Analysis Of The Argentine And Mexican Banking

Published Date: 02 Nov 2017

By

Mine Aysen Doyran, Ph.D.

Emre Erdogan, Ph.D.

Abstract

Purpose- Emerging markets provide a prism through which to view the transformation of countries’ financial and economic systems undergoing globalization. Building on previous research in this area, this paper examines the performance and characteristics of Argentina and Mexico’s banks over the past 20 years.

Design/methodology/approach- Using Arellano-Bond Generalized Method of Moments (GMM) and balanced/firm-level data, this paper examines the effects of firm-level, industry-level, microeconomic and macroeconomic factors on Argentine and Mexican banks’ profits (Return on Assets) over the period 1994-2011.

Findings- A detailed evaluation of bank statements in two countries reveal important similarities in the financial condition and performance of selected institutions. According to firm-level analysis, credit risk (cris) and operating cost (opc) are two common determinants of bank profitability in Argentina and Mexico. As industry-level analysis indicates, in financial systems with higher loans to the private sector, banks become more profitable, as confirmed in the Argentine case. But the picture is reversed for Mexican banks when the macroeconomic environment is controlled. As lending to the private sector increases (decreases), bank profitability decreases (increases). Among macroeconomic variables, the growth rate of money supply (m2gdp) positively affects Argentine bank performance whereas it has a negative impact on Mexican bank returns.

Originality/value--The paper’s value lies in its exploration of differences and similarities in bank performance in two Latin American countries with the largest financial sectors (after Brazil). While firm-level analysis indicates that Mexican banks do not differ significantly from Argentina’s in an operational (behavioral) sense, micro and macro-level analyses reveal that they fail to capitalize on local monetary conditions.

Research limitations/implications- This paper contends that further research of the Latin American financial sectors should highlight the impact of monetary policy choices, most recent regulatory standards (e.g., timing of financial reform) and even countries’ structural characteristics including, for example, degree of financial openness or international financial integration. Firm level differences between banking systems (for example, the impact of loan portfolio diversification on bank returns) can also be compared. The study concludes that if governments have a better understanding of the conditions in which their banking sector operates, they can adopt policies that can help improve its performance. It will also help researchers compare how banks in developing countries fulfill the key function of providing loans to households and firms.

Keywords-Argentina, Mexico, Latin America, Banking Sector Performance, Economic Development, Financial Systems, Emerging Markets

1. Introduction

Latin American banking systems underwent major structural changes over the past 25 years. Financial downturns led to market concentration and an open invitation to foreign bank entry, liberalization, and the merger and acquisitions of large domestic banks. In Argentina, the 2001 crisis affected its banking sector as a whole. In January 2002, the government abandoned the "convertibility plan" that tied the dollar to Argentine peso on a one to one parity. An increase in foreign lending underpinned the system, which led to an unsustainable rise in public debt. As Argentina experienced the quickest recovery since 2003, it became the fastest growing economy in the western hemisphere and in the world (Basu et all, 2004:3; Weisbrot and Sandoval, 2007:1). Followed by booms in two of the larger economies, Brazil and Mexico, the new recovery is driven by political-economic factors—a combination of government intervention, favorable external conditions, easy access to capital markets at low costs as well as "a high level of remittances from migrant workers" (Ocampo, 2010:3).

In contrast to Argentina’s, the Mexican banking sector has limited lending capacity, high level of concentration and less diversified commercial services (Gonzalez-Anaya, 1995). While the liberalization reform of the early 1990s sought to consolidate markets by eliminating inefficient banks, it provided "unrestricted" access of foreign banks to the local market (Haber and Musacchio, 2005:2). Regulatory authorities relaxed all barriers to foreign direct investment because Mexican banks needed extensive recapitalization after the 1995 peso crisis (Murillo, 2007:1). The crisis staggered Mexico’s economy and financial system performance, with major impacts on the banking sector. Despite recapitalization and improved financial regulation, commercial banks experienced a major deterioration of lending to the private sector. For example, despite an average annual growth (GDP) rate of 5.4% from 1996 to 2000, domestic credit to the private sector—lending from commercial banks to the non-financial private sector—decreased from 10% of GDP in 1994 to 0.3% of GDP in 2000 (Gonzalez-Anaya, 1995:1-2).

Connecting with an extensive body of world literature on bank performance, especially in the US and Europe, there is a growing number of empirical studies on the Argentine and Mexican banking sectors (Gruben and McComb, 1997; Burdisso et al 1998; Burdisso, 1999; Cull et al, 1999; Basu et al, 2004; Clarke and Cull, 2005; Haber and Musacchio, 2005; Schulz, 2006; Hernandez-Murillo, 2007; Solis and Maudos, 2008; Bebczuk and Galindo, 2008; Chortareas et al, 2009). The purpose of this paper is to extend earlier research to the comparison of Argentine and Mexican banking sector performance. We analyze the banking-profitability nexus using a dynamic panel approach –Arellano-Bond Generalized Method of Moments. We choose the 1994 to 2011 period because it coincides with major changes in both countries’ financial systems—a period of powerful external shocks as well as economic restructuring. We measure bank performance using return on assets (profitability) as a dependent variable and regress it with three independent variables: macroeconomic (external), microeconomic (market structures/financial environment) and firm level (internal/managerial). In addition, market structure variable, as measured by concentration index (hhi), is used to highlight the competitive conditions in banking markets.

Our results indicate that Mexican banks do not differ significantly from Argentine banks in an operational (or behavioral) sense. Credit risk (cris) and operational cost (opc) are two common determinants of bank profitability in both countries. With respect to Mexico, however, previous years’ profitability and market share (ms) affect performance negatively at firm-level analysis. This may be related to the pro-cyclical impact of global financial crisis on Mexico’s bank performance, especially after the Lehman bankruptcy in 2008. Previous research indicates that structural characteristics of countries such as differences in financial openness and international financial integration impacted bank credit growth. Political factors such as counter-cyclical monetary policy were also instrumental in relieving the "credit crunch" in the period after the crisis (Aisen and Franken, 2010).

