) (%)

3.2

4.0

4.0

3.0

-1.2

1.2

4.2

2.9

5.0

6.0

Inflation (%)

2.5

5.2

3.2

1.6

1.4

-1.4

2.9

4.2

2.0

3.0

1514.7

1552.0

1527.0

1508.0

1507.5

1507.5

1507.5

1507.5

1507.5

1507.5

Exchange rate (LBP/USD) Interbank rate, Lebanese lira (LBP) (%)

8.4

11.2

13.0

11.2

7.5

7.6

9.7

7.6

4.0

3.9

428.1

-

-

249.9

237.9

205.2

502.4

928.4

511.9

361.1

Central bank foreign assets to total assets (year-end)

0.5

0.59

0.61

0.62

0.66

0.50

0.34

0.39

0.41

0.36

Bank reserves to total central bank assets (year-end)

0.5

0.28

0.48

0.42

0.40

0.43

0.56

0.62

0.79

0.76

Lebanon eurobond stripped spread (basis points)

Money supply growth, M1 (LBP sight, in %)

7.9

14.2

12.6

-0.4

5.1

9.8

1.1

4.2

13.6

10.8

Money supply growth, M2 (LBP sight + time & saving) (%)

12.2

29.3

24.6

1.7

16.3

12.0

-15.0

-0.1

30.3

10.5

Money supply growth, M3 (M2 + foreign currency) (%)

12.7

17.8

21.8

16.4

11.9

10.6

7.9

5.2

12.4

10.6

Panel B. Quarterly average time series around the 2001 reserve requirement change 2000Q4 2001Q1 2001Q2 2001Q3 2001Q4 2002Q1

2002Q2

2002Q3

2002Q4

2003Q1

Central bank balance sheet and excess reserves Central bank total assets (LBP billion)

16,996

16,781

17,724

17,877

19,047

19,914

20,693

20,832

20,343

23,036

Central bank foreign assets to total assets (%)

0.51

0.52

0.44

0.34

0.32

0.32

0.24

0.25

0.32

0.45

Central bank securities portfolio to total assets (%)

0.15

0.17

0.26

0.33

0.38

0.39

0.44

0.39

0.31

0.21

Bank reserves scaled by total assets of central bank (%) Excess bank reserves scaled by commercial bank assets (%)

0.44

0.44

0.43

0.41

0.53

0.54

0.51

0.50

0.55

0.57

0.076

0.073

0.080

0.075

0.031

0.038

0.037

0.032

0.038

0.076

6.55

5.43

4.22

3.41

2.12

1.78

1.78

1.76

1.49

1.26

Interest rates on deposit accounts Interest rate on eurodollar deposits (1 month) (%) Interest rate on foreign currency deposits (%) Interest rate on lira deposits (%)

6.06

5.81

5.10

4.71

4.31

4.19

4.17

4.19

4.11

3.77

10.43

10.41

10.19

10.12

10.11

10.17

10.14

10.54

10.31

8.89

NOTES: Source: Author's calculations based on data from Banque du Liban , Bilanbanques , IMF Public Information Notices, JP Morgan Chase EMBIG, and Federal Reserve H.15 Release. The benchmark international rate shown is the 1-month Eurodollar deposit rate because the majority of deposits in Lebanon have a maturity of one month or less (Poddar et al, 2006). Note the Paris II Donor Conference concluded in November 2002.

Table 2. Sample Statistics (1996-2004, Bilanbanques, at year end) Mean

Std. dev.

25th perc.

Median

75th perc.

Obs.

Primary characteristics Local currency deposit share (LCD) Foreign currency liquid asset buffer (FCLA) Equity to asset ratio Size (ln assets)

0.374 0.068 0.119 5.718

0.212 0.078 0.120 1.509

0.230 0.005 0.057 4.780

0.365 0.042 0.077 5.651

0.487 0.119 0.131 6.662

525 514 534 534

Other characteristics Loans to assets Growth in loans Local currency loan share Local currency asset share Asset-liability currency mismatch Foreign currency liquid assets (incl. reserves) Risk-based capital ratio Profitability (return on assets, net after tax) Interbank borrowing scaled by assets

0.248 0.079 0.166 0.392 -0.025 0.168 0.317 0.010 0.041

0.117 0.227 0.174 0.179 0.041 0.112 0.319 0.016 0.080

0.166 -0.064 0.056 0.290 -0.031 0.084 0.152 0.004 0.001

0.245 0.066 0.107 0.393 -0.014 0.154 0.210 0.008 0.016

0.320 0.206 0.218 0.477 -0.002 0.243 0.358 0.015 0.042

534 441 505 533 532 514 485 503 534

0.101 0.063 0.043 0.047

0.019 0.014 0.013 0.036

0.062 0.025 0.024 0.061

0.131 0.056 0.049 0.093

55 57 57 56

0.021 0.040 0.025 0.078

-0.006 -0.008 -0.004 0.051

0.001 -0.004 0.005 0.097

0.007 0.001 0.020 0.135

59 57 57 56

Excess reserves scaled by assets, by currency denomination in 2000-2003 Foreign currency 2000 0.088 2001 0.043 2002 0.031 2003 0.067 Local currency (estimate) 2000 0.003 2001 0.002 2002 0.008 2003 0.098

