Description
Banks liquidity holdings are comfortably above legal or prudential requirements in most Central American countries. While good for financial stability, high systemic liquidity may nonetheless hinder monetary policy transmission and financial markets development.
The Determinants of Banks' Liquidity Buffers in
Central America
Corinne Deléchat, Camila Henao, Priscilla Muthoora,
Svetlana Vtyurina
WP/12/301
© 2012 International Monetary Fund WP/12/301
IMF Working Paper
Western Hemisphere Department
The Determinants of Banks' Liquidity Buffers in Central America
Prepared by Corinne Deléchat, Camila Henao, Priscilla Muthoora, Svetlana Vtyurina
1
Authorized for distribution by Przemek Gajdeczka
December 2012
Abstract
Banks’ liquidity holdings are comfortably above legal or prudential requirements in most
Central American countries. While good for financial stability, high systemic liquidity
may nonetheless hinder monetary policy transmission and financial markets development.
Using a panel of about 100 commercial banks from the region, we find that the demand
for precautionary liquidity buffers is associated with measures of bank size, profitability,
capitalization, and financial development. Deposit dollarization is also associated with
higher liquidity, reinforcing the monetary policy and market development challenges in
highly dollarized economies. Improvements in supervision and measures to promote
dedollarization, including developing local currency capital markets, would help enhance
financial systems’ efficiency and promote intermediation in the region.
JEL Classification Numbers: E44, G21, O16
Keywords: Central America, bank liquidity, credit, dollarization, excess liquidity, foreign banks
Authors’ E-Mail Addresses: [email protected], [email protected], [email protected],
[email protected]
1
The authors would like to thank Miguel Savastano for his guidance. Przemek Gajdeczka, Fernando Delgado,
Mario Mansilla, Stephanie Medina Cas, Andy Swiston and other IMF colleagues, as well as staff from the Central
Bank of Guatemala, Rudy Loo-Kung and Mario Cuevas at the Inter-American Development Bank, and Miquel
Dijkman at the World Bank provided useful comments.
This Working Paper should not be reported as representing the views of the IMF.
The views expressed in this Working Paper are those of the author(s) and do not necessarily
represent those of the IMF or IMF policy. Working Papers describe research in progress by the
author(s) and are published to elicit comments and to further debate.
2
Contents Page
I. Introduction ................................................................................................................................ 3
II. Determinants of Banks’ Liquidity Buffers: Literature Review ................................................. 5
III. Some Background on CAPDR Banking Systems .................................................................... 7
IV. Determinants of Banks Liquidity Buffers in CAPDR ........................................................... 12
A. Data and Variable Definitions..................................................................................... 12
B. Stylized Facts .............................................................................................................. 17
C. Empirical Analysis ...................................................................................................... 19
D. Results ......................................................................................................................... 20
V. Conclusions and Policy Lessons ............................................................................................. 26
References .................................................................................................................................... 29
Tables
1. CAPDR: Banking System Indicators, 2010 ..........................................................................9
2. Variables Used in the Empirical Estimation ..........................................................................15
3. Dependent Variables Means by Liquidity Ratio Quartiles ....................................................18
4a. Determinants of Banks’ Liquidity Buffer in CAPDR-Financial Depth ...............................23
4b. Determinants of Banks’Liquidity Buffer in CAPDR-Credit Institutions .............................24
5. CAPDR: Largest Deposit Drop (December 2001–September 2011) ..................................26
Figures
1. CAPDR Liquidity Buffers (2010) .........................................................................................3
2. CAPDR: Statutory Reserves and Liquidity Requirements by Currency .................................8
3. CAPDR: Financial Soundness Indicators, 2005–10 .............................................................11
4. Bankscope Sample Coverage of Total Banking System’s Assets ........................................12
5. Liquidity Ratios at the System Level and in the Bankscope Sample ...................................12
6. Financial Development Indicators in CAPDR .......................................................................14
7. CAPDR Liquidity Indicators ................................................................................................17
Box
1. Financial Systems Supervision and Development in CAPDR .............................................10
Appendix I. Reserve and Liquidity Requirements in CAPDR ....................................................33
Appendix Tables
Appendix I. Table 1. Legal Reserve and Liquidity Requirements in CAPDR (2010) ............. 36
Appendix I. Table 2. Evolution of Reserve and Liquidity Ratios in CAPDR (2005–10) ........ 37
Appendix II. Table 1. Correlation Between Liquidity Ratios, Credit to GDP and Credit
Institutions in CAPDR ............................................................................ 38
Appendix II. Table 2. CAPDR Financial Safety Nets ................................................................ 39
Appendix III. Table 1. Descriptive Statistics ............................................................................... 40
Appendix IV. Table 1. Determinants of Liquidity Buffers in CAPDR– Robustness Checks ..... 41
Appendix IV. Table 2. Determinants of Liquidity Buffers in CAPDR– Robustness Checks ..... 42
3
I. INTRODUCTION
This paper studies the determinants of banks’ liquidity buffers in Central America,
1
Panama
and the Dominican Republic (CAPDR) using a panel of about 100 commercial banks over
2006-10. In particular, the paper examines whether CAPDR banks’ liquidity buffers, defined
as the liquid assets to deposits ratio, can be explained by bank and country-level
characteristics, as predicted by theory and documented in some empirical studies. Of
particular interest for the region is whether liquidity holdings are related to bank ownership
(are public or foreign banks different?) or the economy’s degree of dollarization.
CAPDR banking systems are highly liquid. Holdings of liquid assets as a share of total
deposits averaged about 28 percent for the region in 2010, while reserve requirements were
set at about 17 percent on average (Figure 1). Liquidity ratios are also high compared to
larger South-American countries; liquidity ratios averaged about 15 percent for Brazil, Chile
and Colombia in 2010. For monetary and supervisory authorities, ensuring that banks hold
adequate amounts of high-quality liquid assets is essential for financial stability, as
highlighted during the recent global financial crisis. However, if liquidity holdings are much
above legal requirements, this may be costly in terms of foregone financial intermediation.
Excess liquidity also hinders the development of interbank and money markets in all
countries, and acts as “sand in the wheels” of the monetary transmission mechanism in
countries with a monetary policy (Gray, 2011).
Figure 1. CAPDR Liquidity Buffers¹ (2010)
1
Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua.
Sources: CAPDR Central Banks and Superintendencies' websites; and Fund staff calculations.
¹ Liquid assets to deposits ratio. Liquid assets include cash, central bank reserves and deposits abroad.
² Liquidity requirements.
³ Observed liquidity. Legal liquidity requirement (30%) is calculated as the ratio of liquid assets including
securities to short term deposits.
0
5
10
15
20
25
30
35
40
Costa Rica Dominican
Republic
El Salvador ² Guatemala Honduras Nicaragua Panama³
(
p
e
r
c
e
n
t
)
Required liquidity holdings (percent of deposits) Excess liquidity (percent of deposits)
4
From individual banks’ point of view, holding sufficient liquidity is necessary to insure
against liquidity risk (Diamond and Dybvig, 1983, Diamong and Rajan, 2001). As loans are
relatively illiquid, large and unexpected deposit withdrawals can lead to insolvency as it may
be too costly or not possible to raise liquidity on short notice, due to capital market
imperfections. Instead of self-insuring, banks could resort to other forms of financing, such
as accessing interbank markets, central bank liquidity windows, or external credit lines.
However, asymmetric information may lead to coordination failures on the interbank market,
and external credit lines may freeze (as seen during the recent financial crisis), so that solvent
but illiquid banks would still fail, absent a Lender of Last Resort (LOLR) (Rochet and Vives,
2004). Thus banks hold a buffer of liquid assets as self-insurance, equating the marginal
benefit of holding liquid assets to the marginal cost of alternative investments.
A priori, one would expect the self-insurance motive to be especially important in CAPDR.
Capital markets are underdeveloped, interbank markets are thin, and LOLR arrangements
remain limited or nonexistent. For the five partially dollarized economies, the high share of
foreign currency assets and liabilities limits the ability of the central bank to act as a LOLR,
while the two fully dollarized economies in the region, Panama and El Salvador, did not have
a LOLR as of end-2010 (El Salvador formally approved the regulations to establish a
liquidity facility in June 2012). Furthermore, while the region’s predominant reliance on
customer deposits is a likely reason for its resilience during the global financial crisis, it is
also a potential source of vulnerability and calls for the holding of adequate liquidity buffers.
In line with theory and results from related empirical studies, we find that CAPDR banks’
liquidity buffers are persistent, consistent with the notion that banks target an optimal amount
of liquid asset coverage of deposits or total assets. Liquidity ratios depend on idiosyncratic
factors such as bank size, profitability, and efficiency. Measures of financial development
also matter. We find some evidence that foreign banks in the region tend to hold less
liquidity than domestic banks, possibly reflecting their access to more diversified funding
sources or less risky business models. Furthermore, we uncover a robust positive impact of
the degree of deposit dollarization on the size of liquidity buffers. This suggests that not only
deposit dollarization in itself,
2
but also its association with higher liquidity holdings,
contributes to weaken the monetary transmission mechanism in partially dollarized
economies and hinders the development of money and interbank markets.
The reminder of the paper is structured as follows. Section II briefly reviews the theoretical
and empirical literature on the determinants of liquidity holdings. Section III provides some
background information on CAPDR banking systems. Section IV describes the data and
presents stylized facts on the distribution of banks’ liquidity holdings. Section V discusses
the econometric methodology and estimation results, and Section VI concludes.
2
See Medina Cas, Carrion and Frantischek, 2011a.
5
II. DETERMINANTS OF BANKS’ LIQUIDITY BUFFERS: LITERATURE REVIEW
The determinants of banks’ liquidity buffers, as identified in the theoretical and empirical
literature, can be classified into four broad categories. These are the opportunity costs of and
shocks to funding, bank characteristics, macroeconomic fundamentals, and moral hazard
motives, as discussed below.
Opportunity cost and shocks to funding
The early literature on bank liquidity uses the firm’s theory of inventory decisions as a
starting point. The cost of holding liquid assets (with low returns compared with other types
of investments) is compared to the benefits of reducing risks of “running out” (Baltensperger,
1980, and Santomero, 1984). These models predict that the size of liquidity buffers should
reflect the opportunity cost of holding liquid assets rather than loans. It should also relate to
the distribution of liquidity shocks that the bank may face, and in particular be positively
related to the volatility of the funding basis as well as the cost of raising additional funds.
Using aggregate time-series data for Thailand, Agénor, Aizenmann and Hoffmaister (2000)
find that banks’ demand for precautionary reserves (measured as the log of excess reserves
over total deposits), is positively related to the penalty rate, proxied by either the discount or
the money market rate, as well as to the volatility of the cash to deposit ratio. Dinger (2009)
finds in a panel of Eastern European banks that liquidity buffers are negatively related to the
real deposit rate but positively related to the interbank rate.
Bank characteristics
The newer generation of models explaining firms’ (including banks’) liquidity demand relies
on some form of market imperfection to explain why banks cannot raise instantaneous and
unlimited amounts of liquidity (financial frictions). The market imperfection is asymmetric
information, either in the form of moral hazard (Holmstrom and Tirole, 1998) or adverse
selection (Kiyotaki and Moore, 2008). Financially constrained banks would thus tend to hold
more liquidity.
3,4
Based on these models, bank characteristics affecting their ability to raise non-deposit forms
of finance, such as bank size (small banks have more difficulties in accessing capital
markets), profitability (more profitable banks can more readily raise capital and are thus less
3
Holmstrom and Tirole (1998) and Kiyotaki and Moore (2008) make this argument for firms in general:
Liquidity constraints, together with liquidity shocks, result in entrepreneurs not being able to raise the entire
cost of their desired investment externally, so that they have to hold enough liquid assets to make a down
payment for each unit of investment (there are also limits on the amount of equity that can be resold). This
explains why, although the rate of return on cash is very low, entrepreneurs will choose to hold some in their
portfolio. Liquidity shocks reduce the price of equity and increase the desired holdings of liquid assets.
4
See for example Almeida, Campello and Weisbach, 2004, Kashyap, Rajan and Stein (2002), Kashyap and
Stein (1997), Repullo (2003), and Rochet and Vives (2004).
6
liquidity constrained), ownership (both public banks and foreign banks should be less
liquidity-constrained than private and domestic banks, respectively, as public banks may
have an implicit guarantee and foreign banks would have access to support from
headquarters)
5
would affect banks’ precautionary demand for liquidity buffers.
Aspachs, Nier and Tiesset (2005) find that banks’ liquidity buffers are related to bank
characteristics such as loan growth and the net interest margin, with the coefficients on size
and profitability being not significant. Kashyap and Stein (1997) and Kashyap, Rajan and
Stein (2002), using a large panel of U.S. banks, find a strong effect of bank size on holdings
of liquid assets, with smaller banks being more liquid as they face constraints in accessing
capital markets. Dinger (2009) also finds that smaller Eastern European banks hold more
liquidity, but with nonlinearities, and that foreign banks hold less liquidity.
Bank ownership may not only exert a direct influence on liquidity holdings, but may also
affect the regression slope through interactions with other explanatory variables. In
particular, Aspach, Nier and Thiesset (2005) find that, for the UK, foreign banks’ liquid asset
holdings are not affected by the availability of a domestic lender of last resort, while local
banks are. Furthermore, in their sample foreign banks’ liquidity holdings tend to react less to
changes in the domestic policy rate and GDP growth, suggesting overall that they are subject
to a somewhat different set of constraints than their local counterparts.
Macroeconomic fundamentals
The models mentioned above also have implications for the cyclical behavior of liquidity
demand. If capital markets are imperfect, the demand for liquidity should be countercyclical,
as banks would hoard liquid assets during recessions and offload them in good times given
more opportunities to lend. This suggests that liquidity buffers would be negatively related to
measures of the output gap or real GDP growth, credit cycle, and policy interest rates.
6
The counter-cyclicality of liquidity buffers limits the effectiveness of monetary policy in
trying to inject liquidity to stimulate the economy in a recession: liquidity buffers would
remain stable or increase but credit would not necessarily pick-up. Moreover, financial
frictions in terms of capital market imperfections should be expected to vary with structural
factors such as the degree of financial development and the quality of financial institutions.
Aspach, Nier and Tiesset (2005) find that UK banks’ liquidity buffers are negatively related
to real GDP growth and the policy rate. Agénor, Aizenmann and Hoffmaister (2000) and
Saxegaard (2006) find that excess reserves are negatively related to the output gap and the
policy rate in Thailand and in sub-Saharan Africa, respectively. Dinger (2009) finds that
liquidity holdings are negatively related to real GDP growth and real per capita GDP.
5
Freixas and Holthausen (2005).
6
Almeida, Campello and Weisbach (2004) develop and estimate on a large sample of U.S. manufacturing firms
a model where financially constrained firms have a higher propensity to save cash out of cash flows.
7
Moral hazard and safety nets
In theory, the strength of the financial safety net and in particular the availability of a LOLR
arrangement, should reduce the banks’ incentives to hold liquidity buffers (Repullo, 2003).
Empirical studies of UK and Argentinian banks, where LOLR support is measured as the
Fitch support rating and the availability of external credit lines in the context of the currency
board, respectively, support this prediction (Aspach, Nier and Thiesset, 2005, and Gonzalez-
Eiras, 2003).
Dollarization or credit and/or deposits reduces the effectiveness of the domestic LOLR, as
partially dollarized economies are subject to currency and liquidity risk, but the central bank
cannot issue foreign currency (Gulde et al., 2004 and Levy-Yeyati and Broda, 2002). One
would thus expect banks to hold higher liquidity buffers, the higher the degree of deposit
dollarization, though the incentives to hold such buffers would diminish in the presence of a
large stock of central bank international reserves or external credit lines, as these would be a
ready source of dollar liquidity in the case of a run on dollar deposits (Ize, Kiguel and Levy-
Yeyati, 2005). Using a sample of about 100 countries, De Nicoló, Honohan and Ize (2005)
find that deposit dollarization is associated with higher solvency and liquidity risk measured
by deposit volatility. To our knowledge no empirical study has focused on the effects of
deposit dollarization on banks’ liquidity.
III. SOME BACKGROUND ON CAPDR BANKING SYSTEMS
As a group of small open economies with strong linkages to the U.S. economy and various
levels of dollarization, most countries maintain fairly stable de facto exchange rates against
the U.S. dollar—although their formal monetary and exchange arrangements are quite
heterogeneous. Two economies (Panama and El Salvador) are fully dollarized, three
countries (Costa-Rica, Dominican Republic and Guatemala) have recently adopted or are
moving toward inflation targeting, and Honduras and Nicaragua maintain crawling or tightly
managed pegs.
7
The monetary policy frameworks in the five partially dollarized economies
rely on rules-based instruments such as standing facilities and reserve requirements, with
only partial money market operations, and all but Nicaragua have a monetary policy interest
rate (Medina Cas, Carrion-Menendez and Frantischek, 2011b).
Reserve requirements in CAPDR are in line with those of the Latin America region, and
average about 15 percent for local currency deposits and 15.5 percent for foreign currency
ones (Appendix I, Table 1 and Figure 2)
. The two officially dollarized economies rely on
prudential liquidity requirements, held at the central bank in the case of El Salvador and by
individual banks in the case of Panama (Appendix I).
8
Although they are potentially useful
7
Costa Rica maintains a managed exchange rate regime with pre-defined bands for intervention but a
substantial margin to fluctuate within the bands. Nonetheless, since December 2010 the exchange rate has
remained within 3 percent of the floor of the band.
8
Honduras also imposes specific liquidity requirements to avoid maturity mismatches (Appendix I).
8
policy instruments, reserve/liquidity requirements are not actively managed in most
countries, with the exception of El Salvador and Honduras (Appendix I, Table 2).
9
Figure 2. CAPDR: Statutory Reserves and Liquidity Requirements by Currency
With the exception of Panama, the region’s banking systems are relatively small, highly
concentrated and dollarized to various degrees (Table 1). Panama stands out of the group in
terms of the size of the system, which is four times greater than the sample average in terms
of assets to GDP (Panama’s offshore banks’ assets represented only 50 percent of GDP at
end-2010). In four countries (Honduras, Nicaragua, El Salvador and Panama) the share of
foreign bank assets in total assets is more than 50 percent suggesting higher potential
vulnerabilities from cross border linkages. While the presence of state banks is quite small in
terms of number of banks and share of system’s assets for the whole sample, state banks have
a very strong presence in Costa Rica, with their assets accounting for 55 percent of total
assets and 60 percent of deposits. Customer deposits are the main source of funding and
show a high degree of dollarization, particularly in Nicaragua and Costa Rica. The share of
short-term deposits is also relatively high in the region, although not in Panama.
9
Given this, excess liquidity is probably best analyzed in the context of single country time-series studies. In
the panel context, our preferred definition of liquidity buffers for the empirical analysis in section IV.C is the
liquid assets to deposit ratio. The liquid assets to total assets ratio is used for robustness checks.
Source: Central Banks and Superindendencies.
Note: Reserve requirements for all countries excluding Panama and Salvador (liquidity requirements). Liquidity requirement for Panama is def ined as the ratio of liquid assets
including securities and obligations payable to banks within 186 days, as a share of short-termdeposits.
15
17
14.6
12
16.25
17
26
30
0
5
10
15
20
25
30
35
C
o
s
t
a
R
i
c
a
D
o
m
.
R
e
p
.
G
u
a
t
e
m
a
l
a
H
o
n
d
u
r
a
s
N
i
c
a
r
a
g
u
a
L
a
t
i
n
A
m
e
r
i
c
a
E
l
S
a
l
v
a
d
o
r
P
a
n
a
m
a
Local Currency Reserve or Liquidity
Requirements (percent of local currency
deposits)
15
20
14.6
12
16.25
14
0
5
10
15
20
25
30
35
C
o
s
t
a
R
i
c
a
D
o
m
.
R
e
p
.
G
u
a
t
e
m
a
l
a
H
o
n
d
u
r
a
s
N
i
c
a
r
a
g
u
a
L
a
t
i
n
A
m
e
r
i
c
a
Foreign Currency Reserve Requirements
(percent of foreign currency deposits)
9
Recent Financial Sector Assessments indicate that, overall, banking sectors in the region are
well-capitalized, liquid and profitable (Figure 3). Financial systems remained resilient in the
face of the 2009 global financial crisis mostly due to their strong initial positions. Despite
rapid credit growth, the region did not experience any excessive credit booms and there was
very limited exposure to toxic asset-backed securities, as well as to wholesale funding. Stress
tests of liquidity risk suggested that banks had adequate coverage of their liquid liabilities,
and could withstand deposit withdrawal shocks of 15–20 percent during a 30 day period.
However, although banking supervision has improved over the past decade, compliance with
Basel Core Principles remains uneven and below that of LA6. Financial safety nets remain
incomplete, and financial markets are underdeveloped, including interbank markets (Box 1).
Number of
banks
5
Number of
State
banks ²
State bank
assests in total
assets
Number of
foreign
banks ³
Foreign
bank
assets in
total assets
Percent of
assets in 5
largest
banks
Assets to
GDP
Credit to
GDP
Assets in
foreign
currency in
total assets
Credit in
foreign
currency in
total credit
Deposits in
foreign
currency in
total
deposits
Demand
deposits
in total
deposits
Costa Rica 16 3 55 9 26 78 60 46 46 47 41 53
Guatemala 18 1 2 7 13 79 44 30 23 30 24 41
Honduras 17 1 1 9 50 75 68 50 24 28 30 22
Nicaragua 6 1 0.01 4 67 97 62 34 72 90 73 31
Dominican Republic 15 2 31 4 29 87 32 18 26 21 30 18
El Salvador ¹ 12 2 6 10 83 85 61 40 … … … 26
Panama
1, 4
49 2 14 28 57 57 200 91 … … … 15
Source: Central American Monetary Council (SECMCA), International Financial Statistics, IMF staff calculations.
¹ Officially dollarized economies. ² State share of more than 50 percent. ³ Banks with 50 percent of capital in foreign hands, excludes offshore.
4
Domestic banking system;
5
April 2011.
Table 1. CAPDR: Banking System Indicators, 2010
(in percent, unless otherwise indicated)
10
Box 1. Financial Systems Supervision and Development in CAPDR
Despite substantial progress over the past decade, financial supervision in CAPDR lags behind,
both compared to larger emerging countries in Latin America and international best practices.
Following a string of banking crises in the 1990s - Costa Rica (1987-91, 1994-95), El Salvador
(1989-90), Nicaragua (1990-93), and Panama (1988-89)
,1/ a system-wide banking crisis in the
Dominican Republic (2003-04), and the failure a large and a medium-sized bank in Guatemala
(2006-07), the Central American countries strived to strengthen their supervisory frameworks. By
2010, CAPDR complied with more than half of the Basel Core Principles, though the performance
was uneven across the region, with Panama showing the highest compliance and Costa Rica the
lowest. The strengthening of financial supervision during the last decade has been relatively
homogeneous between categories, but unequal by category within each country. The weakest areas
in the region are risk-based supervision, cross-border consolidated supervision, and the fact that
nonbank institutions frequently are outside the supervisory perimeter (Delgado and Meza, 2011).
Financial safety nets remain incomplete. All countries, except Panama and El Salvador, have a
LOLR (though El Salvador is in the process of formally implementing one), and deposit insurance
schemes are in place in all but two countries (Costa Rica and Panama). Weaknesses in most
countries are observed in providing legal protection to supervisors and regulating systemic risk.
The region’s money and capital markets remain underdeveloped.2/ Interbank markets are thin,
segmented, and often lack proper collateral. Despite the high level of liquidity in the banking
sector, the absence of collateral in inter-bank transactions increases systemic vulnerability to bank
failures. A lack of active repo operations also stalls the standardization of transactions in the
interbank market (see FSSAs for the region). The number of equity and corporate bond listings are
generally in the single digits and market-capitalization-to-GDP ratios are quite low. Due to legal
and other obstacles, institutional investors intermediate a relatively small share of national
savings. Thus, the principal investors in government securities are commercial banks and public
sector entities such as pension or social security funds. Secondary public debt markets are shallow
as government securities are mostly held to maturity.
____________________
1/ Panama did not experience a typical banking crisis as the closure of the banking system for several
months was due to the invasion of the country by the U.S. armed forces confronting the Noriega regime.
There was no deposit runs prior or after the closure of the banks and there were no bank failures. However,
this contributed to a significant output loss, estimated by some at 85 percent (Laeven and Valencia (2012)).
2/ See Shah et al. (2007a), and Shah et al. (2007b).
11
Figure 3. CAPDR: Financial Soundness Indicators, 2005-11/1
Source: Central American Monetary Council (SECMCA).
1/ Shaded area represents the 2008-09 global financial crisis.
2/ Liquid assets include short term investments.
30.0
35.0
40.0
45.0
50.0
55.0
60.0
65.0
Jan-05Jan-06Jan-07Jan-08Jan-09Jan-10Jan-11
Liquid Assets /Total Assets/ 2
1.0
1.5
2.0
2.5
3.0
3.5
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Return on Assets (ROA)
9.0
14.0
19.0
24.0
29.0
34.0
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Return on Equity (ROE)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Non Performing Loans (NPLs) / Total
Loans
... a deterioration in the quality of loan portfolios.
50.0
75.0
100.0
125.0
150.0
175.0
200.0
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Provisions/NPLs
Despite the global crisis, risk-adjusted capital
ratios remain adequate...
Bank profitability declined in 2008, with falls in the
return on assets...
..and by 2010 banks had rebuilt their liquidity
and return on equity, reflecting in part...
Provisioning has generally lagged the growth
in NPLs, but there are signs of recovery.
6.0
9.0
12.0
15.0
Jan-05Jan-06Jan-07Jan-08Jan-09Jan-10Jan-11
Capital Adequacy Ratio (CAR)
3decile
CAPDR-median
7decile
12
Figure 5. Liquidity Ratios at the System Level and in the Bankscope Sample
(2006-10 average, in percent)
Source: BankScope database, CAPDR Central Banks and Superintendencies' websites;
IMF staff calculations.
0
5
10
15
20
25
30
35
40
45
Costa Rica Dominican
Rep.
El
Salvador
Guatemala Honduras Nicaragua Panama
Banking System's Liquid Assets to Deposits Ratio
Bankscope Sample: Liquid Assets to Customer Deposits and Short Term Funding
Figure 4. Bankscope Sample Coverage of Total Banking System's Assets
Source: BankScope database, CAPDR Central Banks and Superintendencies' websites,
IMF staff calculations.
