Financial Integration in Latin America: Background Paper: Issues on Financial Intergration in Latin America

Financial Integration in Latin America

Abstract

Financial Integration in Latin America

The Largest Private Banks

1. Banking assets of the largest banks in Latin America (LA) are heavily concentrated in Brazil. Brazil accounts for nearly two-thirds of LA’s banking assets, while Mexico contributes just one-tenth. Looking just at private banks, Brazil’s share is closer to its share in regional GDP (45% as against 40%). A few Brazilian banks have the strength and interest to become major players and establish a significant presence across the region. However, not all Brazilian banks are looking to expand abroad, stressing the potential still available domestically.

Figure 1.
Figure 1.

Indicators of Banking Market and Asset Size

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

2. Bank Itaú, based in Saõ Paulõ, has the ambition to further expand to all major markets in LA. The bank is close to the size of the entire Mexican banking system (US$420 billion in assets) and has already important cross-border operations in the region. The bank has grown through mergers and acquisitions (M&A)s in Brazil and cross border, being in corporate and investment banking in Colombia, Mexico, and Peru, and in retail and wholesale in Argentina, Chile, Paraguay, and Uruguay. With its most recent acquisition of Chilean Corpbanca (and merger with Corpbanca Colombia), the bank’s cross-border business will reach 13 percent now from 7 percent in 2011. Its strategy is to diversify its retail and corporate portfolio to other markets. It considers it more challenging to develop a retail business abroad, given the need for funding, while easier to follow corporate clients abroad. Its ultimate corporate vision is to go global. In Mexico, the bank is trying to develop the investment banking business using broker dealers, given it had encountered difficulties developing the credit card business. In Colombia, given the consolidated banking market, the bank will try to develop in investment banking.

Figure 2.
Figure 2.

Largest Private Banks in the LA-7

(Total assets, billions of US dollars)

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Note: Bradesco’s assets include the assets of HSBC which the bank is in the process of acquiring.

3. Investment bank BTG Pactual, based in Saõ Paulõ, aspires to be the largest investment bank of the region. It is easier for investments banks to establish operations abroad compared to retail banking, owing to the lower cost structure and initial investment involved in their operations. BTG Pactual started expanding to LA following the global financial crisis (GFC) when the “global bank” model was under severe challenge. It merged with Celfin Capital, a brokerage and asset manager with operations in Chile, Colombia and Peru, and set up a Greenfield brokerage in Mexico. It sees profit opportunities in these countries given their underdeveloped capital markets and the capital markets integration initiative within the Pacific Alliance (PA). As a 100 percent wholesale funding bank, it sees its expansion through LA as providing a more diversified wholesale funding base and does not plan to enter into retail, given high retail funding costs. Most recently, however, following the arrest of its CEO on corruption charges in November 2015, BTG Pactual has experienced bouts of market pressure, has been facing significant outflows from its asset management unit, and has debt coming due in the coming months. Its high reliance on wholesale funding has translated into noticeable swings in liquidity. Any further expansion of its operations in the region in the near term will depend on its ability to ensure funding at a reasonable cost.

4. Colombia’s three largest banks—Bancolombia (based in Medellin), Banco de Bogotá, and Davivienda (based in Bogotá) expanded aggressively to Central America. The assets of Colombian banks’ subsidiaries abroad reached US$50 billion, accounting for 24 percent of the total assets of the Colombian banking system. Colombian banks have attained a significant market position in Central America (on average: 22 percent). The aim of the expansion was to follow Colombian clients abroad as a first step and then diversify their portfolios to serve other clients too, using acquisition opportunities arising from the withdrawal of foreign banks from Central America and the high capitalization of Colombian banks. Banco de Bogotá has a very different loan portfolio in Colombia (leadership in corporate banking) than that in its BAC subsidiary in Central America/Panama (which has a much more important consumer credit and mortgage lending portfolio). This has enabled it to have a more diversified portfolio overall, as well as to benefit from cross-complementarities: bringing credit card technology from Panama to Colombia, and exporting corporate lending knowhow from Colombia to Panama. Bancolombia’s portfolio in Central America/Panama includes all core banking products (corporate lending, as well as consumer credit and mortgage lending), while mostly corporate banking in Colombia. Davivienda’s portfolio in Central America/Panama is now increasingly focusing on consumer lending, while withdrawing from corporate lending (which had been previously HSBC’s main portfolio focus), notwithstanding the high share of corporate lending, and smaller shares of consumer and mortgage lending in its loan portfolio in Colombia.

5. BBVA, based in Madrid, has a well diversified portfolio across LA. 50 percent of the bank’s profits are derived from its business in Mexico (30 percent) and South America. BBVA’s business model is retail banking, with market shares of 22 percent in Mexico and Peru, and 9.4 percent and 6.7 percent in Colombia, and Chile respectively. BBVA is the only bank in all four countries of the PA, which it views as a huge opportunity for growing its business. The bank sees room for expansion in LA, given low banking intermediation rates. Its model is to establish subsidiaries autonomous in capital and liquidity, which limits contagion risk between the group’s units and reduces systemic risk. The bank is potentially affected by the special ring fencing’ rules issued in Mexico in 2014, according to which not only can the Mexican authorities request full information on parent companies, but they can also stop dividends if they believe the parent company is in trouble. A further impediment for operating in LA is that in consolidation with the parent bank in Europe, home country assets might receive a lower rating (since they are in Mexican pesos for example). In general, more regulatory stringency by the European supervisory authorities, may constrain BBVA’s regional activities.