Although a business friendly environment with abundant credit (growth of money supply as a percentage of GDP) leads to profitable banking in Argentina, our study reveals that Mexico’s banking system fails to capitalize on local monetary conditions. In financial systems with higher loans to the private sector, banks become more profitable, as confirmed in the Argentine case. But the picture is reversed for Mexican banks when the macroeconomic environment is controlled. As lending to the private sector increases (decreases), bank profitability decreases (increases). The reasons for this divergence are not quantified in this paper but they may be attributed to differences in monetary policy, regulatory standards and even countries’ structural characteristics including, for example, the degree of financial openness or international financial integration—issues of further research. Argentina’s economic model requires further examination because a controlled inflation rate in the context of an increased money supply is atypical of developing countries. Although short-lived, this finding is consistent with current government’s goal to increase bank liquidity and keep interest rates low in order to spur economic growth.

2. Latin American Banking Systems: A Comparative Analysis

This section presents a brief summary of the literature on banking sector performance in developing countries. Latin American banks show large variations in terms of managerial and organizational performance. Most recently, there has been substantial research linking these variations to firm-level, microeconomic and macroeconomic environment of banking. Connecting with an extensive body of literature on profitability (Bourke, 1989; Molyneux and Thornton, 1992; Berger 1995; Goldberg and Rai, 1996; Goddard et al, 2007, to name a few), there has been a growing number of empirical studies on Latin American banks (Barajas et al, 2000; Brock and Rojas-Suarez, 2000a; Demirguc-Kunt and Huizinga, 2000; Martinez-Peria and Mody, 2004; Naceur and Ghazouani, 2007; Athasanoglu et al 2008; Chortareas et al, 2011).

Macroeconomic variables establish a context for banking sector analysis (Levine 1997; Levine and Zervos, 1998; Beck et al 2000; Demirguc-Kunt and Maksimovic, 1998; Demirguc-Kunt and Huizinga, 2000; Rousseau and Wachtel, 2000; Naceur and Ghazouani, 2007). For example, methodologies emphasizing the link between financial markets and long-run economic growth are influenced by "endogenous growth" theory. There has been considerable support for the view that banks are central to the allocation of credit in a modern economy. They can stimulate innovation and demand for services by channeling investment funds to firms and households. In addition, financial markets are an integral part of the growth process that constitute the structure of the financial system, including stock markets, financial institutions, capital markets, legal, political and corporate governance structures (Christopolous and Tsionas, 2004:56). For example, in a study of cross-cross country data for 49 countries from 1976 through 1993, Levine and Zervos (1998) studied the long-run empirical relationship between stock market development, banking development, and long-run economic growth. They showed that "well-functioning" (or more liquid) stock markets were associated with banking sector development, which was "robustly" and "positively" correlated with "output growth", "capital accumulation" and "productivity growth".

Demirguc-Kunt and Huizinga (2000) used both cross-sectional (firm-level/cross-country) and time-series data for the period of 1990 to 1997 to examine the link between bank performance and financial structure variables. They calculated financial ratios for each bank in 44 developed and developing countries (including Argentina and Mexico) and then averaged them over 1990-97. Their results indicated that that causality proceeded from macroeconomic and financial structure variables to bank profitability. For example, greater bank development lowered profit margins but increased efficiency and market competition. In countries with underdeveloped financial systems, the transition to a more developed financial structure (captured through stock market size and liquidity), bank profits improved. By increasing access to different sources of capital, stock markets improve borrowing capacity of firms. Furthermore, stock markets provide information that can be used to evaluate credit risk. This can lead to increase in bank profits. It was no longer possible to observe such complementarities, however, at higher levels of financial development as seen in the US and elsewhere. As their analysis showed, marginal impact of financial markets for bank performance was exhausted in advanced industrialized economies.

In another study, Demirguc-Kunt and Maksimovic (1998) used firm-level data for 30 countries for the period 1980-1995 to analyze the effects of legal and financial systems on firm growth. Firm-growth was measured by the firms’ use of "external financing to fund growth". The size of banking sector and legal efficiency index were proxied for financial market and legal development. They concluded that in countries that ranked high on legal efficiency index, a larger proportion of firms relied on "long-term external financing"; a large banking sector and an active stock market were also correlated with externally funded firm growth. They also showed that there was a relationship between lower profit rates and "increased reliance on external financing" in countries with "well-functioning institutions". On the other hand, government subsidies did not increase the level of external financing (Demirguc-Kunt and Maksimovic, 1998:1). This finding is consistent with Demirguc-Kunt and Huzinga (2000) reporting lower return on assets for banks in well-developed financial markets.

When comparing different banking systems, it is necessary to consider the impact of micro-level factors, such as banking market structures (e.g., competitive conditions, entry barriers) and firm-level characteristics. It is often argued that greater market competition leads to more efficient banking systems. Efficient banks generate higher profits, which may give them the ability to capture higher market shares. A recent panel study by Chortareas et al (2011) tested three theories of banking performance —the Structure-Conduct-Performance (SCP), Relative Market Power Hypothesis (RMPH) and Efficient Structure Hypothesis (ESH). Their sample included 2500 bank observations in nine Latin American countries (Argentina, Brazil, Chile, Colombia, Costa Rica, Paraguay, Peru, Uruguay, and Venezuela) over the period 1997-2005. They employed a non-parametric method known as Data Envelopment Analysis (DEA). This method allows better estimation of efficiency scores and production function with the use of mathematical programming. An earlier study utilizing Thick Frontier Approach lent some support to increasing returns to scale in Argentine banking sector (Dick, 1996). Their results lent support to the ESH and negated market power theories that market concentration (as measured by decrease in the number of banks) was associated with higher interest margins and bank profitability. Their findings indicated that capital ratios (degree of capitalization) and banks asset size (logarithm of total assets) were robustly and significantly related to higher than normal profits. While efficiency (particularly scale efficiency) was the main driving force of profitability, profits were unrelated to competitive conditions. Macroeconomic control variables had mixed results and some were less significant than expected. While in Argentina there was a positive relationship between the GDP growth and bank profitability, in Venezuela, inflation seemed to bolster profits. The policy implication derived from their analysis is that regulatory authorities should not limit bank mergers if efficiency gains are high and local markets generate enough competition.

