WP/06/44 Banking Spreads in Latin America

WP/06/44


Banking Spreads in Latin America

R. Gaston Gelos

---

---

© 2006 International Monetary Fund
WP/06/44

IMF Working Paper

Western Hemisphere Department

Banking Spreads in Latin America

Prepared by R. Gaston Gelos1

Authorized for distribution by Charles Collyns

February 2006

Abstract

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.

Intermediation spreads in Latin America are high by international standards. This paper
examines the determinants of bank interest margins in that region using bank- and country-
level data from 85 countries, including 14 Latin American economies. The results suggest
that Latin America has higher interest rates, less efficient banks, and larger reserve
requirements than other regions and that these factors have a significant impact on spreads.
However, Latin American countries do not differ markedly from their peers in other aspects
that are found important in determining the cost of financial intermediation, such as inflation
and bank profit taxation.

JEL Classification Numbers: E43, E44, G21, O54

Keywords: Banking spreads, financial intermediation

Author(s) E-Mail Address: ggelos@imf.org




1 The author would like to thank Ricardo Adrogué, Adolfo Barajas, Agnès Belaisch, Martin
Cerisola, Ana Corbacho, Charles Collyns, Marcio Nakane, David Owen, Steven Phillips, and
Chris Towe for helpful comments and Giovanni Dell’Ariccia, Luc Laeven, and Laura Kodres
for sharing data.

---

- 2 -



Contents Page

I. Introduction.........................................................................................................................3

II. The Determinants of Spreads: Literature Review ..............................................................6


III. Descriptive Evidence ..........................................................................................................8

A. Creditor Rights and the Legal Framework ................................................................11
B.
Macroeconomic
Volatility.........................................................................................11

C. Competition in the Banking Sector ...........................................................................12
D.
Reserve
Requirements ...............................................................................................14

E. The Level of Interest Rates........................................................................................15

F. Taxation of the Financial Sector................................................................................15

G. Availability of Information about Borrowers............................................................16


IV. Econometric Estimations ..................................................................................................16
A.
Alternative
Specifications and Robustness ...............................................................20

B. What Explains the Difference Between the Level of Spreads in Latin America and


Other Developing Countries...................................................................................21

V. Conclusions.......................................................................................................................23

Figures
1. Bank Credit to the Private Sector, 2003 .............................................................................3
2. Ex-Ante Banking Spreads Worldwide................................................................................4
3. Net
Interest
Margins ...........................................................................................................4
4. Ex-Ante Banking Spreads in Latin America ......................................................................5
5. Legal Protection and Interest Margins..............................................................................11
6. Country Risk and Net Interest Margins ............................................................................12
7. Average Bank Overhead and Net Interest Margins..........................................................13
8. Reserve Requirements on Demand Deposits and Net Interest Margins...........................14
9. Deposit Rates and Net Interest Margins ...........................................................................15
10. Availability of Information about Companies and Net Interest Margins .........................16
11. Net Interest Margins: Contribution of Different Factors in Explaining Difference

Between Latin America’s Average and the Average for Developing Countries...........21

Tables
1. Factors Influencing Level of Spreads ...............................................................................10
2. Determinants of Net Interest Margins: Cross-Country Regressions, 2002 ......................19
3. Determinants of Net Interest Margins: Panel Regressions for 1999–2002 ......................22

Appendix
Table AI.
Determinants of Net Interest Margins: Cross-Country Regressions, 2002

(excluding
Brazil,
Malawi, and Venezuela) ....................................................28
Table AII.
Determinants of Spreads: Cross-Country Regressions, 2002

(alternative
spread
definition) ..........................................................................29

References ..............................................................................................................................24