6. Similarly, Santander, based in Madrid, has been diversifying its portfolio in the region. 38 percent of the bank’s profits are derived from its business in Brazil (19 percent), Mexico (8 percent) and the rest of South America. Santander’s business model is mainly retail banking, focused on a few countries where it aims to reach at least a 10 percent market share (in Chile: 17 percent, Mexico: 14 percent, and Brazil: 8 percent). In Brazil, the objective is to grow its retail as well as corporate businesses. In Mexico, the objective is to grow more than the market, particularly with high income clients and SMEs, be one of the leading banks in financing the government’s infrastructure plans. In 2012 Santander sold a 24.9% stake in its Mexican bank through an IPO, following an IPO in 2009 of part of its Brazilian subsidiary, the proceeds of these sales are used to reinforce the group’s core capital. Santander and BBVA seem to have largely divided the LA markets between themselves, and where Santander is present, BBVA generally is not. Santander’s subsidiaries are completely autonomous in capital and liquidity, which limits contagion risk. There is also limited cross-funding between subsidiaries in LA, and excess liquidity cannot be moved easily (deposits cannot be shared across countries). Regarding Brazil, as Santander Brazil is not allowed to lend in dollars to Brazilian corporates, the branch in the Cayman Islands is used (as by other Brazilan banks) for foreign currency lending to Brazilian corporates.

7. Corpbanca, based in Santiago, was the first Chilean bank to expand abroad. The bank invested in Colombia in view of the high banking penetration rates and low opportunities to continue growing profitably in Chile. The bank had a 5 percent market share after 10 years in the Chilean market; the large difference between the profitability and cost of funding of the biggest three banks and the rest of the market made it hard for a bank with a smaller market share to sustain profits. After searching for a jurisdiction with similar policies to Chile, it bought Santander Colombia, which had a market share of 3 percent, and then also Colombian Helmbank, since it felt it would be hard to have a profitable operation with such a small market share. Both banks were retail focused, and together achieved a 6.5 percent market share in the Colombian market.

8. Banco de Crédito del Peru, based in Lima, has a small presence regionally. The bank has retail businesses only in Bolivia and Central America, and entered Chile and Colombia as an investment bank (and offering a microcredit business in Colombia), given that it did not have the capital to expand to those markets as a universal bank. The consolidated nature of the market in Brazil has been seen as a deterrent for entry. The other large domestic private bank, Interbank, has a strong domestic focus, given high interest rate spreads and ROE in Peru, as well as very low bank intermediation rates.

Table 1.

LA-7: Largest Banks Operating in Each Country

(Share of total assets in each country)

article image
Sources: National authorities and Bankscope.

Implementation of Basel Standards

Figure 3.
Figure 3.

Basel 2 Implementation Progress

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Data for G20 countries from BCBS progress report on implementation of the Basel Regulatory Framework (October 2015). Data for other countries from BIS FSI BASEL II, 2.5, and III Implementation Survey (June 2015).
Figure 4.
Figure 4.

Basel 2.5 Implementation Progress

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Data for G20 countries from BCBS progress report on implementation of the Basel Regulatory Framework (October 2015). Data for other countries from BIS FSI BASEL II, 2.5, and III Implementation Survey (June 2015).
Figure 5.
Figure 5.

Basel 3 Implementation Progress

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Data for G20 countries from BCBS progress report on implementation of the Basel Regulatory Framework (October 2015). Data for other countries from BIS FSI BASEL II, 2.5, and III Implementation Survey (June 2015).

Capital Definitions and Capital Ratios Across Latin America

Background

A robust bank capital framework helps ensure financial stability and sustain bank lending during economic downturns. Bank provisions and profits are an important buffer with provisions in particular able to absorb expected losses. However, in the event that losses exceed earnings capital provides banks with a cushion to absorb unexpected losses to reduce the risk of bank failures and prevent interruption of banking services and financing to the real economy. Unfortunately loss absorbency elements like provisions, capital definitions and actual regulatory capital levels across Latin America are not easily comparable even after using harmonized market-based measures.

Capital Definition and Adequacy

Capital definitions differ across Latin American countries and comparisons must be made with utmost caution. Some cross country differences in the computation of capital include, for example, the treatment of the revaluation of fixed assets; the accounting of profits from current or past accounting periods; treatment of investments in capital instruments or requirements on donated capital; and treatment of some deductions from capital (goodwill, intangibles and deferred tax assets); grandfathering of some capital (debt) components. Moreover, capital differs depending on the degree of consolidation undertaken, whether at individual (solo) bank level, banking group level, or at even higher at financial conglomerate level. Furthermore there can be sizeable differences in regulatory risk weights applied to the same asset classes across jurisdictions. Differences in the national definition of capital the Basel framework in use and accounting standards across Latin America imply that any direct comparison of total regulatory capitalization should be interpreted with caution.

uA02fig01

Regulatory Capital Requirement and Total Capital Above Requirement

(In Percent)

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Bankscope, company filings, country Financial Sector Stability Assessments, and Article IV reports.
uA02fig02

RAC Ratio for the Largest Rated Latin America Banks

(In Percent)

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Source: Standard & Poor’s.