Subsequent research included theory and data on performance measures other than return on assets, such as operating efficiency, net interest margins and interest rate spreads (Demirguc-Kunt and Huizinga, 1999; Saunders and Schumacher, 2000; Brock and Suarez, 2000a; Martinez-Peria and Mody, 2004; Pasiouras and Kosmidou, 2007, Athasanoglu et al, 2008). Technically speaking, it is common to argue that high interest margin is associated with poor performance by leading to disintermediation in the banking system. Whereas low deposit rates generate lower returns on deposit accounts by discouraging savings, high rates raise cost of external financing for firms, thus stifling investment activity. A high level is indicative of inefficiency, high cost of banking services and lack of competition. Conversely, high interest margins may bolster profits in the presence of strong capital base (capital-to-asset ratio). This is a "policy trade off" that Saunders and Schumacher (2000) and Barajas et al (1999) pointed in their analyses. If earnings from high interest rates are channeled into the capital base of the banking system, banks can become profitable (Barajas et al, 1999:199). Yet this relationship holds true in less competitive markets, indicating the trade off between consumer welfare, monopolistic profits and strong capital base. For example, high spreads, non-performing loans and strong capital ratios were shown among Colombian banks in Barajas et al (1999).

Research by Basu et al (2004), and by Gelos and Roldos (2002) highlighted the effects of market consolidation on bank performance and market structure evolution respectively. Focusing on Argentina’s experience from 1995-2000, Basu et al (2004) examined the impact of consolidation on two types of performance indicators: Return on equity (ROE) and insolvency risk. Market consolidation was measured in terms of privatizations and mergers/acquisition activity within a panel of 100 banks. Overall, they found a beneficial and statistically significant effect of consolidation on bank performance. While mergers and privatizations strengthened bank soundness by reducing insolvency risks, acquisitions decreased return on equity (ROE). The acquisition of "weak" and "underperforming banks" by healthy banks may be the cause of this effect—a topic considered for future research. Among bank specific variables, bank’s leverage and "loan portfolio diversification" were the most important factors.

On the other hand, Gelos and Roldos (2002) examined the development of competitive conditions in Argentina, Chile and Mexico. They used share in total deposits of the largest banks and Hirshman-Herfindahl Index (HHI) as proxies for market competition. Additionally, their use of estimates based on Panzar-Rosse statistics indicated that level of concentration has not decreased in these markets. While Argentina and Mexico had the highest consolidation score due to foreign acquisitions of largest banks, the ultimate effect of consolidation on competition was marginal. According to their findings, foreign bank entry and liberalization policies offset a decline in competitive pressures despite significant rise in takeovers. In the case of Mexico, however, Haber and Mussachio (2005) observed that Mexican banks began to restrict private credit in "both absolute and relative terms" in the post-liberalization era, but they found no relationship between reduced lending and foreign bank entry. Their evidence did indicate, however that, foreign banks charged lower interest margins and were able to better evaluate borrowers’ risk than domestic banks.

Later research on banking markets took several directions. The studies testing market power and efficiency hypotheses explored how market structures (e.g., consolidation, privatization; liberalization) affected the efficiency of financial intermediation as well as market power. Those using a balanced panel of banks, however, did not include a larger sample size, thereby effectively excluding mergers and acquisitions (Chortareas et al, 2011). Studies using sophisticated techniques (e.g., "stochastic frontier model") examined managerial efficiency (X-efficiency), scale and scope economies (Carvallo and Kasman, 2005; Wong, 2004). Wong (2004) studied the relationship between market structure, competition, and "efficiency in intermediation" in six Latin American banking systems for the period of 1995 to 2000. He concluded that most banking systems did not operate efficiently but high concentration was not the reason behind low efficiency. For example, Chilean banking industry was concentrated but operated more efficiently than other banking systems. In Argentina and Colombia, bank efficiency decreased with macroeconomic instability. Their results negated the Structure-Conduct-Performance hypothesis because finding a measure of market power (or concentration) that is "comparable" across banking systems is difficult. Ultimately, efficiency differences arise from "the differences in cost and revenue structures across countries, which might be due to different management and accounting practices, and also to very different economic environments" (Wong, 2004:29).

A high level of industry concentration and foreign ownership has characterized Mexico’s banking system. Following the 1995 peso crisis, foreign banks acquired more than 80% of Mexican banking assets. While critics blamed foreign banks for earning collusive profits by eliminating competition in local markets, proponents saw them as a critical source of "capital as well as skills, technology, and management know-how" (Schultz, 2004:2). Against this background, Schultz (2004) examined the impact of foreign bank entry on Mexico’s banking system for the period 1997-2004. Using a balance sheet and income statement for all commercial banks, he showed that foreign banks had a "positive" but "limited" impact on banking sector performance. For example, foreign banks significantly contributed to "recapitalization" of the Mexican banking sector and "reduction of bad debt" (or improvement in asset quality). These were the primary reasons for the government’s decision to ease limitations on foreign ownership after the peso crisis. While foreign banks increased the "fee income" and "interest rate margins" of their domestic acquisitions, however, they had no significant effect on "domestic credit provision", "administrative cost" or "employment levels" of Mexican banks. Despite an increase in total "productivity", foreign banks failed to capitalize on FDI or to improve "operating efficiency" often associated with foreign investment. Overall, these problems are related to the low level of competition and institutional incentives in the Mexican banking system (Schultz, 2004:34).

Further studies provided crucial insights into the diversity and significance of the impact of financial market changes (e.g., liberalization, globalization, deregulation) on local banking systems. Some can be broadly interpreted as favoring "market power" theories. Chile is an exception because it is known to have a well-developed banking sector. Berstein and Fuentes (2003) reported that Chile had inflexible interest rate policies for the period of 1995 to 2002. High level of concentration was the main reason for the "rigidity" in bank deposit rates. Gruben and McComb (2003) studied the relationship between the level of competitive intensity and privatization in the Mexican banking industry focusing on all commercial banks from 1987 to 1993. They estimated competition with "maximum likelihood method". Their results indicated that market competition decreased and bank risk increased with an introduction of subsidized deposit insurance in the post-liberalization era.

In a paper titled "Market structure, profits, and spreads in the Mexican banking industry", Chortareas et al (2009) provided evidence of a strong relationship between profitability (dependent variable) and bank capital ratios. They did not identify high industry concentration and market shares as the cause of high interest rate spreads (another dependent variable) while their inefficiency measure (cost/income ratio) was negatively and significantly related to profits thereby lending strong support to efficiency hypothesis. By implication, their findings suggested that Mexico should adopt liberalization policies directed towards mergers/acquisitions. Foreign investments should not be discouraged on the grounds of weakening competition. Solis and Maudos (2011) examined the changing nature of competition in the Mexican banking industry following the period of deregulation, privatization and liberalization (1993-2005). They measured competition using the Lerner Index and Panzar and Rosse’s H statistics.  Their results indicated that banks used a "cross-subsidization strategy" that decreased competition in the deposit market but increased in the loan market.