---

- 3 -

I. INTRODUCTION
Financial intermediation in Latin America is low by international standards. Financial
systems in Latin America are still largely bank-based, but after a short period of strong credit
growth sparked by financial liberalization in the early 1990s, bank lending has not yet
recovered from the collapse following the banking crises of the mid-1990s (Singh et al.,
2005). As a result, bank lending as a percentage of GDP is low compared to industrialized
countries and other emerging markets (Figure 1).
This lack of financial intermediation is seen as an important obstacle to growth: there is a
considerable body of evidence indicating that financial intermediation is not only correlated
with growth but a causal factor in explaining economic performance (see, for example,
Levine, 2004).
Figure 1. Bank Credit to the Private Sector, 2003
(In percent of GDP)
180
180
150
150
120
120
90
90
60
60
30
30
0
0
y
s
a
b
ia
l
e
dor
r
u
z
i
l
c
a
e
p
.
es
c
a
r
e
a
gua
x
i
c
o
m
Pe
i
vi
fri
Bra
e
r
a
g
e
.
R
Chi
t
at
a
i
l
a
n
d
Ko
a
ra
Me
c
ua
t
a
Ri
Bol
S
i
ngdom
P
o
lo
E
m
ondura
h A
C
.
Av
Th
Cos
Do
H
n
it
ed
out
U
S
n
i
t
e
d
K
a
t
.
Am
L
U

Source: IFS.
The low levels of lending in Latin America appear related to the prevalence of high
intermediation costs in the region. Intermediation spreads in Latin America are high
compared to those in other banking markets. This is true when measuring spreads either as
ex-ante bank spreads (the difference between weighted averages of lending rates and bank
funding costs, Figure 2) or ex-post net interest margins (defined as the bank’s total interest
income minus total interest expense, divided by the sum of interest bearing assets, Figure 3).

---

- 4 -
Figure 2. Ex-Ante Banking Spreads Worldwide
(In percentage points, 138 countries, 2003)
50
45
40
35
30
25
20
Latin American
15
Median
Overall Median
10
5
0


Note: Simple average of difference between deposit and lending rates. Source: IFS. Latin American countries
are marked in light color.

Figure 3. Net Interest Margins
(Average, 1999–2002)

25
20
15
10
5
0
ia
l
a
a

a
s
ru
r
azi
b
ia
e
p
.
el
x
i
c
o
B
Chile
.
R
ur
g
u
a
y
Pe
i
can
e
r
a
g
e
)
Boliv
u
a
d
o
r
er
o
lom
s
t
a
Ric
m
Me
ra
ag
C
Ec
o
nd
e
n
ezu
m
Co
Do
H
Pa
V
er
v
e
l
o
p
i
ng
av
s
(
a
ve
ie
a
t
i
n
A
r
de
tr
L
t
he
O
c
o
un

Note: Total interest income minus total interest expense, divided by the sum of interest bearing assets. Source:
Bankscope

---

- 5 -

The high levels of spreads are not a recent phenomenon, and unlikely to be a result of the
ongoing process of bank consolidation. As in other regions, in the 1990s Latin America saw
a trend toward bank consolidation and concentration. To a large extent, consolidation was
driven by financial crises and subsequent regulatory tightening (IDB, 2004). While this
resulted in a marked reduction of the number of banks in most countries, it does not seem to
have yielded a decline in competition intensity, partly because it was accompanied by a
lowering of barriers of entry to foreign banks (see Gelos and Roldós, 2004, and IDB, 2004).
In fact, banking spreads have fallen somewhat over the last decade although the pace of the
reduction has been slow (Figure 4).2
Figure 4. Ex-Ante Banking Spreads in Latin America
(In percentage points)
50
40
30
+1 standard deviation
20
10
0
-10
-1 standard deviation
-20
-30
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Note: Simple average of difference between lending and deposit rates for 12 Latin American
economies. Source: IFS.
This paper takes a closer look at the determinants of bank interest margins in Latin America
using bank- and country-level data from 85 countries, including 14 Latin American
economies. We assess the role of bank- and country-specific factors in determining banking
spreads using micro data from over 2,200 banks. Taking an international perspective allows
us to delineate the key dimensions in which Latin America differs from other regions. The
paper differs from existing studies in further dimensions apart from the Latin American
specific focus in a cross-country approach. Among other aspects, we make use of new or
expanded data sets (including comprehensive information on reserve requirements and
foreign ownership) and investigate the relationship between competition and bank spreads by
using behavioral measures of bank competition suggested by the new industrial organization
literature.