Market Based Estimates of Capital

Some systemic Colombian banks have lower levels of capital in excess of the regulatory minimum than some regional peers. Regulatory capital requirements differ across Latin American countries with some higher—Brazil (11 percent), Peru, Guatemala and Uruguay (10 percent)—and some lower than Colombia’s (9 percent)—Chile, Argentina (8 percent) and Mexico (10.5 percent). The decision to choose a given level of national minimum regulatory capital reflects a series of factors, including supervisory judgment and discretion. The four largest banks in Colombian have lower capital than the large banks in some other Latin American countries. Total capital ratios in excess of the regulatory minimum requirement, stood at 2.9 percent, the lower end of regional peer comparisons. Attempts to obtain a more consistent harmonized measure of capital across Latin America have been tried by rating agencies but again depending on the measure used comparisons on quantity and quality of capital vary. For example, Colombian banks have lower levels of capital according to the Standard and Poor’s risk-adjusted capital (RAC) measure—which deducts all goodwill on the balance sheet from banks’ respective total adjusted capital. This measure is important inasmuch as Colombia has seen a large number the mergers and acquisitions following the financial crisis of the late 1990s that, together with the geographic expansion of the largest banks over the last few years, has created large amounts of goodwill assets. Using the Fitch Core Capital (FCC) measure Brazil, Chile and Colombia have much lower capital due to higher leverage of the system, sizeable investments in insurance companies and high levels of goodwill and deferred tax assets which are all deducted from equity to reach FCC levels.

uA02fig03

2013 Fitch Core Capital

(Generally Adequate Capital Ratios in Latin America, in percent)

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Source: Fitch Ratings.

Additional Loss Absorbency

While capital ratios on market based measures may seem low for some Latin America countries these countries have additional loss absorbency in the banking system. Many banks hold high levels of provisions (Brazil, Colombia), have lower NPLs (Colombia) and have more conservative risk weights (Colombia, Chile).

Basel III

Many Latin American countries are adopting Basel III standards, albeit at different paces. The adoption of Basel III standards should help address inconsistency of capital definitions with recent work by the Basel Committee ensuring harmonization of risk weights. Notwithstanding the move to Basel III actual implementation may still see differentiation of capital stem from differences in adoption of above minimum capital (Pillar 2 and conservation, countercyclical and D-SIB buffers). This may reflect the need to address supervisory failings and the desire to tailor capital to bank risks across Latin America which may well be above Basel III voluntary minimums in some countries.

Conclusion

The addition of further consistency to the already robust capital framework across Latin America will help to ensure financial stability and sustain bank lending during economic downturns. Challenges from moderating economic growth, low yield environment, volatility around US monetary policy normalization, cross border risks and conglomerate expansion require that Latin American banks adopt a more conservative long-term capital planning approach. Moving to Basel III should help and is attainable for most Latin American banking systems as current capital is sufficient to support transition and additional loss absorbency exists beyond capital.

Basel Implementation Progress

article image

Quantifying the Impact of Regional Integration

A. Measuring the Degree of Integration of LA Countries

9. The measurement of financial integration can be refined. Simple cross-country comparisons may paint a distorted picture of the degree of integration of LA markets relative to other regions, for instance, because countries that are less advanced economically often have shallower financial markets. This section attempts to quantify the extent to which LA markets are “under-integrated” given their economic fundamentals by controlling for factors such as the level of economic development (proxied by GDP per capita in PPP dollars), trade openness (exports plus imports divided by GDP), the past history of financial crises (measured by the Reinhart and Rogoff database indicators), the level of public debt-to-GDP ratio (which, as a stock variable, cannot be easily modified by the government), and the quality of the institutional framework (measured by the investment profile subcomponent of the International Country Risk Guide Index). Variables that are more directly and immediately affected by economic policy, such as the extent of capital controls, are not included, as the purpose of the econometric analysis is not to provide the best fit of the data but to control for exogenous factors.

10. The models relate financial integration to a set of control variables. In each specification, a measure of financial integration is regressed (either the baseline or alternative composite indices of financial integration presented in Box 2 or their subcomponent of openness) on its macroeconomic determinants and fixed effects. The degree of under or over-integration is then calculated as the difference between the estimated country (or region) fixed effect and the sample average of all country (or region) fixed effects. As the purpose of the regressions is to filter out the effect of certain fundamentals and not to interpret a causal model, the endogeneity problem, inherent to this type of analysis, is less of an issue. The following equation is estimated over a sample of 67 countries between the mid-1980s and 2014:
FIit=β.Xit+αi+ϵit

where FIit denotes the financial integration indicator, Xit are control variables, and αi is the fixed effect.

Building a Synthetic Index of Financial Integration

Our baseline composite index combines information from two main dimensions of integration: financial openness and financial convergence. The first component is the de facto openness of the financial account measured by the sum of stocks of foreign assets and liabilities as a share of GDP. The inclusion of this variable follows directly from the definition of financial integration (see section II in the main paper. The second component is the regional dispersion of stock market returns measured by the standard deviation of returns of Morgan Stanley Capital Interactions (MSCI) indices across countries of the same region (lower standard deviations would imply greater convergence). Although this indicator of regional convergence is widely used in the literature (Baele and others, 2004), it presents obvious drawbacks (in particular, differences in returns may be related to idiosyncratic risks) but the analysis is limited by data availability. To combine the two indicators, a principal component analysis is used, where the standardized variables’ weights are the squared factor loadings. The objective is to reduce the number of variables of interest into a single factor, which captures most of their variances (for the three indices constructed in this exercise, the first component explains more than 50 percent of the total variance).