(2006-10 average, in percent)
98.38 97.44
90.68
85.91
63.24
56.98
54.94
0
20
40
60
80
100
120
Costa Rica Dominican Rep. Guatemala Honduras Panama El Salvador Nicaragua
IV. DETERMINANTS OF BANKS LIQUIDITY BUFFERS IN CAPDR
A. Data and Variable Definitions
Our sample combines annual data for 96 CAPDR banks over 2006–10 from the BankScope
data base,
10
with country-level
macroeconomic fundamentals and
structural variables drawn from
regional monetary and supervisory
authorities’ websites and other
publicly-available databases as
described in Table 2 below and
Appendix III, Table 1. It covers
72 percent of all commercial banks in
the region and about 80 percent of
total banking system assets, though
admittedly the coverage is not
homogeneous across all countries in
the region (Figure 4).
11
Definition of liquidity buffers
Liquidity buffers are measured by the
ratio of liquid assets to customer
deposits and short-term funding.
Liquid assets include cash and cash-
like assets, quoted or listed
government bonds, and short-term
claims on other banks. Although the
breakdown of the numerator
components is not available, there are
relatively few listed government
securities in the region (Shah et al.,
2007b). The denominator includes
banks’ customer deposits and short-
term interbank deposits. Customer deposits are the main source of funding in the region, with
very low reliance on short-term funding (the share of customer deposits in the denominator is
93 percent for the whole sample). Overall, the ratio of liquid assets to customer deposits and
short-term funding is close to system-wide liquidity ratios, defined as liquid assets (cash and
10
A financial database supplied by Bureau van Djik.
11
The information on coverage averaged over banks/years. A caveat is that missing institutions may not be
random. The time period is restricted to the interval for which data for most CAPDR banks were available.
13
cash-like, excluding securities) to deposits (Figure 5). We use it as our main dependent
variable, and use the ratio of liquid assets to total assets for robustness checks.
12
Choice of explanatory variables
Opportunity cost, liquidity shocks and bank characteristics
We use the spread between the lending and the deposit rate as a measure of the opportunity
cost. The probability of a liquidity shock can be proxied by a measure of the volatility of total
deposits at the system level (we can calculate a monthly coefficient of variation of total
deposits for each country, but have only annual bank-level data), or by the volatility of
inflation. Past liquidity shocks may also matter: a history of banking crisis could lead banks
to become more risk-averse and hold more liquidity.
Given the importance of public and foreign banks in Central America’s banking systems, we
are particularly interested in testing whether liquidity buffers vary systematically according
to bank ownership (public/private and foreign/domestic).
We control for other bank characteristics such as size, measured as the log of total assets. The
squared value of this variable captures possible non-linearities in the impact of bank size on
liquid asset holdings (Dinger, 2009). Capitalization is measured as the ratio of equity to total
assets. Profitability is measured by the ratio of the net interest margin to interest-earning
assets. The loan-loss reserves to gross loans ratio should capture the banks’ degree of risk
aversion or perceived riskiness of the loan portfolio.
Macroeconomic fundamentals
We use output growth in CAPDR to capture the economic cycle. Financial development is
captured by private sector credit to GDP ratio, a traditional proxy for financial depth. We
also construct a quality of credit institutions index which combines four variables from the
World Bank’s Doing Business database
13
focused on credit and legal rights, ?strength of
legal rights index (0-10); depth of credit information index (0-6); private bureau coverage
(percent of adults), and the inverse of the cost of enforcing contracts (cost as percent of
claim). Since these variables are correlated, we use principal components analysis to
construct an index which captures the underlying common variance of those four variables
(See pair-wise correlations in Appendix II. Table 1).
14
The index is normalized so a value of
12
Empirical studies use both ratios, see Aspach, Nier and Thiesset (2005) and Dinger (2009). The ratio of liquid
assets to liabilities is the most consistent with the notion of CAPDR banks self-insuring against deposit shocks,
though banking theory also emphasizes asset-side liquidity problems (Diamond and Rajan, 2005).
13
Available on the internet at www.doingbusiness.org.
14
Principal components analysis is a multivariate statistical technique that uses n correlated variables, and
reduces their dimension such that each component is a linear combination of the initial variables, and
components are orthogonal with each other. The components are ordered so that the first component accounts
for as much of the variability in the original data as possible. We use the first component as our index of the
(continued…)
14
zero represents the lowest quality of credit institutions, and 100 the highest. Although both
measures of financial development are highly correlated in some CAPDR countries (e.g.
Panama, Nicaragua), they differ quite radically in others (e.g. the Dominican Republic, El
Salvador) (Figure 6). Whereas credit-to-GDP is an outcome variable that captures both
supply and demand factors, the quality of credit institutions is more likely associated with
obstacles to the supply of credit. In this case, the lack of alternative investment opportunities,
rather than pure precautionary motives, would also explain high levels of bank liquidity.
Figure 6. Financial Development Indicators in CAPDR
Moral hazard and safety nets
We construct an index variable categorizing the existence of each of the four pillars of a
safety net in each country (regulation and supervision, banking resolution, LOLR facility,
and deposit insurance scheme). We attribute a value of 1 if the pillar is present and
functioning in a country, so the index can vary between zero and four, though in practice all
countries have banking regulation and supervision systems in place (Appendix II, Table 2).
Dollarization is measured as the share of dollar deposits in total system deposits (no currency
breakdown is available for bank-level data in BankScope). Net international reserves
holdings capture the dollar LOLR function of central banks.
quality of credit institutions in CAPDR, which we normalize to bound its values between 0 and 100 (see
Cárdenas and Henao, 2010, for an application).
Sources: WEO database, Doing Business database; IMF staff calculations
0
10
20
30
40
50
60
70
80
90
100
Costa Rica Dominican
Republic
El Salvador Guatemala Honduras Nicaragua Panama
Total credit to GDP (percent) Credit Institutional Index (0 worst; 100 best)
Total credit to GDP (percent) and credit institutional index (0-100)
15
Data description
Table 2 below presents the variable definitions, expected sign and data sources, and
Appendix II, Table 1 in annex describes the data. Overall there is significant variation in
liquidity holdings in the sample. Liquidity holdings in terms of customer deposits and short-
term funding are on average 25 percent in our sample, and represent about 18 percent of total
assets. Average capitalization is relatively high at about 13 percent, as noted in Basso,
Delgado and Meza (2012). Foreign banks represent 45 percent of observations, and private
banks about 90 percent. Deposit dollarization amounts to about 50 percent, though with wide
variations across countries (see Table 1).
Table 2. Variables Used in the Empirical Estimation
Variable Name
(expected sign)
Concept Measurement Data
source
Dependent variable
Liquidity ratio Liquid assets to customer
deposits and short-term
funding
(Cash, short-term claims on
other banks (including CDs)
and where appropriate the
trading portfolio)/Customer
deposits and short-term
funding.
BankScope
Liquid assets to total
assets
(Cash, short-term claims on
other banks (including CDs)
and where appropriate the
trading portfolio)/total
assets.
BankScope
Explanatory variables
Bank Characteristics
Lagged liquidity ratio
(+)
Liquidity buffers should be
persistent over time
See above for definition BankScope
Capitalization (-) Better capitalized banks
should have easier
access to markets and
thus hold less liquidity.
Ratio of equity to total
assets.
BankScope
Net Interest Income
to Average Earning
Assets (-)
Profitability: more
profitable banks should
hold less liquidity.
(Interest income-interest
paid)/ interest earning
assets.
BankScope
Loan-loss reserves
ratio (+)
Perceived riskiness by
banks of their loan
portfolio: banks
anticipating higher losses
should hold higher
liquidity buffers.
Ratio of loan-loss reserves
to gross loans.
BankScope
Size (-) If small banks are
financially constrained
than they should hold
more liquidity.
Natural logarithm of total
assets.
BankScope
Foreign ownership
(-)
Foreign banks should be
less financially-
constrained than
domestic banks and thus
hold lower levels of liquid
assets.
Dichotomous variable (1 for
foreign; 0 for domestic).
BankScope,
country
desks
A distinction is also made
between foreign subsidiaries
and branches.
16
Private ownership
(+)
Public banks could have
less incentives to hold
liquid assets as they
benefit from an implicit
state guarantee
Dichotomous variable (1 for
private; 0 for public).
BankScope,
country
desks.
Macroeconomic fundamentals
Real GDP growth (-) Imperfect capital markets
imply that liquidity buffers
should be countercyclical.
Annual growth rate of real
GDP per capita.
IMF country
desks
Interest Rate Spread
(-)
Measure of the
opportunity cost of
holding liquid assets.
Difference between average
lending and deposit rate
CAPDR
Monitor, IMF
country
desks.
Country Characteristics
Deposit volatility (+) Higher aggregate deposit
volatility forces banks to
hold more liquid assets to
hedge against
unanticipated deposit
withdrawals.
Coefficient of variation of
monthly system-wide
deposits during one year.
IMF country
desks.
Inflation volatility (+) High inflation volatility is a
proxy for macroeconomic
instability.
Coefficient of variation of
monthly inflation during one
year.
WEO.
History of banking
crisis (+)
Banks in countries that
have experienced a
banking crisis should hold
higher liquidity buffers.
Dichotomous variable (1 if
country experienced a
banking crisis in the past 50
years).
Laeven and
Valencia
(2012)
Credit-to-GDP ratio
(-)
Captures financial
development. More
developed economies
should have less financial
constraints.
Total credit to the private
sector as percent of GDP.
SECMCA
Quality of credit
institutions index (-)
Financial constraints
should be lower in
countries with better
credit institutions
Index capturing the quality of
credit institutions (0 is worst,
100 best).
World Bank
Doing
Business
Database
Moral hazard and safety nets
Deposit dollarization
(+)
The higher the
dollarization, the lower the
effectiveness of the
domestic lender of last
resort.
Share of foreign-exchange
deposits in total deposits
(system-wide).
IMF country
desks
Net international
reserves (-)
In partially dollarized
economies, NIR capture
the capacity of the central
bank to act as a lender of
last resort in case of a
foreign currency shock.
Natural logarithm of net
international reserves
IFS
Lender of last resort
(-)
Incentives to hold liquidity
decline with a strong
safety net.
Index variable categorizing
the existence of each of the
4 pillars of a safety net
(regulation, banking
resolution, LOLR and
deposit insurance.
Bolzico et
al.(2010),
Guerrero et
al. (2010)
17
B. Stylized Facts
At the country level, CAPDR banks’ liquidity holdings are above liquidity ratios in larger
South American countries, with the exception of Paraguay. Liquidity ratios have remained
fairly stable during the past decade including the global financial crisis (though with
variations across countries) – and have moved countercyclically, increasing when lending
opportunities were low and decreasing when they were high, consistently with modern
banking models’ predictions. Lower credit demand and implementation of countercyclical
fiscal policies during the crisis also led to an increase in banks’ holdings of government
securities over the past few years (Figure 7).
Figure 7. CAPDR: Liquidity Indicators
Moving to our bank sample, Table 3 summarizes the relationships between key explanatory
variables and liquidity ratios. More specifically, it shows the sample mean of explanatory
variables by liquidity quartile (with the quartiles being ranked from low to high). We are
interested in whether the characteristics of banks with high liquidity buffers (in the fourth
quartile) are different from those with low liquidity buffers (first quartile). Thus the last
column of the table shows whether the difference in the means of the variables for banks in
the 4
th
liquidity quartile versus the first quartile is statistically significant.
Source: CAPDR Central Banks, Superintedencies' websites, WEO; Fund staff calculations.
-5
0
5
10
15
20
25
30
35
40
2005 2006 2007 2008 2009 2010 2011
Government Bond Holdings Liquidity Credit to Private Sector
Liquidity and Domestic Government Assets
(annual percent change)
CRI
DR
SLV
GTM
HND
NIC
PAN
PAR
COL
ECU
CHI
R² = 0.61
0
2
4
6
8
10
12
14
10 20 30 40 50 60 70
¹ Liquid assets (including securities) to deposits ratio (percent)
Liquidity Ratios in CAPDR and Selected LA Countries ¹
(2010)
G
D
P
p
e
r
c
a
p
i
t
a
(
c
o
n
s
t
a
n
t
2
0
0
0
U
S
D
,
a
v
g
.
2
0
0
8
-
1
1
)
Liquidity Ratios in CAPDR and Selected LA Countries ¹
(2010)
BRZ
15
17
19
21
23
25
27
29
31
33
35
J
a
n
-
0
5
S
e
p
-
0
5
M
a
y
-
0
6
J
a
n
-
0
7
S
e
p
-
0
7
M
a
y
-
0
8
J
a
n
-
0
9
S
e
p
-
0
9
M
a
y
-
1
0
J
a
n
-
1
1
S
e
p
-
1
1
Percentile 30%
Median
Percentile 70%
Liquidity to Deposits, median for CAPDR
(in percent)
-10
-5
0
5
10
15
M
a
r
-
0
3
N
o
v
-
0
3
J
u
l
-
0
4
M
a
r
-
0
5
N
o
v
-
0
5
J
u
l
-
0
6
M
a
r
-
0
7
N
o
v
-
0
7
J
u
l
-
0
8
M
a
r
-
0
9
N
o
v
-
0
9
J
u
l
-
1
0
M
a
r
-
1
1
Liquidity to Deposits
Index of economic activity
Evolution of Liquidity Buffers and Economic Activity
(Median, y/y percent change, 3 mo. moving av.)
18
Overall most proposed explanatory variables exhibit the predicted relationship to liquidity
buffers, though not all the differences between the first and last quartile are significant. As
compared to those in the first liquidity quartile, observations (bank/years) for which observed
liquidity is high also tend to be less profitable, smaller, and private. They also have a lower
probability to be in a country with a relatively low interest rate spread, low deposit
dollarization, a history of banking crisis, an incomplete financial safety net. High liquidity
also seems associated with lower quality of financial institutions.
15
However, high liquidity
also seems to be associated with lower inflation volatility and higher financial depth.
15
Given that our time dimension is relatively short, and that it covers the period of the global financial crisis
period, we are also interested in testing whether the behavior of the main explanatory variables was different
pre-, during- and post-crisis. We find that for the crisis years (2008-09) the main relationships identified for the
whole sample continue to hold (Tabulations available upon request).
1
st
2
nd
3
rd
4
th
p-value 1/
Mean of Liquidity to customer and short-term
funding ratio 11.93 18.91 25.38 46.87
Loan Loss Reserves to Gross Loans 3.47 3.11 3.14 3.16 -0.32 0.46
Net Interest Margin 9.73 8.71 8.53 8.09 -1.64 0.26
Capitalization (equity to asset ratio) 13.69 11.52 11.88 15.55 1.86 0.23
Bank Size (log of total assets) 12.87 13.00 13.02 12.36 -0.51 0.03
Foreign ownership dummy (=1 if foreign bank) 0.41 0.38 0.53 0.43 0.02 0.74
Private ownership dummy (=1 if private bank) 0.88 0.89 0.94 0.94 0.07 0.07
Real GDP growth 3.24 4.40 3.50 4.04 0.80 0.32
Net International Reserves 2946.79 3443.16 2947.53 2515.13 -431.67 0.00
CPI volatility 2.27 2.10 2.01 1.87 -0.40 0.01
Interest rate spread 9.41 8.32 8.12 7.28 -2.13 0.00
Deposit volatility 3.91 4.06 3.53 3.94 0.03 0.91
Deposit dollarization 37.56 44.21 52.39 64.93 27.36 0.00
Banking crisis dummy (=1 if country had banking
crisis after 1970) 0.53 0.55 0.72 0.88 0.35 0.00
Financial safety net (categorical variable, 1-4;
4=comprehensive safety net) 3.08 2.63 2.66 2.50 -0.58 0.00
Credit to GPD (%) 43.86 44.52 48.86 54.81 10.94 0.00
Quality of credit institutions index (0=worst;
100=best) 66.27 58.46 71.75 78.44 12.17 0.00
Memorandum item
Number of observations 120 109 112 107
Source: Authors' calculations
Table 3. Dependent Variables Means by Liquidity Ratio Quartiles
1/ P-value from a test of statistical difference of the means of the 4
th
quartile versus the 1
st
quartile.
Liquidity ratio quartile
(4
th
- 1
st
quartile)
19
C. Empirical Analysis
Baseline Specification
In line with the discussion in the previous section and similar to other studies (Aspach et al.,
2005, Barajas et al., 2010, and Dinger, 2009), we estimate the determinants of the liquidity
buffers based on bank characteristics, macroeconomic fundamentals and country specific-
characteristics.
The baseline specification can be represented by equation (1):
I
ìt
= [
0
+ [
1
I
ì],t-1
+ [
2
bonk
ì]t
+ [
3
mocro
]t
+ [
4
country
]t
+ p - ] + v - t + ç
ì]t
(1)
Where the subscripts i, j and t refer to bank, country and time (year) respectively. L
represents bank level liquidity buffers. We include a lagged dependent variable: if, as
predicted by theory, banks target an optimum level of liquidity holdings, then we should
expect these holdings to be persistent over time, as shown by Opler et al. (1999) in the case
of U.S. firms. Bank denotes variables measuring bank fundamentals and are derived from
the balance sheets of banks. Macro represents the macroeconomic determinants of individual
banks’ liquidity buffers such as real GDP growth and interest rates, and country are
observable country level characteristics, including, for ease of presentation, the moral hazard
and safety net variables presented in the previous section and Table 2. Unobservable country
and time effects are captured by country (j) and time (t) dummy variables.
Hypotheses of interest
Based on our review of the theoretical and empirical literature as well as stylized facts on
liquidity data for CAPDR countries, we will pay particular attention to the following:
(i) Does ownership matter? We test separately for the effect of private vs. public
ownership, and domestic vs. foreign. As discussed in Section III, ownership may not only on
exert a direct influence on liquidity holdings, as discussed in section of ownership on
liquidity holdings, but may also affect the regression slope through interactions with other
explanatory variables. To test this hypothesis, we interact the relevant ownership dummy
variable (own
ijt
) with the other explanatory variables as shown in Equation (2):
I
ìt
= [
0
+ [
1
I
ì],t-1
+ [
2
bonk
ì]t
+ [
3
mocro
]t
+ [
4
country
]t
+ ([
5
bonk
ì]t
-
own
ì]t
) + ([
6
mocro
]t
- own
ì]t
)+([
7
country
]t
- own
ìt
) + p - ] + v - t + ç
ì]t
(2)
(ii) We use the same framework to test whether liquidity buffers (banks perceived need
for self-insurance) are higher in countries with more dollarized banking systems, as
measured by the share of foreign currency deposits in total deposits. We also test whether
liquidity buffers might be higher in countries with less comprehensive financial safety nets.
20
Estimation Methodology
Equations (1) and (2) are first estimated by Ordinary Least Squares (OLS) using the Least
Squares Dummy Variable approach. This methodology enables us to introduce and identify
bank, country and time effects on bank level liquidity. However, there may also be
unobserved bank-specific and/or country specific time-invariant heterogeneity, which could
bias our estimates if not properly accounted for. The error term may contain time varying
bank or country-specific characteristics which may be correlated with banks’ liquidity ratios.
Another issue is potential endogeneity of some of the explanatory variables such as credit to
GDP.
To address these concerns, we also estimate equations (1) and (2) using the Generalized
Methods of Moments (GMM) developed by Blundell and Bond (2000) and Bond (2002).
GMM estimators are particularly appropriate to address the dynamic panel bias that arises in
the presence of lagged dependent variables in samples with a large number of groups (N) and
a relatively small number of time periods (T), such as ours. Given persistent liquidity ratios,
our preferred estimator is the Systems GMM as it helps overcome the weak instrument
problem (past changes do contain information about current levels), and results in
improvements in the efficiency of the estimates (Arellano and Bond, 1991, Roodman,
2006).
16
D. Results
Tables 4a and 4b present estimation results of a parsimonious robust specification of
equations 1 and 2 above, using the ratio of liquid assets to customer and short-term funding
as a dependent variable.
17
Table 4a includes the credit-to-GDP ratio as a proxy for financial
depth, while Table 4b presents estimation results including the index of the quality of credit
institutions.
16
This was implemented in STATA using Roodman’s (2006) xtabond2 routine. To avoid instrument
proliferation, we restrict the number of lags for the GMM instruments to 2 (Roodman, 2009). We treat the bank
size, country and year dummy variables as predetermined and the rest as endogenous. In addition to OLS
(whose estimate of the lagged dependent variable coefficient is biased upward) we also estimated the model
with robust fixed effects (with the coefficient on the lagged dependent variable is biased downward). Results of
the fixed effects estimations are in line with those from OLS and GMM estimates (shown for the baseline
specification in Appendix IV, Table 2).
17
Given the limited time span of our panel, the coefficients on the macroeconomic variables (real GDP growth,
interest rate spread) were overall consistent with predictions but not significant nor very robust, as part of the
effect of these variables on liquidity buffers was likely captured by the country and time dummies. GMM
estimation of the full model also became difficult as the number of instruments was becoming too large relative
to available observations.
21
Baseline specification
Estimation results from the baseline specification are very robust to the choice of financial
development variable (Tables 4a and 4b, columns 1 and 2). They show that liquidity buffers
in CAPDR are persistent: the coefficient on the lagged dependent variable is positive and
significant. This is consistent with the view that banks target an optimal or desired level of
precautionary liquidity holdings, but could also be attributed to the presence of structural
obstacles to credit that lead banks to hold higher liquidity buffers.
Liquidity ratios are related to bank size, though with non-linearities: liquidity holdings
increase with bank size, but there is a point at which bank size begins exhibiting a marginal
decreasing effect on liquidity. This is the opposite of what is found by Dinger (2009) in
Eastern Europe, and may be explained by differences in the distribution of bank size in both
regions. In CAPDR, the distribution of banks is highly skewed with quite high concentration
of assets in a few large banks, as indicated in Table 1.
18
Liquidity holdings are also negatively related to the loan-loss reserve ratio, indicating that
banks with higher savings against potential losses or riskier loan portfolios also tend to have
lower liquidity buffers in CAPDR. They are negatively associated (though the relationship is
not as robust as for the previous two variables) with the net interest margin, as expected. The
coefficient on capitalization is negative and significant in the baseline, indicating that better
capitalized banks would tend to hold less liquidity (the coefficient remains negative but is
generally no longer significant in the specifications with interaction terms). This is somewhat
counterintuitive, as the expectation would be that better capitalized banks would also hold
more liquidity buffers, if higher capitalization is indicative of a prudent business model. In
Table 4a, the credit-to-GDP ratio is negatively related to liquidity buffers, in line with
predictions (though the coefficient is not significant). In table 4b, the coefficient on the credit
institutions variable is positive and significant in the GMM regression.
19
Specifications with interaction terms—the role of bank ownership
Results indicate that ownership has some effect on liquidity holdings, though mostly through
the interaction terms. Our results do not show any significant evidence that private ownership
does affect liquidity buffers, though the coefficient on private ownership is positive in the
GMM specification (consistent with Table 3). Foreign banks tend to hold less liquidity, but
the coefficient on ownership is not statistically significant either. Foreign banks with riskier
loan portfolios or which are more conservative regarding expected loan losses do tend to
18
In estimations without the quadratic term the coefficient on bank size is negative and robust across
specifications as expected from theory and found in related empirical studies (results available upon request).
19
Estimating a more comprehensive model including all macroeconomic fundamentals presented in Table 2
indicate that the results on bank characteristics and variables of interest remain robust. However, as indicated
before the macroeconomic variables are not very precisely estimated in our short panel and although
coefficients on these variables are overall in line with predictions they are neither always consistent nor
significant.
22
have higher liquidity buffers (Table 4a, column 6). This is consistent with predictions and
findings in Detragiache, Tressel and Gupta (2008), which show that foreign banks tend to be
more prudent and lend to less risky customers.
Specifications with interaction terms—deposit dollarization
As indicated in columns 7–8 of Tables 4a and 4b, deposit dollarization is robustly and
significantly associated with higher liquidity buffers, both by itself (in both tables) and
through interactions with other bank-level variables (Table 4b). The individual effect is quite
large: a one standard deviation (34 percent) increase in deposit dollarization leads to a 70 to
100 percent increase in the liquidity to deposit ratio.
20
The strong positive association
between deposit dollarization and liquidity buffers may however not necessarily indicate a
direct causal relationship. The same factors that cause households and firms to hold more
dollar deposits could very well also lead banks to hold more precautionary liquidity.
21
Nonetheless, the positive relationship between dollarization and high liquidity holdings
would provide yet another reason why the monetary transmission mechanism is slower in
more dollarized economies (as in Medina Cas, Carrion-Menendez and Frantischek, 2011).
The interaction with the loan-loss reserve ratio also indicates that prudent banks or banks
with risky loan portfolios in dollarized economies tend to hold more liquidity (though the
coefficient is not significant in the GMM specification). More profitable banks in dollarized
economies tend to hold less liquidity. The relation between dollarization and the quality of
credit institutions is interesting: whereas there is a positive and significant association
between the quality of credit institutions and liquidity buffers, the sign of the coefficient
switches when institutions are interacted with the degree of deposit dollarization. As
mentioned above, this negative association may indicate either that high dollarization itself is
a result of lower quality of domestic credit institutions, or that sound credit institutions in
dollarized economies help reduce bank liquidity.
Specifications with interaction terms—financial safety net
Columns 9 and 10 of Tables 4a and 4b show the interactions with the financial safety net
variable. Though not significant, the coefficient on the GMM estimation is negative,
indicating that banks in countries with a more comprehensive financial safety net tend to hold
20
Given that reserve requirements are set at the same rate for local and foreign currency deposits in most
countries, and that actual liquidity holdings are held above requirements, it is unlikely that this result is driven
mechanically by reserve requirements. However, the large standard deviation is in part due to the fact that the
share of foreign deposits in total is 100 percent in El Salvador and Panama.
21
De Nicoló, Honahan and Ize (2005) find in a large cross-country sample that the credibility of
macroeconomic policy and the quality of institutions are key determinants of deposit dollarization.