NOTES: Local Currency Deposit Share (LCD) is the fraction of a bank's deposits from customers denominated in lira (LBP). Foreign Currency Liquid Asset Buffer (FCLA) is the sum of securities (Lebanese and other government treasury bills, bonds, financial instruments and other marketable securities) denominated in foreign currency, scaled by total assets. Equity to Asset Ratio is book value of shareholders' equity to assets. Size is measured by the natural log of total assets, where assets are in USD million. Loans to Assets is the ratio of total loans (excluding substandard and doubtful loans) to total assets. Growth in Loans is the log difference in these loans. Local Currency Loan (Asset) Share is the fraction of a bank's total loans (assets) denominated in local currency. Asset-Liability Currency Mismatch is the difference between the fraction of total assets and total liabilities (incl. equity) denominated in local currency. Foreign Currency Liquid Assets (incl. reserves) is an alternative FCLA measure which is the sum of securities and cash and central bank (reserves) scaled by total assets. Risk-Based Capital Ratio is the Basel risk asset ratio. Profitability is measured by the ratio of net profits after tax to assets. Interbank Borrowing is borrowing from banks and financial institutions (including repos), scaled by assets. Excess Reserves are constructed as a bank's total reserves in the respective currency denomination less its computed required reserves, scaled by assets. Computed required reserves are 0 prior to 2001 and 15% of foreign currency customer deposits beginning in 2001. And are 13% of local currency deposits prior to 2001 and estimated at 15.6% beginning in 2001 (no breakdown by sight/time at the currency level is reported in Bilanbanques; therefore 15.6 is 25% of sight (roughly 6% of aggregate local currency deposits) + 15% of term deposits).

Table 3. The 2001 Reserve Requirement Change and its Effect on Bank Credit (The dependent variable is the annual growth of loans, 1996-2004 bank panel) (1) (2) (3) (4) (5) Pre-announcement foreign currency Domestic bank Average effect All banks reserves to asset ratio

(6) Foreign bank

(7) Testing sensitivity differences by bank type

Below median

Above median

Local currency deposit share (LCDt-1)

0.092 (0.199)

-0.189 (0.356)

0.279 (0.203)

0.121 (0.250)

-0.207 (0.359)

-0.084 (0.243)

Foreign currency liquid asset buffer (FCLAt-1)

-0.229 (0.310)

-0.472 (0.322)

0.197 (0.468)

-0.700** (0.260)

0.370 (0.647)

-0.587* (0.278)

-0.376* (0.177)

-1.041** (0.325)

-0.022 (0.242)

-0.616* (0.260)

0.090 (0.189)

-0.490* (0.222) 0.028 (0.096)

Interaction terms of interest (ex ante sensitivity to reserve requirement change and ability to offset) Year 2001 × LCDt-1 0.423** 1.131** (0.158) (0.375) Year 2001 × LCDt-1 × Foreign

0.185 (0.158)

0.400† (0.224)

0.689* (0.295)

0.515** (0.190) 0.024 (0.229)

Year 2001

-0.088** (0.026)

Year 2001 × Foreign

0.522† (0.305)

0.823† (0.462)

0.689 (0.584)

0.190 (0.455)

0.450 (0.607)

0.282 (0.418) 0.209 (0.667)

0.078 (0.302)

-0.242 (0.267)

0.713 (0.697)

-0.197 (0.326)

0.650 (0.803)

0.051 (0.304)

Year 2001 × Equity to asset ratiot-1

-0.145 (0.352)

0.384 (0.416)

-0.549 (0.623)

0.399 (0.310)

-1.846** (0.628)

-0.141 (0.303)

Size (ln assets)t-1

-0.106* (0.047)

-0.142† (0.074)

-0.068 (0.064)

-0.085† (0.046)

-0.190 (0.144)

-0.096* (0.047)

Year 2001 × Sizet-1

0.019 (0.022)

0.062* (0.030)

-0.016 (0.037)

0.054† (0.031)

-0.049† (0.024)

0.031 (0.026)

0.0001 (0.005)

0.013 (0.009)

-0.008 (0.006)

0.003 (0.007)

0.001 (0.006)

0.000 (0.005)

-0.008 (0.011)

-0.031 (0.022)

0.005 (0.010)

-0.011 (0.014)

-0.013 (0.015)

-0.009 (0.011)

-1.531 (1.143)

-5.309* (1.932)

-0.074 (1.336)

-0.275 (1.260)

-2.388 (1.993)

-1.693 (1.140)

4.455 (2.967)

14.892** (5.168)

-0.026 (3.039)

2.844 (3.355)

2.533 (6.663)

4.721 (2.994)

424 64 0.19

186 28 0.30

232 32 0.19

294 41 0.21

130 23 0.27

424 64 0.20

Year 2001 × FCLAt-1 Year 2001 × FCLAt-1 × Foreign Other bank controls Equity to asset ratiot-1

Other macroeconomic controls Interbank rate (change) Interbank rate × LCDt-1 Economic activity (coincident indicator, % change)

Economic activity × LCDt-1 Observations Number of banks R2

441 64 0.02

NOTES: Robust standard errors are in parentheses, clustered at the bank level. **, *, † indicate 1%, 5%, and 10% significance, respectively. Panel regressions are estimated with bank fixed effects and the reported R2 is the within R2 (unless otherwise noted). Also controlled for in col (7) are the double interaction terms Foreign×LCD and Foreign×FCLA (insignificant effects). Regressions are estimated after trimming the top and bottom 2.5 percent of observations based on the dependent variable.