The analysis also uses three alternative integration indices:

  • The first alternative index replaces the traditional broad indicator of external openness (stock of external assets plus liabilities as a ratio to GDP) with the narrower external liability-to-GDP ratio. Indeed, some countries may hold large proportions of financial assets abroad, while having a low level of de facto integration. These assets, which may coexist with capital controls, may reflect past capital outflows (e.g., Argentina) or large current account surpluses (e.g., China). Limiting the measure of openness to include only external liabilities is one way of circumventing this problem and testing whether the results still hold.

  • In the second alternative indicator, the first two components are identical to those used in the baseline index but a third component is added, which is the ratio of private sector credit by banks to GDP. There are two reasons why a measure of financial depth may enter the integration index. First, since financial integration allows savers to invest in a broader range of investment and risk-sharing instruments, while enabling borrowers to tap a broader range of financing and risk management instruments, at home and abroad, the concepts of integration and depth are closely related. Second, to reap the full benefits of integration and be a meaningful contributor to an integrated playing field, individual markets need to have a certain size. Thus, the depth criterion excludes markets that are too small even if they meet the other two criteria.

  • The third alternative index provides a better picture of regional integration by including a measure of relative regional openness (ratio of regional assets and liabilities to total foreign assets and liabilities of a given country), alongside global financial openness and regional convergence. The intuition is that countries are regionally integrated when they fulfill three conditions: they have to be (i) open globally, (ii) relatively more open to their neighbors, (iii) and present signs of financial convergence.1 Several variants of the regional openness concept are developed as defining regions can be tricky. The first is an 8 region world (advanced economies, Africa, Asia, emerging Europe, Latin America and the Caribbean, Middle East and North Africa, Commonwealth of Independent States, and other small states) based on IMF WEO classifications that emphasizes regionalism among emerging/developing economies. The second approach consolidates to just 4 regions (Asia, Europe, Western Hemisphere, and rest of the world) and captures the observed behavior that emerging/developing countries tend to integrate with nearby advanced economies (for instance, Mexico with the United-States, or Eastern with Western Europe). The third version replaces pre-determined regions with distance-based weights whose values rise when countries are geographically close. Bilateral financial positions are then weighted with this distance, so that the regional openness indicator increases when countries are more financially open to neighbors.

Box Table 1.

Principle Components for Financial Integration (FI) Indicators

article image

Country specific indicators combined into a single measure that captures global and regional integration through principal component analysis.

Observes integration of emerging markets by dividing the world into 1 “region” of advanced economies and 7 emerging market regions: Africa, Asia, Europe, Latin America, Middle East and North Africa, Commonwealth of Independent States, and other small states.

Broader measure that captures the integration of both emerging and advanced economies within one of 4 large geographic regions: Asia, Europe, Western hemisphere and other countries.

Observes the degree to which international financial partner countries are close (values near one are more regional) or distant (values near 0 are less regional). Constructed as 1 minus the normalized, weighted average distance between the reporting and all partner countries. Distances are normalized by dividing all distances by the maximum distance between any 2 countries. Normalized distances are then weighted by the reporting country’s share of either external assets + external liabilities or just the external liabilities vis-a-vis each of its partners.

11. The econometric results confirm that the LA-7 countries are under-integrated as a whole, although there are important differences between countries, as well as across the various dimensions of financial integration. In each model, the sign of the control variables is consistent with priors. The main result is that although the LA-7 countries do not appear under-integrated from the perspective of international cross-border capital flows, once broader measures of integration are used through the composite integration indexes, these countries do appear to be under-integrated, even after controlling for fundamentals. The extent of under-integration varies across the countries, and there is one notable exception, Panama, which is well-integrated across most specifications.

  • Table 2 shows the outcomes of various models explaining the degree of financial openness (measured either as the ratio of gross external assets and liabilities to GDP or as the liability ratio). The results suggest that LA-7 countries are relatively well integrated from an openness perspective compared to the sample average, but this result is partly driven by Panama and Chile, which clearly show a greater degree of openness than the others.

    Table 2.

    Financial Market Integration: Financial Openness

    article image
    Notes: Time dummies have been incorporated in all specifications.

    The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 67 countries from 1986-2011. Selected country and/or regional dummies are included.

    The FE regressions estimate country fixed effects for all countries in the sample, but only the LA7 results are reported in this table.

    The investment profile subcomponent of the International Country Risk Guide political risk index is used to gauge institutional quality.

    Reinhart and Rogoff indicator of past banking crises.

    Demeaned estimates: fixed effect estimates minus a sample average of fixed effects.

    Robust T-statistics are in italics. *** p<0.01, ** p<0.05, * p<0.1

  • Table 3 presents the results using the baseline consolidated index of financial integration (described in Box 2). After combining the dimensions of financial openness and financial convergence, it appears that the LA-7 countries are indeed under-integrated, with the exception of Panama, which shows a level of integration in line with the sample average after controlling for fundamentals. This result suggests that the relatively high degree of openness of countries such as Chile and Peru in Table 2, is dominated by the lack of regional convergence exhibited by their financial markets.