In Argentina, the impact of market structures on bank performance differs from Mexico’s and foreign ownership of banks is less significant as well. Most recently, researching the impact of "loan portfolio diversification" on bank returns during the business cycle (before and after crisis) in Argentina, Bebczuk and Galindo (2008) drew attention to the effect of bank size and ownership. They measured diversification using the Herfindahl index and "the share of total corporate loans to tradable producing firms" (Bebczuk and Galindo, 2008:210). Their analysis indicated that more diversified banks with a "higher tradable share" had stronger return on assets and lower nonperforming loans. Big and foreign banks had more diversified portfolios than small banks concentrated on "relationship based lending". Diversification, however, was more valuable at the onset of the financial crisis. Using firm-level data for Argentina, Chile, Colombia, Mexico, and Peru, Martinez-Peria and Mody (2004) found that foreign banks (especially those entering the market as "de novo" banks) charged lower interest than domestic banks; thus, financial systems with greater foreign entry became more competitive (in price terms) and cost efficient. Overall, their results lend support to the Efficient Structure Hypothesis that can foreign banks can assist countries in rebuilding banking systems that are more competitive and efficient by introducing new standards and practices. On the other hand, Levy-Yeyati and Micco (2007) proposed evidence that contradicted this view. They reported that foreign bank entry (in most cases) resulted in more concentrated markets, as evident in higher profits and higher charter value for foreign banks. Moreover, bank risk increased with low level of competition after foreign entry, acquisitions and mergers between large banks.

Overall, institutional and firm-level complexities challenge conventional interpretations based on the efficiency theory of banking. In addition, terms like "efficiency" or "profitability" cannot be easily measured when there are significant variations in the institutional make up of financial systems. For example, in a rapidly liberalizing but inadequately regulated financial systems, it is expected that banks perform poorly. International financial integration and liberalization may carry risks for some countries, especially in the absence of adequate capitalization of banks. These risks generate macroeconomic imbalances that may in return influence banking sector behavior. As comparative analysis by Brock and Suarez (2000a; 2000b) showed, both microeconomic and macroeconomic factors contributed to high interest rates in Latin America. In Mexico, Argentina, Bolivia, Chile, Colombia and Peru, the level of GDP (as a proxy for financial instability), inflation, and bank capitalization and provisioning for non-performing loans were most important determinants of bank spreads. In weaker financial systems like Peru and Bolivia, spreads declined when non-preforming loans increased. On the other hand, in robust economies like Argentina, Chile and Colombia, low spreads were associated with higher capital adequacy. Overall, the legal and regulatory environment, rather than microeconomic (competitive conditions) and firm-specific factors alone, had a major impact on bank spreads.

3. Institutional Background

Both Mexico and Argentina rapidly restructured their banking systems following the financial crises in 1995 and 2001-2002 respectively. Unlike Mexico, Argentine domestic banks have played an increasingly significant role in financial intermediation since then. State and privately owned domestic banks rose to top ranking in terms of total assets. Currently, there are three types of industry groups that represent bank ownership. The Association of Argentinian Banks (Asociación de Bancos Argentinos ADEBA) consists of private domestic banks; the Association of Banks of Argentina (Asociación de Bancos de la Argentina ABA) consists of private foreign banks; and the Association of Public and Private Banks of the Argentine Republic (Asociación de Bancos Públicos y Privados de la República Argentina ABAPPRA) mostly includes public and co-operative banks (EIU, 2011:10-11).

According to the Association of Argentinian Banks (Asociación de Bancos de la Argentina ABA), the top ten domestic banks –private, state-owned, provincial, co-operative—accounted for 60.93% of assets in the banking sector in April 2011 (EIU, 2011:11). Our sample includes only commercial banks in all categories of ownership as indicated in Table 1 that ranks banks in terms of total assets and latest accounts to date, December 2011. Top 10 banks account for 79%-80% of total industry deposits; market share provides an aggregate measure of concentration in the Argentine financial system, which is marked by a larger concentration of deposits in domestically owned banks. Among top 10 commercial banks, 3 are public, 7 are private, and Citi Bank (US), HSBC (UK) Banco Frances BBVA (Spain) and Banco Santander Río (Spain) are foreign-owned. Unlike Mexico, foreign banks are less important players in Argentina’s banking sector in the aftermath of the 2001-02 financial crisis.

According to the central bank (Banco Central de la República Argentina BCRA) and the Association of Banks of Argentina (Asociación de Bancos de la Argentina ABA), the top 10 foreign banks held 25.5% of total banking sector assets as of April 2011. Banco Santander Río  (Spain), BBVA Banco Francés (BBVA Spain), HSBC (UK), Citibank (US) and Standard Bank (South Africa) are the top five foreign banks that are most prominent in terms of market share (EIU, 2011:13).  In our sample, South African owned bank Standard Bank, was not ranked among top 10 banks at the time of observation, December 2010; it ranked 12 among all country banks, therefore was not included in our analysis of market share in the top 10 list. As Table 1 indicates, foreign banks held 19.5 % of banking sector deposits in 2010.

Overall, public banks are prominent players in Argentina. They are banking companies owned by the central or provincial governments. As the largest bank in the counry, Banco de la Nación Argentina "manages the government’s finances, receiving all federal tax revenue and making revenue-sharing payments to the provincial administrations on behalf of the federal government" (EIU, 2011:12). Banco de la Provincia de Buenos Aires is the second largest domestic bank, owned by the province of Buenos Aires; it channels funds from the central government to the province. Privately owned Banco de Galicia –the third largest domestic bank in the country, follows foreign-owned Banco Santander (Spain). Galicia is a diversified financial services company, consisting of non-bank businesses, such as insurance firm, mutual fund, factoring and leasing firm, brokerage house and warrant company (EUI, 2011:12).