Table 4. The Adjustment to the 2001 Reserve Requirement Change: Bank Credit and Other Borrowing in 2001-2002 (1) Δ ln (loans)t

(1996-2004 bank panel) (2) (3) Δ ln (loans)t Δ ln (loans)t

All banks

Domestic

Foreign

Local currency deposit share (LCDt-1)

0.105 (0.182)

-0.012 (0.236)

Foreign currency liquid asset buffer (FCLAt-1)

-0.365 (0.325)

Years 2001-02

0.044 (0.143)

(5) Δ ln (other financing)t Testing sensitivity Foreign differences

(6) Δ ln (deposits from affiliates) t Foreign

(7) Δ ln (deposits from affiliates) t Foreign

0.081 (0.322)

-0.110 (0.227)

0.151 (0.486)

2.210 (1.268)

3.642 (1.436)

-0.774** (0.274)

0.048 (0.667)

-0.664* (0.290)

0.524 (0.395)

1.333 (1.471)

1.140 (1.406)

-0.147 (0.173)

0.468 (0.284)

-0.045 (0.149) 0.131 (0.079)

0.441 (0.368)

0.384 (2.622)

0.887 (0.578)

0.323* (0.126)

-0.486** (0.161)

-0.978 (1.077)

-1.867† (1.050)

-3.015† (1.730)

Years 2001-02 × Foreign Interaction terms of interest (ex ante sensitivity to reserve requirement change and ability to offset) 0.218† Years 2001-02 × LCDt-1 0.336** 0.082 (0.115) (0.117) (0.197) Years 2001-02 × LCDt-1 × Foreign

(4) Δ ln (loans)t

-0.264† (0.155)

0.633* (0.293)

0.669† (0.354)

0.471 (0.566)

0.681† (0.352) -0.199 (0.559)

-0.506 (0.924)

-2.110 (2.585)

0.182 (0.311)

-0.131 (0.293)

0.808 (0.910)

0.179 (0.311)

2.148* (0.989)

-1.515 (4.923)

Years 2001-02 × Equity to asset ratiot-1

-0.680* (0.317)

-0.297 (0.224)

-1.805† (1.030)

-0.673* (0.301)

-0.259 (1.004)

-1.556 (5.109)

Size (ln assets)t-1

-0.106* (0.048)

-0.089† (0.050)

-0.171 (0.143)

-0.098* (0.049)

-0.097 (0.142)

0.015 (0.532)

Years 2001-02 × Sizet-1

-0.030 (0.020)

-0.017 (0.025)

-0.069† (0.038)

-0.025 (0.020)

-0.051 (0.050)

0.075 (0.293)

0.001 (0.005)

0.006 (0.007)

-0.003 (0.007)

0.001 (0.005)

-0.017** (0.005)

0.011 (0.056)

-0.009 (0.011)

-0.016 (0.014)

-0.004 (0.016)

-0.009 (0.011)

0.031* (0.012)

-0.079 (0.114)

0.317 (0.764)

1.158 (0.899)

0.091 (1.592)

0.267 (0.778)

2.022 (2.044)

-8.750 (6.688)

Economic activity × LCDt-1

0.917 (2.124)

-0.005 (2.239)

-1.562 (5.772)

0.988 (2.119)

1.812 (5.308)

21.466 (13.528)

Observations Number of banks R2

424 64 0.21

294 41 0.25

130 23 0.26

424 64 0.22

135 23 0.17

99 21 0.12

Years 2001-02 × FCLAt-1 Years 2001-02 × FCLAt-1 × Foreign Other bank controls Equity to asset ratiot-1

Other macroeconomic controls Interbank rate (change) Interbank rate × LCDt-1 Economic activity (coincident indicator, % change)

99 21 0.08

NOTES: Robust standard errors are in parentheses, clustered at the bank level. **, *, † indicate 1%, 5%, and 10% significance, respectively. Panel regressions are estimated with bank fixed effects and the reported R2 is the within R2. Also controlled for in col (4) are the double interaction terms Foreign×LCD and Foreign×FCLA (insignificant effects). Regressions are estimated after trimming the top and bottom 2.5 percent of observations based on the dependent variable.

Table A1. Composition of Bank Balance Sheets: Average 1996-2004 All years Number of banks

Domestic banks 1996 2000

2001

2004

All years

Foreign banks 1996 2000

2001

2004

42

40

39

40

40

23

21

21

21

18

Mean assets (1996 USD millions)

912.0

500.1

903.7

936.4

1379.7

486.6

353.5

518.6

536.3

540.2

Median assets (1996 USD millions)

312.8

151.5

350.1

376.0

461.1

134.8

95.4

146.3

178.2

172.8

Fraction of total system assets

0.792

0.739

0.773

0.795

0.850

0.208

0.261

0.227

0.205

0.150

0.157

0.103

0.110

0.149

0.283

0.171

0.149

0.144

0.166

0.239

Fraction of total assets in balance sheet category, mean bank Assets, of which: Cash and Central Bank (Reserves) in Lebanese lira (LBP)

0.066

0.038

0.038

0.040

0.159

0.050

0.053

0.035

0.035

0.065

in Foreign Currency (FC)