23
2
3
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM
Liquid assets ratio (-1) 0.347*** 0.189*** 0.355*** 0.218* 0.371*** 0.231** 0.336*** 0.223** 0.345*** 0.290***
(0.064) (0.044) (0.065) (0.114) (0.072) (0.099) (0.067) (0.092) (0.064) (0.086)
Bank size 7.401*** 7.994*** 7.327*** 8.545*** 7.635*** 10.381** 8.460*** 5.639** 8.326*** 7.281***
(1.448) (1.875) (1.472) (2.299) (1.609) (4.137) (1.629) (2.635) (1.354) (1.546)
Bank size squared -0.350*** -0.371*** -0.351*** -0.392*** -0.366*** -0.483** -0.401*** -0.244* -0.392*** -0.325***
(0.064) (0.092) (0.065) (0.126) (0.073) (0.203) (0.072) (0.129) (0.061) (0.073)
Capitalization -0.355*** -0.321** -0.318*** -0.505 -0.288*** -0.316 -0.219 -0.017 -0.375 -0.238
(0.060) (0.123) (0.043) (0.336) (0.058) (0.305) (0.220) (0.542) (0.299) (0.928)
Net interest margin -0.156** -0.123 0.718 -0.089 -0.278* -0.593 -0.210 0.404 0.577 1.278*
(0.064) (0.076) (0.569) (1.067) (0.164) (1.199) (0.133) (0.331) (0.359) (0.690)
Loan-loss reserve ratio -0.224 -0.282 -0.505** -0.035 -0.419*** -0.550 -1.252*** -0.799 1.738 0.780
(0.160) (0.252) (0.216) (0.588) (0.160) (0.506) (0.394) (0.624) (1.092) (1.743)
Credit to GDP ratio -0.213 -0.323 -0.337 0.404 -0.277 -0.441 -0.374 -0.041 0.342 -0.593
(0.272) (0.292) (0.283) (0.664) (0.275) (0.344) (0.756) (0.679) (0.717) (0.729)
Variable -2.623 42.500 -3.002 -13.249 1.068** 1.491*** 7.345 -32.043
(6.514) (36.406) (6.185) (18.512) (0.516) (0.470) (7.880) (33.069)
Capitalization*variable -0.105 0.077 -0.332 0.161 -0.003 -0.001 0.022 -0.001
(0.128) (0.616) (0.240) (0.647) (0.004) (0.014) (0.142) (0.301)
Net interest margin*variable -0.868 0.036 0.083 0.309 0.001 -0.022 -0.224** -0.401*
(0.576) (1.128) (0.180) (1.058) (0.005) (0.015) (0.112) (0.218)
Loan-loss reserve ratio*variable 0.346 -0.077 1.585* 2.858** 0.043*** 0.027 -0.613* -0.259
(0.347) (0.827) (0.817) (1.291) (0.015) (0.022) (0.344) (0.568)
Credit to GDP ratio*variable 0.190 -1.283 0.063 0.169 -0.004 -0.012 -0.257 0.199
(0.119) (0.880) (0.064) (0.282) (0.012) (0.009) (0.285) (0.297)
Observations 321 321 321 321 321 321 321 321 321 321
R-squared 0.55 0.55 0.57 0.58 0.57
No. of groups 96 96 96 96 96
No. of instruments . 64 54 54 64 54
Hansen test p-value 0.348 0.192 0.132 0.232 0.249
A-B AR(2) test 1.283 1.027 1.040 1.562 1.283
A-B AR(2) test p-value 0.199 0.305 0.298 0.118 0.200
Source: Authors' calculations.
Dependent variable is the ratio of total liquid assets to customer deposits and short-term funding. GMM is two-step system GMM estimator with Windmeijer standard error correction.
Columns (3) through (10 test the hypotheses that ownership (foreign/domestic and public/private), degree of dollarization and coverage of the financial safety net affect banks' liquidity buffers
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the country level. Safety net
is an index variable categorizing the existence of each of the 4 components of a safety net in each country (Appendix II. Table 1).
All regressions include time and country dummies. Constant estimated but not reported.
Table 4a. Determinants of Banks' Liquidity Buffers in CAPDR-Financial Depth
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Dependent variable is the ratio of
total liquid assets to customer
deposits and short-term funding
Variable = Private ownership Foreign ownership Dollarization Baseline Safety net
24
2
4
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM
Liquid assets ratio (-1) 0.346*** 0.180*** 0.355*** 0.350** 0.377*** 0.298** 0.342*** 0.280** 0.346*** 0.267**
(0.064) (0.047) (0.065) (0.134) (0.069) (0.148) (0.067) (0.114) (0.064) (0.107)
Bank size 7.441*** 8.027*** 7.301*** 5.349** 7.142*** 7.070** 8.318*** 6.768*** 8.287*** 7.904***
(1.486) (2.059) (1.536) (2.099) (1.539) (2.752) (1.550) (2.548) (1.381) (1.661)
Bank size squared -0.352*** -0.384*** -0.348*** -0.238** -0.343*** -0.325** -0.395*** -0.306** -0.389*** -0.354***
(0.065) (0.096) (0.068) (0.108) (0.068) (0.137) (0.069) (0.127) (0.061) (0.084)
Capitalization -0.356*** -0.336*** -0.316*** -0.423 -0.283*** -0.180 -0.217 -0.026 -0.390 -0.461
(0.061) (0.106) (0.050) (0.424) (0.061) (0.393) (0.220) (0.537) (0.300) (0.956)
Net interest margin -0.160** -0.141* 1.115** 1.945 -0.235 -0.546 -0.191 0.489 0.573 1.293*
(0.063) (0.083) (0.552) (1.851) (0.158) (0.968) (0.132) (0.317) (0.349) (0.678)
Loan-loss reserve ratio -0.211 -0.299 -0.292 0.415 -0.351** -0.252 -1.292*** -1.236 1.879* 1.283
(0.159) (0.300) (0.178) (0.639) (0.153) (0.526) (0.404) (1.033) (1.054) (2.196)
Credit Institutions 1/ 0.051 0.053* 0.113 0.174 0.047 0.001 0.188** 0.153* 0.685*** 0.788**
(0.035) (0.028) (0.073) (0.161) (0.035) (0.051) (0.089) (0.086) (0.209) (0.355)
Variable 11.918* 27.989 0.579 6.271 0.774** 0.718** 17.220*** 33.601*
(6.497) (28.631) (4.080) (18.141) (0.304) (0.303) (2.981) (19.225)
Capitalization*variable -0.109 -0.054 -0.367 -0.034 -0.003 -0.005 0.033 0.120
(0.129) (0.849) (0.236) (0.716) (0.004) (0.013) (0.141) (0.320)
Net interest margin*variable -1.269** -1.987 0.032 0.205 0.000 -0.025* -0.225** -0.419*
(0.558) (1.862) (0.173) (0.927) (0.005) (0.015) (0.109) (0.211)
Loan-loss reserve ratio*variable 0.132 -0.407 1.296* 1.910 0.047*** 0.050 -0.640* -0.419
(0.298) (0.764) (0.682) (1.466) (0.015) (0.038) (0.332) (0.609)
Credit Institutions*variable 1/ -0.061 -0.148 0.018 -0.074 -0.007** -0.006** -0.186*** -0.212**
(0.066) (0.156) (0.039) (0.118) (0.003) (0.002) (0.061) (0.087)
Observations 321 321 321 321 321 321 321 321 321 321
R-squared 0.55 0.55 0.57 0.59 0.57
No. of groups 96 96 96 96 96
No. of instruments . 64 54 54 64 54
Hansen test p-value 0.338 0.0836 0.134 0.125 0.290
A-B AR(2) test 1.470 1.481 0.914 1.115 1.264
A-B AR(2) test p-value 0.141 0.139 0.360 0.265 0.206
Source: Authors' calculations.
Dependent variable is the ratio of total liquid assets to customer deposits and short-term funding. GMM is two-step system GMM estimator with Windmeijer standard error correction.
Columns (3) through (10 test the hypotheses that ownership (foreign/domestic and public/private), degree of dollarization and coverage of the financial safety net affect banks' liquidity buffers
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the country level. Safety net
is an index variable categorizing the existence of each of the 4 components of a safety net in each country (Appendix II. Table 1).
All regressions include time and country dummies. Constant estimated but not reported.
1/ Credit institutions are proxied by the principal components analysis of the Doing Business variables measuring the quality of credit institutions.
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Table 4b. Determinants of Banks' Liquidity Buffers in CAPDR-Credit Institutions
Dependent variable is the ratio of
total liquid assets to customer
deposits and short-term funding
Baseline Variable = Private ownership Foreign ownership Dollarization Safety net
25
less liquidity. The interaction between the safety net and net interest margin is negative,
indicating as mentioned above that profitable banks in countries with more comprehensive
safety nets do hold less liquidity. This is in contrast with the coefficient on the interest
margin (without interactions), which is significant and positive in column 10 of Table 4a.
Compared with the baseline, it seems that the absence of a safety net would lead even
profitable banks to hold an extra liquidity cushion. Table 4b, columns 9-10 show estimation
results with the financial safety net variable and interactions, including the quality of credit
institutions variable. As in the case of deposit dollarization, whereas the sign of the
coefficients for credit institutions and safety net variables are now positive and significant,
the coefficient on the interaction term between safety net and credit institutions is negative:
when the safety net is less comprehensive and the quality of institutions is low, banks would
hold more precautionary liquidity buffers. Further, as in Table 4a, if the safety net is
comprehensive, more profitable banks or banks with higher loan loss reserves tend to hold
less liquidity.
Robustness checks
As a main robustness check, we estimate our model using the ratio of liquid assets to total
assets as our dependent variable (Appendix IV. Table 1). Overall results are broadly
consistent with the findings presented in Tables 4a and 4b, at least in terms of signs of
coefficients. The coefficient on the lagged dependent variable is about twice as large, and the
coefficient of the dollarization variable remains significant and relatively large.
Appendix IV, Table 2 presents further robustness checks. These include running the fixed
effects regressions as discussed in footnote 16 above (column 3), and looking into the
interactions of foreign ownership, dollarization and safety net only for the private banks of
the sample (columns 4-9). One caveat is that limiting the number of observations increases
the risk of overfitting the model due to too many instruments. Nonetheless, the Hansen
statistic’s p-value remains reasonable for all specifications.
These additional regressions support our main findings. The relative size of the coefficient on
the lagged dependent variable in the pooled OLS, fixed effects and GMM is consistent with
expectations: in OLS this coefficient is correlated with the error term and biased upward,
while in the fixed effects specification it is the opposite. Good estimates of the true parameter
should lie in between or near these values, which is the case here (see column 2 of Appendix
IV. Table 2). Previous results on ownership, dollarization and safety net hold in the private
banks sample, particularly as regards the role of credit institutions in highly dollarized
economies or economies with a less comprehensive safety net.
Our results still need to be considered against the caveat of data limitations. The uneven
coverage of individual countries’ banking systems and short estimation time frame may
affect the coefficient estimates from the regressions. Nevertheless, some useful policy
lessons already emerge from our analysis. These are discussed in the following section.
26
V. CONCLUSIONS AND POLICY LESSONS
Our study of liquidity buffers in CAPDR finds that they are comfortably above legal and
prudential requirements. With average liquidity ratios of about 25 percent of deposits, banks
in the region have handled and are able to handle historic deposit volatility levels outside of
crisis episodes (Table 5). Therefore, the adoption of the new Basel III liquidity requirements
in the region should not have much impact on banks’ balance sheet.
23
Indeed, Basso, Delgado
and Meza (2012) find that liquidity holdings in the region already meet or exceed the new
Basel III ratios.
A closer look at the reasons for which banks would want to hold liquidity buffers above legal
or prudential requirements indicates that CAPDR banks appear guided at least in part by
rational precautionary motives. As found in other countries or regions, bank characteristics
that influence their ability to raise additional funding on demand play an important role:
smaller, lower-capitalized, less efficient and less profitable banks tend to hold higher
liquidity buffers. Foreign banks tend to hold less liquidity, possibly because they have access
to emergency lines from headquarters. Surprisingly, banks with riskier loan portfolios also
hold less liquidity overall, though this is not the case for foreign banks and banks in highly
dollarized economies.
A first policy lesson stemming from these results would be to continue with ongoing efforts
to strengthen financial sector supervision, enhance financial safety nets and develop financial
markets. Greater confidence in the system and more opportunities for investment and
23
The Basel III liquidity requirements are to be in effect in 2015. The Basel Committee on Banking Supervision
has defined two minimum standards for funding liquidity: (i) the liquidity coverage ratio, which aims at
promoting short-term resilience of a bank’s liquidity profile by ensuring that it has sufficient high-quality liquid
assets (cash or cash-equivalent) to survive a significant stress scenario lasting for one month; and (ii) the net
stable funding ratio, in order to promote resilience over a longer time period by matching long-term assets with
stable funding sources over a one-year horizon (BIS, 2010).
1 3 6 1 3 6 1 3 6 1 3 6 1 3 6
Costa Rica 3.3 4.9 4.7 5.6 6.8 5.6 7.4 9.9 12.1 6.7 9.6 15.2 3.8 5.4 5.2
Dominican Republic 34.4 23.7 13.5 54.5 40.1 27.6 30.0 20.0 10.6 19.9 30.6 35.0 45.5 31.7 22.5
El Salvador 2.7 3.0 3.9 17.8 17.4 17.8 2.2 3.2 4.1 2.7 3.0 3.9
Guatemala 3.1 1.8 0.6 6.2 6.7 5.6 5.3 5.2 3.1 5.6 4.1 1.1 3.5 3.0 1.6
Honduras 3.0 3.8 3.3 10.3 8.8 10.2 2.3 3.2 3.7 7.4 9.4 10.0 0.3 5.3 5.4
Nicaragua 1.7 4.2 3.0 10.8 11.2 11.4 2.9 4.9 3.1 3.4 5.1 6.1 6.6 4.2 4.5
Panama 3.7 6.3 7.7 8.9 8.1 8.7 3.1 2.0 2.1 3.7 6.3 7.7
Total 1.7 1.8 0.6 5.6 6.7 5.6 2.2 2.0 2.1 3.4 4.1 1.1 0.3 3.0 1.6
Source: CAPDR Central Banks and superintendencies’ websites and Fund staff calculations.
Notes: Largest decline in bank deposits at the one, three and six months horizon, between December 2001 and September 2011.
Foreign Currency Local Currency
Table 5. CAPDR: Largest deposit drop (December 2001-September 2011)
(Percent)
Deposits
Total Demand Time & Savings
no. of months no. of months no. of months no. of months no. of months
27
intermediation (through stronger credit institutions) could help lower banks’ precautionary
liquidity buffers without compromising financial stability.
Strengthened supervision would help address the issue of the negative relationship between
the loan-loss ratio and liquidity buffers, which may indicate that domestic banks, in contrast
with foreign banks which are likely subject to strict internal guidelines, may not fully
internalize the costs of riskier lending practices. As mentioned, further progress in risk-based
supervision would be especially warranted: in spite of notable progress, CAPDR countries
still do not meet minimum international standards and lag behind larger South American
countries.
Another important lesson relates to the dollarization of CAPDR economies and banking
systems and calls for strengthening the credibility of macroeconomic policy and institutions,
as well as the coverage of financial safety nets. Our findings show that, in our sample, banks’
precautionary demand for liquidity is associated to the degree of deposit dollarization, and
the safety net, in each country. Given the lack of dollar LOLR in all countries and the lack of
a LOLR in the two fully dollarized economies, our findings suggest that continuing with
ongoing efforts to strengthen financial safety nets would be efficient. El Salvador has
approved legislation to provide emergency liquidity support to banks, and Panama is
considering establishing a similar facility.
Furthermore, maintaining higher liquidity buffers because of dollarization also has negative
implications for the development of financial markets, and for the adequate functioning of
the monetary policy transmission mechanism. For the countries in the region that aim at
transitioning to inflation targeting, tackling the root causes of deposit dollarization should be
an important part of their strategy. Studies on deposit dollarization find that it is associated
with lower credibility of macroeconomic policy, high inflation, and poor quality of
institutions; and also find that financial instability is higher in dollarized economies (De
Nicoló, Honohan and Ize, 2005, Gulde et al., 2004). Arguably, most CAPDR countries have
been successful over the past decade at macroeconomic stabilization, lowering inflation and
reducing output volatility (except during the 2008–09 global crisis). Nonetheless, financial
dollarization tends to persist even when macroeconomic stability is restored and inflation
relatively low if institutions are still perceived to be weak.
24
Even in tightly managed
exchange rate regimes, partially dollarized economies are subject to higher liquidity risks,
and our results tend to indicate that, at least to some extent, CAPDR banks do internalize
these risks.
With causality likely running both from policies to dollarization and back, measures that
would help create a “virtuous cycle” of de-dollarization and lower precautionary liquidity
holdings could be informed by the experience of de-dollarization in South America. In
particular, in a study of financial de-dollarization in Bolivia, Paraguay, Peru and Uruguay,
24
See Reinhart, Rogoff and Savastano (2003), Galindo and Leiderman (2005), and Erasmus et al. (2009).
28
Garcia-Escribano and Sosa (2011) find that successful, market-driven de-dollarization was
associated with (i) stronger macroeconomic policies and institutions, credibly and
consistently implemented over time, (ii) active management of reserve requirement
differentials and the introduction of other prudential measures; and (iii) domestic currency
capital market development. As discussed in this paper, there is ample room for more active
liquidity management on the part of the CAPDR monetary and prudential authorities. Finally,
measures to develop local currency capital markets, starting with public domestic debt
markets, would enhance financial systems’ efficiency, diversify sources of funding and
investment opportunities.
29
REFERENCES
Agénor, Pierre-Richard, Joshua Aizenman, and Alexander W. Hoffmaister, 2004, “The Credit
Crunch in East Asia: What Can Bank Excess Liquid Assets Tell Us?” Journal of International
Money and Finance, Vol. 23, pp. 27–49.
Almeida, Heitor, Murrillo Campello, and Michael S. Weisbach, 2004, “The Cash Flow Sensitivity of
Cash,” The Journal of Finance, Vol. 59, No. 4, pp. 1777–1804.
Arellano, Manuel, and Stephen Bond, 1991, “Some Tests of Specification for Panel Data: Monte
Carlo Evidence and an Application to Employment Equations”, Review of Economic Studies
Vol. 58, pp. 277–97.
Aspachs, Oriol, Erlend Nier, and Muriel Tiesset, 2005, “Liquidity, Banking Regulation and the
Macroeconomy,” Mimeo (London: London School of Economics).
Baltensperger, Ernst (1980). “Alternative Approaches to the Theory of the Banking Firm,”Journal of
Monetary Economics, Vol. 6, pp.1–37.
Barajas, Adolfo, Ralph Chami, Raphael Espinoza, and Heiko Hesse, 2010, “Recent Credit stagnation
in the MENA Region: What to Expect? What Can be done?”, IMF Working Paper 10/219
(Washington: International Monetary Fund).
Basel III: International Framework for Liquidity Risk Measurement, Standards and Monitory, 2010,
(Switzerland: Bank for International Settlements).
Basso, Oscar, Fernando L. Delgado and Mynor Meza, 2012, “Strengthening Bank Capital and
Liquidity in Central America: the Road to Basel III”, mimeo.
Bolzico, Javier, Eugenia Gozzi and Florencia Rossini, 2010, “Redes de Seguridad
Financiera,” Fit & Proper, Available via the Internet: http://www.fitproper.com/
Cárdenas, Mauricio and Camila Henao, 2010, “Latin America and the Caribbean Economic
Recovery”, The Brookings Institution, Available via the Internet:
http://www.brookings.edu/~/media/research/files/articles/2010/7/latin%20america%20econo
my%20cardenas/07_latin_america_economy_cardenas
Clarke, George R. G., Robert Cull, and María Soledad Martínez Pería, 2006, “Foreign Bank
Participation and Access to Credit Across Firms in Developing Countries,” Journal of
Comparative Economics, Vol. 34, pp. 774–95.
Delgado, Fernando and Mynor Meza, 2011, “Developments in Financial Supervision and the Use of
Macroprudential Measures in Central America”, IMF Working Paper 11/299 (Washington:
International Monetary Fund).
30
Detragiache, Enrica, Thierry Tressel, and Poonam Gupta, 2008, “Foreign Banks in Poor Countries:
Theory and Evidence,” The Journal of Finance, Vol. 63, No5 (October), pp. 2123–60.
Diamond, Douglas W., and Raghuram G. Rajan, 1999, “Liquidity Risk, Liquidity Creation and
Financial Fragility: A theory of Banking,” NBER Working Paper 7430 (Cambridge: National
Bureau of Economic Research)
_______, and Raghuram G. Rajan, 2005, “Liquidity Shortages and Banking Crises,” Journal of
Finance Vol. 30, No. 2, pp. 615–47.
Diamond, Douglas W., and Philip H. Dybvig, 1983, “Bank Runs, Deposit Insurance, and
Liquidity,” Journal of Political Economy, Vol. 91, No. 3, pp. 401–19.
De Nicoló, Gianni, Patrick Honohan, and Alain Ize, 2004, “Dollarization of Bank Deposits: Causes
and Consequences,” Journal of Banking & Finance, Vol. 29, pp. 1697–1727.
Dinger, Valeriya, 2009, “Do Foreign-owned Banks Affect Banking System Liquidity Risk?” Journal
of Comparative Economics, Vol. 37, pp. 647–57.
Douglas W. Diamond & Raghuram G. Rajan, 2001, “Liquidity Risk, Liquidity Creation, and
Financial Fragility: A Theory of Banking,” Journal of Political Economy, University of
Chicago Press, Vol. 109, No. 2, pp. 287–327.
Freixas, Xavier, and Cornelia Holthasusen, 2004, “Interbank Market Integration under Asymmetric
Information,” The Review of Financial Studies”,Vol. 18, No. 2. pp. 459–90.
______, and Leonardo Leiderman, 2005, “Living With Dollarization and the Route to
Dedollarization,” IDB Working Paper 526 (Washington: Inter-American Development
Bank).
Galindo, Arturo J., Alejandro Izquierdo, and Liliana Rojas-Suárez, 2010, “Financial Integration and
Foreign Banks in Latin America,” IDB Working Paper 116 (Washington: Inter-American
Development Bank)
García-Escribano, Mercedes, and Sebastián Sosa, 2011, “What is Driving Financial De- dollarization
in Latin America?” IMF Working Paper 11/10 (Washington: International Monetary Fund).
Gelos, R. Gaston, and Jorge Roldós, 2004, “Consolidation and Market Structure in Emerging Market
Banking Systems,” Emerging Markets Review, Elsevier, Vol. 5, No. 1 (March), pp. 39–
59.
Gonzalez-Eiras, Martín, 2003, “Banks’ Liquidity Demand in the Presence of a Lender of Last
Resort,” Documento de Trabajo 61 (Buenos Aires: Universidad de San Andrés).
31
Gray, Simon, 2011, “Central Bank Balances and Reserve Requirements,” IMF Working Paper 11/36
(Washington: International Monetary Fund).
Guerrero, Rosa Matilde, Kurt Focke, and Florencia Rossini, 2010, “Redes de Seguridad Financiera,”
IDB Working Paper No. 121 (Washington: Inter-American Development Bank).
Gulde, Anne-Marie, David Hoelscher, Alain Ize, David Marston, and Gianni De Nicoló, 2004,
“Financial Stability in Dollarized Economies,” IMF Occasional Paper 230 (Washington:
International Monetary Fund).
Holmström, Bengt, and Jean Tirole, 1998, “Private and Public Supply of Liquidity,” Journal of
Political Economy Vol. 106, No. 1, pp. 1–40.
Ize, Alan, Miguel A. Kiguel, and Eduardo Levy Yeyati, 2005, “Managing Systemic Liquidity Risk
in Financially Dollarized Economies,” IMF Working Paper 05/188 (Washington:
International Monetary Fund).
Kashyap, Anil K., and Jeremy C. Stein, 1995, “The Impact of Monetary Policy on Bank Balance
Sheets,” Carnegie-Rochester Conference Series on Public Policy, Vol. 42, pp. 151–95.
______, Raghuram Rajan, and Jeremy C. Stein, 2002, “Banks and Liquidity Providers: An
Explanation for the Coexistence of Lending and Deposit-Taking,” The Journal of Finance
Vol. 57, No. 1 (February), pp. 33–73.
Kiyotaki, Nobuhiro and John Moore, 2008, “Liquidity, Business Cycles and Monetary Policy,”
Technical Paper for Clarendon Lecture 2.
Laeven, Luc, and Fabián Valencia, 2012, “Systemic Banking Crises Database: An Update,” IMF
Working Paper 12/163 (Washington: International Monetary Fund).
Levy-Yeyati, Eduardo, Christian Broda, 2002, “Dollarization and Lender of Last Resort,”
(Massachussets Institute of Technology; Universidad Torcuato De Tella).
Medina Cas, Stephanie, Alejandro Carrión-Menéndez, and Florencia Frantischek, 2011, “The Policy
Interest-Rate Pass-Through in Central America,” IMF Working Paper 11/240 (Washington:
International Monetary Fund).
______, Alejandro Carrión-Menéndez, and Florencia Frantischek, 2011, “Improving the Monetary
Policy Frameworks in Central America,” IMF Working Paper 11/245 (Washington:
International Monetary Fund).
Micco, Alejandro, Ugo Panizza and Mónica Yañez, 2004, “Bank Ownership and Performance,”
IDB Working Paper No. 518 (Washington: Inter-American Development Bank).
32
_____, and Ugo Panizza, 2006, “Bank Ownership and Lending Behavior,” Economic Letters,
Elsevier, Vol. 93, No. 2 (November), pp. 248–54.
Opler, Tim, Lee Pinkowitz, René Stulz, and Rohan Williamson, “The Determinants and Implications
of Corporate Cash Holdings,” RBER Working Paper 6234 (Cambridge National Bureau of
Economic Research).
Reinhart, Carmen M., Kenneth S. Rogoff, and Miguel, 2003, “Addicted to Dollars,” NBER Working
Paper 10015 (Cambridge: National Bureau of Economic Research).
Repullo, Rafael, 2005, “Liquidity, Risk-Taking and The Lender of Last Resort,” CEMFI Working
Paper No. 0504 (Madrid: Centro de Estudios Monetarios y Financierios).
Rochet, Jean-Charles, and Xavier Vives, 2004, “Coordination Failures and the Lender of Last
Resort: Was Bagehot Right After All?” Journal of the European Economic Association Vol.2
No. 6, pp. 1116–47.
Roodman, David, 2009, “Practitioner’s Corner: A Note on the Theme of Too Many Instruments”,
Oxford Bulletin of Economics and Statistics, Vol. 71, No.1, pp. 135-58.
____, 2006, “How to Do xtabond2: An Introduction to ‘Difference’ and ‘System’ GMM in Stata,”
CGD Working Paper No. 103 (Washington: Center for Global Development).
Santomero, Anthony M., 1984, “Modeling the Banking Firm”, Journal of Money, Credit, and
Banking, Part 2, Vol. 16, pp. 576–602.
Saxegaard, Magnus, 2006, “Excess Liquidity and Effectiveness of Monetary Policy: Evidence from
Sub-Saharan Africa,” IMF Working Paper 06/115 (Washington International Monetary
Fund).
Shah, Hemant, Andreas A. Jobst, Laura Valderrama-Ferrando, Ivan Guerra, 2007, “Public Debt
Markets in Central America, Panama, and the Dominican Republic,” IMF Working Paper
07/147 (Washington: International Monetary Fund).
____, Ana Carvajal, Geoffrey Bannister, Jorge Cha-Lau, and Ivan Guerra, 2007, “Equity and Private
Debt Markets in Central America, Panama, and the Dominican Republic,” IMF Working
Paper 07/288 (Washington: International Monetary Fund).
33
APPENDIX I. RESERVE AND LIQUIDITY REQUIREMENTS IN CAPDR
Most central banks or supervisory authorities require depository institutions to hold
minimum reserves or liquidity against their deposits.