2 See also Singh et al. (2005).

---

- 6 -
The results suggest that Latin America has higher interest rate levels, less efficient banks, and
larger reserve requirements than other regions and that these factors have a significant impact
on spreads. However, Latin American countries do not differ markedly from their peers in
other aspects that are found important in determining the cost of financial intermediation,
such as inflation, bank profit taxation, average bank size, or average equity as a percentage of
total bank assets.

II. THE DETERMINANTS OF SPREADS: LITERATURE REVIEW
The extensive literature on bank behavior suggests that, at the country level, the following
factors are likely to influence the cost of credit:
• Creditor rights and the quality of the legal framework. Higher recovery rates and
shorter times to repossess collateral in countries with better legal environments are
expected to reduce bank spreads. Some studies have found a significant impact of
judicial efficiency on ex-ante spreads and net interest margins (see Laeven and
Majnoni, 2003, and Demirguç-Kunt, Laeven, and Levine, 2004, henceforth DLL).

• The degree of banking competition. More intense competition intensity should
yield lower spreads. Empirically, however, a strong correlation between direct
measures of competition or concentration and spreads is hard to find.3 Nevertheless,
administrative expenses have been found to be positively associated with net interest
margins, and the presence of inefficiencies in a country’s banking system in turn is
likely to reflect a lack of competitive pressure.

• The macroeconomic environment. There is no generally accepted model relating
macroeconomic performance to spreads between borrowing and lending rates.
However, macroeconomic volatility may raise the risk of default and therefore bank
spreads. In the dealership model of banks developed by Ho and Saunders (1981),
interest margins rise with the variance of interest rates as a result of the
intermediation risk faced by banks. This is supported empirically by Saunders and
Schumacher (2000), among others. Moreover, if inflation shocks are not passed
through to both borrowing and lending rates equally rapidly, bank spreads may be
correlated with inflation rates, and indeed various studies find a positive correlation
between spreads and inflation (see Honohan, 2003). Similarly, theory predicts that the
riskiness of borrowers is likely to rise with the level of interest rates, possibly in a
nonlinear way. Banks will typically want to be compensated for higher risk, which
yields a positive relationship between the level of interest rates and spreads. Lastly,
an increase in economic activity may raise the net worth of borrowers and lower
spreads (see Bernanke and Gertler, 1989).


3 DLL do not find a robust correlation between measures of concentration and net interest
margins. However, Martínez Peria and Mody (2004), focusing on a small number of
countries, report a positive relationship between bank concentration and spreads. For a
discussion of competition in Brazilian banking, see Belaisch (2003) and Nakane (2003).

---

- 7 -
• Taxation, including reserve requirements. Taxation of financial intermediation can
take different forms, such as financial transaction taxes, bank profit and revenue
taxes, or reserve requirements that are remunerated at below-market rates. Taxation
drives a wedge between borrowing and lending rates. Effective tax rates usually rise
with inflation and the level of short-term interest rates, and depending on the
behavioral response of banks and depositors, the same form of taxation will have a
larger effect on bank spreads at higher inflation levels. See Honohan (2003).

• Availability of information about borrowers. Higher availability of information
about potential borrowers will reduce default risk and lower spreads. For example, the
existence of credit information registries and good accounting standards are widely
seen as important for improving risk assessment and reducing ex-ante spreads (and,
through risk premia, also ex-post margins). See Chu and Schechtman (2003).

• Banking regulations and mandated lending. Regulatory barriers to entry will
reduce competition and increase intermediation margins. Regulations may also take
the form of mandated lending programs which requires lending to particular sectoral
groups, often at low or subsidized rates. The costs associated with these mandated
lending programs tend to be borne by other borrowers.