0.092

0.065

0.074

0.113

0.125

0.123

0.096

0.113

0.131

0.174 0.208

Securities

a

0.350

0.385

0.367

0.363

0.297

0.234

0.257

0.250

0.223

in LBP

0.258

0.356

0.275

0.241

0.154

0.172

0.231

0.175

0.139

0.105

in FC

0.069

0.011

0.068

0.094

0.117

0.067

0.020

0.113

0.072

0.112

Interbank assets Total loans

b

0.134

0.141

0.135

0.126

0.133

0.116

0.108

0.112

0.124

0.099

0.274

0.288

0.307

0.282

0.196

0.275

0.300

0.268

0.275

0.245

Total liabilities, of which: Total deposits

0.741

0.756

0.741

0.743

0.761

0.691

0.728

0.659

0.708

0.725

in LBP

0.318

0.406

0.318

0.270

0.305

0.183

0.254

0.161

0.156

0.143

in FC

0.434

0.369

0.423

0.474

0.471

0.512

0.474

0.498

0.552

0.582

0.041

0.026

0.045

0.041

0.046

0.055

0.035

0.046

0.072

0.068

Interbank liabilities (including to BDL) Subordinated & long-term debt

0.016

0.016

0.018

0.017

0.012

0.002

0.000

0.002

0.002

0.000

0.034

0.033

0.042

0.035

0.032

0.078

0.080

0.109

0.069

0.050

0.049

0.070

0.044

0.037

0.037

0.053

0.070

0.057

0.043

0.044

0.118

0.100

0.110

0.121

0.111

0.121

0.087

0.126

0.106

0.114

Provisions for doubtful loans (% of assets)

0.061

0.061

0.054

0.057

0.062

0.073

0.059

0.072

0.079

0.106

Profitability (return on assets)

0.011

0.016

0.009

0.009

0.007

0.008

0.010

0.009

0.007

0.005

0.103

0.120

0.117

0.114

0.066

0.100

0.122

0.096

0.094

0.091

0.075

0.093

0.085

0.072

0.045

0.052

0.071

0.055

0.052

0.027

0.033

0.045

0.031

0.027

0.023

0.033

0.041

0.033

0.035

0.027

Deposits from head office and affiliates Other financing

c

c

Equity Footnotes (mean)

Interest rate receivable on loans & advances (%) Interest rate payable on deposits (%)

d

d

Overall interest margin on interest earning assets d NOTES: Source: Author's calculations based on Bilanbanques. a

Securities are the sum of Lebanese Treasury bills and other government bills, bonds and financial instruments with fixed income, and securities with variable income. Total loans are the sum of commercial loans, other loans to customers, overdraft accounts, net debtor/creditor accounts and cash collateral, loans and advances to related parties, and doubtful loans. Current loans used in the analysis deduct substandard and doubtful loans.

b

c d

These are current accounts and time deposits placed by head office, branches, parent company, sister institutions and subsidiaries. Other financing includes liabilities under financial instruments such as CDs and debenture bonds.

Interest rate on loans calculated as ratio of interest income on loans & advances to total loans and advances to customers. Interest rate on deposits calculated as ratio of interest expenses on deposits from customers and other creditor accounts to deposits from customers. Overall interest margin is calculated as the difference between interest received and interest paid as a ratio of average interest-earning assets.

The next set of tables are in the Online Appendix only

Table A2. Robustness Checks: The 2001 Reserve Requirement Change and its Effect on Bank Credit (The dependent variable is the annual growth of loans, 1996-2004 bank panel) (1) (2) (3) (4) Average effect Average effect Average effect Average effect Year 2001 Year 2001 Years 2001-02 Years 2001-02 No bank F.E. No bank F.E.

(5) Robustness check Table 3(2) No bank F.E.

(6) Robustness check Table 3(2), bank F.E. and all year effects

Local currency deposit share (LCDt-1)

-0.117 (0.108)

0.059 (0.199)

Foreign currency liquid asset buffer (FCLAt-1)

-0.511* (0.230)

0.141 (0.373)

-0.227 (0.193)

-0.427* (0.210)

0.370* (0.164)

0.406* (0.165)

0.636* (0.318)

0.356 (0.314)

0.072 (0.205)

0.261 (0.298)

Year 2001 × Equity to asset ratiot-1

-0.575 (0.393)

-0.176 (0.356)

Size (ln assets)t-1

-0.009 (0.011)

-0.015 (0.044)

Year 2001 × Sizet-1

0.002 (0.024)

0.020 (0.023)

0.001 (0.005)

-

-0.013 (0.010)

-0.009 (0.011)

-1.803 (1.141)

-

3.772 (2.913)

3.261 (2.906)

424 64 0.12

424 64 0.25

Year 2001 Year 2001 × Foreign

-0.113** (0.029) 0.079 (0.058)

-0.077* (0.032) -0.016 (0.062)

-0.141** (0.021) 0.060 (0.053)

-0.107** (0.025) -0.034 (0.052)

Interaction terms of interest (ex ante sensitivity to reserve requirement change and ability to offset) Year 2001 × LCDt-1 Year 2001 × FCLAt-1 Other bank controls Equity to asset ratiot-1

Other macroeconomic controls Interbank rate (change) Interbank rate × LCDt-1 Economic activity (coincident indicator, % change)

Economic activity × LCDt-1 Observations Number of banks 2 R

441 64 0.02

441 64 0.01

441 64 0.07

441 64 0.05

NOTES: Robust standard errors are in parentheses, clustered at the bank level. **, *, † indicate 1%, 5%, and 10% significance, respectively. Panel regressions are estimated with bank fixed effects and the reported R2 is the within R2 (except in columns (2), (4), and (5)). The indicator for Year 2001 is replaced with Years 2001-02 in columns (3) and (4). Note that the year effects in column (6) subsume the time-varying interbank rate and economic activity variables. Regressions are estimated after trimming the top and bottom 2.5 percent of observations based on the dependent variable.