Required reserves
1
In systems with central banks, there are typically two main reasons to require the holdings of
reserves:
? Prudential: Reserve requirements serve as a safeguard against both liquidity and
solvency risk in a case of a sudden and inordinate demand for withdrawals (as in a
run on a bank) or as a result of a need to make sudden large payments abroad, for
example. In highly dollarized economies, the prudential purpose of the reserve
requirements perhaps carries more prominence as central banks have a very limited
ability to control monetary policy in the environment of an overwhelming presence of
foreign currency.
? Monetary control and liquidity management: Adjusting reserve requirements is a
policy tool for injecting cash (liquidity) into, or withdrawing it from, an economy.
This helps to control domestic liquidity, i.e., commercial bank balance sheet growth.
Likewise, adjusting reserve requirements could influence the spread between deposit
and lending rates, thereby impacting the growth of monetary aggregates and inflation.
Reserve requirements are calculated in reference to commercial banks’ deposits.
2
They can
be mandated on both local and foreign currency deposit liabilities at similar or differentiated
rates, and on demand, savings and time horizons of these deposits.
3
The reserve requirement
is usually defined as a minimum amount of liquid assets, cash or cash-equivalents
4
(computed as a percentage of demand and time deposits), that banks and other depository
institutions (credit unions, insurance companies) are required by law to keep on hand, and
which may not be used for lending or investing. In countries with underdeveloped financial
1
This section is based on Gray (2011).
2
Composition of required reserves depends on individual country requirements. Some countries only allow
deposits with the central bank to count as legal reserves and do not allow averaging (calculating the level of
deposits on average in one month versus a requirement to hold going forward the end of the month balance,
which limits banks ability to manage liquidity).
3
In cases where the authorities are trying to discourage dollarization, higher reserve requirements may be
imposed on foreign liabilities.
4
Liquidity is defined as the ability of an asset, other than cash, to be converted into cash quickly and without
any price discount. In the U.S., e.g., these constitute certificates of deposit (CD), marketable securities,
negotiable financial instruments (such as a cashier's checks), etc., that has a very high degree of convertibility
into cash (liquidity).
34
markets, reserve requirements are usually cash in vault, banks’ deposits with the central bank
or abroad.
In many cases, banks hold liquidity above the required levels at the central bank, especially if
they are remunerated.
5
This liquidity could be truly voluntary, implying that banks’ hold
them for precautionary reasons, and involuntary, reflecting idle balances that banks hold
above their demanded levels.
6
In countries with monetary policies, this genuine voluntary
liquidity in the banking system may undermine both the central bank’s capacity to affect
banks’ cost of funding and its ability to restrain any rapid unwinding of these reserves, and
possible triggering of inflation. This may also induce excessive exchange rate volatility under
an abrupt change in expectations, and ultimately jeopardize financial stability. The presence
of voluntary reserves could also be an indication of rigidities in the system as excess liquidity
creates opportunity costs.
Given the banks’ ability to influence monetary conditions through holdings of excess
reserves, the monetary authorities ought to carefully monitor these levels. For example, they
should ensure that excess liquidity, if needed, could be counteracted by their own provision
of reserves to the system. In cases where reserves are remunerated, changing the
remuneration rate – rather than conducting open market operations - can signal the stance of
the desired monetary policy.
Liquidity requirements
Liquidity requirements are commonly defined as the ratio of liquid assets to liabilities,
although definitions can vary by country. Liquid assets generally include cash in vault,
deposits abroad, and deposits held at the central bank, as well as liquid securities (central
bank paper, quoted government securities). The definition of liquidity requirements differs
sometimes quite substantially across countries, as supervisor include eligible securities based
on the existence of a well-functioning market guaranteeing that they could be converted into
cash quickly and at no or little loss of value.
Application in CAPDR
Reserve and liquidity requirements are relatively homogeneously defined in CAPDR,
although there could be differences in definitions of liquid assets. Reserve requirements in
five countries consist mostly of cash, cash in vault, and central bank and government
5
Level of (voluntary) reserves could be higher if they are remunerated by the central bank. Remuneration of
reduces the cost of having idle resources and reduces incentives for the financial system to avoid liabilities that
require reserve provision.. At the same time, out of 121 central banks surveyed by Gray (2011), 70 percent did
not remunerate reserves; and 20 percent remunerated at below primary rate.
6
Agenor et al (2004).
35
securities. In El Salvador, deposits abroad are also permitted as a proportion of a reserve
requirement.
7
Eligible liabilities on which requirement is calculated are mostly all deposits, although there
are some variations. Some countries limit eligible liabilities to demand and foreign deposits
only (Dominican Republic and Honduras). Many countries exclude savings deposits mostly
because they are of longer maturity and also because they are not large comparing to demand
and time deposits. Reserve requirements in partially dollarized economies tend to be different
for local or foreign currency deposits. In CAPDR these average 15 and 15.5 percent,
respectively. Dominican Republic has the highest reserve requirement on foreign deposits, at
20 percent (Appendix 1. Table 1).
Reserve requirements are stipulated in central bank laws and form a part of a monetary
policy toolkit. Costa Rica and Guatemala have kept ratios constant over the past 5 years,
indicating that the requirement serves mostly a prudential objective rather than a monetary
policy tool (Appendix I.Table 2). In Nicaragua a downward adjustment in reserve
requirement was taken years before the 2008 crisis and the ratio was kept stable during the
crisis. In Dominican Republic there was a downward adjustment in local currency reserve
requirement in 2009, to inject liquidity. Honduras appears to be the only country where there
was an active management of reserves. Only in Nicaragua, total reserves at the central bank
continuously exceed the required amounts.
El Salvador, Honduras and Panama also impose prudential liquidity requirements. In
Honduras, the liquidity requirement is aimed at avoiding maturity mismatches. In El
Salvador, there is a prudential requirement of maintaining at the central bank a liquid reserve
amounting to 3 percent of deposits. For Panama, the liquidity requirement (30 percent)
includes a vast array of instruments as liquid assets, even obligations payable to banks in 186
days, but the numerator only includes short-term deposits, which raises the ratio
significantly, as these constitute only a small share of total deposits.
7
Perhaps, as a relic of the non-officially dollarized economy, El Salvador preserved a liquid asset requirement
of 3 percent of deposits (funds to be deposited with the former central bank or with a reputable bank abroad),
however, it eliminated it during the 2008 global financial crisis in an attempt to free up liquidity.
3
6
Reserve requirements
Costa Rica 15 15 n.a Demand, foreign
currency, time,
interbank,
government.
Excludes
interbank
deposits.
Deposits at the central
bank (only those at the
reserve account) in the
some currency than the
deposits
15 days
maintenance
period
Interest rate of
discount
window over the
reserve
deficiency
Monetary
policy
Guatemala 14.6 14.6 0.6 n.a All deposits Cash in vault and deposits
at the central bank in the
some currency than the
deposits
Monthly n.a Monetary
policy
n.a
El Salvador 23 n.a n.a 3 All deposits 25 % in demand deposits
at CB or foreign bank, 25%
in deposits or CB
securities, 50% in CB
securities issued for
liquidity purposes
Over a two-
week period
n.a. Prudential
Panama n.a n.a n.a n.a Prudential
1/ Honduras also imposes specif ic liquidity requirements, based on temporal bands f or maturity mismatches. For the f irst band, the maturity mismatch in cash f lows f or the next month must be lower than
the amount of liquid assets, while the f or the second band the maturity mismatch in cash f lows f or the next three months must be lower than 1.5 times the liquid assets.
Sources: CAPDR Central banks and Superintendencies websites.
Cash of CB securities n.a Interest
charged based
on internbank
int rate (greater
than 1%)
Monetary
policy/2005-06
n.a 30; 20 (Applies to
all onshore general
license banks and
state-owned Banks
at 30.0 for general
licence banks;
20.0 for general
licence banks that
maintain average
interbank quarterly
deposits
exceeding 80
percent of total
deposits.
Demand, term
deposits up to 186
days (unless
portion that
guarantees loans
in the bank itself),
savings deposits.
Deposits received
from the parent
company, branch,
subsidiary or
affiliated abroad
are excluded.
Legal tender in Panama,
bank deposits in Panama,
bank deposits abroad,
obligations issued by
foreign governments,
obligations issued by
foreign private and
government agencies,
banking obligations
payable in Panama up to
186 days, installments of
payable obligations up to
186 days, other Liquid
Assets.
Nicaragua 16.25 16.25 on excess
reserves (n.a.)
All deposits
Demand, foreign
currency, time,
interbank,
government.
Excludes
interbank
deposits.
18 % in deposits with
central bank and 2% cash
in vault allowed.
Weekly,
holding
period ends
on Friday
n.a Monetary
Policy/2009
Deposits, expired
term deposits,
reduced capital
contracts and
savings stamps
and others
Cash in vault, deposits at
the central bank, and
government bonds in the
case of compulsory
investment in local
currency in the some
currency than the deposits
Over a two-
week period
Penalties
depend on the
currency of
denomination
and the type of
institution.
Monetary
policy/ 2008-09
Dominican
Republic
17 20 Foreign
currency
reserves are
remunerated at
Feds overnight
rate - 200bps
n.a
Yes; 2 1/
Appendix I. Table 1: Legal Reserve and Liquidity Requirements in CAPDR (2010)
Country Remuneration
(percent)
Liquidity
requirement
(percent)
Eligible Liabilities Compliance Assets Averaging Penalty Purpose/ Last
change
Local
Currency
(percent)
Foreign
Currency
(percent)
Honduras 6 (unrem),
12
(remun)
12(unrem),
10(remun)
Only
compulsory
investments are
remunerated at
½ of the policy
rate.
3
7
Country Currency 2005 2006 2007 2008 2009 2010
Costa Rica LCU RR 15 15 15 15 15 15
FCY RR 15 15 15 15 15 15
Guatemala LCU RR 14.6 14.6 14.6 14.6 14.6 14.6
FCY RR 14.6 14.6 14.6 14.6 14.6 14.6
Dominican Rep. LCU RR 20 20 20 20 17.5 (May -Dec) 17
FCY RR
20 20 20 20 20 20
Nicaragua LCU RR 16.25 (Jan-
May); 19.25
19.25 (Jan-Sep);
16.25 (Oct-Dec)
16.25 16.25 16.25 16.25
FCY RR 16.25 (Jan-
May); 19.25
(June-Dec)
19.25 (Jan-Sep);
16.25 (Oct-Dec)
16.25 16.25 16.25 16.25
Honduras LCU RR unremun. 12 12 12 12(Jan-Feb); 10.2
(Mar); 10.1 (Apr);
9.4 (May); 8.8
(June-Sept); 7.5
(Oc-Nov); 4.3
(Dec)
2.6 (Jan); 1.1 (Feb);
1.2 (May-July); 6.1
(Aug)
6
LCU RR remun. 2 (Jan); 1 (Oct) 1 3 (Jan-Jun); 4
(Jul-Dec)
4 (Jan-Feb); 6
(Mar); 7 (Apr(; 8
(May); 8.8 (June-
Dec)
8.8 (Jan.-Jul); 11 (Aug);
12 (Sep-Dec)
12
FCY RR unrem. 12 12 12 12 12 12
FCY RR remun.
2 (Apr); 4 (May); 6 (Jun-
Jul); 8 (Aug); 81. (Sep);
10 (Oct-Dec)
10
FCY LR 2
Panama LR 30; 20 30; 20 30; 20 30; 20 30; 20 30; 20
El Salvador RR 19.2 (Jan), 19.6
(Feb); 19.8
(Mar-Apr); 19.7
(May); 19.3
(Jun-Jul); 19.7
(Aug);19.4
(Sep);19.6
(Oct); 20.1
(Nov);20.3
(Dec)
20.2 (Jan-Feb);
20.3 (Mar);19.9
(Apr--Jun);19.8
(Jul);20.6
(Aug);19.9
(Sep);20.4
(Oct);20.2
(Nov);20.5 (Dec)
20.4(Jan-
Feb);20.6
(Mar);20.7
(Apr);20.8
(May);20.5
(Jun);21.2
(Jul);21.8
(Aug);21.4
(Sep);21.0
(Oct);21.2
20.1 (Jan); 20.8
(Feb);20.4
(Mar);20.6
(Apr);20.2
(May);20.0
(june)20.1
(Jul);20.2
(Aug);20.0
(Sep);20.1(Oct);20
.3 (Nov);20.4 (Dec)
20.9(Jan);20.7
(Feb);21.9(Mar);22.7(Ap
r);23.4(May);23.5(Jun);2
3.9 (Jul);24.3 (Aug);24.1
(Sep);23.3 (Oct);22.8
(Nov);23.7(Dec)
23.7(Jan);23.4(Feb);23.1(Mar
);23.0(Apr);22.6
(May);22.7(June);23.1 (July)
; 23.5(Aug);23.2 (Sep);23.2
(Oct);23.3 (Nov);23.1 (Dec)
LR 3.0 3.0 3.0
Source: CAPDR Central banks and Superintendencies Websites.
Notes: RR is reserve requirement, LR liquidity requirement. LCU is local currency, FCY foreign currency.
Appendix I. Table 2: Evolution of Reserve and Liquidity Ratios in CAPDR (2005-2010)
(In percent)
3
8
Liquid assets to
customer
deposits and
short term
funding
Liquid assets to
total assets
Total credit to
GDP
Getting Credit -
Strength of legal
rights index (0-
10)
Getting Credit -
Depth of credit
information
index (0-6)
Getting Credit -
Private bureau
coverage (% of
adults)
Enforcing
Contracts - Cost
(% of claim)
Institutional
Credit index (0-
100)
Liquid assets to customer
deposits and short term funding
1
Liquid assets to total assets 0.7598 1
(0)
Total credit to GDP 0.1565 0.19 1
-0.0009 -0.0001
Getting Credit - Strength of legal
rights index (0-10) 0.0577 0.0101 0.3555 1
(0.2232) (0.8318) (0)
Getting Credit - Depth of credit
information index (0-6) 0.1085 0.157 0.333 0.2571 1
(0.0216) (0.0009) (0) (0)
Getting Credit - Private bureau
coverage (% of adults) 0.0178 0.0596 0.2076 0.1231 0.5076 1
(0.7073) (0.2082) (0) (0.0069) (0)
Enforcing Contracts - Cost (% of
claim) -0.197 -0.2523 -0.5936 -0.1544 -0.5436 -0.0342 1
(0) (0) (0) (0.0007) (0) (0.4545)
Institutional Credit index 0.1447 0.1882 0.534 0.4873 0.9051 0.6104 -0.6698 1
(0.0021) (0.0001) (0) (0) (0) (0) (0)
Source: WEO database, Doing Business database; IMF staff calculations.
¹ P-values in parentheses.
Appendix II. Table 1. Correlations between liquidity ratios, credit to GDP and credit institutions in CAPDR ¹
3
9
Systemic
risk
regulation
Safety net
variable 2/
In place Coverage Legal
protection of
supervisor
Partial
transfer of
uninsured
deposits
In place In place Coverage Contributes
to bank
resolution
process
Target
fund
Costa-Rica yes 50% of liquid
assets
low n/a no no n/a n/a n/a
2
Dominican
Republic
yes 1.5 times
paid capital
high yes yes yes 13,900 yes 5% of
total
deposits 4
Guatemala yes 50% of
regulatory
capital
low yes no yes 2,400 yes 5% of
total
deposits
3
Honduras yes 100% of
capital and
reserves
high yes yes yes 9,600 yes 5% of
total
deposits 4
Nicaragua yes not
established
by law
medium yes yes yes 10,000 yes no
4
El Salvador no/1 n/a medium yes no yes 9,000 yes no
2
Panama no n/a low n/a no no
10,000
n/a n/a
1
Source: Bolzico, J., Gozzi E., and F. Rossini, 2010, Financial Saf ety Nets in American Countries: A Comparative Analysis;IMF staf f calculations.
1/ While by law (2010) the Central Bank is allowed to provide liquidity as a LOLR, in practice this f acility is just being established.
2/ Index variable categorizing the existence of each of the 4 pillars of a saf ety net (regulation, banking resolution, LOLR and deposit insurance),
with 4 assigned in presence of all f our pillars.
Lender of last resort Bank resolution Deposit Insurance
Appendix II. Table 2. CAPDR Financial Safety Nets
4
0
Variable No. observations Mean Std. Dev. Min. Max.
Liquid assets to customer deposits and short-term funding ratio 448 25.3 18.3 2.0 191.0
Liquid assets to total assets ratio 448 18.9 10.4 0.9 75.6
Capitalization (equity to asset ratio) 448 13.2 9.3 2.6 83.0
Loan Loss Reserves to Gross Loans 417 3.2 3.1 0.0 25.0
Net Interest Income to Average Earning Assets 428 8.8 10.5 1.0 87.0
Loan to Asset ratio 448 58.9 17.1 3.3 90.6
Loan growth (y/y, percent) 350 24.0 55.6 -66.2 594.5
Bank Size (log of total assets) 448 12.8 1.7 4.1 16.5
Foreign ownership dummy (=1 if foreign bank) 480 0.4 0.5 0.0 1.0
Private ownership dummy (=1 if private bank) 480 0.9 0.3 0.0 1.0
Interest rate spread 480 8.3 2.9 3.1 16.8
Real GDP growth 384 3.9 5.7 -7.9 15.3
Net international reserves 480 2931.0 1113.3 832.9 5594.9
Deposit dollarization 480 50.2 34.2 13.5 100.0
Banking crisis dummy (=1 if country had banking crisis after 1970) 480 0.7 0.5 0.0 1.0
Financial safety net (categorical variable, 1-4; 4=comprehensive safety
net) 480 2.7 1.2 1.0 4.0
Credit to GPD (%) 480 48.8 25.9 17.2 93.7
Quality of credit institutions index (0=worst; 100=best) 480 68.3 28.2 0.0 100.0
Sources: International Financial Statistics, World Economic Outlook Database, Bankscope Database,
CAPDR Central banks and Banking Supervision websites, Laeven and Valencia (2010), Bolzico et al. (2010); IMF staff calculations.
Appendix III. Table 1: Descriptive Statistics
4
1
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Credit Financial Credit Financial Credit Financial Credit Financial Credit Financial
Institutions Depth Institutions Depth Institution Depth Institutions Depth Institutions Depth
Liquid assets ratio (-1) 0.484*** 0.557*** 0.539*** 0.567*** 0.398*** 0.483*** 0.563*** 0.519*** 0.543*** 0.534***
(0.112) (0.098) (0.101) (0.089) (0.118) (0.101) (0.088) (0.093) (0.106) (0.114)
Bank size 4.321*** 3.861*** 3.666*** 4.077*** 4.850*** 5.484*** 3.596*** 3.815*** 4.146*** 4.023***
(0.979) (0.866) (1.192) (1.048) (1.330) (1.908) (0.959) (1.257) (0.889) (0.977)
Bank size squared -0.201*** -0.180*** -0.171*** -0.197*** -0.232*** -0.262*** -0.167*** -0.179*** -0.191*** -0.187***
(0.049) (0.044) (0.060) (0.056) (0.065) (0.085) (0.048) (0.061) (0.044) (0.045)
Capitalization -0.226*** -0.211*** -0.098 -0.192 -0.188 -0.175* -0.100 -0.063 -0.292* -0.249
(0.071) (0.070) (0.215) (0.124) (0.158) (0.099) (0.233) (0.215) (0.162) (0.275)
Net interest margin -0.047 -0.037 0.096 -0.799 -0.570* -0.593* 0.197** 0.216 0.203 0.163
(0.033) (0.033) (2.354) (1.193) (0.331) (0.311) (0.099) (0.151) (0.323) (0.337)
Loan-loss reserve ratio 0.015 -0.036 0.025 0.277 -0.080 -0.292* -0.467* -0.178 -0.056 -0.346
(0.168) (0.145) (0.608) (0.280) (0.163) (0.166) (0.251) (0.395) (1.433) (1.230)
Credit 1/ 0.033 -0.181 0.114 0.117 0.044* -0.295 0.147*** 0.249 0.464*** -0.135
(0.024) (0.181) (0.159) (0.355) (0.026) (0.251) (0.047) (0.553) (0.133) (0.636)
Variable 9.567 16.285 0.403 -5.554 0.482*** 1.336*** 21.139*** -8.880
(24.722) (24.600) (6.609) (7.717) (0.174) (0.346) (7.458) (28.601)
Capitalization*variable -0.167 -0.071 -0.044 -0.161 -0.002 -0.003 0.053 0.027
(0.335) (0.251) (0.210) (0.212) (0.005) (0.004) (0.060) (0.119)
Net interest margin*variable -0.147 0.756 0.421 0.487 -0.011*** -0.012 -0.080 -0.067
(2.352) (1.186) (0.312) (0.328) (0.004) (0.008) (0.099) (0.109)
Loan-loss reserve ratio*variable 0.082 -0.291 0.948 1.340 0.019** 0.005 0.016 0.091
(0.681) (0.456) (0.878) (0.903) (0.008) (0.017) (0.425) (0.377)
Credit *variable 1/ -0.084 -0.444 -0.016 0.100 -0.005** -0.013 -0.131*** 0.006
(0.155) (0.424) (0.058) (0.130) (0.002) (0.009) (0.039) (0.267)
Observations 321 321 321 321 321 321 321 321 321 321
R-squared 0.72
No. of groups 96 96 96 96 96 96 96 96 96
No. of instruments 64 64 67 67 67 77 77 67 67
Hansen test p-value 0.357 0.337 0.302 0.267 0.283 0.747 0.448 0.546 0.388
A-B AR(2) test 1.191 1.075 1.236 1.152 0.891 1.026 1.427 1.012 1.075
A-B AR(2) test p-value 0.234 0.282 0.216 0.249 0.373 0.305 0.154 0.311 0.282
Source: Authors' calculations.
Dependent variable is the ratio of liquid assets to total assets. GMM is two-step system GMM estimator with Windmeijer standard error correction.
Columns (3) through (10 test the hypotheses that ownership (foreign/domestic and public/private), degree of dollarization and coverage of the financial safety net affect banks
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the country lev
All regressions include time and country dummies. Constant estimated but not reported.
1/ Credit is captured either by a measure of credit institutions, proxied by the principal components analysis of the Doing Business variables measuring the
quality of credit institutions. Financial depth is proxied by the credit-to-GDP ratio.
Appendix IV. Table 1. Determinants of Banks' Liquidity Buffers in CAPDR - GMM Estimates
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Dependent variable is the ratio of
liquid assets to total assets
Baseline Variable = Private Foreign ownership Dollarization Safety net
4
2
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Credit Financial Credit Financial Credit Financial
Pooled OLS GMM Fixed Effects Institutions Depth Institutions Depth Institutions Depth
Liquid assets ratio (-1) 0.347*** 0.189*** 0.169*** 0.171** 0.164** 0.256*** 0.227*** 0.252*** 0.221***
(0.064) (0.044) (0.025) (0.070) (0.069) (0.074) (0.072) (0.083) (0.063)
Bank size 7.401*** 7.994*** 16.804 9.227* 11.172** 7.615** 7.848** 8.551*** 8.543***
(1.448) (1.875) (10.703) (4.806) (5.221) (2.972) (3.038) (2.081) (2.126)
Bank size squared -0.350*** -0.371*** -0.886* -0.469** -0.553** -0.362** -0.381** -0.410*** -0.414***
(0.064) (0.092) (0.476) (0.204) (0.218) (0.150) (0.152) (0.107) (0.106)
Capitalization -0.355*** -0.321** -0.502*** -0.711 -0.453 -0.281 -0.424 -0.386 -0.262
(0.060) (0.123) (0.184) (0.557) (0.516) (0.495) (0.477) (0.528) (0.457)
Net interest margin -0.156** -0.123 0.011 -0.774 -1.169 0.452 0.332 -0.056 -0.107
(0.064) (0.076) (0.325) (0.752) (0.833) (0.276) (0.335) (0.704) (0.706)
Loan-loss reserve ratio -0.224 -0.282 0.221 -0.448 -0.984 -1.071 -1.032 1.536 1.277
(0.160) (0.252) (0.290) (0.553) (0.810) (0.678) (0.704) (1.691) (1.627)
Credit 3/ -0.213 -0.323 -0.337* 0.044 -0.904** 0.240*** -0.860 0.625** -0.123
(0.272) (0.292) (0.199) (0.046) (0.387) (0.088) (0.675) (0.250) (0.794)
Variable -6.238 -18.363 0.666** 1.201*** 27.313** -10.493
(12.128) (15.445) (0.293) (0.412) (13.739) (35.604)
Capitalization*variable 0.339 -0.070 -0.001 0.000 -0.037 -0.092
(0.608) (0.711) (0.007) (0.008) (0.191) (0.164)
Net interest margin*variable 0.441 0.973 -0.025** -0.021 -0.023 -0.000
(0.750) (0.891) (0.012) (0.015) (0.217) (0.218)
Loan-loss reserve ratio*variable 2.167* 3.021* 0.042* 0.036 -0.429 -0.339
(1.222) (1.552) (0.021) (0.022) (0.549) (0.451)
Credit to GDP ratio*variable 1/ 0.010 0.212 -0.008** -0.003 -0.169** -0.231
(0.091) (0.214) (0.003) (0.009) (0.070) (0.262)
Observations 321 321 321 289 289 289 289 289 289
R-squared 0.55 0.19
No. of groups 96 96 88 88 88 88 88 88
No. of instruments . 64 67 67 77 77 67 67
Hansen test p-value 0.348 0.105 0.117 0.135 0.135 0.187 0.243
A-B AR(2) test 1.283 1.262 0.770 1.600 1.574 1.231 1.322
A-B AR(2) test p-value 0.199 0.207 0.442 0.110 0.116 0.218 0.186
Source: Authors' calculations.
Dependent variable is the ratio of liquid assets to customer deposits and short-term funding. Two-step system GMM estimator with Windmeijer standard error correction.
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the
country level. Safety net is an index variable categorizing the existence of each of the 4 components of a safety net in each country (Appendix II. Table 1).
All regressions include time and country dummies. Constant estimated but not reported.
1/ Columns (1) and (2) as in Table 4a. Column (3) reports results of the fixed effects estimation.
2/ Columns (4)-(9) show GMM estimation results for the sample of private banks (excluding public banks).
3/ In columns (1)-(3) credit is captured by a measure of financial depth, proxied by the credit-to-GDP ratio. In columns (4)-(9), Credit institutions are proxied
by the principal components analysis of the Doing Business variables measuring the quality of credit institutions.
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Appendix IV. Table 2. Determinants of Banks' Liquidity Buffers in CAPDR - GMM Estimates (unless otherwise specified)
Dependent variable is the ratio of
liquid assets to customer deposits
and short-term funding
Baseline 1/ Foreign ownership 2/ Dollarization 2/ Safety net 2/
doc_170349835.pdf
Banks liquidity holdings are comfortably above legal or prudential requirements in most Central American countries. While good for financial stability, high systemic liquidity may nonetheless hinder monetary policy transmission and financial markets development.