Table 1: Main Characteristics of Top 10 Commercial Banks in Argentina, December 2010

Total Assets

($)

Net Income ($)

Net Interest Margin %

Return on Avrg. Assets %

Return on Avrg. Equity%

Market Share of Deposits

Banco de la Nación Argentina

35,622,877

589,687

0.55

2.02

22.9

0.294

Banco de la Provincia de Buenos Aires

10,315,319

136,198

5.33

1.49

27.78

0.095

Banco Santander Rio (Spain)

9,041,254

404,676

6.71

4.93

46.08

0.068

Banco de Galicia

8,922,877

138,423

9.18

1.75

21.46

0.076

Banco Macro

8,474,317

255,410

5.51

3.35

23.97

0.061

BBVA Banco Frances (Spain)

8,235,642

302,856

6.2

4.06

33.91

0.059

HSBC Bank Argentina (US)

4,893,667

144,328

6.96

2.7

24.36

0.037

Banco de la Ciudad de Buenos Aires

4,470,576

161,603

11.52

4.09

33.57

0.039

Banco de San Juan

4,058,215

115,293

6.94

3.31

29.99

0.034

Citibank NA (US)

3,736,350

138,069

7.04

3.7

26.17

0.031

Source: Bankscope Database of Fitch/IBCA/ Bureau Van Dijk ($ in thousands). Economist Intelligence Unit, Argentina: Country Finance, November, 2011. Market shares are author’s own calculation as ratio of each bank’s total deposits to total deposits of industry.

Compared to Argentina, Mexico’s banking industry is characterized by a high level of concentration and greater foreign bank penetration. According to one IMF study, Mexican banks rank among the lowest in Latin America and other emerging economies of similar GNI in terms of "financial intermediation" and "credit to the private sector" (IMF, 2011:13). Soon after the 1995 peso crisis, the Mexican government introduced a number of measures aiming to bolster bank capitalization. As part of the bailout program, foreign investors purchased three largest banks that were in danger of failing (EIU, 2012a: 15). In 1998, a new law required foreign banks to increase the maximum amount of bank capital from 9% (first introduced by NAFTA in 1994) to 25% (Hernandez-Murillo 2007:424). Nevertheless, trend toward industry concentration has risen, mainly as a consequence of bank mergers and takeovers while the number of foreign banks increased substantially. Chortareas et al (2009) noted that over the period 1996-2003 Mexico had a small number of large mergers and acquisitions (M&As) that decreased the total number of banks in the system by 22% and increased industry concentration (HHI Index) by 29%.

In June 2011, the three largest banks controlled 55% of total bank assets while 42 commercial banks accounted for more than half of the financial system’s assets. Seven large banks dominate the banking system, controlling 82% of bank assets; five of them are foreign-owned subsidiaries of prominent international banks. The remaining 34 banks are a mixed group offering "corporate and consumer lending as well as niche banking, creating a two-tiered industrial organization" (IMF, 2011:13-15). Domestic banks control a small share of the market in terms of total assets. For example, the top ten domestic banks at end-2011 owned only 24.6% of the banking system’s assets. Four of the five largest banks were foreign owned and controlled 61.04% of total assets. Overall, the top ten foreign banks held 71.5% of assets (EIU, 2012a:13-15).

Table 2: Main Characteristics of Top 10 Commercial Banks in Mexico, December 2010

Total Assets

($)

Net Income

($)

Net Interest Margin %

Return on Avrg. Assets %

Return on Avrg. Equity%

Market Share of Deposits

BBVA Bancomer (Spain)

83,728,929

1,412,545

5.31

1.64

17.12

0.179

Banco Nacional de Mexico, BANAMEX (Citigroup-US)

83,220,532

1,137,862

5.71

1.45

12.18

0.256

Banco Santander (Spain)

44,303,643

856,441

4.78

1.74

16.09

0.082

Banco Mercantil del Norte-BANORTE

42,073,865

392,995

4.36

0.92

10.67

0.151

HSBC Mexico (UK)

29,569,696

10,514

5.89

0.03

0.37

0.084

Banco Inbursa

14,666,697

370,328

4.89

2.41

12.03

0.042

Scotiabank Inverlat (Canada)

12,605,083

156,065

6.76

1.26

8.08

0.039

Banco del Bajio

5,868,279

54,500

4.18

0.98

8.10

0.019

ING Bank (Netherlands)

4,846,884

65,933

5.88

1.02

12.32

0.003

Banco Ixe

4,846,256

276

2.38

0.01

0.07

0.017

Source: Bankscope Database of Fitch/IBCA/ Bureau Van Dijk ($ in thousands); Economist Intelligence Unit, Mexico: Country Finance, March, 2012. Market shares are author’s own calculation as ratio of each bank’s total deposits to total deposits of industry.

Table 2 indicates that foreign banks are dominant players in Mexican banking. At the end of December 2010, six of the ten largest banks were foreign owned and controlled 64.3% of total deposits. The largest banks are subsidiaries of international banks that are competing for a larger share of the local banking sector, such as BBVA (Spain) and BANAMEX (US). The US-based financial services group Citigroup has owned BANAMEX since 2011. The largest domestic bank is Monterrey-based Mercantil del Norte (Banorte), which is the third-largest bank in the country by total assets. Banorte actively seeks to gain a bigger share of the remittances market in Mexico and the US. Its "strategy has included forging cross-border alliances and acquisitions, including the purchase of 70% of US-based Inter National Bank for US$259m in November 2006, a move designed to target Latinos living along the Texas- Mexico border" (EIU, 2012a:14).

Overall, although the Mexican banking system appears to be liquid, profitable and well capitalized (for example, capital asset ratio stood at 16.5% as of June 2011, according to IMF’s financial stability assessment report), there are concerns that high level of concentration may potentially harm lending and market competition if not regulated properly. In addition, despite a high level of foreign investment in Mexican banks, local firms still face problems obtaining credit, often leading them to acquire funds in international capital markets. In Argentina, banking sector faces other problems. According to Moody’s (2011), 28.5% of total bank assets in 2010 include public loans and government securities, indicating exposure to government risk. At the same time, a lower ratio of non-performing loans and higher capitalization indicate low level of credit risk in both banking systems. Only in the scenario of adequate lending and more credit availability (not for private sector only but also for households), banks can foster financial development, although it is difficult to forecast accurately.

4. Methodology and Data

We collected the data from the Bureau of Van Dijk’s Bankscope database—an electronic Publishing maintained by Fitch/IBCA/ Bureau Van Dijk (Bankscope, 2011). Bankscope is the primary source of information for all banking firms and firm-level proxies. The data were drawn from yearly income statements for the commercial banks operating in Argentina and Mexico during the period 1994-2011. In an effort to compare banking systems, we sampled 62 banks from Argentine and 37 from Mexico. In the final estimation (Table 8), this came to a total of 498 (Argentina) and 161 (Mexico) units of observations. Since our sample of banks was observed for each year, we estimated our model using a balanced panel of data. While unbalanced data better allows for "heterogeneity" between units of observation and time-effects, as indicated in previous research (Chortareas et al, 2011), we chose balanced data for one reason: In order to account for the longest period as well as the most recent changes that took place in both countries for the period considered, we excluded units of observation (banks) that were missing for those years. In addition, time-series for financial and macroeconomic variables were obtained from the database of Development Indicators & Global Development Finance of the World Bank.