Table A3. Robustness Checks: Controlling for Additional Bank Characteristics and Economic Shocks (The dependent variable is the annual growth of loans, 1996-2004 bank panel) (1) (2) (3) (4) Alternative Alternative Alternative FCLA risk-based economic measure capital ratio activity

(5)

(6)

Local currency deposit share (LCDt-1)

0.122 (0.195)

0.095 (0.245)

0.064 (0.263)

0.202 (0.209)

-0.067 (0.204)

0.109 (0.208)

Foreign currency liquid asset buffer (FCLAt-1)

0.091 (0.232)

-0.274 (0.369)

-0.523 (0.326)

-0.238 (0.305)

-0.196 (0.326)

-0.075 (0.339)

Year 2001

-0.533** (0.169)

-0.427** (0.150)

-0.471** (0.167)

-0.320† (0.175)

-0.411* (0.176)

-0.387* (0.175)

0.301* (0.125)

0.494** (0.151)

0.403* (0.182)

Interaction terms of interest (ex ante sensitivity to reserve requirement change and ability to offset) Year 2001 × LCDt-1 0.515** 0.421* 0.430** (0.163) (0.206) (0.154) 0.612* (0.271)

0.609† (0.317)

0.626 (0.341)

0.535 (0.315)

0.487 (0.307)

0.437 (0.314)

0.082 (0.307)

0.034 (0.039)

-0.173 (0.270)

0.119 (0.299)

0.307 (0.297)

0.047 (0.307)

Year 2001 × Equity to asset ratiot-1

-0.012 (0.290)

0.100 (0.162)

0.114 (0.404)

-0.166 (0.356)

-0.170 (0.348)

-0.020 (0.346)

Size (ln assets)t-1

-0.121* (0.049)

-0.086† (0.051)

-0.076† (0.043)

-0.095* (0.044)

-0.121* (0.051)

-0.103* (0.048)

Year 2001 × Sizet-1

0.028 (0.020)

0.021 (0.020)

0.030 (0.021)

0.020 (0.023)

0.018 (0.022)

0.019 (0.022)

Year 2001 × FCLAt-1 Other bank controls Equity to asset ratiot-1

†

Profitability (return on assets, net after tax)t-1

-1.038 (1.985)

Year 2001 × Profitabilityt-1

1.695 (1.604)

Interbank borrowing (wholesale funding)t-1

-0.286 (0.343)

Year 2001 × interbank borrowingt-1

-1.063† (0.573)

Other macroeconomic controls Interbank rate (change)

†

0.000 (0.005)

-0.001 (0.006)

0.002 (0.005)

-0.002 (0.005)

0.005 (0.005)

0.000 (0.006)

Interbank rate × LCDt-1

-0.008 (0.011)

-0.006 (0.012)

-0.012 (0.010)

0.000 (0.010)

-0.015 (0.011)

-0.008 (0.012)

Economic activity (coincident indicator, % change)

-1.679 (1.199)

-2.312* (1.112)

-1.638 (1.146)

-0.528 (1.283)

-2.641* (1.075)

-1.685 (1.344)

Economic activity × LCDt-1

4.426 (2.948)

6.850* (2.912)

4.214 (2.976)

4.661 (2.957)

5.764* (2.838)

4.308 (3.224)

Foreign exchange reserves at central bank (% change)

0.141* (0.060)

FX reserves × LCDt-1

-0.123 (0.156)

Year 2003 (Paris II donor conference)

0.039 (0.174)

Year 2003 × LCDt-1

-0.209 (0.147)

Year 2003 × FCLAt-1

-0.311 (0.342)

Observations Number of banks 2 R

424 64 0.20

390 61 0.19

402 62 0.20

424 64 0.21

424 64 0.21

424 64 0.21

NOTES: Robust standard errors are in parentheses, clustered at the bank level. **, *, † indicate 1%, 5%, and 10% significance, respectively. Panel regressions are estimated with bank fixed effects and the reported R2 is the within R2. Regressions are estimated after trimming the top and bottom 2.5 percent of observations based on the dependent variable. Column (1) includes foreign currency cash and bank reserves to the liquid asset measure. Column (2) uses the Basel risk-based capital ratio instead of the book equity-to-asset ratio. Column (4) uses the GDP growth estimate reported by the IMF instead of the coincident indicator. Column (6) also controls for the respective interaction terms of Year 2003 with equity to asset ratio and asset size.