The Determinants of Banks' Liquidity Buffers in
Central America
Corinne Deléchat, Camila Henao, Priscilla Muthoora,
Svetlana Vtyurina
WP/12/301
© 2012 International Monetary Fund WP/12/301
IMF Working Paper
Western Hemisphere Department
The Determinants of Banks' Liquidity Buffers in Central America
Prepared by Corinne Deléchat, Camila Henao, Priscilla Muthoora, Svetlana Vtyurina
1
Authorized for distribution by Przemek Gajdeczka
December 2012
Abstract
Banks’ liquidity holdings are comfortably above legal or prudential requirements in most
Central American countries. While good for financial stability, high systemic liquidity
may nonetheless hinder monetary policy transmission and financial markets development.
Using a panel of about 100 commercial banks from the region, we find that the demand
for precautionary liquidity buffers is associated with measures of bank size, profitability,
capitalization, and financial development. Deposit dollarization is also associated with
higher liquidity, reinforcing the monetary policy and market development challenges in
highly dollarized economies. Improvements in supervision and measures to promote
dedollarization, including developing local currency capital markets, would help enhance
financial systems’ efficiency and promote intermediation in the region.
JEL Classification Numbers: E44, G21, O16
Keywords: Central America, bank liquidity, credit, dollarization, excess liquidity, foreign banks
Authors’ E-Mail Addresses: [email protected], [email protected], [email protected],
[email protected]
1
The authors would like to thank Miguel Savastano for his guidance. Przemek Gajdeczka, Fernando Delgado,
Mario Mansilla, Stephanie Medina Cas, Andy Swiston and other IMF colleagues, as well as staff from the Central
Bank of Guatemala, Rudy Loo-Kung and Mario Cuevas at the Inter-American Development Bank, and Miquel
Dijkman at the World Bank provided useful comments.
This Working Paper should not be reported as representing the views of the IMF.
The views expressed in this Working Paper are those of the author(s) and do not necessarily
represent those of the IMF or IMF policy. Working Papers describe research in progress by the
author(s) and are published to elicit comments and to further debate.
2
Contents Page
I. Introduction ................................................................................................................................ 3
II. Determinants of Banks’ Liquidity Buffers: Literature Review ................................................. 5
III. Some Background on CAPDR Banking Systems .................................................................... 7
IV. Determinants of Banks Liquidity Buffers in CAPDR ........................................................... 12
A. Data and Variable Definitions..................................................................................... 12
B. Stylized Facts .............................................................................................................. 17
C. Empirical Analysis ...................................................................................................... 19
D. Results ......................................................................................................................... 20
V. Conclusions and Policy Lessons ............................................................................................. 26
References .................................................................................................................................... 29
Tables
1. CAPDR: Banking System Indicators, 2010 ..........................................................................9
2. Variables Used in the Empirical Estimation ..........................................................................15
3. Dependent Variables Means by Liquidity Ratio Quartiles ....................................................18
4a. Determinants of Banks’ Liquidity Buffer in CAPDR-Financial Depth ...............................23
4b. Determinants of Banks’Liquidity Buffer in CAPDR-Credit Institutions .............................24
5. CAPDR: Largest Deposit Drop (December 2001–September 2011) ..................................26
Figures
1. CAPDR Liquidity Buffers (2010) .........................................................................................3
2. CAPDR: Statutory Reserves and Liquidity Requirements by Currency .................................8
3. CAPDR: Financial Soundness Indicators, 2005–10 .............................................................11
4. Bankscope Sample Coverage of Total Banking System’s Assets ........................................12
5. Liquidity Ratios at the System Level and in the Bankscope Sample ...................................12
6. Financial Development Indicators in CAPDR .......................................................................14
7. CAPDR Liquidity Indicators ................................................................................................17
Box
1. Financial Systems Supervision and Development in CAPDR .............................................10
Appendix I. Reserve and Liquidity Requirements in CAPDR ....................................................33
Appendix Tables
Appendix I. Table 1. Legal Reserve and Liquidity Requirements in CAPDR (2010) ............. 36
Appendix I. Table 2. Evolution of Reserve and Liquidity Ratios in CAPDR (2005–10) ........ 37
Appendix II. Table 1. Correlation Between Liquidity Ratios, Credit to GDP and Credit
Institutions in CAPDR ............................................................................ 38
Appendix II. Table 2. CAPDR Financial Safety Nets ................................................................ 39
Appendix III. Table 1. Descriptive Statistics ............................................................................... 40
Appendix IV. Table 1. Determinants of Liquidity Buffers in CAPDR– Robustness Checks ..... 41
Appendix IV. Table 2. Determinants of Liquidity Buffers in CAPDR– Robustness Checks ..... 42
3
I. INTRODUCTION
This paper studies the determinants of banks’ liquidity buffers in Central America,
1
Panama
and the Dominican Republic (CAPDR) using a panel of about 100 commercial banks over
2006-10. In particular, the paper examines whether CAPDR banks’ liquidity buffers, defined
as the liquid assets to deposits ratio, can be explained by bank and country-level
characteristics, as predicted by theory and documented in some empirical studies. Of
particular interest for the region is whether liquidity holdings are related to bank ownership
(are public or foreign banks different?) or the economy’s degree of dollarization.
CAPDR banking systems are highly liquid. Holdings of liquid assets as a share of total
deposits averaged about 28 percent for the region in 2010, while reserve requirements were
set at about 17 percent on average (Figure 1). Liquidity ratios are also high compared to
larger South-American countries; liquidity ratios averaged about 15 percent for Brazil, Chile
and Colombia in 2010. For monetary and supervisory authorities, ensuring that banks hold
adequate amounts of high-quality liquid assets is essential for financial stability, as
highlighted during the recent global financial crisis. However, if liquidity holdings are much
above legal requirements, this may be costly in terms of foregone financial intermediation.
Excess liquidity also hinders the development of interbank and money markets in all
countries, and acts as “sand in the wheels” of the monetary transmission mechanism in
countries with a monetary policy (Gray, 2011).
Figure 1. CAPDR Liquidity Buffers¹ (2010)
1
Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua.
Sources: CAPDR Central Banks and Superintendencies' websites; and Fund staff calculations.
¹ Liquid assets to deposits ratio. Liquid assets include cash, central bank reserves and deposits abroad.
² Liquidity requirements.
³ Observed liquidity. Legal liquidity requirement (30%) is calculated as the ratio of liquid assets including
securities to short term deposits.
0
5
10
15
20
25
30
35
40
Costa Rica Dominican
Republic
El Salvador ² Guatemala Honduras Nicaragua Panama³
(
p
e
r
c
e
n
t
)
Required liquidity holdings (percent of deposits) Excess liquidity (percent of deposits)
4
From individual banks’ point of view, holding sufficient liquidity is necessary to insure
against liquidity risk (Diamond and Dybvig, 1983, Diamong and Rajan, 2001). As loans are
relatively illiquid, large and unexpected deposit withdrawals can lead to insolvency as it may
be too costly or not possible to raise liquidity on short notice, due to capital market
imperfections. Instead of self-insuring, banks could resort to other forms of financing, such
as accessing interbank markets, central bank liquidity windows, or external credit lines.
However, asymmetric information may lead to coordination failures on the interbank market,
and external credit lines may freeze (as seen during the recent financial crisis), so that solvent
but illiquid banks would still fail, absent a Lender of Last Resort (LOLR) (Rochet and Vives,
2004). Thus banks hold a buffer of liquid assets as self-insurance, equating the marginal
benefit of holding liquid assets to the marginal cost of alternative investments.
A priori, one would expect the self-insurance motive to be especially important in CAPDR.
Capital markets are underdeveloped, interbank markets are thin, and LOLR arrangements
remain limited or nonexistent. For the five partially dollarized economies, the high share of
foreign currency assets and liabilities limits the ability of the central bank to act as a LOLR,
while the two fully dollarized economies in the region, Panama and El Salvador, did not have
a LOLR as of end-2010 (El Salvador formally approved the regulations to establish a
liquidity facility in June 2012). Furthermore, while the region’s predominant reliance on
customer deposits is a likely reason for its resilience during the global financial crisis, it is
also a potential source of vulnerability and calls for the holding of adequate liquidity buffers.
In line with theory and results from related empirical studies, we find that CAPDR banks’
liquidity buffers are persistent, consistent with the notion that banks target an optimal amount
of liquid asset coverage of deposits or total assets. Liquidity ratios depend on idiosyncratic
factors such as bank size, profitability, and efficiency. Measures of financial development
also matter. We find some evidence that foreign banks in the region tend to hold less
liquidity than domestic banks, possibly reflecting their access to more diversified funding
sources or less risky business models. Furthermore, we uncover a robust positive impact of
the degree of deposit dollarization on the size of liquidity buffers. This suggests that not only
deposit dollarization in itself,
2
but also its association with higher liquidity holdings,
contributes to weaken the monetary transmission mechanism in partially dollarized
economies and hinders the development of money and interbank markets.
The reminder of the paper is structured as follows. Section II briefly reviews the theoretical
and empirical literature on the determinants of liquidity holdings. Section III provides some
background information on CAPDR banking systems. Section IV describes the data and
presents stylized facts on the distribution of banks’ liquidity holdings. Section V discusses
the econometric methodology and estimation results, and Section VI concludes.
2
See Medina Cas, Carrion and Frantischek, 2011a.
5
II. DETERMINANTS OF BANKS’ LIQUIDITY BUFFERS: LITERATURE REVIEW
The determinants of banks’ liquidity buffers, as identified in the theoretical and empirical
literature, can be classified into four broad categories. These are the opportunity costs of and
shocks to funding, bank characteristics, macroeconomic fundamentals, and moral hazard
motives, as discussed below.
Opportunity cost and shocks to funding
The early literature on bank liquidity uses the firm’s theory of inventory decisions as a
starting point. The cost of holding liquid assets (with low returns compared with other types
of investments) is compared to the benefits of reducing risks of “running out” (Baltensperger,
1980, and Santomero, 1984). These models predict that the size of liquidity buffers should
reflect the opportunity cost of holding liquid assets rather than loans. It should also relate to
the distribution of liquidity shocks that the bank may face, and in particular be positively
related to the volatility of the funding basis as well as the cost of raising additional funds.
Using aggregate time-series data for Thailand, Agénor, Aizenmann and Hoffmaister (2000)
find that banks’ demand for precautionary reserves (measured as the log of excess reserves
over total deposits), is positively related to the penalty rate, proxied by either the discount or
the money market rate, as well as to the volatility of the cash to deposit ratio. Dinger (2009)
finds in a panel of Eastern European banks that liquidity buffers are negatively related to the
real deposit rate but positively related to the interbank rate.
Bank characteristics
The newer generation of models explaining firms’ (including banks’) liquidity demand relies
on some form of market imperfection to explain why banks cannot raise instantaneous and
unlimited amounts of liquidity (financial frictions). The market imperfection is asymmetric
information, either in the form of moral hazard (Holmstrom and Tirole, 1998) or adverse
selection (Kiyotaki and Moore, 2008). Financially constrained banks would thus tend to hold
more liquidity.
3,4
Based on these models, bank characteristics affecting their ability to raise non-deposit forms
of finance, such as bank size (small banks have more difficulties in accessing capital
markets), profitability (more profitable banks can more readily raise capital and are thus less
3
Holmstrom and Tirole (1998) and Kiyotaki and Moore (2008) make this argument for firms in general:
Liquidity constraints, together with liquidity shocks, result in entrepreneurs not being able to raise the entire
cost of their desired investment externally, so that they have to hold enough liquid assets to make a down
payment for each unit of investment (there are also limits on the amount of equity that can be resold). This
explains why, although the rate of return on cash is very low, entrepreneurs will choose to hold some in their
portfolio. Liquidity shocks reduce the price of equity and increase the desired holdings of liquid assets.
4
See for example Almeida, Campello and Weisbach, 2004, Kashyap, Rajan and Stein (2002), Kashyap and
Stein (1997), Repullo (2003), and Rochet and Vives (2004).
6
liquidity constrained), ownership (both public banks and foreign banks should be less
liquidity-constrained than private and domestic banks, respectively, as public banks may
have an implicit guarantee and foreign banks would have access to support from
headquarters)
5
would affect banks’ precautionary demand for liquidity buffers.
Aspachs, Nier and Tiesset (2005) find that banks’ liquidity buffers are related to bank
characteristics such as loan growth and the net interest margin, with the coefficients on size
and profitability being not significant. Kashyap and Stein (1997) and Kashyap, Rajan and
Stein (2002), using a large panel of U.S. banks, find a strong effect of bank size on holdings
of liquid assets, with smaller banks being more liquid as they face constraints in accessing
capital markets. Dinger (2009) also finds that smaller Eastern European banks hold more
liquidity, but with nonlinearities, and that foreign banks hold less liquidity.
Bank ownership may not only exert a direct influence on liquidity holdings, but may also
affect the regression slope through interactions with other explanatory variables. In
particular, Aspach, Nier and Thiesset (2005) find that, for the UK, foreign banks’ liquid asset
holdings are not affected by the availability of a domestic lender of last resort, while local
banks are. Furthermore, in their sample foreign banks’ liquidity holdings tend to react less to
changes in the domestic policy rate and GDP growth, suggesting overall that they are subject
to a somewhat different set of constraints than their local counterparts.
Macroeconomic fundamentals
The models mentioned above also have implications for the cyclical behavior of liquidity
demand. If capital markets are imperfect, the demand for liquidity should be countercyclical,
as banks would hoard liquid assets during recessions and offload them in good times given
more opportunities to lend. This suggests that liquidity buffers would be negatively related to
measures of the output gap or real GDP growth, credit cycle, and policy interest rates.
6
The counter-cyclicality of liquidity buffers limits the effectiveness of monetary policy in
trying to inject liquidity to stimulate the economy in a recession: liquidity buffers would
remain stable or increase but credit would not necessarily pick-up. Moreover, financial
frictions in terms of capital market imperfections should be expected to vary with structural
factors such as the degree of financial development and the quality of financial institutions.
Aspach, Nier and Tiesset (2005) find that UK banks’ liquidity buffers are negatively related
to real GDP growth and the policy rate. Agénor, Aizenmann and Hoffmaister (2000) and
Saxegaard (2006) find that excess reserves are negatively related to the output gap and the
policy rate in Thailand and in sub-Saharan Africa, respectively. Dinger (2009) finds that
liquidity holdings are negatively related to real GDP growth and real per capita GDP.
5
Freixas and Holthausen (2005).
6
Almeida, Campello and Weisbach (2004) develop and estimate on a large sample of U.S. manufacturing firms
a model where financially constrained firms have a higher propensity to save cash out of cash flows.
7
Moral hazard and safety nets
In theory, the strength of the financial safety net and in particular the availability of a LOLR
arrangement, should reduce the banks’ incentives to hold liquidity buffers (Repullo, 2003).
Empirical studies of UK and Argentinian banks, where LOLR support is measured as the
Fitch support rating and the availability of external credit lines in the context of the currency
board, respectively, support this prediction (Aspach, Nier and Thiesset, 2005, and Gonzalez-
Eiras, 2003).
Dollarization or credit and/or deposits reduces the effectiveness of the domestic LOLR, as
partially dollarized economies are subject to currency and liquidity risk, but the central bank
cannot issue foreign currency (Gulde et al., 2004 and Levy-Yeyati and Broda, 2002). One
would thus expect banks to hold higher liquidity buffers, the higher the degree of deposit
dollarization, though the incentives to hold such buffers would diminish in the presence of a
large stock of central bank international reserves or external credit lines, as these would be a
ready source of dollar liquidity in the case of a run on dollar deposits (Ize, Kiguel and Levy-
Yeyati, 2005). Using a sample of about 100 countries, De Nicoló, Honohan and Ize (2005)
find that deposit dollarization is associated with higher solvency and liquidity risk measured
by deposit volatility. To our knowledge no empirical study has focused on the effects of
deposit dollarization on banks’ liquidity.
III. SOME BACKGROUND ON CAPDR BANKING SYSTEMS
As a group of small open economies with strong linkages to the U.S. economy and various
levels of dollarization, most countries maintain fairly stable de facto exchange rates against
the U.S. dollar—although their formal monetary and exchange arrangements are quite
heterogeneous. Two economies (Panama and El Salvador) are fully dollarized, three
countries (Costa-Rica, Dominican Republic and Guatemala) have recently adopted or are
moving toward inflation targeting, and Honduras and Nicaragua maintain crawling or tightly
managed pegs.
7
The monetary policy frameworks in the five partially dollarized economies
rely on rules-based instruments such as standing facilities and reserve requirements, with
only partial money market operations, and all but Nicaragua have a monetary policy interest
rate (Medina Cas, Carrion-Menendez and Frantischek, 2011b).
Reserve requirements in CAPDR are in line with those of the Latin America region, and
average about 15 percent for local currency deposits and 15.5 percent for foreign currency
ones (Appendix I, Table 1 and Figure 2)
. The two officially dollarized economies rely on
prudential liquidity requirements, held at the central bank in the case of El Salvador and by
individual banks in the case of Panama (Appendix I).
8
Although they are potentially useful
7
Costa Rica maintains a managed exchange rate regime with pre-defined bands for intervention but a
substantial margin to fluctuate within the bands. Nonetheless, since December 2010 the exchange rate has
remained within 3 percent of the floor of the band.
8
Honduras also imposes specific liquidity requirements to avoid maturity mismatches (Appendix I).
8
policy instruments, reserve/liquidity requirements are not actively managed in most
countries, with the exception of El Salvador and Honduras (Appendix I, Table 2).
9
Figure 2. CAPDR: Statutory Reserves and Liquidity Requirements by Currency
With the exception of Panama, the region’s banking systems are relatively small, highly
concentrated and dollarized to various degrees (Table 1). Panama stands out of the group in
terms of the size of the system, which is four times greater than the sample average in terms
of assets to GDP (Panama’s offshore banks’ assets represented only 50 percent of GDP at
end-2010). In four countries (Honduras, Nicaragua, El Salvador and Panama) the share of
foreign bank assets in total assets is more than 50 percent suggesting higher potential
vulnerabilities from cross border linkages. While the presence of state banks is quite small in
terms of number of banks and share of system’s assets for the whole sample, state banks have
a very strong presence in Costa Rica, with their assets accounting for 55 percent of total
assets and 60 percent of deposits. Customer deposits are the main source of funding and
show a high degree of dollarization, particularly in Nicaragua and Costa Rica. The share of
short-term deposits is also relatively high in the region, although not in Panama.
9
Given this, excess liquidity is probably best analyzed in the context of single country time-series studies. In
the panel context, our preferred definition of liquidity buffers for the empirical analysis in section IV.C is the
liquid assets to deposit ratio. The liquid assets to total assets ratio is used for robustness checks.
Source: Central Banks and Superindendencies.
Note: Reserve requirements for all countries excluding Panama and Salvador (liquidity requirements). Liquidity requirement for Panama is def ined as the ratio of liquid assets
including securities and obligations payable to banks within 186 days, as a share of short-termdeposits.
15
17
14.6
12
16.25
17
26
30
0
5
10
15
20
25
30
35
C
o
s
t
a
R
i
c
a
D
o
m
.
R
e
p
.
G
u
a
t
e
m
a
l
a
H
o
n
d
u
r
a
s
N
i
c
a
r
a
g
u
a
L
a
t
i
n
A
m
e
r
i
c
a
E
l
S
a
l
v
a
d
o
r
P
a
n
a
m
a
Local Currency Reserve or Liquidity
Requirements (percent of local currency
deposits)
15
20
14.6
12
16.25
14
0
5
10
15
20
25
30
35
C
o
s
t
a
R
i
c
a
D
o
m
.
R
e
p
.
G
u
a
t
e
m
a
l
a
H
o
n
d
u
r
a
s
N
i
c
a
r
a
g
u
a
L
a
t
i
n
A
m
e
r
i
c
a
Foreign Currency Reserve Requirements
(percent of foreign currency deposits)
9
Recent Financial Sector Assessments indicate that, overall, banking sectors in the region are
well-capitalized, liquid and profitable (Figure 3). Financial systems remained resilient in the
face of the 2009 global financial crisis mostly due to their strong initial positions. Despite
rapid credit growth, the region did not experience any excessive credit booms and there was
very limited exposure to toxic asset-backed securities, as well as to wholesale funding. Stress
tests of liquidity risk suggested that banks had adequate coverage of their liquid liabilities,
and could withstand deposit withdrawal shocks of 15–20 percent during a 30 day period.
However, although banking supervision has improved over the past decade, compliance with
Basel Core Principles remains uneven and below that of LA6. Financial safety nets remain
incomplete, and financial markets are underdeveloped, including interbank markets (Box 1).
Number of
banks
5
Number of
State
banks ²
State bank
assests in total
assets
Number of
foreign
banks ³
Foreign
bank
assets in
total assets
Percent of
assets in 5
largest
banks
Assets to
GDP
Credit to
GDP
Assets in
foreign
currency in
total assets
Credit in
foreign
currency in
total credit
Deposits in
foreign
currency in
total
deposits
Demand
deposits
in total
deposits
Costa Rica 16 3 55 9 26 78 60 46 46 47 41 53
Guatemala 18 1 2 7 13 79 44 30 23 30 24 41
Honduras 17 1 1 9 50 75 68 50 24 28 30 22
Nicaragua 6 1 0.01 4 67 97 62 34 72 90 73 31
Dominican Republic 15 2 31 4 29 87 32 18 26 21 30 18
El Salvador ¹ 12 2 6 10 83 85 61 40 … … … 26
Panama
1, 4
49 2 14 28 57 57 200 91 … … … 15
Source: Central American Monetary Council (SECMCA), International Financial Statistics, IMF staff calculations.
¹ Officially dollarized economies. ² State share of more than 50 percent. ³ Banks with 50 percent of capital in foreign hands, excludes offshore.
4
Domestic banking system;
5
April 2011.
Table 1. CAPDR: Banking System Indicators, 2010
(in percent, unless otherwise indicated)
10
Box 1. Financial Systems Supervision and Development in CAPDR
Despite substantial progress over the past decade, financial supervision in CAPDR lags behind,
both compared to larger emerging countries in Latin America and international best practices.
Following a string of banking crises in the 1990s - Costa Rica (1987-91, 1994-95), El Salvador
(1989-90), Nicaragua (1990-93), and Panama (1988-89)
,1/ a system-wide banking crisis in the
Dominican Republic (2003-04), and the failure a large and a medium-sized bank in Guatemala
(2006-07), the Central American countries strived to strengthen their supervisory frameworks. By
2010, CAPDR complied with more than half of the Basel Core Principles, though the performance
was uneven across the region, with Panama showing the highest compliance and Costa Rica the
lowest. The strengthening of financial supervision during the last decade has been relatively
homogeneous between categories, but unequal by category within each country. The weakest areas
in the region are risk-based supervision, cross-border consolidated supervision, and the fact that
nonbank institutions frequently are outside the supervisory perimeter (Delgado and Meza, 2011).
Financial safety nets remain incomplete. All countries, except Panama and El Salvador, have a
LOLR (though El Salvador is in the process of formally implementing one), and deposit insurance
schemes are in place in all but two countries (Costa Rica and Panama). Weaknesses in most
countries are observed in providing legal protection to supervisors and regulating systemic risk.
The region’s money and capital markets remain underdeveloped.2/ Interbank markets are thin,
segmented, and often lack proper collateral. Despite the high level of liquidity in the banking
sector, the absence of collateral in inter-bank transactions increases systemic vulnerability to bank
failures. A lack of active repo operations also stalls the standardization of transactions in the
interbank market (see FSSAs for the region). The number of equity and corporate bond listings are
generally in the single digits and market-capitalization-to-GDP ratios are quite low. Due to legal
and other obstacles, institutional investors intermediate a relatively small share of national
savings. Thus, the principal investors in government securities are commercial banks and public
sector entities such as pension or social security funds. Secondary public debt markets are shallow
as government securities are mostly held to maturity.
____________________
1/ Panama did not experience a typical banking crisis as the closure of the banking system for several
months was due to the invasion of the country by the U.S. armed forces confronting the Noriega regime.
There was no deposit runs prior or after the closure of the banks and there were no bank failures. However,
this contributed to a significant output loss, estimated by some at 85 percent (Laeven and Valencia (2012)).
2/ See Shah et al. (2007a), and Shah et al. (2007b).
11
Figure 3. CAPDR: Financial Soundness Indicators, 2005-11/1
Source: Central American Monetary Council (SECMCA).
1/ Shaded area represents the 2008-09 global financial crisis.
2/ Liquid assets include short term investments.
30.0
35.0
40.0
45.0
50.0
55.0
60.0
65.0
Jan-05Jan-06Jan-07Jan-08Jan-09Jan-10Jan-11
Liquid Assets /Total Assets/ 2
1.0
1.5
2.0
2.5
3.0
3.5
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Return on Assets (ROA)
9.0
14.0
19.0
24.0
29.0
34.0
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Return on Equity (ROE)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Non Performing Loans (NPLs) / Total
Loans
... a deterioration in the quality of loan portfolios.
50.0
75.0
100.0
125.0
150.0
175.0
200.0
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11
Provisions/NPLs
Despite the global crisis, risk-adjusted capital
ratios remain adequate...
Bank profitability declined in 2008, with falls in the
return on assets...
..and by 2010 banks had rebuilt their liquidity
and return on equity, reflecting in part...
Provisioning has generally lagged the growth
in NPLs, but there are signs of recovery.
6.0
9.0
12.0
15.0
Jan-05Jan-06Jan-07Jan-08Jan-09Jan-10Jan-11
Capital Adequacy Ratio (CAR)
3decile
CAPDR-median
7decile
12
Figure 5. Liquidity Ratios at the System Level and in the Bankscope Sample
(2006-10 average, in percent)
Source: BankScope database, CAPDR Central Banks and Superintendencies' websites;
IMF staff calculations.
0
5
10
15
20
25
30
35
40
45
Costa Rica Dominican
Rep.
El
Salvador
Guatemala Honduras Nicaragua Panama
Banking System's Liquid Assets to Deposits Ratio
Bankscope Sample: Liquid Assets to Customer Deposits and Short Term Funding
Figure 4. Bankscope Sample Coverage of Total Banking System's Assets
Source: BankScope database, CAPDR Central Banks and Superintendencies' websites,
IMF staff calculations.
(2006-10 average, in percent)
98.38 97.44
90.68
85.91
63.24
56.98
54.94
0
20
40
60
80
100
120
Costa Rica Dominican Rep. Guatemala Honduras Panama El Salvador Nicaragua
IV. DETERMINANTS OF BANKS LIQUIDITY BUFFERS IN CAPDR
A. Data and Variable Definitions
Our sample combines annual data for 96 CAPDR banks over 2006–10 from the BankScope
data base,
10
with country-level
macroeconomic fundamentals and
structural variables drawn from
regional monetary and supervisory
authorities’ websites and other
publicly-available databases as
described in Table 2 below and
Appendix III, Table 1. It covers
72 percent of all commercial banks in
the region and about 80 percent of
total banking system assets, though
admittedly the coverage is not
homogeneous across all countries in
the region (Figure 4).