    Table 3.

    Financial Market Integration: Composite Financial Integration Index with Two Components1/

    article image
    Notes: Time dummies have been incorporated in all specifications.

    Principle component from 2 variables: openness (external assets+liabilities as a ratio to GDP) and convergence.

    The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 67 countries from 1986-2011. Selected country and/or regional dummies are included.

    The FE regression estimates country fixed effects for all countries in the sample; only LA7 results are reported.

    The investment profile subcomponent of the International Country Risk Guide political risk index is used to gauge institutional quality.

    Reinhart and Rogoff indicator of past banking crises.

    Demeaned estimates: fixed effect estimates minus a sample average of fixed effects. Robust T-statistics are in italics. *** p<0.01, ** p<0.05, * p<0.1

  • Table 4 reports the results using the first alternative consolidated index of financial integration, which combines convergence and external liabilities-to-GDP, as a measure of openness (described in Box 2). The results using this narrower measure of openness, which helps preclude cases where large external assets do not correspond to integration, corroborate the findings of the baseline index. With the exception of Panama, the LA-7 countries show a degree of under-integration virtually identical to the baseline results presented in Table 3. In this case, Panama stands out as the one LA-7 country whose level of integration is above the sample average.

    Table 4.

    Financial Market Integration: Composite Financial Integration Index with Two Components1/

    article image
    Notes: Time dummies have been incorporated in all specifications.

    Principle component from 2 variables: external liabilities as a ratio to GDP and convergence.

    The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 67 countries from 1986-2011. Selected country and/or regional dummies are included.

    The FE regression estimates country fixed effects for all countries in the sample; only LA7 results are reported.

    The investment profile subcomponent of the International Country Risk Guide political risk index is used to gauge institutional quality.

    Reinhart and Rogoff indicator of past banking crises.

    Demeaned estimates: fixed effect estimates minus a sample average of fixed effects. Robust T-statistics are in italics. *** p<0.01, ** p<0.05, * p<0.1

  • Table 5 presents the findings using the second alternative consolidated index of integration, incorporating three components: openness, convergence and depth. The results support the outcomes of Tables 3 and 4, confirming that even with the added dimension of depth, the LA-7 countries—excluding Panama—are under-integrated relative to the sample average, after controlling for fundamentals. An interesting nuance of these results is that after adding depth, the integration outcomes worsened for all LA-7 countries except for Panama and Chile. Panama’s result was not only above the sample average but significantly stronger than its outcomes using the two-component indexes of integration. Regarding Chile, while the integration outcome was still negative, the magnitude of under-integration was halved relative to Chile’s results using the two-component indexes, suggesting a relatively deep market. Combined with Chile’s positive result in the openness models presented in Table 2, one could surmise that Chile’s under-integration comes largely from a lack of convergence with the region rather than other factors. For the remaining five LA-7 countries, the onus of under-integration falls on the lack of convergence and depth of their financial markets.

    Table 5.

    Financial Market Integration: Composite Financial Integration Index with Three Components1/

    article image
    Notes: Time dummies have been incorporated in all specifications.

    Principle component from 3 variables: openness, convergence and depth.

    The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 66 countries from 1986-2011. Selected country and/or regional dummies are included.

    The FE regression estimates country fixed effects for all countries in the sample; only LA7 results are reported.

    The investment profile subcomponent of the International Country Risk Guide political risk index is used to gauge institutional quality.

    Reinhart and Rogoff indicator of past banking crises.

    Demeaned estimates: fixed effect estimates minus a sample average of fixed effects. Robust T-statistics are in italics. *** p<0.01, ** p<0.05, * p<0.1

  • Table 6 displays the findings of the third alternative consolidated index of integration, which includes a measure for relative regional openness, in addition to the measure for global financial openness and regional convergence. The results including the regional measure stand out from the previous findings in that all LA-7 countries, including Panama, exhibit underintegration relative to the sample average. That said, Panama still shows the lowest degree of underintegration among the LA-7 countries. This may suggest that Panama’s high degree of financial integration, demonstrated in the previous results, largely reflects extra-rather than intra-regional integration. Another interesting finding that emerges is that Brazil, Colombia, Peru and Uruguay are less underintegrated relative to the sample than Chile and Mexico, using this index. Mexico’s result may reflect its higher degree of integration with the U.S.A. (which was not included in the same regional grouping as Mexico) relative to with the region. In the case of Chile, which showed a relatively high degree of openness compared to other LA7 countries in the previous indexes, the results suggest that the interconnections of its relatively deep financial markets principally stem outside the region rather than within.

Table 6.

Financial Market Integration: Composite Financial Integration Index with Three Components

article image
Notes: Time dummies have been incorporated in all specifications.

Principle component from 3 variables: global openness, regional convergence and regional integration based on 8 regions.

The OLS regressions are ordinary least squares regressions with standard errors adjusted for clustering at the country level for a panel of 172 countries from 1986-2011. Selected country and/or regional dummies are included.

The FE regression estimates country fixed effects for all countries in the sample; only LA7 results are reported.

The investment profile subcomponent of the International Country Risk Guide political risk index is used to gauge institutional quality.