Table A4. Matching industry statistics from Kompass company database to banks

Fraction of total system assets (Bilanbanques) Fraction of total system loans (Bilanbanques) Fraction of firms with primary banker (Kompass)

All sample

Domestic bank

Foreign bank

0.792 0.752 0.723

0.208 0.248 0.273

Kompass sample Domestic Foreign bank bank

Big bank Small bank 0.645 0.628 0.666

0.355 0.372 0.329

Big bank Small bank

Distribution of firms across Kompass sample, and by primary bank type: Size Small firms ( 5 or less employees) 0.265 0.276 Medium-size (6-24) 0.476 0.479 Big (25-59) 0.157 0.153 Very big (60 or more) 0.102 0.092

0.213 0.511 0.206 0.071

0.284 0.448 0.177 0.091

0.209 0.564 0.151 0.076

Age Old (established <= 1981) Middle (established > 1981 & < 1997) Young (established >= 1997 & < 2000) Very young (established >= 2000)

0.259 0.468 0.161 0.112

0.237 0.485 0.164 0.114

0.262 0.459 0.180 0.098

0.234 0.493 0.171 0.101

0.262 0.448 0.163 0.127

Region Beirut Metn Baabda, Aley, and Chouf Kesrouane, Batroun, Jbeil and Bcharreh Tripoli and Akkar Saida and Tyr Zahle

0.432 0.306 0.103 0.114 0.019 0.012 0.014

0.439 0.318 0.093 0.126 0.012 0.008 0.004

0.421 0.306 0.093 0.104 0.022 0.022 0.033

0.432 0.306 0.101 0.121 0.016 0.011 0.013

0.439 0.330 0.077 0.118 0.014 0.014 0.009

Sector Agriculture, hunting, fishing & forestry Mining & quarrying Electricity, gas, and water Manufacturing Construction & public works Wholesale/retail trade (commerce) Services

0.005 0 0 0.194 0.054 0.445 0.302

0.006 0 0 0.198 0.056 0.468 0.272

0.005 0 0 0.208 0.071 0.443 0.273

0.009 0 0 0.181 0.065 0.450 0.295

0.000 0 0 0.240 0.050 0.484 0.226

Source: Author's calculations using Kompass Lebanon 2005 for a random 10% of the companies (804 companies out of the database's 8245 companies). Data on banks from Bilanbanques, various issues.

Table A5. Foreign Currency Deposits to Total Deposits (Average 1995 - 2001) and Different Reserve Requirement Regulation

If different reserve requirement (a), then LC > FC: 1 FC > LC: 2

Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

Cambodia Bolivia Uruguay Bosnia-Herzegovina Lao PDR Armenia Croatia Nicaragua Peru Angola Georgia Azerbaijan Belarus Argentina Lebanon Tadjikistan Macedonia FYR Paraguay Kyrgyz Republic Congo Dem. Rep. of Bulgaria Yemen Turkey Mozambique Kazakhstan Latvia Sao Tome & Principe Hong Kong Bahrain Lithuania Costa Rica Moldova Romania Ecuador Vietnam Mongolia Zambia Slovenia Ukraine Haiti Russia Turkmenistan Philippines Albania Honduras Ghana Tanzania Greece Guinea Pakistan

93.2 90.0 80.0 73.6 73.0 71.4 70.4 67.4 65.7 65.1 64.6 63.6 63.2 61.3 60.4 57.6 56.8 55.4 51.9 50.6 50.5 49.6 49.0 48.3 47.5 47.5 46.3 45.7 40.6 40.6 40.4 38.3 37.7 37.6 37.5 37.4 36.3 35.3 35.2 35.1 34.0 31.3 30.7 28.6 28.4 28.1 28.0 27.4 25.4 24.9

1

2

1 1 2

2

2

2

1

2 1 2 1 1

2

Rank, cont'd 51 Egypt 52 St. Kitts and Nevis 53 Hungary 54 Trinidad and Tobago 55 United Arab Emirates 56 Uganda 57 Indonesia 58 Jamaica 59 Poland 60 Saudi Arabia 61 Estonia 62 Israel 63 Netherlands 64 Malawi 65 Guinea-Bissau 66 Slovak Republic 67 United Kingdom 68 Uzbekistan 69 Barbados 70 Czech Republic 71 Kenya 72 China 73 Mexico 74 El Salvador 75 Japan 76 Chile 77 Antigua and Barbuda 78 Nigeria 79 Netherlands Antilles 80 Italy 81 Denmark 82 Norway 83 New Zealand 84 Finland 85 South Africa 86 Bhutan 87 Korea 88 Malaysia 89 Dominica 90 Belize 91 Bahamas, The 92 Austria 93 St. Vincent&Grenadine 94 Spain 95 Sweden 96 Thailand 97 Guatemala 98 Comoros 99 Bangladesh 100 Switzerland 101 Venezuela

If different reserve requirement, then LC > FC: 1 FC > LC: 2 24.8 1 24.8 24.3 23.7 1 22.6 21.9 21.4 21.0 21.0 1 20.2 19.1 19.0 17.5 17.3 15.5 14.8 13.4 11.4 1 11.4 11.1 10.9 8.2 8.1 8.0 6.7 6.6 2 5.4 4.2 1 4.1 3.8 3.8 3.7 3.5 3.2 3.1 2.8 2.7 2.4 2.4 2.2 2.0 1.9 1.9 1.8 1.4 1.0 0.8 0.5 0.3 0.3 0.2

Source: Author's calculations based on foreign currency deposit share (1990 - 2001) provided in De Nicolo et al, Journal of Banking and Finance, "Dollarization of bank deposits: Causes and consequences", 29, 2005, Appendix A, Table 4. And author's interpretation of reserve requirements based on Barth, Caprio, and Levine (2001, updated 2003) World Bank, "The Regulation and Supervision of Banks around the World: A New Database", Balino et al (1999), and Gulde et al (2004). See Table A6 for more information. Note: a The following additional six countries have higher reserve requirements on local currency than on foreign currency deposits (these countries are not listed in De Nicolo et al (2005) but are in the other cited references): Colombia, Cyprus, Kuwait, Morocco, Sri Lanka, and Tunisia. And Vanuatu has a higher reserve requirement on foreign currency deposits.