11
Definition of liquidity buffers
Liquidity buffers are measured by the
ratio of liquid assets to customer
deposits and short-term funding.
Liquid assets include cash and cash-
like assets, quoted or listed
government bonds, and short-term
claims on other banks. Although the
breakdown of the numerator
components is not available, there are
relatively few listed government
securities in the region (Shah et al.,
2007b). The denominator includes
banks’ customer deposits and short-
term interbank deposits. Customer deposits are the main source of funding in the region, with
very low reliance on short-term funding (the share of customer deposits in the denominator is
93 percent for the whole sample). Overall, the ratio of liquid assets to customer deposits and
short-term funding is close to system-wide liquidity ratios, defined as liquid assets (cash and
10
A financial database supplied by Bureau van Djik.
11
The information on coverage averaged over banks/years. A caveat is that missing institutions may not be
random. The time period is restricted to the interval for which data for most CAPDR banks were available.
13
cash-like, excluding securities) to deposits (Figure 5). We use it as our main dependent
variable, and use the ratio of liquid assets to total assets for robustness checks.
12
Choice of explanatory variables
Opportunity cost, liquidity shocks and bank characteristics
We use the spread between the lending and the deposit rate as a measure of the opportunity
cost. The probability of a liquidity shock can be proxied by a measure of the volatility of total
deposits at the system level (we can calculate a monthly coefficient of variation of total
deposits for each country, but have only annual bank-level data), or by the volatility of
inflation. Past liquidity shocks may also matter: a history of banking crisis could lead banks
to become more risk-averse and hold more liquidity.
Given the importance of public and foreign banks in Central America’s banking systems, we
are particularly interested in testing whether liquidity buffers vary systematically according
to bank ownership (public/private and foreign/domestic).
We control for other bank characteristics such as size, measured as the log of total assets. The
squared value of this variable captures possible non-linearities in the impact of bank size on
liquid asset holdings (Dinger, 2009). Capitalization is measured as the ratio of equity to total
assets. Profitability is measured by the ratio of the net interest margin to interest-earning
assets. The loan-loss reserves to gross loans ratio should capture the banks’ degree of risk
aversion or perceived riskiness of the loan portfolio.
Macroeconomic fundamentals
We use output growth in CAPDR to capture the economic cycle. Financial development is
captured by private sector credit to GDP ratio, a traditional proxy for financial depth. We
also construct a quality of credit institutions index which combines four variables from the
World Bank’s Doing Business database
13
focused on credit and legal rights, ?strength of
legal rights index (0-10); depth of credit information index (0-6); private bureau coverage
(percent of adults), and the inverse of the cost of enforcing contracts (cost as percent of
claim). Since these variables are correlated, we use principal components analysis to
construct an index which captures the underlying common variance of those four variables
(See pair-wise correlations in Appendix II. Table 1).
14
The index is normalized so a value of
12
Empirical studies use both ratios, see Aspach, Nier and Thiesset (2005) and Dinger (2009). The ratio of liquid
assets to liabilities is the most consistent with the notion of CAPDR banks self-insuring against deposit shocks,
though banking theory also emphasizes asset-side liquidity problems (Diamond and Rajan, 2005).
13
Available on the internet at www.doingbusiness.org.
14
Principal components analysis is a multivariate statistical technique that uses n correlated variables, and
reduces their dimension such that each component is a linear combination of the initial variables, and
components are orthogonal with each other. The components are ordered so that the first component accounts
for as much of the variability in the original data as possible. We use the first component as our index of the
(continued…)
14
zero represents the lowest quality of credit institutions, and 100 the highest. Although both
measures of financial development are highly correlated in some CAPDR countries (e.g.
Panama, Nicaragua), they differ quite radically in others (e.g. the Dominican Republic, El
Salvador) (Figure 6). Whereas credit-to-GDP is an outcome variable that captures both
supply and demand factors, the quality of credit institutions is more likely associated with
obstacles to the supply of credit. In this case, the lack of alternative investment opportunities,
rather than pure precautionary motives, would also explain high levels of bank liquidity.
Figure 6. Financial Development Indicators in CAPDR
Moral hazard and safety nets
We construct an index variable categorizing the existence of each of the four pillars of a
safety net in each country (regulation and supervision, banking resolution, LOLR facility,
and deposit insurance scheme). We attribute a value of 1 if the pillar is present and
functioning in a country, so the index can vary between zero and four, though in practice all
countries have banking regulation and supervision systems in place (Appendix II, Table 2).
Dollarization is measured as the share of dollar deposits in total system deposits (no currency
breakdown is available for bank-level data in BankScope). Net international reserves
holdings capture the dollar LOLR function of central banks.
quality of credit institutions in CAPDR, which we normalize to bound its values between 0 and 100 (see
Cárdenas and Henao, 2010, for an application).
Sources: WEO database, Doing Business database; IMF staff calculations
0
10
20
30
40
50
60
70
80
90
100
Costa Rica Dominican
Republic
El Salvador Guatemala Honduras Nicaragua Panama
Total credit to GDP (percent) Credit Institutional Index (0 worst; 100 best)
Total credit to GDP (percent) and credit institutional index (0-100)
15
Data description
Table 2 below presents the variable definitions, expected sign and data sources, and
Appendix II, Table 1 in annex describes the data. Overall there is significant variation in
liquidity holdings in the sample. Liquidity holdings in terms of customer deposits and short-
term funding are on average 25 percent in our sample, and represent about 18 percent of total
assets. Average capitalization is relatively high at about 13 percent, as noted in Basso,
Delgado and Meza (2012). Foreign banks represent 45 percent of observations, and private
banks about 90 percent. Deposit dollarization amounts to about 50 percent, though with wide
variations across countries (see Table 1).
Table 2. Variables Used in the Empirical Estimation
Variable Name
(expected sign)
Concept Measurement Data
source
Dependent variable
Liquidity ratio Liquid assets to customer
deposits and short-term
funding
(Cash, short-term claims on
other banks (including CDs)
and where appropriate the
trading portfolio)/Customer
deposits and short-term
funding.
BankScope
Liquid assets to total
assets
(Cash, short-term claims on
other banks (including CDs)
and where appropriate the
trading portfolio)/total
assets.
BankScope
Explanatory variables
Bank Characteristics
Lagged liquidity ratio
(+)
Liquidity buffers should be
persistent over time
See above for definition BankScope
Capitalization (-) Better capitalized banks
should have easier
access to markets and
thus hold less liquidity.
Ratio of equity to total
assets.
BankScope
Net Interest Income
to Average Earning
Assets (-)
Profitability: more
profitable banks should
hold less liquidity.
(Interest income-interest
paid)/ interest earning
assets.
BankScope
Loan-loss reserves
ratio (+)
Perceived riskiness by
banks of their loan
portfolio: banks
anticipating higher losses
should hold higher
liquidity buffers.
Ratio of loan-loss reserves
to gross loans.
BankScope
Size (-) If small banks are
financially constrained
than they should hold
more liquidity.
Natural logarithm of total
assets.
BankScope
Foreign ownership
(-)
Foreign banks should be
less financially-
constrained than
domestic banks and thus
hold lower levels of liquid
assets.
Dichotomous variable (1 for
foreign; 0 for domestic).
BankScope,
country
desks
A distinction is also made
between foreign subsidiaries
and branches.
16
Private ownership
(+)
Public banks could have
less incentives to hold
liquid assets as they
benefit from an implicit
state guarantee
Dichotomous variable (1 for
private; 0 for public).
BankScope,
country
desks.
Macroeconomic fundamentals
Real GDP growth (-) Imperfect capital markets
imply that liquidity buffers
should be countercyclical.
Annual growth rate of real
GDP per capita.
IMF country
desks
Interest Rate Spread
(-)
Measure of the
opportunity cost of
holding liquid assets.
Difference between average
lending and deposit rate
CAPDR
Monitor, IMF
country
desks.
Country Characteristics
Deposit volatility (+) Higher aggregate deposit
volatility forces banks to
hold more liquid assets to
hedge against
unanticipated deposit
withdrawals.
Coefficient of variation of
monthly system-wide
deposits during one year.
IMF country
desks.
Inflation volatility (+) High inflation volatility is a
proxy for macroeconomic
instability.
Coefficient of variation of
monthly inflation during one
year.
WEO.
History of banking
crisis (+)
Banks in countries that
have experienced a
banking crisis should hold
higher liquidity buffers.
Dichotomous variable (1 if
country experienced a
banking crisis in the past 50
years).
Laeven and
Valencia
(2012)
Credit-to-GDP ratio
(-)
Captures financial
development. More
developed economies
should have less financial
constraints.
Total credit to the private
sector as percent of GDP.
SECMCA
Quality of credit
institutions index (-)
Financial constraints
should be lower in
countries with better
credit institutions
Index capturing the quality of
credit institutions (0 is worst,
100 best).
World Bank
Doing
Business
Database
Moral hazard and safety nets
Deposit dollarization
(+)
The higher the
dollarization, the lower the
effectiveness of the
domestic lender of last
resort.
Share of foreign-exchange
deposits in total deposits
(system-wide).
IMF country
desks
Net international
reserves (-)
In partially dollarized
economies, NIR capture
the capacity of the central
bank to act as a lender of
last resort in case of a
foreign currency shock.
Natural logarithm of net
international reserves
IFS
Lender of last resort
(-)
Incentives to hold liquidity
decline with a strong
safety net.
Index variable categorizing
the existence of each of the
4 pillars of a safety net
(regulation, banking
resolution, LOLR and
deposit insurance.
Bolzico et
al.(2010),
Guerrero et
al. (2010)
17
B. Stylized Facts
At the country level, CAPDR banks’ liquidity holdings are above liquidity ratios in larger
South American countries, with the exception of Paraguay. Liquidity ratios have remained
fairly stable during the past decade including the global financial crisis (though with
variations across countries) – and have moved countercyclically, increasing when lending
opportunities were low and decreasing when they were high, consistently with modern
banking models’ predictions. Lower credit demand and implementation of countercyclical
fiscal policies during the crisis also led to an increase in banks’ holdings of government
securities over the past few years (Figure 7).
Figure 7. CAPDR: Liquidity Indicators
Moving to our bank sample, Table 3 summarizes the relationships between key explanatory
variables and liquidity ratios. More specifically, it shows the sample mean of explanatory
variables by liquidity quartile (with the quartiles being ranked from low to high). We are
interested in whether the characteristics of banks with high liquidity buffers (in the fourth
quartile) are different from those with low liquidity buffers (first quartile). Thus the last
column of the table shows whether the difference in the means of the variables for banks in
the 4
th
liquidity quartile versus the first quartile is statistically significant.
Source: CAPDR Central Banks, Superintedencies' websites, WEO; Fund staff calculations.
-5
0
5
10
15
20
25
30
35
40
2005 2006 2007 2008 2009 2010 2011
Government Bond Holdings Liquidity Credit to Private Sector
Liquidity and Domestic Government Assets
(annual percent change)
CRI
DR
SLV
GTM
HND
NIC
PAN
PAR
COL
ECU
CHI
R² = 0.61
0
2
4
6
8
10
12
14
10 20 30 40 50 60 70
¹ Liquid assets (including securities) to deposits ratio (percent)
Liquidity Ratios in CAPDR and Selected LA Countries ¹
(2010)
G
D
P
p
e
r
c
a
p
i
t
a
(
c
o
n
s
t
a
n
t
2
0
0
0
U
S
D
,
a
v
g
.
2
0
0
8
-
1
1
)
Liquidity Ratios in CAPDR and Selected LA Countries ¹
(2010)
BRZ
15
17
19
21
23
25
27
29
31
33
35
J
a
n
-
0
5
S
e
p
-
0
5
M
a
y
-
0
6
J
a
n
-
0
7
S
e
p
-
0
7
M
a
y
-
0
8
J
a
n
-
0
9
S
e
p
-
0
9
M
a
y
-
1
0
J
a
n
-
1
1
S
e
p
-
1
1
Percentile 30%
Median
Percentile 70%
Liquidity to Deposits, median for CAPDR
(in percent)
-10
-5
0
5
10
15
M
a
r
-
0
3
N
o
v
-
0
3
J
u
l
-
0
4
M
a
r
-
0
5
N
o
v
-
0
5
J
u
l
-
0
6
M
a
r
-
0
7
N
o
v
-
0
7
J
u
l
-
0
8
M
a
r
-
0
9
N
o
v
-
0
9
J
u
l
-
1
0
M
a
r
-
1
1
Liquidity to Deposits
Index of economic activity
Evolution of Liquidity Buffers and Economic Activity
(Median, y/y percent change, 3 mo. moving av.)
18
Overall most proposed explanatory variables exhibit the predicted relationship to liquidity
buffers, though not all the differences between the first and last quartile are significant. As
compared to those in the first liquidity quartile, observations (bank/years) for which observed
liquidity is high also tend to be less profitable, smaller, and private. They also have a lower
probability to be in a country with a relatively low interest rate spread, low deposit
dollarization, a history of banking crisis, an incomplete financial safety net. High liquidity
also seems associated with lower quality of financial institutions.
15
However, high liquidity
also seems to be associated with lower inflation volatility and higher financial depth.
15
Given that our time dimension is relatively short, and that it covers the period of the global financial crisis
period, we are also interested in testing whether the behavior of the main explanatory variables was different
pre-, during- and post-crisis. We find that for the crisis years (2008-09) the main relationships identified for the
whole sample continue to hold (Tabulations available upon request).
1
st
2
nd
3
rd
4
th
p-value 1/
Mean of Liquidity to customer and short-term
funding ratio 11.93 18.91 25.38 46.87
Loan Loss Reserves to Gross Loans 3.47 3.11 3.14 3.16 -0.32 0.46
Net Interest Margin 9.73 8.71 8.53 8.09 -1.64 0.26
Capitalization (equity to asset ratio) 13.69 11.52 11.88 15.55 1.86 0.23
Bank Size (log of total assets) 12.87 13.00 13.02 12.36 -0.51 0.03
Foreign ownership dummy (=1 if foreign bank) 0.41 0.38 0.53 0.43 0.02 0.74
Private ownership dummy (=1 if private bank) 0.88 0.89 0.94 0.94 0.07 0.07
Real GDP growth 3.24 4.40 3.50 4.04 0.80 0.32
Net International Reserves 2946.79 3443.16 2947.53 2515.13 -431.67 0.00
CPI volatility 2.27 2.10 2.01 1.87 -0.40 0.01
Interest rate spread 9.41 8.32 8.12 7.28 -2.13 0.00
Deposit volatility 3.91 4.06 3.53 3.94 0.03 0.91
Deposit dollarization 37.56 44.21 52.39 64.93 27.36 0.00
Banking crisis dummy (=1 if country had banking
crisis after 1970) 0.53 0.55 0.72 0.88 0.35 0.00
Financial safety net (categorical variable, 1-4;
4=comprehensive safety net) 3.08 2.63 2.66 2.50 -0.58 0.00
Credit to GPD (%) 43.86 44.52 48.86 54.81 10.94 0.00
Quality of credit institutions index (0=worst;
100=best) 66.27 58.46 71.75 78.44 12.17 0.00
Memorandum item
Number of observations 120 109 112 107
Source: Authors' calculations
Table 3. Dependent Variables Means by Liquidity Ratio Quartiles
1/ P-value from a test of statistical difference of the means of the 4
th
quartile versus the 1
st
quartile.
Liquidity ratio quartile
(4
th
- 1
st
quartile)
19
C. Empirical Analysis
Baseline Specification
In line with the discussion in the previous section and similar to other studies (Aspach et al.,
2005, Barajas et al., 2010, and Dinger, 2009), we estimate the determinants of the liquidity
buffers based on bank characteristics, macroeconomic fundamentals and country specific-
characteristics.
The baseline specification can be represented by equation (1):
I
ìt
= [
0
+ [
1
I
ì],t-1
+ [
2
bonk
ì]t
+ [
3
mocro
]t
+ [
4
country
]t
+ p - ] + v - t + ç
ì]t
(1)
Where the subscripts i, j and t refer to bank, country and time (year) respectively. L
represents bank level liquidity buffers. We include a lagged dependent variable: if, as
predicted by theory, banks target an optimum level of liquidity holdings, then we should
expect these holdings to be persistent over time, as shown by Opler et al. (1999) in the case
of U.S. firms. Bank denotes variables measuring bank fundamentals and are derived from
the balance sheets of banks. Macro represents the macroeconomic determinants of individual
banks’ liquidity buffers such as real GDP growth and interest rates, and country are
observable country level characteristics, including, for ease of presentation, the moral hazard
and safety net variables presented in the previous section and Table 2. Unobservable country
and time effects are captured by country (j) and time (t) dummy variables.
Hypotheses of interest
Based on our review of the theoretical and empirical literature as well as stylized facts on
liquidity data for CAPDR countries, we will pay particular attention to the following:
(i) Does ownership matter? We test separately for the effect of private vs. public
ownership, and domestic vs. foreign. As discussed in Section III, ownership may not only on
exert a direct influence on liquidity holdings, as discussed in section of ownership on
liquidity holdings, but may also affect the regression slope through interactions with other
explanatory variables. To test this hypothesis, we interact the relevant ownership dummy
variable (own
ijt
) with the other explanatory variables as shown in Equation (2):
I
ìt
= [
0
+ [
1
I
ì],t-1
+ [
2
bonk
ì]t
+ [
3
mocro
]t
+ [
4
country
]t
+ ([
5
bonk
ì]t
-
own
ì]t
) + ([
6
mocro
]t
- own
ì]t
)+([
7
country
]t
- own
ìt
) + p - ] + v - t + ç
ì]t
(2)
(ii) We use the same framework to test whether liquidity buffers (banks perceived need
for self-insurance) are higher in countries with more dollarized banking systems, as
measured by the share of foreign currency deposits in total deposits. We also test whether
liquidity buffers might be higher in countries with less comprehensive financial safety nets.
20
Estimation Methodology
Equations (1) and (2) are first estimated by Ordinary Least Squares (OLS) using the Least
Squares Dummy Variable approach. This methodology enables us to introduce and identify
bank, country and time effects on bank level liquidity. However, there may also be
unobserved bank-specific and/or country specific time-invariant heterogeneity, which could
bias our estimates if not properly accounted for. The error term may contain time varying
bank or country-specific characteristics which may be correlated with banks’ liquidity ratios.
Another issue is potential endogeneity of some of the explanatory variables such as credit to
GDP.
To address these concerns, we also estimate equations (1) and (2) using the Generalized
Methods of Moments (GMM) developed by Blundell and Bond (2000) and Bond (2002).
GMM estimators are particularly appropriate to address the dynamic panel bias that arises in
the presence of lagged dependent variables in samples with a large number of groups (N) and
a relatively small number of time periods (T), such as ours. Given persistent liquidity ratios,
our preferred estimator is the Systems GMM as it helps overcome the weak instrument
problem (past changes do contain information about current levels), and results in
improvements in the efficiency of the estimates (Arellano and Bond, 1991, Roodman,
2006).
16
D. Results
Tables 4a and 4b present estimation results of a parsimonious robust specification of
equations 1 and 2 above, using the ratio of liquid assets to customer and short-term funding
as a dependent variable.
17
Table 4a includes the credit-to-GDP ratio as a proxy for financial
depth, while Table 4b presents estimation results including the index of the quality of credit
institutions.
16
This was implemented in STATA using Roodman’s (2006) xtabond2 routine. To avoid instrument
proliferation, we restrict the number of lags for the GMM instruments to 2 (Roodman, 2009). We treat the bank
size, country and year dummy variables as predetermined and the rest as endogenous. In addition to OLS
(whose estimate of the lagged dependent variable coefficient is biased upward) we also estimated the model
with robust fixed effects (with the coefficient on the lagged dependent variable is biased downward). Results of
the fixed effects estimations are in line with those from OLS and GMM estimates (shown for the baseline
specification in Appendix IV, Table 2).
17
Given the limited time span of our panel, the coefficients on the macroeconomic variables (real GDP growth,
interest rate spread) were overall consistent with predictions but not significant nor very robust, as part of the
effect of these variables on liquidity buffers was likely captured by the country and time dummies. GMM
estimation of the full model also became difficult as the number of instruments was becoming too large relative
to available observations.
21
Baseline specification
Estimation results from the baseline specification are very robust to the choice of financial
development variable (Tables 4a and 4b, columns 1 and 2). They show that liquidity buffers
in CAPDR are persistent: the coefficient on the lagged dependent variable is positive and
significant. This is consistent with the view that banks target an optimal or desired level of
precautionary liquidity holdings, but could also be attributed to the presence of structural
obstacles to credit that lead banks to hold higher liquidity buffers.
Liquidity ratios are related to bank size, though with non-linearities: liquidity holdings
increase with bank size, but there is a point at which bank size begins exhibiting a marginal
decreasing effect on liquidity. This is the opposite of what is found by Dinger (2009) in
Eastern Europe, and may be explained by differences in the distribution of bank size in both
regions. In CAPDR, the distribution of banks is highly skewed with quite high concentration
of assets in a few large banks, as indicated in Table 1.
18
Liquidity holdings are also negatively related to the loan-loss reserve ratio, indicating that
banks with higher savings against potential losses or riskier loan portfolios also tend to have
lower liquidity buffers in CAPDR. They are negatively associated (though the relationship is
not as robust as for the previous two variables) with the net interest margin, as expected. The
coefficient on capitalization is negative and significant in the baseline, indicating that better
capitalized banks would tend to hold less liquidity (the coefficient remains negative but is
generally no longer significant in the specifications with interaction terms). This is somewhat
counterintuitive, as the expectation would be that better capitalized banks would also hold
more liquidity buffers, if higher capitalization is indicative of a prudent business model. In
Table 4a, the credit-to-GDP ratio is negatively related to liquidity buffers, in line with
predictions (though the coefficient is not significant). In table 4b, the coefficient on the credit
institutions variable is positive and significant in the GMM regression.
19
Specifications with interaction terms—the role of bank ownership
Results indicate that ownership has some effect on liquidity holdings, though mostly through
the interaction terms. Our results do not show any significant evidence that private ownership
does affect liquidity buffers, though the coefficient on private ownership is positive in the
GMM specification (consistent with Table 3). Foreign banks tend to hold less liquidity, but
the coefficient on ownership is not statistically significant either. Foreign banks with riskier
loan portfolios or which are more conservative regarding expected loan losses do tend to
18
In estimations without the quadratic term the coefficient on bank size is negative and robust across
specifications as expected from theory and found in related empirical studies (results available upon request).
19
Estimating a more comprehensive model including all macroeconomic fundamentals presented in Table 2
indicate that the results on bank characteristics and variables of interest remain robust. However, as indicated
before the macroeconomic variables are not very precisely estimated in our short panel and although
coefficients on these variables are overall in line with predictions they are neither always consistent nor
significant.
22
have higher liquidity buffers (Table 4a, column 6). This is consistent with predictions and
findings in Detragiache, Tressel and Gupta (2008), which show that foreign banks tend to be
more prudent and lend to less risky customers.
Specifications with interaction terms—deposit dollarization
As indicated in columns 7–8 of Tables 4a and 4b, deposit dollarization is robustly and
significantly associated with higher liquidity buffers, both by itself (in both tables) and
through interactions with other bank-level variables (Table 4b). The individual effect is quite
large: a one standard deviation (34 percent) increase in deposit dollarization leads to a 70 to
100 percent increase in the liquidity to deposit ratio.
20
The strong positive association
between deposit dollarization and liquidity buffers may however not necessarily indicate a
direct causal relationship. The same factors that cause households and firms to hold more
dollar deposits could very well also lead banks to hold more precautionary liquidity.
21
Nonetheless, the positive relationship between dollarization and high liquidity holdings
would provide yet another reason why the monetary transmission mechanism is slower in
more dollarized economies (as in Medina Cas, Carrion-Menendez and Frantischek, 2011).
The interaction with the loan-loss reserve ratio also indicates that prudent banks or banks
with risky loan portfolios in dollarized economies tend to hold more liquidity (though the
coefficient is not significant in the GMM specification). More profitable banks in dollarized
economies tend to hold less liquidity. The relation between dollarization and the quality of
credit institutions is interesting: whereas there is a positive and significant association
between the quality of credit institutions and liquidity buffers, the sign of the coefficient
switches when institutions are interacted with the degree of deposit dollarization. As
mentioned above, this negative association may indicate either that high dollarization itself is
a result of lower quality of domestic credit institutions, or that sound credit institutions in
dollarized economies help reduce bank liquidity.
Specifications with interaction terms—financial safety net
Columns 9 and 10 of Tables 4a and 4b show the interactions with the financial safety net
variable. Though not significant, the coefficient on the GMM estimation is negative,
indicating that banks in countries with a more comprehensive financial safety net tend to hold
20
Given that reserve requirements are set at the same rate for local and foreign currency deposits in most
countries, and that actual liquidity holdings are held above requirements, it is unlikely that this result is driven
mechanically by reserve requirements. However, the large standard deviation is in part due to the fact that the
share of foreign deposits in total is 100 percent in El Salvador and Panama.
21
De Nicoló, Honahan and Ize (2005) find in a large cross-country sample that the credibility of
macroeconomic policy and the quality of institutions are key determinants of deposit dollarization.