Demeaned estimates: fixed effect estimates minus a sample average of fixed effects. Robust T-statistics are in italics. *** p<0.01, ** p<0.05, * p<0.1

B. Macroeconomic Gains from Regional Integration in Latin America

12. To quantify the benefits of further integration in LA, a model relating financial integration to economic growth is estimated. The specification, which follows Beck and Levine (2004) and Sahay and others (2015), includes the standard control variables of growth equations: initial income per capita, trade openness, inflation, the government expenditure-to-GDP ratio, investment-to-GDP ratio, population growth, and several measures of institutional framework quality (proxied by the ICRG indicators of country risk). The sample is similar to the one used in the previous exercise, and includes 76 countries between the mid-1980s and 2014. In light of the endogeneity of the integration variable with respect to growth, the baseline model uses an instrumental variable (IV) panel estimator with the following instruments: the first lag of the integration variable; the capital controls indicator by Fernández and others (2015); the occurrence of a banking crisis 10 years earlier; and a subcomponent of the ICRG political risk index, which describes the extent to which profits can be transferred or repatriated out of a country. All the instruments are assumed to impact integration directly but affect growth indirectly.2 The estimated equation is therefore:
yit=αi+β1*FIit+β2.Xit+ϵit

where yit denotes GDP growth, FIit the financial integration indicator defined in Box 2, Xit the control variables, and αi is the fixed effect. Time dummies are also included in some specifications.

13. Instrumental variables indicate that financial integration is found to be positively correlated with growth. In models without correction, integration is either statistically insignificant or has a negative effect on growth. IV estimates indicate that the elasticity is clearly positive regardless of the number of control variables (Table 7, columns 1–3), or when the equation is saturated with time dummies (column 4), or whether real growth or real growth per capita is used as a dependent variable (column 5). Results are also robust to removing the banking crisis instrument, which presents the disadvantage of reducing the sample size as the variable denotes the existence of a crisis 10 years earlier and is not available for some countries (column 6). The results of a dynamic model estimated by Arellano–Bond GMM with the lagged GDP growth as explanatory variable are also presented, and the financial integration variable coefficient is broadly unchanged (column 7).3

Table 7.

Relationship of Financial Integration to GDP Growth (Baseline Results)

article image
Notes: all specification estimated with panel IV estimator except for specifications 7 and 9 that use GMM.

Principle component of 2 variables: global openness and regional asset price convergence.

Exports plus imports, ratio to GDP

Private and public investment, ratio to GDP

Current and capital expenditures of the general government, ratio to GDP

ICRG composite index of political, economic and financial country risks.

This specification is saturated with time dummies, which are not presented in the table.

14. Another potential issue is that the lagged GDP-per-capita level is endogenous in growth equations (Bond and others, 2001). To circumvent this problem, an equation is presented excluding the variable and finds that the integration coefficient is broadly unchanged (column 8). The endogeneity of both the integration variable and the lagged GDP level are corrected by rewriting the growth regression as a dynamic model in levels4 and estimating it with the first-differenced GMM estimator of Arellano Bond (2001)—alongside the lagged (first-differenced) variables, the additional instruments mentioned above (capital controls indicator, occurrence of a banking crisis 10 years earlier, and profit repatriation rule) are included. The effect of financial integration is again positive and significant (column 9), but the regression suffers from the traditional GMM shortcomings, including a high sensitivity to the number of lags used for the instruments. Finally, the possibility of non-linear relationships was accounted for through interaction terms and a quadratic form of the integration indicator. However, the non-linear models did not produce robust results.

15. Table 8 reports the results of specifications with alternative measures of integration. The alternative indicators described in Box 2 are used: a two-component index with the ratio of external liabilities-to-GDP (column 1); a three-component index that adds a measure of financial depth (column 2); and variants of the three-component index including regional openness (column 3-6). Column 3 measures regional integration as the ratio of a country’s regional assets and liabilities to total foreign assets and liabilities in a 8-region framework. Column 4 replicates the indicator with a 4-region split. Column 5 (resp. 6) measures regional integration by weighting the sum of assets and liabilities (resp. liabilities only) with the distance between countries. In all these specifications, the effect of integration remains positive and significant.

Table 8.

Impact of Financial Integration on Growth Using Alternative Financial Integration Indicators

article image
Notes: all specification estimated with panel IV estimator except for final 3 specifications that use GMM.

Principle component of 2 variables: global integration of external liabilities and (narrow) regional asset price convergence. See box A1.

Principle component of 3 variables: global integration of external assets and liabilities, banking system credit to the private sector, and (narrow) regional asset price convergence. See box A1.

Principle component of 3 variables: global integration of external assets and liabilities, (narrow) regional asset price convergence, and (narrow) regional integration of external assets and liabilties. See box A1.

Principle component of 3 variables: global integration of external assets and liabilities, (broad) regional asset price convergence, and (broad) regional integration of external assets and liabilties. See box A1.

Principle component of 3 variables: global integration of external assets and liabilities, (narrow) regional asset price convergence, and average proximity of external asset and liability partners. See box A1.

Principle component of 3 variables: global integration of external liabilities, (narrow) regional asset price convergence, and average proximity of external liability partners. See box A1.