Table A6. Comparing reserve requirement regulation across countries Income Level (Grouping according to Caprio et al 2001). High Income = 0; Other = 1

Author interpretation of whether there are different What percent of the What percent of the FC/LC reserve requirements If different, then commercial banking commercial banking (based on responses to LC > FC: 1 and system’s assets is foreign- system’s liabilities is foreignquestions 7.3-7.6 in Caprio et if FC > LC: 2 currency denominated? currency denominated? al): Different? 1 Equal? 0

Summary statistics on the data

38 high income countries and 111 other

Median: 26.8%

Median: 24.5%

Lebanon

1

66.30%

64.50%

28 Countries

1

RR on LC

10 countries with LC > FC req and 18 countries with FC > LC req.

1

25% on demand; 15% on time

Albania

1

37.6%

37.2%

0

10%

Algeria

1

2.35%

10.71%

0

6.25%

14.30%

61.10%

0

6%

Antigua and Barbuda

1

17.13%

14.20%

0

6%

Argentina

1

74.70%

73.20%

0

17%

Anguilla

Armenia

1

60.10%

80.80%

0

8%

Aruba

0

26.80%

21.50%

0

7%

Australia

0

7.20%

39.10%

0

0

Austria

0

N/A

N/A

0

Azerbaijan

1

N/A

N/A

0

N/A 10%

Bahrain

1

52%

43%

0

5%

Belarus

1

50.30%

N/A

0

10%

Belgium

0

28%

30%

0

0

Belize

1

11.30%

15.20%

0

6%

Benin

1

0

0

0

Bhutan

1

37.35%

0.08%

0

N/A 20% 12%

Bolivia

1

83.20%

91.10%

0

Bosnia and Herzegovina

1

50.00%

64.00%

0

5%

Botswana

1

1%

18%

0

3.25%

Brazil

1

22.40%

21.40% 0

60% demand; 20% savings; 15% time Lump sum

British Virgin Islands

0

N/A

N/A

0

Bulgaria

1

55.10%

56.60%

0

8%

Burkina Faso

1

0

0

0

Burundi

1

20%

20%

0

N/A 8.50%

Cambodia

1

98%

98%

0

8%

Cameroon

1 0

7.75% on demand; 5.75% fixed term

Canada

0

Central African Republic

1

Chad

1

Chile

1

42%

45%

0

0

1

2

9% on demand; 3.6% on time deposits

13.6 - 19%

1

1

N/A

0

32.70%

Colombia

1

6.20%

6.00%

Commonwealth of Dominica

1

10.50%

2.39%

Congo

1

15%

0 5% on demand; 3% fixed term 5% on demand; 3% fixed term

0

32.60%

RR on FC if different

0

6%

0

7.75% on demand; 5.75% fixed term

Costa Rica

1

57.16%

54.38%

0

5%

Côte d'Ivoire

1

0

0

0

Croatia

1

28.83%

71.66%

0

N/A 19%

Cyprus

0

34%

34%

1

Czech Republic

1

18%

17%

Denmark

0

47.71%

48.86%

0

Ecuador

1

100%

100%

0

1

0

6.50% 2% for up to 2 year maturity

0

0 4%

Egypt

1

29.80%

29.60%

1

El Salvador

1

100%

100%

0

N/A

Equatorial Guinea

1 0

7.75% on demand; 5.75% fixed term 13%

1

14%

Estonia

1

63.60%

45.60%

0

Fiji

1

10.50%

9.52%

0

5%

Finland

0

29%

26%

0

N/A

France

0

16.30%

15.00%

0

N/A

10%

Gabon

1

0

5% on demand; 3% fixed term

Gambia

1

2.89%

0.40%

0

16%

Germany

0

6.50%

5.70%

0

N/A

Ghana

1

30%

27%

1

9% primary reserve; 25% secondary reserve (these earn t-bill interest)