The emerging market literature provided the foundations for the variables that measure commercial banking performance (Demirguc-Kunt and Huizinga, 2000; Athasanoglu et al 2008; Naceur and Kandil, 2008; Chortareas et al, 2011). The performance measure (profitability) is represented by Roa, namely the ratio of net income to total assets. Roa indicates how efficiently banks generate higher earnings from their asset base and financial resources. While some analysts use Roe (return on shareholder’s equity) as a primary measure of performance, Roa (return on assets) is a better metric because it illustrates bank management’s ability to use resources in order to generate profits. Furthermore, Roa indicates not only managerial decisions but also economic factors and government policy (Sufian and Chong, 2010:95). Hagel et al (2010) argue that Roa is a better measurement of performance since Roe may hide excessive leverage that companies take in order to maintain a " healthy" Roe and thus "obscure" under-performance.

Following the literature, we employed four types of independent variables as a proxy for financial performance—firm level, market structure (industry-related), financial structure and macroeconomic. The firm-level variables included in the regressions are bank risk (brisk), tstdep, credit risk (cris), operating cost (opc) and market share (ms). Brisk, which is also an indicator of bank’s capital adequacy or degree of capitalization, is calculated as the ratio of equity to total assets. Higher capital ratio is an indicator of lower leverage and risk, which therefore leads to lower expectations of bankruptcy and less borrowing costs. Furthermore, by providing additional liquidity to the bank, strong capitalization offers safety net during a financial crisis. Hence, we expect that profitability level is higher for better-capitalized bank. While it is true that higher capital ratios are associated with more profitable banking institutions (positive relationship), the opposite relationship can also be true. In the presence of higher capital, banks take less risk and therefore earn less profit (Chortareas et al, 2011).

The variable, tstdep, (time and savings deposits/total deposits), reflects the streams of income that banks have in terms of risk adversity (Heggestad, 1977). Therefore the higher ratio, the more risk-adverse a bank is, reflecting more reliance on deposit based funding and less reliance on more volatile, non-deposit borrowings. Cris is incorporated into the model as a proxy for "credit risk" and is measured as loan loss provisions/total assets. This ratio measures how much a bank is setting aside for bad loans each year relative to its total assets. Higher bank provisioning for bad loans indicates "higher loan provisioning expenses" and thus "higher reserve coverage of non-performing loans" – leading to lower return on assets (Crystal et al 2002: 4). According to Miller and Noulas (1997) a higher ratio of credit risk, as indicated by increases in loan loss provisions, is the main reason for lower return on assets or decreased profitability. This is related to the provisions for losses from high-risk or poor-quality loans. Banks with a higher level of loan loss provisioning are typically reported having higher initial problem loan levels. The amount of provisions or allowance set aside for doubtful debt, for example, is an important indicator of a bank’s asset quality. Thus, higher amount may also indicate that banks are more affirmative in addressing quality deterioration. Therefore, the sign of the relationship between cris and roa can be both positive and negative.

Finally, the operating cost ratio (total non-interest expense/total assets) is expected to negatively impact profitability— the higher the operational efficiency level (or lower operational cost to asset ratio), the higher the profit level of a bank (since profitable banks aim to minimize their operating costs). This variable represents the costs that a bank must incur in order to operate, such as total amount of employee salaries and benefits, property taxes, bad debt allowances as well as the use of technology, equipment and service fees. Higher ratio implies a less efficient administration in regards to expense management.

Furthermore, market share (ms) is included as a proxy for "relative" market power and is expected to positively impact profitability. Market share is measured as bank i’s total deposits at time t divided by total industry deposit during that time. A positive sign of market share in regression output would provide support for the "Relative Market Power Hypothesis"— therefore banks with a relatively larger market share are expected to enjoy market power—that is to set prices as they think "fit" and earn high (non-competitive) profits without confronting the usual market restraints. Despite the positive relationship between ms and roa, there is no role for hhi in the relative market power hypothesis (Chortareas, 2011). The HHI index is an industry-specific or market structure variable and is included as a proxy for market concentration of all firms operating in the chosen industry during this period. Measured as a sum of squared market shares of all banks (in terms of total deposits), it constitutes a direct measure of the degree of monopoly in banking markets. Under the "Structure-Conduct-Paradigm", a positive sign of hhi suggests "collusive" profits and non-competitive market structures. Accordingly, a concentrated market will result in higher profits for banks as they have the power to capture a higher market share. If efficiency theory holds true, however, both hhi and ms would lose their explanatory power and is insignificantly associated with profitability (Gilbert 1984; Molyneux, 1992; Berger, 1995; Lloyd-Williams et al, 1994; Naceur and Kandil, 2008; Chortareas, 2011).

Following Demirguc-Kunt and Huizinga (2000); Pasiouras and Kosmidou (2007), Naceur and Kandil (2008), Kosmidou (2008) and Chortareas et al (2011), we included country-specific, macroeconomic variables—inflation (ir), money supply (m2gdp), real interest rates (ri), GDP per capita income growth (gdpergr) as well financial structure indicators—market capitalization of listed companies (mcapgdp), total value of stocks traded (tvtgdp), domestic credit to the private sector (dcp) to control for the effects of external variables on banks’ performance during this period. We expect a positive relationship between gdpercgr and roa since bank profitability is pro-cyclical—meaning that it is sensitive to macroeconomic conditions and higher per capita income encourages banks to lend more.  It is also hypothesized that inflation negatively affects bank profitability. As a growing body of literature suggests, inflation "interferes with the ability of the financial sector to allocate resources effectively" (Boyd et al, 2011). Finally, the sign of the relationship between interest rates (ri) and roa may vary; by restricting economic activity in the form of higher borrowing costs, high interest rates allow banks to charge higher margins and thus earn higher profits.