23
2
3
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM
Liquid assets ratio (-1) 0.347*** 0.189*** 0.355*** 0.218* 0.371*** 0.231** 0.336*** 0.223** 0.345*** 0.290***
(0.064) (0.044) (0.065) (0.114) (0.072) (0.099) (0.067) (0.092) (0.064) (0.086)
Bank size 7.401*** 7.994*** 7.327*** 8.545*** 7.635*** 10.381** 8.460*** 5.639** 8.326*** 7.281***
(1.448) (1.875) (1.472) (2.299) (1.609) (4.137) (1.629) (2.635) (1.354) (1.546)
Bank size squared -0.350*** -0.371*** -0.351*** -0.392*** -0.366*** -0.483** -0.401*** -0.244* -0.392*** -0.325***
(0.064) (0.092) (0.065) (0.126) (0.073) (0.203) (0.072) (0.129) (0.061) (0.073)
Capitalization -0.355*** -0.321** -0.318*** -0.505 -0.288*** -0.316 -0.219 -0.017 -0.375 -0.238
(0.060) (0.123) (0.043) (0.336) (0.058) (0.305) (0.220) (0.542) (0.299) (0.928)
Net interest margin -0.156** -0.123 0.718 -0.089 -0.278* -0.593 -0.210 0.404 0.577 1.278*
(0.064) (0.076) (0.569) (1.067) (0.164) (1.199) (0.133) (0.331) (0.359) (0.690)
Loan-loss reserve ratio -0.224 -0.282 -0.505** -0.035 -0.419*** -0.550 -1.252*** -0.799 1.738 0.780
(0.160) (0.252) (0.216) (0.588) (0.160) (0.506) (0.394) (0.624) (1.092) (1.743)
Credit to GDP ratio -0.213 -0.323 -0.337 0.404 -0.277 -0.441 -0.374 -0.041 0.342 -0.593
(0.272) (0.292) (0.283) (0.664) (0.275) (0.344) (0.756) (0.679) (0.717) (0.729)
Variable -2.623 42.500 -3.002 -13.249 1.068** 1.491*** 7.345 -32.043
(6.514) (36.406) (6.185) (18.512) (0.516) (0.470) (7.880) (33.069)
Capitalization*variable -0.105 0.077 -0.332 0.161 -0.003 -0.001 0.022 -0.001
(0.128) (0.616) (0.240) (0.647) (0.004) (0.014) (0.142) (0.301)
Net interest margin*variable -0.868 0.036 0.083 0.309 0.001 -0.022 -0.224** -0.401*
(0.576) (1.128) (0.180) (1.058) (0.005) (0.015) (0.112) (0.218)
Loan-loss reserve ratio*variable 0.346 -0.077 1.585* 2.858** 0.043*** 0.027 -0.613* -0.259
(0.347) (0.827) (0.817) (1.291) (0.015) (0.022) (0.344) (0.568)
Credit to GDP ratio*variable 0.190 -1.283 0.063 0.169 -0.004 -0.012 -0.257 0.199
(0.119) (0.880) (0.064) (0.282) (0.012) (0.009) (0.285) (0.297)
Observations 321 321 321 321 321 321 321 321 321 321
R-squared 0.55 0.55 0.57 0.58 0.57
No. of groups 96 96 96 96 96
No. of instruments . 64 54 54 64 54
Hansen test p-value 0.348 0.192 0.132 0.232 0.249
A-B AR(2) test 1.283 1.027 1.040 1.562 1.283
A-B AR(2) test p-value 0.199 0.305 0.298 0.118 0.200
Source: Authors' calculations.
Dependent variable is the ratio of total liquid assets to customer deposits and short-term funding. GMM is two-step system GMM estimator with Windmeijer standard error correction.
Columns (3) through (10 test the hypotheses that ownership (foreign/domestic and public/private), degree of dollarization and coverage of the financial safety net affect banks' liquidity buffers
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the country level. Safety net
is an index variable categorizing the existence of each of the 4 components of a safety net in each country (Appendix II. Table 1).
All regressions include time and country dummies. Constant estimated but not reported.
Table 4a. Determinants of Banks' Liquidity Buffers in CAPDR-Financial Depth
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Dependent variable is the ratio of
total liquid assets to customer
deposits and short-term funding
Variable = Private ownership Foreign ownership Dollarization Baseline Safety net
24
2
4
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM Pooled OLS GMM
Liquid assets ratio (-1) 0.346*** 0.180*** 0.355*** 0.350** 0.377*** 0.298** 0.342*** 0.280** 0.346*** 0.267**
(0.064) (0.047) (0.065) (0.134) (0.069) (0.148) (0.067) (0.114) (0.064) (0.107)
Bank size 7.441*** 8.027*** 7.301*** 5.349** 7.142*** 7.070** 8.318*** 6.768*** 8.287*** 7.904***
(1.486) (2.059) (1.536) (2.099) (1.539) (2.752) (1.550) (2.548) (1.381) (1.661)
Bank size squared -0.352*** -0.384*** -0.348*** -0.238** -0.343*** -0.325** -0.395*** -0.306** -0.389*** -0.354***
(0.065) (0.096) (0.068) (0.108) (0.068) (0.137) (0.069) (0.127) (0.061) (0.084)
Capitalization -0.356*** -0.336*** -0.316*** -0.423 -0.283*** -0.180 -0.217 -0.026 -0.390 -0.461
(0.061) (0.106) (0.050) (0.424) (0.061) (0.393) (0.220) (0.537) (0.300) (0.956)
Net interest margin -0.160** -0.141* 1.115** 1.945 -0.235 -0.546 -0.191 0.489 0.573 1.293*
(0.063) (0.083) (0.552) (1.851) (0.158) (0.968) (0.132) (0.317) (0.349) (0.678)
Loan-loss reserve ratio -0.211 -0.299 -0.292 0.415 -0.351** -0.252 -1.292*** -1.236 1.879* 1.283
(0.159) (0.300) (0.178) (0.639) (0.153) (0.526) (0.404) (1.033) (1.054) (2.196)
Credit Institutions 1/ 0.051 0.053* 0.113 0.174 0.047 0.001 0.188** 0.153* 0.685*** 0.788**
(0.035) (0.028) (0.073) (0.161) (0.035) (0.051) (0.089) (0.086) (0.209) (0.355)
Variable 11.918* 27.989 0.579 6.271 0.774** 0.718** 17.220*** 33.601*
(6.497) (28.631) (4.080) (18.141) (0.304) (0.303) (2.981) (19.225)
Capitalization*variable -0.109 -0.054 -0.367 -0.034 -0.003 -0.005 0.033 0.120
(0.129) (0.849) (0.236) (0.716) (0.004) (0.013) (0.141) (0.320)
Net interest margin*variable -1.269** -1.987 0.032 0.205 0.000 -0.025* -0.225** -0.419*
(0.558) (1.862) (0.173) (0.927) (0.005) (0.015) (0.109) (0.211)
Loan-loss reserve ratio*variable 0.132 -0.407 1.296* 1.910 0.047*** 0.050 -0.640* -0.419
(0.298) (0.764) (0.682) (1.466) (0.015) (0.038) (0.332) (0.609)
Credit Institutions*variable 1/ -0.061 -0.148 0.018 -0.074 -0.007** -0.006** -0.186*** -0.212**
(0.066) (0.156) (0.039) (0.118) (0.003) (0.002) (0.061) (0.087)
Observations 321 321 321 321 321 321 321 321 321 321
R-squared 0.55 0.55 0.57 0.59 0.57
No. of groups 96 96 96 96 96
No. of instruments . 64 54 54 64 54
Hansen test p-value 0.338 0.0836 0.134 0.125 0.290
A-B AR(2) test 1.470 1.481 0.914 1.115 1.264
A-B AR(2) test p-value 0.141 0.139 0.360 0.265 0.206
Source: Authors' calculations.
Dependent variable is the ratio of total liquid assets to customer deposits and short-term funding. GMM is two-step system GMM estimator with Windmeijer standard error correction.
Columns (3) through (10 test the hypotheses that ownership (foreign/domestic and public/private), degree of dollarization and coverage of the financial safety net affect banks' liquidity buffers
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the country level. Safety net
is an index variable categorizing the existence of each of the 4 components of a safety net in each country (Appendix II. Table 1).
All regressions include time and country dummies. Constant estimated but not reported.
1/ Credit institutions are proxied by the principal components analysis of the Doing Business variables measuring the quality of credit institutions.
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Table 4b. Determinants of Banks' Liquidity Buffers in CAPDR-Credit Institutions
Dependent variable is the ratio of
total liquid assets to customer
deposits and short-term funding
Baseline Variable = Private ownership Foreign ownership Dollarization Safety net
25
less liquidity. The interaction between the safety net and net interest margin is negative,
indicating as mentioned above that profitable banks in countries with more comprehensive
safety nets do hold less liquidity. This is in contrast with the coefficient on the interest
margin (without interactions), which is significant and positive in column 10 of Table 4a.
Compared with the baseline, it seems that the absence of a safety net would lead even
profitable banks to hold an extra liquidity cushion. Table 4b, columns 9-10 show estimation
results with the financial safety net variable and interactions, including the quality of credit
institutions variable. As in the case of deposit dollarization, whereas the sign of the
coefficients for credit institutions and safety net variables are now positive and significant,
the coefficient on the interaction term between safety net and credit institutions is negative:
when the safety net is less comprehensive and the quality of institutions is low, banks would
hold more precautionary liquidity buffers. Further, as in Table 4a, if the safety net is
comprehensive, more profitable banks or banks with higher loan loss reserves tend to hold
less liquidity.
Robustness checks
As a main robustness check, we estimate our model using the ratio of liquid assets to total
assets as our dependent variable (Appendix IV. Table 1). Overall results are broadly
consistent with the findings presented in Tables 4a and 4b, at least in terms of signs of
coefficients. The coefficient on the lagged dependent variable is about twice as large, and the
coefficient of the dollarization variable remains significant and relatively large.
Appendix IV, Table 2 presents further robustness checks. These include running the fixed
effects regressions as discussed in footnote 16 above (column 3), and looking into the
interactions of foreign ownership, dollarization and safety net only for the private banks of
the sample (columns 4-9). One caveat is that limiting the number of observations increases
the risk of overfitting the model due to too many instruments. Nonetheless, the Hansen
statistic’s p-value remains reasonable for all specifications.
These additional regressions support our main findings. The relative size of the coefficient on
the lagged dependent variable in the pooled OLS, fixed effects and GMM is consistent with
expectations: in OLS this coefficient is correlated with the error term and biased upward,
while in the fixed effects specification it is the opposite. Good estimates of the true parameter
should lie in between or near these values, which is the case here (see column 2 of Appendix
IV. Table 2). Previous results on ownership, dollarization and safety net hold in the private
banks sample, particularly as regards the role of credit institutions in highly dollarized
economies or economies with a less comprehensive safety net.
Our results still need to be considered against the caveat of data limitations. The uneven
coverage of individual countries’ banking systems and short estimation time frame may
affect the coefficient estimates from the regressions. Nevertheless, some useful policy
lessons already emerge from our analysis. These are discussed in the following section.
26
V. CONCLUSIONS AND POLICY LESSONS
Our study of liquidity buffers in CAPDR finds that they are comfortably above legal and
prudential requirements. With average liquidity ratios of about 25 percent of deposits, banks
in the region have handled and are able to handle historic deposit volatility levels outside of
crisis episodes (Table 5). Therefore, the adoption of the new Basel III liquidity requirements
in the region should not have much impact on banks’ balance sheet.
23
Indeed, Basso, Delgado
and Meza (2012) find that liquidity holdings in the region already meet or exceed the new
Basel III ratios.
A closer look at the reasons for which banks would want to hold liquidity buffers above legal
or prudential requirements indicates that CAPDR banks appear guided at least in part by
rational precautionary motives. As found in other countries or regions, bank characteristics
that influence their ability to raise additional funding on demand play an important role:
smaller, lower-capitalized, less efficient and less profitable banks tend to hold higher
liquidity buffers. Foreign banks tend to hold less liquidity, possibly because they have access
to emergency lines from headquarters. Surprisingly, banks with riskier loan portfolios also
hold less liquidity overall, though this is not the case for foreign banks and banks in highly
dollarized economies.
A first policy lesson stemming from these results would be to continue with ongoing efforts
to strengthen financial sector supervision, enhance financial safety nets and develop financial
markets. Greater confidence in the system and more opportunities for investment and
23
The Basel III liquidity requirements are to be in effect in 2015. The Basel Committee on Banking Supervision
has defined two minimum standards for funding liquidity: (i) the liquidity coverage ratio, which aims at
promoting short-term resilience of a bank’s liquidity profile by ensuring that it has sufficient high-quality liquid
assets (cash or cash-equivalent) to survive a significant stress scenario lasting for one month; and (ii) the net
stable funding ratio, in order to promote resilience over a longer time period by matching long-term assets with
stable funding sources over a one-year horizon (BIS, 2010).
1 3 6 1 3 6 1 3 6 1 3 6 1 3 6
Costa Rica 3.3 4.9 4.7 5.6 6.8 5.6 7.4 9.9 12.1 6.7 9.6 15.2 3.8 5.4 5.2
Dominican Republic 34.4 23.7 13.5 54.5 40.1 27.6 30.0 20.0 10.6 19.9 30.6 35.0 45.5 31.7 22.5
El Salvador 2.7 3.0 3.9 17.8 17.4 17.8 2.2 3.2 4.1 2.7 3.0 3.9
Guatemala 3.1 1.8 0.6 6.2 6.7 5.6 5.3 5.2 3.1 5.6 4.1 1.1 3.5 3.0 1.6
Honduras 3.0 3.8 3.3 10.3 8.8 10.2 2.3 3.2 3.7 7.4 9.4 10.0 0.3 5.3 5.4
Nicaragua 1.7 4.2 3.0 10.8 11.2 11.4 2.9 4.9 3.1 3.4 5.1 6.1 6.6 4.2 4.5
Panama 3.7 6.3 7.7 8.9 8.1 8.7 3.1 2.0 2.1 3.7 6.3 7.7
Total 1.7 1.8 0.6 5.6 6.7 5.6 2.2 2.0 2.1 3.4 4.1 1.1 0.3 3.0 1.6
Source: CAPDR Central Banks and superintendencies’ websites and Fund staff calculations.
Notes: Largest decline in bank deposits at the one, three and six months horizon, between December 2001 and September 2011.
Foreign Currency Local Currency
Table 5. CAPDR: Largest deposit drop (December 2001-September 2011)
(Percent)
Deposits
Total Demand Time & Savings
no. of months no. of months no. of months no. of months no. of months
27
intermediation (through stronger credit institutions) could help lower banks’ precautionary
liquidity buffers without compromising financial stability.
Strengthened supervision would help address the issue of the negative relationship between
the loan-loss ratio and liquidity buffers, which may indicate that domestic banks, in contrast
with foreign banks which are likely subject to strict internal guidelines, may not fully
internalize the costs of riskier lending practices. As mentioned, further progress in risk-based
supervision would be especially warranted: in spite of notable progress, CAPDR countries
still do not meet minimum international standards and lag behind larger South American
countries.
Another important lesson relates to the dollarization of CAPDR economies and banking
systems and calls for strengthening the credibility of macroeconomic policy and institutions,
as well as the coverage of financial safety nets. Our findings show that, in our sample, banks’
precautionary demand for liquidity is associated to the degree of deposit dollarization, and
the safety net, in each country. Given the lack of dollar LOLR in all countries and the lack of
a LOLR in the two fully dollarized economies, our findings suggest that continuing with
ongoing efforts to strengthen financial safety nets would be efficient. El Salvador has
approved legislation to provide emergency liquidity support to banks, and Panama is
considering establishing a similar facility.
Furthermore, maintaining higher liquidity buffers because of dollarization also has negative
implications for the development of financial markets, and for the adequate functioning of
the monetary policy transmission mechanism. For the countries in the region that aim at
transitioning to inflation targeting, tackling the root causes of deposit dollarization should be
an important part of their strategy. Studies on deposit dollarization find that it is associated
with lower credibility of macroeconomic policy, high inflation, and poor quality of
institutions; and also find that financial instability is higher in dollarized economies (De
Nicoló, Honohan and Ize, 2005, Gulde et al., 2004). Arguably, most CAPDR countries have
been successful over the past decade at macroeconomic stabilization, lowering inflation and
reducing output volatility (except during the 2008–09 global crisis). Nonetheless, financial
dollarization tends to persist even when macroeconomic stability is restored and inflation
relatively low if institutions are still perceived to be weak.
24
Even in tightly managed
exchange rate regimes, partially dollarized economies are subject to higher liquidity risks,
and our results tend to indicate that, at least to some extent, CAPDR banks do internalize
these risks.
With causality likely running both from policies to dollarization and back, measures that
would help create a “virtuous cycle” of de-dollarization and lower precautionary liquidity
holdings could be informed by the experience of de-dollarization in South America. In
particular, in a study of financial de-dollarization in Bolivia, Paraguay, Peru and Uruguay,
24
See Reinhart, Rogoff and Savastano (2003), Galindo and Leiderman (2005), and Erasmus et al. (2009).
28
Garcia-Escribano and Sosa (2011) find that successful, market-driven de-dollarization was
associated with (i) stronger macroeconomic policies and institutions, credibly and
consistently implemented over time, (ii) active management of reserve requirement
differentials and the introduction of other prudential measures; and (iii) domestic currency
capital market development. As discussed in this paper, there is ample room for more active
liquidity management on the part of the CAPDR monetary and prudential authorities. Finally,
measures to develop local currency capital markets, starting with public domestic debt
markets, would enhance financial systems’ efficiency, diversify sources of funding and
investment opportunities.
29
REFERENCES
Agénor, Pierre-Richard, Joshua Aizenman, and Alexander W. Hoffmaister, 2004, “The Credit
Crunch in East Asia: What Can Bank Excess Liquid Assets Tell Us?” Journal of International
Money and Finance, Vol. 23, pp. 27–49.
Almeida, Heitor, Murrillo Campello, and Michael S. Weisbach, 2004, “The Cash Flow Sensitivity of
Cash,” The Journal of Finance, Vol. 59, No. 4, pp. 1777–1804.
Arellano, Manuel, and Stephen Bond, 1991, “Some Tests of Specification for Panel Data: Monte
Carlo Evidence and an Application to Employment Equations”, Review of Economic Studies
Vol. 58, pp. 277–97.
Aspachs, Oriol, Erlend Nier, and Muriel Tiesset, 2005, “Liquidity, Banking Regulation and the
Macroeconomy,” Mimeo (London: London School of Economics).
Baltensperger, Ernst (1980). “Alternative Approaches to the Theory of the Banking Firm,”Journal of
Monetary Economics, Vol. 6, pp.1–37.
Barajas, Adolfo, Ralph Chami, Raphael Espinoza, and Heiko Hesse, 2010, “Recent Credit stagnation
in the MENA Region: What to Expect? What Can be done?”, IMF Working Paper 10/219
(Washington: International Monetary Fund).
Basel III: International Framework for Liquidity Risk Measurement, Standards and Monitory, 2010,
(Switzerland: Bank for International Settlements).
Basso, Oscar, Fernando L. Delgado and Mynor Meza, 2012, “Strengthening Bank Capital and
Liquidity in Central America: the Road to Basel III”, mimeo.
Bolzico, Javier, Eugenia Gozzi and Florencia Rossini, 2010, “Redes de Seguridad
Financiera,” Fit & Proper, Available via the Internet: http://www.fitproper.com/
Cárdenas, Mauricio and Camila Henao, 2010, “Latin America and the Caribbean Economic
Recovery”, The Brookings Institution, Available via the Internet:
http://www.brookings.edu/~/media/research/files/articles/2010/7/latin%20america%20econo
my%20cardenas/07_latin_america_economy_cardenas
Clarke, George R. G., Robert Cull, and María Soledad Martínez Pería, 2006, “Foreign Bank
Participation and Access to Credit Across Firms in Developing Countries,” Journal of
Comparative Economics, Vol. 34, pp. 774–95.
Delgado, Fernando and Mynor Meza, 2011, “Developments in Financial Supervision and the Use of
Macroprudential Measures in Central America”, IMF Working Paper 11/299 (Washington:
International Monetary Fund).
30
Detragiache, Enrica, Thierry Tressel, and Poonam Gupta, 2008, “Foreign Banks in Poor Countries:
Theory and Evidence,” The Journal of Finance, Vol. 63, No5 (October), pp. 2123–60.
Diamond, Douglas W., and Raghuram G. Rajan, 1999, “Liquidity Risk, Liquidity Creation and
Financial Fragility: A theory of Banking,” NBER Working Paper 7430 (Cambridge: National
Bureau of Economic Research)
_______, and Raghuram G. Rajan, 2005, “Liquidity Shortages and Banking Crises,” Journal of
Finance Vol. 30, No. 2, pp. 615–47.
Diamond, Douglas W., and Philip H. Dybvig, 1983, “Bank Runs, Deposit Insurance, and
Liquidity,” Journal of Political Economy, Vol. 91, No. 3, pp. 401–19.
De Nicoló, Gianni, Patrick Honohan, and Alain Ize, 2004, “Dollarization of Bank Deposits: Causes
and Consequences,” Journal of Banking & Finance, Vol. 29, pp. 1697–1727.
Dinger, Valeriya, 2009, “Do Foreign-owned Banks Affect Banking System Liquidity Risk?” Journal
of Comparative Economics, Vol. 37, pp. 647–57.
Douglas W. Diamond & Raghuram G. Rajan, 2001, “Liquidity Risk, Liquidity Creation, and
Financial Fragility: A Theory of Banking,” Journal of Political Economy, University of
Chicago Press, Vol. 109, No. 2, pp. 287–327.
Freixas, Xavier, and Cornelia Holthasusen, 2004, “Interbank Market Integration under Asymmetric
Information,” The Review of Financial Studies”,Vol. 18, No. 2. pp. 459–90.
______, and Leonardo Leiderman, 2005, “Living With Dollarization and the Route to
Dedollarization,” IDB Working Paper 526 (Washington: Inter-American Development
Bank).
Galindo, Arturo J., Alejandro Izquierdo, and Liliana Rojas-Suárez, 2010, “Financial Integration and
Foreign Banks in Latin America,” IDB Working Paper 116 (Washington: Inter-American
Development Bank)
García-Escribano, Mercedes, and Sebastián Sosa, 2011, “What is Driving Financial De- dollarization
in Latin America?” IMF Working Paper 11/10 (Washington: International Monetary Fund).
Gelos, R. Gaston, and Jorge Roldós, 2004, “Consolidation and Market Structure in Emerging Market
Banking Systems,” Emerging Markets Review, Elsevier, Vol. 5, No. 1 (March), pp. 39–
59.
Gonzalez-Eiras, Martín, 2003, “Banks’ Liquidity Demand in the Presence of a Lender of Last
Resort,” Documento de Trabajo 61 (Buenos Aires: Universidad de San Andrés).
31
Gray, Simon, 2011, “Central Bank Balances and Reserve Requirements,” IMF Working Paper 11/36
(Washington: International Monetary Fund).
Guerrero, Rosa Matilde, Kurt Focke, and Florencia Rossini, 2010, “Redes de Seguridad Financiera,”
IDB Working Paper No. 121 (Washington: Inter-American Development Bank).
Gulde, Anne-Marie, David Hoelscher, Alain Ize, David Marston, and Gianni De Nicoló, 2004,
“Financial Stability in Dollarized Economies,” IMF Occasional Paper 230 (Washington:
International Monetary Fund).
Holmström, Bengt, and Jean Tirole, 1998, “Private and Public Supply of Liquidity,” Journal of
Political Economy Vol. 106, No. 1, pp. 1–40.
Ize, Alan, Miguel A. Kiguel, and Eduardo Levy Yeyati, 2005, “Managing Systemic Liquidity Risk
in Financially Dollarized Economies,” IMF Working Paper 05/188 (Washington:
International Monetary Fund).
Kashyap, Anil K., and Jeremy C. Stein, 1995, “The Impact of Monetary Policy on Bank Balance
Sheets,” Carnegie-Rochester Conference Series on Public Policy, Vol. 42, pp. 151–95.
______, Raghuram Rajan, and Jeremy C. Stein, 2002, “Banks and Liquidity Providers: An
Explanation for the Coexistence of Lending and Deposit-Taking,” The Journal of Finance
Vol. 57, No. 1 (February), pp. 33–73.
Kiyotaki, Nobuhiro and John Moore, 2008, “Liquidity, Business Cycles and Monetary Policy,”
Technical Paper for Clarendon Lecture 2.
Laeven, Luc, and Fabián Valencia, 2012, “Systemic Banking Crises Database: An Update,” IMF
Working Paper 12/163 (Washington: International Monetary Fund).
Levy-Yeyati, Eduardo, Christian Broda, 2002, “Dollarization and Lender of Last Resort,”
(Massachussets Institute of Technology; Universidad Torcuato De Tella).
Medina Cas, Stephanie, Alejandro Carrión-Menéndez, and Florencia Frantischek, 2011, “The Policy
Interest-Rate Pass-Through in Central America,” IMF Working Paper 11/240 (Washington:
International Monetary Fund).
______, Alejandro Carrión-Menéndez, and Florencia Frantischek, 2011, “Improving the Monetary
Policy Frameworks in Central America,” IMF Working Paper 11/245 (Washington:
International Monetary Fund).
Micco, Alejandro, Ugo Panizza and Mónica Yañez, 2004, “Bank Ownership and Performance,”
IDB Working Paper No. 518 (Washington: Inter-American Development Bank).
32
_____, and Ugo Panizza, 2006, “Bank Ownership and Lending Behavior,” Economic Letters,
Elsevier, Vol. 93, No. 2 (November), pp. 248–54.
Opler, Tim, Lee Pinkowitz, René Stulz, and Rohan Williamson, “The Determinants and Implications
of Corporate Cash Holdings,” RBER Working Paper 6234 (Cambridge National Bureau of
Economic Research).
Reinhart, Carmen M., Kenneth S. Rogoff, and Miguel, 2003, “Addicted to Dollars,” NBER Working
Paper 10015 (Cambridge: National Bureau of Economic Research).
Repullo, Rafael, 2005, “Liquidity, Risk-Taking and The Lender of Last Resort,” CEMFI Working
Paper No. 0504 (Madrid: Centro de Estudios Monetarios y Financierios).
Rochet, Jean-Charles, and Xavier Vives, 2004, “Coordination Failures and the Lender of Last
Resort: Was Bagehot Right After All?” Journal of the European Economic Association Vol.2
No. 6, pp. 1116–47.
Roodman, David, 2009, “Practitioner’s Corner: A Note on the Theme of Too Many Instruments”,
Oxford Bulletin of Economics and Statistics, Vol. 71, No.1, pp. 135-58.
____, 2006, “How to Do xtabond2: An Introduction to ‘Difference’ and ‘System’ GMM in Stata,”
CGD Working Paper No. 103 (Washington: Center for Global Development).
Santomero, Anthony M., 1984, “Modeling the Banking Firm”, Journal of Money, Credit, and
Banking, Part 2, Vol. 16, pp. 576–602.
Saxegaard, Magnus, 2006, “Excess Liquidity and Effectiveness of Monetary Policy: Evidence from
Sub-Saharan Africa,” IMF Working Paper 06/115 (Washington International Monetary
Fund).
Shah, Hemant, Andreas A. Jobst, Laura Valderrama-Ferrando, Ivan Guerra, 2007, “Public Debt
Markets in Central America, Panama, and the Dominican Republic,” IMF Working Paper
07/147 (Washington: International Monetary Fund).
____, Ana Carvajal, Geoffrey Bannister, Jorge Cha-Lau, and Ivan Guerra, 2007, “Equity and Private
Debt Markets in Central America, Panama, and the Dominican Republic,” IMF Working
Paper 07/288 (Washington: International Monetary Fund).
33
APPENDIX I. RESERVE AND LIQUIDITY REQUIREMENTS IN CAPDR
Most central banks or supervisory authorities require depository institutions to hold
minimum reserves or liquidity against their deposits.