Exports plus imports, ratio to GDP

Private and public investment, ratio to GDP

Current and capital expenditures of the general government, ratio to GDP

ICRG composite index of political, economic and financial country risks.

16. Using the measure of under-integration calculated in the previous section, the econometric analysis suggests that closing the integration gap in LA-7 countries may raise GDP growth by 0.25 to 0.75 percentage point. The various specifications return integration elasticities of 0.01-0.02. Using the fixed effects estimates of the previous section5, the equations would therefore predict a growth effect in the range of 0.25 to 0.75 percentage point on average, if the gap were to be fully closed. The growth dividend will be lower if progress is partial. These results should be treated with caution, as most variables in growth regressions are endogenous, creating potential estimation biases that IV and GMM estimators cannot always correct.

Market-Implied Interlinkages

17. Financial linkages among financial or banking institutions can be broadly split in two categories: direct and indirect linkages. Direct financial linkages denote explicit balance sheet positions from one financial institution onto another; essentially these are assets or liabilities of financial institutions vis-à-vis each other. Indirect linkages arise when there are no explicit direct linkages among financial institutions, yet market indicators of these financial institutions (for instance, stock prices) tend to exhibit some degree of co-movement or synchronicity. These indirect linkages could be the consequence of having similar business models, or common exposures to related economic sectors, or simply being perceived by the markets as being vulnerable to the same type of shocks (e.g. a change in legislation affecting most banks in one country).

18. The aim of the market-implied interlinkage analysis is to quantify both direct and indirect linkages among different financial institutions in Latin America. The sample includes the largest listed banks from Argentina, Brazil, Chile, Colombia, Mexico, and Peru, over the period 2005–2015, and relies on publically available daily time series of financial variables (e.g. stock prices, CDS spreads, etc).

19. The methodology relies on the computation of empirical distributions characterizing the joint and conditional probabilities of distress among financial institutions. This is largely based on the CIMDO methodology developed by Segoviano (2006).6 In particular, two synthetic measures are used:

  1. (i)The vulnerability index (VI), which measures the susceptibility of a particular institution to fall in distress given distress in other financial institutions (loosely speaking, it measures an institution’s “vulnerability to contagion from other financial institutions”). Algebraically, the vulnerability index of a given financial institution i is given by:
    VI(Ai)=ΣjiNαjP(Ai|Aj)

    Where the weight αj=1NP(Aj) denotes the number of financial institutions in the sample, and P(Aj) is the probability that institution j falls in distress.7

  2. (ii)The contribution to systemic risk, which measures the contribution of a given institution to changes in the vulnerability to contagion of other institutions (i.e. its role as a “source of contagion”). In other words, it is the percent share that a given financial institution represents in the changes in the vulnerability index of all other institutions in the sample.

20. In order to analyze the dynamics of the above two market-based measures in Latin America, both of them are computed for a sample of five Brazilian banks (as Brazil is by far the largest banking sector in the region) (Figure 6); and then separately for a sample of 15 banks from the six LA countries in the country sample (Figure 7).8

Figure 6.
Figure 6.

Market Based Measures of Inter-linkages among Selected Banks in Brazil

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Source: Moodys KMV, Datastream, and IMF staff calculations.
Figure 7.
Figure 7.

Market-based Interlinkages in Selected Latin American Banks

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Moody’s KMV, Datastream and IMF staff calculations.

21. Figure 6 presents the evolution of the vulnerability index of the five large Brazilian banks (left panel), and it also exhibits the percent contribution of each of these banks to the system’s change in vulnerability during three selected periods (right panel). Interestingly, it appears that foreign banks such as Santander (SAN) were perceived by the markets to be relatively safe compared to domestic banks during the CFC (“Period I”). However, in the recent period of falling economic activity, lower commodity prices and heightened political tensions in Brazil (“Period III”), public banks (BDB, RSU) are perceived to be a larger source of risk for the banking system compared to privately owned banks such as Itau (ITA) and Bradesco (BRA). Public banks in Brazil are a important part of the banking system co-movement of their market indicators may reflect more wider economic concerns for Brazil and thus are significant source of risk for the banking system.

22. Regarding the sample of the financial institutions for the six Latin American countries together, Figure 7 presents the evolution of the vulnerability index for the 15 banks included in the sample (top two panels and bottom-left panel). Likewise, Figure 7 also exhibits the percent contribution of each of these banks to the rest of the system’s change in vulnerability during three selected periods (bottom-right panel). In this case, Argentinean banks and Banorte (from Mexico) appear to be the most “vulnerable to contagion” during the GFC (“Period I”). However, these relatively high market-implied interlinkages during that period appear to be important mainly among themselves (Figure 7).

23. In the most recent period (“Period III”), Banco do Brazil (BDOBR) appears to be driving most of the market-implied contagion (Figure 7, bottom-right panel). However, the actual spillovers outside of Brazil appear to be rather small, for instance when comparing the levels of the vulnerability index to those observed during the GFC. In other words, large domestic public banks in Brazil might be very important for the domestic market (in Brazil), but not really for the rest of region. These market-based measures seem to be in line with the limited actual cross-border balance sheet exposures of the banking sectors in LA.

List of Latin-American Banks Included in the Market-Implied Interlinkages Analysis

article image
Table 9.
Table 9.

Contribution to Systemic Risk During Global Financial Crisis

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Moody’s KMV, Datastream, and IMF staff calculations.