5. Econometric Model

In order to identify the determinants of bank profitability in Argentina and Mexico, we conducted a series of analyses based on the Arellano-Bond Generalized Method of Moments (GMM). All computations were carried out with STATA. Analyses were conducted separately for Mexican (37) and Argentina banks (62) combining cross-sectional (banking firms) and time-series (yearly) data. Since banks were observed each year at different levels of analysis (firm-level, microeconomic, macroeconomic), number of observations ranged from 500 to 498 over a 17-year period. We also included a lagged measurement of profitability to refine the model and to control for auto-regression associated with the use of annual time-series.

The advantage of using the Arellano-Bond (1991) estimation is explained in depth in some of the major works on banking. As Naceur and Kandil (2008) explain, panel data analysis is better at correcting "inconsistent estimates" that are common in pure cross-section regressions. In such regressions, the inconsistent estimation results from the "omitted variables bias" and "potential endogeneity of regressors" when the error term contains "unobserved" firm effects and is correlated with the dependent variable. Given that the determinants of bank performance vary over-time, panel data provides depth of information missing in pure cross-sectional regression. Furthermore, firm-specific effects are better controlled in a panel data setting.

Table 3: Measures of Commercial Banking Performance

Variable Description

Hypothesized Relationship with Profitability

Dependent

roa: This is a measure of profitability and is calculated as net income/total assets

NA

Independent

Bank Characteristics

brisk: This is a measure of "bank risk" and capital adequacy/capitalization and calculated as equity capital/total assets. High equity capital asset ratio is an indicator of lower leverage (and therefore lower risk).

+/-

tstdep: This is a measure of deposit composition and calculated as Time Deposits and Savings/Total Deposits. Higher ratio indicates safer (risk-averse) and more stable streams of income for banks.

+

ms: This is a measure of market power and calculated as market share of each bank’s deposits for a given year. Higher market share implies banks with market power and higher levels of profit.

+

cris: This is a measure of "credit risk" and calculated as loan loss provisions/total assets. Higher ratio indicates high level of high-risk loans (doubtful debt, unpaid loans) relative to a bank’s total assets for a given year.

-

opc: This is a measure of bank operating cost and calculated as total non-interest expense/total assets. Higher ratio implies a higher cost of operating and a less efficient management—thus lower profitability.

-

Microeconomic (industry-level) and Financial Structure

hhi: Herschman Herfindahl Index for deposits. This is a proxy for market concentration and calculated as sum of squared market shares of each bank’s deposits and short-term funding for a given year. A concentrated market will confer higher profits for banks as they are able to capture a higher market share.

+

dcp: This is a measure of credit growth in the domestic economy and calculated as domestic credit to the private sector as % of GDP for a given year. Higher domestic credit indicates higher profitability.

mcapgdp: This is a measure of total value of all listed companies’ outstanding shares times the share prices. It calculated as market capitalization of all listed companies/ GDP for a given year. Higher capitalization indicates higher overall market liquidity and volume, and thus higher profitability.

+

+

tvtgdp: This is a measure of stock market size and calculated as total value of stocks traded divided by GDP for a given year. Higher the total value of shares traded indicates higher profitability.

+

Macroeconomic

gdpercgr: GDP per capital growth (annual %). This is measured as "annual (%) percentage growth rate of GDP per capita based on constant local currency" (World Bank, 2011). Higher growth indicates higher demand for banking services.

+

inf: Inflation consumer prices annual (%) as measured by the consumer price index. The sign of inf with roa may vary although it is largely expected to be negative.

-

ri: Real interest rate (%). This is a measure of the "lending interest rate adjusted for inflation as measured by the GDP deflator" (World Bank, 2011).

+

m2gdp: This is a measure of level of money supply in the economy and calculated as the sum of money and quasi money (M2) as % of GDP. Higher money supply indicates higher profitability for banks.

+

As developed in Naceur and Kandil (2008), Arrelano and Bond (1991) suggest using GMM for estimating panels with a limited number of years and a large number of observations, which is the case in our analysis—498 and 161 observations over a 17 years time period. GMM is estimated by first differencing the initial equation and then using lagged values of regressors as instruments in the first differenced equation. This suggests a joint estimation of first difference equation and lagged values of the right-hand side variables. Using GMM in STATA has allowed us to run these equations automatically. Table 3 describes the measures (variables) of banking performance used in the study. Bank performance is measured according to the empirical model grouped into four types of variables, where is represented by Return on Assets (Roa) for the firm I during the period t; are bank specific variables for bank I at time t; are macroeconomic variables, are industry-related variables, and are financial market variables. We introduce a linear regression model of the short form to test the relationship between profitability and the above variables (Athasanoglu, Delis and Staikouras, 2006).

After introducing the short form, we specify the empirical model as follows:

As shown above, our equation has both temporal and cross-sectional dimensions, I, in this case, is the profitability of bank I at time t, with i: 1…..,N ; t=1,……,T; c is a constant term, is the one period lagged profitability and is the "speed of adjustment to equilibrium" (Naceur and Kandil, 2008:7). X are explanatory variables grouped into firm-specific, industry-related (microeconomic), financial structure and country-specific macroeconomic determinants, j, l, m, n, respectively; captures the random error or disturbance in time denoted by white noise (residual).

6. Results and Discussion

6.1. Summary Statistics

Our expectations about the effects of each independent variable on dependent variable (roa) are summarized in Table 3. Profitability of a bank positively correlates with capitalization and available long-term funds and negatively correlates with its credit risk and operational costs. At the sectoral level, a higher concentration of the sector is expected to lead higher profitability for banks due to their ability to capture higher market shares. As financial indicators, credits provided to domestic firms is expected to have a positive effect on profitability and as well as the depth of the financial sector (market capitalization of all listed companies and total volume of traded stocks). Finally it is expected that growth of the economy positively affect bank profitability. Meanwhile increased money supply also is related to higher profitability for the sector as a whole. Bank profits is negatively correlated with inflation rates and positively correlated with real interest rates.

Table 4 reports the bank observations used for each country and the averages of key financial variables used in our analysis. Based on these indicators, it is clear Argentina comprises a larger banking market—around 62 banks on average per year and a maximum of 802 bank level observations (Table 5). In contrast, Mexico presents around 37 banks on average per year, with max of 353 bank level observations. In addition, the average size of banks in terms of total assets indicates the peculiarities of the banking systems in our sample. For example, as seen in Table 5, average total bank assets in Mexico are roughly 2 times larger than that of Argentina. At first blush such a result might be surprising given that Argentina has larger number of banks. The difference can be explained by the 2001 crisis when banks dramatically lost assets, failed, or were taken over by other banks. Since then banks have scaled back and asset recovery has been slow, postponing industry growth. Another reason might be that Mexican banking has a higher concentration of markets and foreign banks with larger asset size than Argentina’s, such as Bancomer and Banomex (Table 1 and Table 2).