Gibraltar

0

N/A

N/A

Greece

0

16.26%

17.20%

1

0

0

0

2% on mostly up to 2 year maturity

Grenada

1

10.80%

21.56%

0

6%

Guatemala

1

15%

14%

0

14.60%

Guernsey

0

66%

67%

0

Guinea

1

28.20%

19.50%

0

0 5.50%

Guinea Bissau

1

0

0

0

Guyana

1

7.40%

2.49%

0

Honduras

1

33.20%

33.70%

1

Hong Kong, China

0

57%

53%

0

Hungary

1

35.30%

34.70%

0

0 5%

Iceland

0

37%

46%

0

1.5-4%

India

1

6.61%

6.73%

0

5.50%

Ireland

0

40.10%

42.90%

0

35%

35%

0

N/A 0

0

6% up to 6 days 3% 1wk - 1yr and 0 for > 1yr

Isle of Man

N/A 12% 2

12%

Israel

0

32%

31%

Italy

0

11.50%

14.50%

0

2%

Japan

0

N/A

N/A

N/A

N/A 0

Jersey

1

Jordan

1

38.50%

37%

0

8%

Kazakhstan

1

65.11%

83.75%

N/A 0

N/A 10%

0

Kenya

1

6%

14%

Kuwait

0

20.20%

20.40%

1

Kyrgyzstan

1

53.60%

64.50%

0

10%

Latvia

1

65.10%

65.80%

0

3% 3%

1

20%

Lesotho

1

22%

3%

0

Liechtenstein

0

N/A

N/A

0

0

Lithuania

1

51.30%

52.30%

0

6%

Luxembourg

0

40%

45%

0

N/A

0

3% demand; 2% liabilities up to 3 months; and 1% liab > 3 mo

1

10% on LC demand and up to 3 mo; and 5% on other term LC deposits

Macau, China

0

68.28%

9% across the board

66.06%

Macedonia, Republic of

1

64%

53%

Madagascar

1

18.70%

18.10%

0

Malaysia

1

7.10%

6.20%

0

4%

Mali

1

0

0

0

1

50%

0

0

24% on demand

Malta

1

50.77%

51.68%

0

N/A 4%

Mauritius

1

10.20%

9.30%

0

5.50%

Mexico

1

17.10%

16.60%

0

0%

Moldova, Republic of

1

39.68%

47.29%

0

8%

Montserrat

1

14.89%

2.33%

0

6%

Morocco

1

4%

5%

1

Namibia

1

5%

6%

0

1

14%

0

1% 2% for up to 2 year maturity

Netherlands

0

N/A

N/A

New Zealand

0

8.42%

22.97%

0

0

Nicaragua

1

51.80%

71.30%

0

19.25%

Niger

1

0

0

0

Nigeria

1

15.21%

4.61%

1

Norway

0

22.20%

32.70%

0

0

Oman

1

25%

21%

0

N/A

Pakistan

1

18.30%

22.20%

1

Panama

1

100%

100%

0

Papua New Guinea

1

3.60%

7.60%

0

Paraguay

1

61.14%

64.98%

1

2

Peru

1

73.41%

72.24%

1

2

0

1

2

N/A 0

5%

20%

0 5% 15% for up to 2 yrs26.5% for up to 2 yrs 6% 6%+20%

Philippines

1

28.57%

32.49%

1

1

17%

Poland

1

23.90%

18%

1

1

4.50%

Portugal

0

8.20%

9.60%

0

Puerto Rico

1

0

0

0

0%

N/A 0

Qatar

0

50%

33%

0

Romania

1

50%

47.07%

1

2

18%

25%

Russia

1

37.90%

32.10%

1

2

10%

Rwanda

1

30%

26%

0

7% 8%

Saint Kitts and Nevis

1

18.70%

5.30%

0

6%

Saint Lucia

1

13.40%

4.60%

0

6%

1

13.00%

8.00%

0

6%

1

17%

25%

0

4.80%

Saudi Arabia

1

26%

24%

0

N/A

Senegal

1

0

0

0

Serbia & Montenegro

1

65.00%

61.60%

0

N/A 20%

Seychelles

1

0.06%

3.75%

0

2.50%

Singapore

0

75%

75%

0

N/A

Slovakia

1

18.50%

17.30%

N/A

Slovenia

0

34.40%

35.30%

1

South Africa

1

10.50%

9.10%

0

2.50%

South Korea

1

10.14%

9.66%

0

Spain

0

9%

11%

N/A 2% for up to 2 year maturity

Sri Lanka

1

Sudan

1

23%

Saint Vincent and The Grenadines Samoa (Western)

2.75%

1

0

N/A 7%

15%

1

30%

0

14% N/A 3%

Suriname

1

25.80%

45%

N/A

Swaziland

1

30%

0.80%

0

1

10%

Sweden

0

N/A

N/A

0

0

Switzerland

0

56%

54%

0

0%

Taiwan

0

0

5%-10.75%

Tajikistan

1

70%

73%

0

18%

Thailand

1

12.20%

4.00%

0

6%

Togo

1

0

0

0

Tonga

1

18%

16%

0

N/A 15%

Trinidad and Tobago

1

27.58%

32.58%

1

2%

0

1

18%

0 11%

Tunisia

1

7.40%

8.80% 1

1

2% for < 3 mo; 1% for 3-2 yrs; 0 otherwise

Turkey

1

45.60%

56.60%

1

2

6%

N/A

95%

0

N/A

0

10%

N/A

Turkmenistan

1

Turks and Caicos Islands

0

Ukraine

1

38.00%

37.40%

United Arab Emirates

1

33%

11%

0

N/A 14% demand; 1% on time

United Kingdom

0

53%

53%

0

N/A

United States

0

N/A

N/A

0

N/A

Uruguay

1

77.64%

89.98% 1

1

Vanuatu

1

53%

8.50%

1

2

Venezuela

1

17%

2%

0

Zimbabwe

1

2.60%

5.70%

0

27.5% < 30 days; 21.5% for 30 - 180 days; 5% for > 180 10% < 180 days; days 4% > 180 days 10%

0

15% 50% demand; 20% saving

50%

Other ctries in Balino et al (1999) but not in above Caprio et al sample: Malawi

0

20%

Maldives

0

35%

Nepal

0

Sao Tome and Principe

1

2

15%

Tanzania

1

1

12%

0%

Lao PDR

0

12%

12%

Uzbekistan

1

25%

0%

Vietnam

0

12%

1

30%

10%

Source: Author's calculations based on Barth, Caprio, and Levine (2001, updated 2003) World Bank, "The Regulation and Supervision of Banks around the World: A New Database", and the IMF's Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER), various issues.