Bank-specific factors, however, also affect the level of interest margins. Banks have different
strategies which may affect their product mix and the pricing of loans (see Dell’Ariccia and
Márquez, 2002). For example, some banks may rely more on fee income than others.
Similarly, in an imperfectly competitive environment, larger banks may be able to exploit
economies of scale and lower interest margins, institutions with larger market share may be
able to charge more, and banks with larger overhead costs may pass these costs on to
borrowers. Well-capitalized banks may face lower funding costs, implying larger net interest
margins (see DLL).
Foreign banks in developing countries are generally thought to be more efficient than
domestic ones. The evidence on the impact of bank ownership on the level of spreads
charged, is, however, mixed. Martínez Peria and Mody (2004) find that foreign banks in
Latin America, in particular new entrants (as opposed to banks acquired by foreigners) were
able to charge lower spreads than domestic ones. On the other hand, Claessens et al (2001)
report that foreign banks tend to have higher net interest margins. Detragiache et al. (2005)
find that in poor countries, a larger foreign bank presence is associated with shallower credit
markets.
Previous studies on Latin American banking spreads have emphasized the importance of
taxes, operating costs, imperfect competition, and macroeconomic volatility in determining
the cost of financial intermediation. Using data from seven Latin American economies,
Brock and Rojas Suárez (2000a) find that administrative and other operating costs as well as
reserve requirements contribute to the prevalence of high spreads in Latin America.4
Moreover, macroeconomic volatility also appears to be a factor behind the high spread
levels. According to the authors’ findings, and contrary to results reported in the literature for

4 See also Brock and Rojas Suárez (2000b).

---

- 8 -
industrial countries, in Latin America, higher shares of non-performing loans tend to be
associated with lower spreads.5 Examining the effect of foreign participation and market
concentration on spreads in five Latin American countries, Martínez Peria and Mody (2000)
do not find a significant effect of the share of nonperforming loans on spreads. The authors
also report that spreads are positively related to concentration measures and administrative
costs. In a study comprising five Central American economies, Dick (1999) also documents
that high operating costs are the most important component of the spread.
Similar results have been found in studies focusing on individual Latin American countries.
Barajas, Steiner, and Salazar (1999) examine the behavior of spreads in Colombia, reporting
that imperfect competition and operating costs, the fraction of nonperforming loans, and
financial taxation all contributed to high spreads. Catão (1988) studies intermediation spreads
in Argentina, singling out high administrative costs, loan-loss provisions, and imperfect
competition as key determinants. Brock and Franken (2003) assess the factors driving
interest spreads in Chile. Among their findings is that macroeconomic volatility, bank size,
and concentration measures are significant spread determinants across different
specifications. A substantial number of studies have focused on Brazil. For example, the
Brazilian central bank regularly publishes a decomposition of ex-ante spreads, and has
traditionally emphasized taxation, administrative costs, and loan-loss provisions as the main
determinants.6 Koyama and Nakane (2002) conclude that risk perceptions accounted for the
largest fraction of total ex-ante spreads, with administrative costs and indirect taxes
(excluding reserve requirements) coming second and third, respectively.7

III. DESCRIPTIVE EVIDENCE
In this section, we use data from a variety of sources to assess how Latin America scores in
international comparisons of the dimensions highlighted above. Data on bank-specific
variables, including net interest margins, are taken from the Bankscope database. (Details on
the construction of bank-specific variables can be found in the appendix). We include data
for the period 1999–2002. We use data on banks from all developing countries and emerging
markets included in Bankscope’s database. Not all variables are available for all banks, and
in most estimations we work with samples of 30–60 countries and data on 1,000–1,300
banks. To measure economy wide characteristics, we resort to a variety of sources, as
discussed below.
The focus on ex-post net interest margins as opposed to ex-ante spreads between deposit and
loan rates allows for a broader examination of the costs of financial intermediation. Net

5 The authors argue that this could be the result of inadequate provisioning or reflect the fact
that banks with a high proportion of bad loans may lower spreads to attract deposits and grow
out of their troubles.
6 See Banco Central do Brasil (1999, 2000, 2001, 2002, and 2003).
7 See also Afanasieff, Lhacer, and Nakane (2002). Cardoso (2003) examines the effects of
reserve requirements on banking spreads through their impact on bank inflationary revenue
(resulting from noninterest bearing demand deposits), finding that spreads decline with
inflationary income. The central bank spread reports contain a number of other studies of
spread determinants that cannot all be surveyed here.

---