In terms of profitability (Roa), the ratios show an average of negative 0.9% (Argentina) and negative 0.1% (Mexico). This is possibly related to the recent financial crises (especially 2001-2002 in Argentina) as well as the global downturn that have taken place since 2008. However, it is also evident that the scores started to recover after 2009 and displayed a steady record of growth in the following years (Figure 1). In both countries, market consolidation due to mergers/acquisitions, bank failures, and prudential regulations (higher capital requirements) increased in the aftermath of financial crises. However, the stable pattern in Mexico’s Roa may be related to the highly concentrated banking sector (as a "shield" against crises), although this is a different topic of research than the one investigated here.

Table 4 also shows that profitability (Roa) and other bank ratios show a higher level of variation between two countries. The coefficient of variation, calculated by dividing standard deviation to mean, gives an idea about the scope of variation. In the Argentine case, the score is -11.18, indicating a high level of variation; in the Mexican case, it presents a very high level of volatility with a score of -65.02. This finding leads us to conclude that these countries demonstrate how the banking and financial systems are far from stable. The market shares of banks are also relatively more volatile in both countries, 2.37 in the Argentine case and 1.84 in the Mexican sector. These higher levels of volatility suggest that market shares and profitability are dependent on external factors.

An unstable banking system is also observable in other indicators employed in our analyses. Credit risk (cris), as measured by Loan Loss Provisions/Total Loans, is the most volatile financial indicator in Argentina with a coefficient of variation of 12.78, meanwhile it is only 2.54 in the Mexican case. There is also large variation in money supply and market concentration (hhi) indicators, which may be attributed to structural differences between these two banking systems. Other bank level variables are relatively stable overtime and variation is due to change in overtime and interbank variation is relatively low.

Table 4: Descriptive Statistics of Data

Argentina

Mexico

Mean

Coefficient of Variation

Observations

Mean

Coefficient of Variation

Observations

roa

-0.009

-11.18

794

-0.001

-65.02

350

brisk

0.230

1.02

802

0.183

1.01

353

tstdep

0.439

0.49

674

0.649

0.41

287

cris

0.097

12.78

696

0.036

2.54

284

ms

0.023

2.37

789

0.047

1.84

343

opc

0.083

0.70

801

0.072

1.35

352

hhi

0.151

0.95

1116

0.205

0.28

592

mcapgdp

0.248

0.66

1054

0.289

0.29

629

dcp

0.173

0.28

1054

0.216

0.27

629

tvtgdp

0.028

0.76

1054

0.092

0.42

629

gdpercgr

0.026

2.24

1054

0.011

3.27

629

inf

0.062

1.09

1054

0.109

0.91

629

ri

0.071

1.21

1054

0.038

1.40

629

m2gdp

0.448

0.49

1054

0.276

0.07

629

Source: Bankscope and World Development Indicators, 2011.

Sectoral level statistics are relatively stable, highest level of variation is observed in market concentration (hhi), with a coefficient of variation of 0.95 in the Argentine case. Mexican statistics presents a much more stable financial sector. None of sectoral indicators has a coefficient of variation greater than 0.42 meaning that Mexican banking and financial sector is significantly stationary. Based on these statistics, it is possible to attribute higher volatility in profits to bank level factors, rather than environmental changes.

Among macroeconomic indicators, gross domestic product per capita growth is significantly volatile over time in both countries, with a coefficient of variation of 2.24 and 3.27 respectively. In the Mexican case, money supply (m2gdp) has a horizontal slope. Other indicators presented relatively stable pattern overtime.

Correlation coefficients between dependent and independent variables are presented in the annexed tables. We can summarize this table by underlining relatively strong relationships. In Argentina, bank profitability is negatively correlated with credit risk (-0.42) and moderately correlated with growth in GDP per capita (0.25) and money supply (0.24). Market capitalization, inflation and real interest rates seem to be negatively correlated with bank profitability. Among independent variables, capitalization seems to be negatively correlated with market share (-0.26) leading us to conclude that bigger banks are more open to risks. Credit risk is positively correlated (0.32) with ratio of operational costs to total assets and real interest rates (0.27). This score is negatively correlated with per capital growth in the economy (-0.40).

Market share is negatively correlated with operational costs, (-0.29) showing that bigger banks (or banks with larger market share) are less efficient in operational terms. Industrial concentration is positively correlated with the depth of the financial market (0.30); as the financial market grows, the sector becomes more concentrated and vice versa. Market capitalization is positively correlated with domestic credit provided to the private sector (0.51) and real interest rates (0.71). Meanwhile its relationship with GDP per capita growth and money supply is negative (-0.67 and -0.54). Domestic credit to the private sector is positively correlated with real interest rates (0.64) and negatively correlated with GDP per capita growth (-0.49), total volume of traded stocks (0.27) and money supply measured as M2 (-0.62). GDP per capita growth is negatively correlated with real interest rates (-0.77) and real interest rates are negatively correlated with money supply (-0.72).

In the Mexican case, correlation between bank profitability and credit risk is also negative (-0.69), whereas negative correlations are visible between bank profitability and operating cost (-0.86). Bank risk variable has a positive correlation with credit risk (0.31) and operating cost (0.55). Credit risk has a positive but less than moderate correlation with ratio of market capitalization to gross domestic product (0.22) and domestic credit volume (0.23). Moreover there is a less than moderate correlation between ratio of total savings to total deposits (tstdep) and market concentration index (0.25).

Table 5: Total Assets of Banks

Year

Argentina (n=62)

Tab

% Change

Mexico (n=37)

Tab

% Change

1994

1,422,343

NA

NA

NA

1995

25,466,260

1690.44%

29,297

NA

1996

36,569,446

43.60%

902,102

2979.16%

1997

56,458,095

54.39%

2,271,942

151.85%

1998

76,519,884

35.53%

89,104,609

3821.96%

1999

100,416,477

31.23%

83,664,953

-6.10%

2000

105,534,659

5.10%

94,413,934

12.85%

2001

86,323,412

-18.20%

123,526,716

30.84%

2002

39,632,108

-54.09%

145,804,017

18.03%

2003

46,043,789

16.18%

147,015,338

0.83%

2004

64,651,775