Required reserves
1
In systems with central banks, there are typically two main reasons to require the holdings of
reserves:
? Prudential: Reserve requirements serve as a safeguard against both liquidity and
solvency risk in a case of a sudden and inordinate demand for withdrawals (as in a
run on a bank) or as a result of a need to make sudden large payments abroad, for
example. In highly dollarized economies, the prudential purpose of the reserve
requirements perhaps carries more prominence as central banks have a very limited
ability to control monetary policy in the environment of an overwhelming presence of
foreign currency.
? Monetary control and liquidity management: Adjusting reserve requirements is a
policy tool for injecting cash (liquidity) into, or withdrawing it from, an economy.
This helps to control domestic liquidity, i.e., commercial bank balance sheet growth.
Likewise, adjusting reserve requirements could influence the spread between deposit
and lending rates, thereby impacting the growth of monetary aggregates and inflation.
Reserve requirements are calculated in reference to commercial banks’ deposits.
2
They can
be mandated on both local and foreign currency deposit liabilities at similar or differentiated
rates, and on demand, savings and time horizons of these deposits.
3
The reserve requirement
is usually defined as a minimum amount of liquid assets, cash or cash-equivalents
4
(computed as a percentage of demand and time deposits), that banks and other depository
institutions (credit unions, insurance companies) are required by law to keep on hand, and
which may not be used for lending or investing. In countries with underdeveloped financial
1
This section is based on Gray (2011).
2
Composition of required reserves depends on individual country requirements. Some countries only allow
deposits with the central bank to count as legal reserves and do not allow averaging (calculating the level of
deposits on average in one month versus a requirement to hold going forward the end of the month balance,
which limits banks ability to manage liquidity).
3
In cases where the authorities are trying to discourage dollarization, higher reserve requirements may be
imposed on foreign liabilities.
4
Liquidity is defined as the ability of an asset, other than cash, to be converted into cash quickly and without
any price discount. In the U.S., e.g., these constitute certificates of deposit (CD), marketable securities,
negotiable financial instruments (such as a cashier's checks), etc., that has a very high degree of convertibility
into cash (liquidity).
34
markets, reserve requirements are usually cash in vault, banks’ deposits with the central bank
or abroad.
In many cases, banks hold liquidity above the required levels at the central bank, especially if
they are remunerated.
5
This liquidity could be truly voluntary, implying that banks’ hold
them for precautionary reasons, and involuntary, reflecting idle balances that banks hold
above their demanded levels.
6
In countries with monetary policies, this genuine voluntary
liquidity in the banking system may undermine both the central bank’s capacity to affect
banks’ cost of funding and its ability to restrain any rapid unwinding of these reserves, and
possible triggering of inflation. This may also induce excessive exchange rate volatility under
an abrupt change in expectations, and ultimately jeopardize financial stability. The presence
of voluntary reserves could also be an indication of rigidities in the system as excess liquidity
creates opportunity costs.
Given the banks’ ability to influence monetary conditions through holdings of excess
reserves, the monetary authorities ought to carefully monitor these levels. For example, they
should ensure that excess liquidity, if needed, could be counteracted by their own provision
of reserves to the system. In cases where reserves are remunerated, changing the
remuneration rate – rather than conducting open market operations - can signal the stance of
the desired monetary policy.
Liquidity requirements
Liquidity requirements are commonly defined as the ratio of liquid assets to liabilities,
although definitions can vary by country. Liquid assets generally include cash in vault,
deposits abroad, and deposits held at the central bank, as well as liquid securities (central
bank paper, quoted government securities). The definition of liquidity requirements differs
sometimes quite substantially across countries, as supervisor include eligible securities based
on the existence of a well-functioning market guaranteeing that they could be converted into
cash quickly and at no or little loss of value.
Application in CAPDR
Reserve and liquidity requirements are relatively homogeneously defined in CAPDR,
although there could be differences in definitions of liquid assets. Reserve requirements in
five countries consist mostly of cash, cash in vault, and central bank and government
5
Level of (voluntary) reserves could be higher if they are remunerated by the central bank. Remuneration of
reduces the cost of having idle resources and reduces incentives for the financial system to avoid liabilities that
require reserve provision.. At the same time, out of 121 central banks surveyed by Gray (2011), 70 percent did
not remunerate reserves; and 20 percent remunerated at below primary rate.
6
Agenor et al (2004).
35
securities. In El Salvador, deposits abroad are also permitted as a proportion of a reserve
requirement.
7
Eligible liabilities on which requirement is calculated are mostly all deposits, although there
are some variations. Some countries limit eligible liabilities to demand and foreign deposits
only (Dominican Republic and Honduras). Many countries exclude savings deposits mostly
because they are of longer maturity and also because they are not large comparing to demand
and time deposits. Reserve requirements in partially dollarized economies tend to be different
for local or foreign currency deposits. In CAPDR these average 15 and 15.5 percent,
respectively. Dominican Republic has the highest reserve requirement on foreign deposits, at
20 percent (Appendix 1. Table 1).
Reserve requirements are stipulated in central bank laws and form a part of a monetary
policy toolkit. Costa Rica and Guatemala have kept ratios constant over the past 5 years,
indicating that the requirement serves mostly a prudential objective rather than a monetary
policy tool (Appendix I.Table 2). In Nicaragua a downward adjustment in reserve
requirement was taken years before the 2008 crisis and the ratio was kept stable during the
crisis. In Dominican Republic there was a downward adjustment in local currency reserve
requirement in 2009, to inject liquidity. Honduras appears to be the only country where there
was an active management of reserves. Only in Nicaragua, total reserves at the central bank
continuously exceed the required amounts.
El Salvador, Honduras and Panama also impose prudential liquidity requirements. In
Honduras, the liquidity requirement is aimed at avoiding maturity mismatches. In El
Salvador, there is a prudential requirement of maintaining at the central bank a liquid reserve
amounting to 3 percent of deposits. For Panama, the liquidity requirement (30 percent)
includes a vast array of instruments as liquid assets, even obligations payable to banks in 186
days, but the numerator only includes short-term deposits, which raises the ratio
significantly, as these constitute only a small share of total deposits.
7
Perhaps, as a relic of the non-officially dollarized economy, El Salvador preserved a liquid asset requirement
of 3 percent of deposits (funds to be deposited with the former central bank or with a reputable bank abroad),
however, it eliminated it during the 2008 global financial crisis in an attempt to free up liquidity.
3
6
Reserve requirements
Costa Rica 15 15 n.a Demand, foreign
currency, time,
interbank,
government.
Excludes
interbank
deposits.
Deposits at the central
bank (only those at the
reserve account) in the
some currency than the
deposits
15 days
maintenance
period
Interest rate of
discount
window over the
reserve
deficiency
Monetary
policy
Guatemala 14.6 14.6 0.6 n.a All deposits Cash in vault and deposits
at the central bank in the
some currency than the
deposits
Monthly n.a Monetary
policy
n.a
El Salvador 23 n.a n.a 3 All deposits 25 % in demand deposits
at CB or foreign bank, 25%
in deposits or CB
securities, 50% in CB
securities issued for
liquidity purposes
Over a two-
week period
n.a. Prudential
Panama n.a n.a n.a n.a Prudential
1/ Honduras also imposes specif ic liquidity requirements, based on temporal bands f or maturity mismatches. For the f irst band, the maturity mismatch in cash f lows f or the next month must be lower than
the amount of liquid assets, while the f or the second band the maturity mismatch in cash f lows f or the next three months must be lower than 1.5 times the liquid assets.
Sources: CAPDR Central banks and Superintendencies websites.
Cash of CB securities n.a Interest
charged based
on internbank
int rate (greater
than 1%)
Monetary
policy/2005-06
n.a 30; 20 (Applies to
all onshore general
license banks and
state-owned Banks
at 30.0 for general
licence banks;
20.0 for general
licence banks that
maintain average
interbank quarterly
deposits
exceeding 80
percent of total
deposits.
Demand, term
deposits up to 186
days (unless
portion that
guarantees loans
in the bank itself),
savings deposits.
Deposits received
from the parent
company, branch,
subsidiary or
affiliated abroad
are excluded.
Legal tender in Panama,
bank deposits in Panama,
bank deposits abroad,
obligations issued by
foreign governments,
obligations issued by
foreign private and
government agencies,
banking obligations
payable in Panama up to
186 days, installments of
payable obligations up to
186 days, other Liquid
Assets.
Nicaragua 16.25 16.25 on excess
reserves (n.a.)
All deposits
Demand, foreign
currency, time,
interbank,
government.
Excludes
interbank
deposits.
18 % in deposits with
central bank and 2% cash
in vault allowed.
Weekly,
holding
period ends
on Friday
n.a Monetary
Policy/2009
Deposits, expired
term deposits,
reduced capital
contracts and
savings stamps
and others
Cash in vault, deposits at
the central bank, and
government bonds in the
case of compulsory
investment in local
currency in the some
currency than the deposits
Over a two-
week period
Penalties
depend on the
currency of
denomination
and the type of
institution.
Monetary
policy/ 2008-09
Dominican
Republic
17 20 Foreign
currency
reserves are
remunerated at
Feds overnight
rate - 200bps
n.a
Yes; 2 1/
Appendix I. Table 1: Legal Reserve and Liquidity Requirements in CAPDR (2010)
Country Remuneration
(percent)
Liquidity
requirement
(percent)
Eligible Liabilities Compliance Assets Averaging Penalty Purpose/ Last
change
Local
Currency
(percent)
Foreign
Currency
(percent)
Honduras 6 (unrem),
12
(remun)
12(unrem),
10(remun)
Only
compulsory
investments are
remunerated at
½ of the policy
rate.
3
7
Country Currency 2005 2006 2007 2008 2009 2010
Costa Rica LCU RR 15 15 15 15 15 15
FCY RR 15 15 15 15 15 15
Guatemala LCU RR 14.6 14.6 14.6 14.6 14.6 14.6
FCY RR 14.6 14.6 14.6 14.6 14.6 14.6
Dominican Rep. LCU RR 20 20 20 20 17.5 (May -Dec) 17
FCY RR
20 20 20 20 20 20
Nicaragua LCU RR 16.25 (Jan-
May); 19.25
19.25 (Jan-Sep);
16.25 (Oct-Dec)
16.25 16.25 16.25 16.25
FCY RR 16.25 (Jan-
May); 19.25
(June-Dec)
19.25 (Jan-Sep);
16.25 (Oct-Dec)
16.25 16.25 16.25 16.25
Honduras LCU RR unremun. 12 12 12 12(Jan-Feb); 10.2
(Mar); 10.1 (Apr);
9.4 (May); 8.8
(June-Sept); 7.5
(Oc-Nov); 4.3
(Dec)
2.6 (Jan); 1.1 (Feb);
1.2 (May-July); 6.1
(Aug)
6
LCU RR remun. 2 (Jan); 1 (Oct) 1 3 (Jan-Jun); 4
(Jul-Dec)
4 (Jan-Feb); 6
(Mar); 7 (Apr(; 8
(May); 8.8 (June-
Dec)
8.8 (Jan.-Jul); 11 (Aug);
12 (Sep-Dec)
12
FCY RR unrem. 12 12 12 12 12 12
FCY RR remun.
2 (Apr); 4 (May); 6 (Jun-
Jul); 8 (Aug); 81. (Sep);
10 (Oct-Dec)
10
FCY LR 2
Panama LR 30; 20 30; 20 30; 20 30; 20 30; 20 30; 20
El Salvador RR 19.2 (Jan), 19.6
(Feb); 19.8
(Mar-Apr); 19.7
(May); 19.3
(Jun-Jul); 19.7
(Aug);19.4
(Sep);19.6
(Oct); 20.1
(Nov);20.3
(Dec)
20.2 (Jan-Feb);
20.3 (Mar);19.9
(Apr--Jun);19.8
(Jul);20.6
(Aug);19.9
(Sep);20.4
(Oct);20.2
(Nov);20.5 (Dec)
20.4(Jan-
Feb);20.6
(Mar);20.7
(Apr);20.8
(May);20.5
(Jun);21.2
(Jul);21.8
(Aug);21.4
(Sep);21.0
(Oct);21.2
20.1 (Jan); 20.8
(Feb);20.4
(Mar);20.6
(Apr);20.2
(May);20.0
(june)20.1
(Jul);20.2
(Aug);20.0
(Sep);20.1(Oct);20
.3 (Nov);20.4 (Dec)
20.9(Jan);20.7
(Feb);21.9(Mar);22.7(Ap
r);23.4(May);23.5(Jun);2
3.9 (Jul);24.3 (Aug);24.1
(Sep);23.3 (Oct);22.8
(Nov);23.7(Dec)
23.7(Jan);23.4(Feb);23.1(Mar
);23.0(Apr);22.6
(May);22.7(June);23.1 (July)
; 23.5(Aug);23.2 (Sep);23.2
(Oct);23.3 (Nov);23.1 (Dec)
LR 3.0 3.0 3.0
Source: CAPDR Central banks and Superintendencies Websites.
Notes: RR is reserve requirement, LR liquidity requirement. LCU is local currency, FCY foreign currency.
Appendix I. Table 2: Evolution of Reserve and Liquidity Ratios in CAPDR (2005-2010)
(In percent)
3
8
Liquid assets to
customer
deposits and
short term
funding
Liquid assets to
total assets
Total credit to
GDP
Getting Credit -
Strength of legal
rights index (0-
10)
Getting Credit -
Depth of credit
information
index (0-6)
Getting Credit -
Private bureau
coverage (% of
adults)
Enforcing
Contracts - Cost
(% of claim)
Institutional
Credit index (0-
100)
Liquid assets to customer
deposits and short term funding
1
Liquid assets to total assets 0.7598 1
(0)
Total credit to GDP 0.1565 0.19 1
-0.0009 -0.0001
Getting Credit - Strength of legal
rights index (0-10) 0.0577 0.0101 0.3555 1
(0.2232) (0.8318) (0)
Getting Credit - Depth of credit
information index (0-6) 0.1085 0.157 0.333 0.2571 1
(0.0216) (0.0009) (0) (0)
Getting Credit - Private bureau
coverage (% of adults) 0.0178 0.0596 0.2076 0.1231 0.5076 1
(0.7073) (0.2082) (0) (0.0069) (0)
Enforcing Contracts - Cost (% of
claim) -0.197 -0.2523 -0.5936 -0.1544 -0.5436 -0.0342 1
(0) (0) (0) (0.0007) (0) (0.4545)
Institutional Credit index 0.1447 0.1882 0.534 0.4873 0.9051 0.6104 -0.6698 1
(0.0021) (0.0001) (0) (0) (0) (0) (0)
Source: WEO database, Doing Business database; IMF staff calculations.
¹ P-values in parentheses.
Appendix II. Table 1. Correlations between liquidity ratios, credit to GDP and credit institutions in CAPDR ¹
3
9
Systemic
risk
regulation
Safety net
variable 2/
In place Coverage Legal
protection of
supervisor
Partial
transfer of
uninsured
deposits
In place In place Coverage Contributes
to bank
resolution
process
Target
fund
Costa-Rica yes 50% of liquid
assets
low n/a no no n/a n/a n/a
2
Dominican
Republic
yes 1.5 times
paid capital
high yes yes yes 13,900 yes 5% of
total
deposits 4
Guatemala yes 50% of
regulatory
capital
low yes no yes 2,400 yes 5% of
total
deposits
3
Honduras yes 100% of
capital and
reserves
high yes yes yes 9,600 yes 5% of
total
deposits 4
Nicaragua yes not
established
by law
medium yes yes yes 10,000 yes no
4
El Salvador no/1 n/a medium yes no yes 9,000 yes no
2
Panama no n/a low n/a no no
10,000
n/a n/a
1
Source: Bolzico, J., Gozzi E., and F. Rossini, 2010, Financial Saf ety Nets in American Countries: A Comparative Analysis;IMF staf f calculations.
1/ While by law (2010) the Central Bank is allowed to provide liquidity as a LOLR, in practice this f acility is just being established.
2/ Index variable categorizing the existence of each of the 4 pillars of a saf ety net (regulation, banking resolution, LOLR and deposit insurance),
with 4 assigned in presence of all f our pillars.
Lender of last resort Bank resolution Deposit Insurance
Appendix II. Table 2. CAPDR Financial Safety Nets
4
0
Variable No. observations Mean Std. Dev. Min. Max.
Liquid assets to customer deposits and short-term funding ratio 448 25.3 18.3 2.0 191.0
Liquid assets to total assets ratio 448 18.9 10.4 0.9 75.6
Capitalization (equity to asset ratio) 448 13.2 9.3 2.6 83.0
Loan Loss Reserves to Gross Loans 417 3.2 3.1 0.0 25.0
Net Interest Income to Average Earning Assets 428 8.8 10.5 1.0 87.0
Loan to Asset ratio 448 58.9 17.1 3.3 90.6
Loan growth (y/y, percent) 350 24.0 55.6 -66.2 594.5
Bank Size (log of total assets) 448 12.8 1.7 4.1 16.5
Foreign ownership dummy (=1 if foreign bank) 480 0.4 0.5 0.0 1.0
Private ownership dummy (=1 if private bank) 480 0.9 0.3 0.0 1.0
Interest rate spread 480 8.3 2.9 3.1 16.8
Real GDP growth 384 3.9 5.7 -7.9 15.3
Net international reserves 480 2931.0 1113.3 832.9 5594.9
Deposit dollarization 480 50.2 34.2 13.5 100.0
Banking crisis dummy (=1 if country had banking crisis after 1970) 480 0.7 0.5 0.0 1.0
Financial safety net (categorical variable, 1-4; 4=comprehensive safety
net) 480 2.7 1.2 1.0 4.0
Credit to GPD (%) 480 48.8 25.9 17.2 93.7
Quality of credit institutions index (0=worst; 100=best) 480 68.3 28.2 0.0 100.0
Sources: International Financial Statistics, World Economic Outlook Database, Bankscope Database,
CAPDR Central banks and Banking Supervision websites, Laeven and Valencia (2010), Bolzico et al. (2010); IMF staff calculations.
Appendix III. Table 1: Descriptive Statistics
4
1
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Credit Financial Credit Financial Credit Financial Credit Financial Credit Financial
Institutions Depth Institutions Depth Institution Depth Institutions Depth Institutions Depth
Liquid assets ratio (-1) 0.484*** 0.557*** 0.539*** 0.567*** 0.398*** 0.483*** 0.563*** 0.519*** 0.543*** 0.534***
(0.112) (0.098) (0.101) (0.089) (0.118) (0.101) (0.088) (0.093) (0.106) (0.114)
Bank size 4.321*** 3.861*** 3.666*** 4.077*** 4.850*** 5.484*** 3.596*** 3.815*** 4.146*** 4.023***
(0.979) (0.866) (1.192) (1.048) (1.330) (1.908) (0.959) (1.257) (0.889) (0.977)
Bank size squared -0.201*** -0.180*** -0.171*** -0.197*** -0.232*** -0.262*** -0.167*** -0.179*** -0.191*** -0.187***
(0.049) (0.044) (0.060) (0.056) (0.065) (0.085) (0.048) (0.061) (0.044) (0.045)
Capitalization -0.226*** -0.211*** -0.098 -0.192 -0.188 -0.175* -0.100 -0.063 -0.292* -0.249
(0.071) (0.070) (0.215) (0.124) (0.158) (0.099) (0.233) (0.215) (0.162) (0.275)
Net interest margin -0.047 -0.037 0.096 -0.799 -0.570* -0.593* 0.197** 0.216 0.203 0.163
(0.033) (0.033) (2.354) (1.193) (0.331) (0.311) (0.099) (0.151) (0.323) (0.337)
Loan-loss reserve ratio 0.015 -0.036 0.025 0.277 -0.080 -0.292* -0.467* -0.178 -0.056 -0.346
(0.168) (0.145) (0.608) (0.280) (0.163) (0.166) (0.251) (0.395) (1.433) (1.230)
Credit 1/ 0.033 -0.181 0.114 0.117 0.044* -0.295 0.147*** 0.249 0.464*** -0.135
(0.024) (0.181) (0.159) (0.355) (0.026) (0.251) (0.047) (0.553) (0.133) (0.636)
Variable 9.567 16.285 0.403 -5.554 0.482*** 1.336*** 21.139*** -8.880
(24.722) (24.600) (6.609) (7.717) (0.174) (0.346) (7.458) (28.601)
Capitalization*variable -0.167 -0.071 -0.044 -0.161 -0.002 -0.003 0.053 0.027
(0.335) (0.251) (0.210) (0.212) (0.005) (0.004) (0.060) (0.119)
Net interest margin*variable -0.147 0.756 0.421 0.487 -0.011*** -0.012 -0.080 -0.067
(2.352) (1.186) (0.312) (0.328) (0.004) (0.008) (0.099) (0.109)
Loan-loss reserve ratio*variable 0.082 -0.291 0.948 1.340 0.019** 0.005 0.016 0.091
(0.681) (0.456) (0.878) (0.903) (0.008) (0.017) (0.425) (0.377)
Credit *variable 1/ -0.084 -0.444 -0.016 0.100 -0.005** -0.013 -0.131*** 0.006
(0.155) (0.424) (0.058) (0.130) (0.002) (0.009) (0.039) (0.267)
Observations 321 321 321 321 321 321 321 321 321 321
R-squared 0.72
No. of groups 96 96 96 96 96 96 96 96 96
No. of instruments 64 64 67 67 67 77 77 67 67
Hansen test p-value 0.357 0.337 0.302 0.267 0.283 0.747 0.448 0.546 0.388
A-B AR(2) test 1.191 1.075 1.236 1.152 0.891 1.026 1.427 1.012 1.075
A-B AR(2) test p-value 0.234 0.282 0.216 0.249 0.373 0.305 0.154 0.311 0.282
Source: Authors' calculations.
Dependent variable is the ratio of liquid assets to total assets. GMM is two-step system GMM estimator with Windmeijer standard error correction.
Columns (3) through (10 test the hypotheses that ownership (foreign/domestic and public/private), degree of dollarization and coverage of the financial safety net affect banks
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the country lev
All regressions include time and country dummies. Constant estimated but not reported.
1/ Credit is captured either by a measure of credit institutions, proxied by the principal components analysis of the Doing Business variables measuring the
quality of credit institutions. Financial depth is proxied by the credit-to-GDP ratio.
Appendix IV. Table 1. Determinants of Banks' Liquidity Buffers in CAPDR - GMM Estimates
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Dependent variable is the ratio of
liquid assets to total assets
Baseline Variable = Private Foreign ownership Dollarization Safety net
4
2
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Credit Financial Credit Financial Credit Financial
Pooled OLS GMM Fixed Effects Institutions Depth Institutions Depth Institutions Depth
Liquid assets ratio (-1) 0.347*** 0.189*** 0.169*** 0.171** 0.164** 0.256*** 0.227*** 0.252*** 0.221***
(0.064) (0.044) (0.025) (0.070) (0.069) (0.074) (0.072) (0.083) (0.063)
Bank size 7.401*** 7.994*** 16.804 9.227* 11.172** 7.615** 7.848** 8.551*** 8.543***
(1.448) (1.875) (10.703) (4.806) (5.221) (2.972) (3.038) (2.081) (2.126)
Bank size squared -0.350*** -0.371*** -0.886* -0.469** -0.553** -0.362** -0.381** -0.410*** -0.414***
(0.064) (0.092) (0.476) (0.204) (0.218) (0.150) (0.152) (0.107) (0.106)
Capitalization -0.355*** -0.321** -0.502*** -0.711 -0.453 -0.281 -0.424 -0.386 -0.262
(0.060) (0.123) (0.184) (0.557) (0.516) (0.495) (0.477) (0.528) (0.457)
Net interest margin -0.156** -0.123 0.011 -0.774 -1.169 0.452 0.332 -0.056 -0.107
(0.064) (0.076) (0.325) (0.752) (0.833) (0.276) (0.335) (0.704) (0.706)
Loan-loss reserve ratio -0.224 -0.282 0.221 -0.448 -0.984 -1.071 -1.032 1.536 1.277
(0.160) (0.252) (0.290) (0.553) (0.810) (0.678) (0.704) (1.691) (1.627)
Credit 3/ -0.213 -0.323 -0.337* 0.044 -0.904** 0.240*** -0.860 0.625** -0.123
(0.272) (0.292) (0.199) (0.046) (0.387) (0.088) (0.675) (0.250) (0.794)
Variable -6.238 -18.363 0.666** 1.201*** 27.313** -10.493
(12.128) (15.445) (0.293) (0.412) (13.739) (35.604)
Capitalization*variable 0.339 -0.070 -0.001 0.000 -0.037 -0.092
(0.608) (0.711) (0.007) (0.008) (0.191) (0.164)
Net interest margin*variable 0.441 0.973 -0.025** -0.021 -0.023 -0.000
(0.750) (0.891) (0.012) (0.015) (0.217) (0.218)
Loan-loss reserve ratio*variable 2.167* 3.021* 0.042* 0.036 -0.429 -0.339
(1.222) (1.552) (0.021) (0.022) (0.549) (0.451)
Credit to GDP ratio*variable 1/ 0.010 0.212 -0.008** -0.003 -0.169** -0.231
(0.091) (0.214) (0.003) (0.009) (0.070) (0.262)
Observations 321 321 321 289 289 289 289 289 289
R-squared 0.55 0.19
No. of groups 96 96 88 88 88 88 88 88
No. of instruments . 64 67 67 77 77 67 67
Hansen test p-value 0.348 0.105 0.117 0.135 0.135 0.187 0.243
A-B AR(2) test 1.283 1.262 0.770 1.600 1.574 1.231 1.322
A-B AR(2) test p-value 0.199 0.207 0.442 0.110 0.116 0.218 0.186
Source: Authors' calculations.
Dependent variable is the ratio of liquid assets to customer deposits and short-term funding. Two-step system GMM estimator with Windmeijer standard error correction.
Ownership is captured by dummy variables (=1 if the bank is private, =1 if the bank is foreign), dollarization by the share of dollar deposits in total deposits at the
country level. Safety net is an index variable categorizing the existence of each of the 4 components of a safety net in each country (Appendix II. Table 1).
All regressions include time and country dummies. Constant estimated but not reported.
1/ Columns (1) and (2) as in Table 4a. Column (3) reports results of the fixed effects estimation.
2/ Columns (4)-(9) show GMM estimation results for the sample of private banks (excluding public banks).
3/ In columns (1)-(3) credit is captured by a measure of financial depth, proxied by the credit-to-GDP ratio. In columns (4)-(9), Credit institutions are proxied
by the principal components analysis of the Doing Business variables measuring the quality of credit institutions.
Notes: Robust standard errors in parentheses. *** Coefficient significant at the 1 percent level; ** at the 5 percent level; * at the 10 percent level
Appendix IV. Table 2. Determinants of Banks' Liquidity Buffers in CAPDR - GMM Estimates (unless otherwise specified)
Dependent variable is the ratio of
liquid assets to customer deposits
and short-term funding
Baseline 1/ Foreign ownership 2/ Dollarization 2/ Safety net 2/
doc_170349835.pdf