IMF Research Department (RES) Bank Contagion Module9

24. Direct linkages (through cross-border lending and purchases of debt securities) among financial institutions can lead to increased spillover risks among countries. These risks may take the form of losses on risk exposures that may cascade across borders through interlinked financial systems.

25. The aim of RES’ Bank Contagion Module is to analyze potential spillover effects arising from the international lending operations of global banks. The main exposure metric in this analysis assesses lender banking systems’ exposure to shocks in borrower countries. Essentially, the framework simulates the propagation of financial shocks across borders through bank losses and deleveraging. The module utilizes the BIS bilateral banking statistics—representing claims of banking systems in BIS reporting countries vis-à-vis all residents (banks, non-banks, and public sector) in reporting and non-reporting countries. There are four BIS reporting countries in Latin America: Brazil, Chile, Mexico, and Panama.

26. Latin American banking systems are not strongly integrated among themselves but have tight links with advanced economy banking systems, from where shocks may emanate. Bank linkages with advanced economies outside the region—in particular, Canada, Spain, UK, and the United States—are relatively important. Results from the RES Bank Contagion Module suggest that an asset-side shock to these advanced economies’ banking systems could have a sizeable impact on the availability of foreign credit to Latin American countries (Table 10). A shock to any of the Latin American banking systems would likely have small direct spillovers onto other countries in the region due to limited intraregional cross-border banking exposures.

Table 10.

Simulated Spillovers in Selected Latin American Countries10

article image

Magnitude denotes the percent of on-balance sheet claims (all borrowing sectors) that default.

Reduction in foreign banks’ credit due to the impact of the shock on their balance sheet, assuming uniform deleveraging across domestic and external claims. All simulations are based on 2014Q3 data.

LA-7 Country Profiles

A. Brazil

27. Brazil’s financial system is by far the largest in LA. Commensurate to the size of its overall economy, Brazil’s total financial sector assets dwarf those of the other countries in the region. Brazil’s nominal GDP amounted to about US$2.35 trillion in 2014 (Figure 8), comparable to that of the largest 5 other economies in LA together. Brazil’s financial sector is not only large in absolute terms, but also relative to its economy.

Figure 8.
Figure 8.

Brazil: Indicators of Regional Size

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: National authorities; United Nations; Haver; and IMF staff calculations.

28. Accordingly, Brazil’s banking system is the largest in absolute terms. Furthermore, with total assets close to US$2.4 trillion, the banking sector is also one of the largest in percent of GDP, representing close to 117 percent (Figure 9).

Figure 9.
Figure 9.

Brazil: Banking Sector Assets Relative to Regional Peers

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: National authorities; Bankscope; and IMF staff calculations.

29. The Brazilian banking system remains dominated by large public banks. Publically owned banks represent about half of the banking system (Figure 10). Furthermore, the banking sector remains highly concentrated, with the 8 largest banks accounting for about 85 percent of the banking sector (Figure 11). The financial system is characterized by a high degree of conglomeration. Interest margins are high, which is partly reflected in high profitability, particularly for the large banks (c.f. FSI’s chart). However, the system appears to be stuck in a “high interest rate and short duration” equilibrium, which limits capital market development, and thus potential growth.

Figure 10.
Figure 10.

Brazil: Assets and Deposits by Nationality of Bank Ownership

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: National authorities; Bankscope; and IMF staff calculations.
Figure 11.
Figure 11.

Brazil: Ownership of Major Banks

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Note: In 2015, Bradesco purchased HSBC’s operations in Brazil to increase their market share to about 14%.Sources: National authorities; Bureau van Dijk, Bankscope; and IMF staff calculations.

30. Regarding non-banks, the insurance sector is performing well. Profitability in the insurance sector has been relatively high over the past few years, likely benefiting from high interest rates, which have translated into solid solvency ratios. Mutual funds and banks are highly interconnected through repo operations and the holding of deposits and bank-issued bonds by the funds. Pension funds are sizeable in Brazil, with assets under management close to US$280 billion. Essentially all these assets are invested domestically (Figure 12).

Figure 12.
Figure 12.

LA-7: Pension Fund Assets Under Management1

(Billions of US dollars)

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: National authorities; Bureau van Dijk; and IMF staff calculations.1 Year-end 2014 or latest available.

31. Itau is the only universal Brazilian bank with a significant presence across the region. Most Brazilian banks tend to be inward looking. This reflects in large part the significant share of publically-owned banks, as well as the large domestic market. Itau, which is the largest privately-owned bank in Brazil, has nevertheless sizeable stakes in the region, representing almost 10 percent of the banks’ total assets. The bank is present in Argentina, Chile, Colombia, Mexico, Paraguay, and Uruguay. BTG Pactual is trying to position itself as a regional investment bank. Investment banks have the advantage of operating with smaller balance sheets, hence the potential ability to be profitable without the need for large scale. This is also reflected in terms of their capital costs.

32. Brazilian foreign claims remain concentrated in a few advanced economies. Brazilian claims on countries such as the U.S. and the U.K. dwarf those on other LA countries (Figure 13). The only exception is Chile, where Itau has a significant presence. Cayman Island has a notable share of Brazilian foreign claims; most Brazilian banks establish operations there in order to offer their Brazilian clients investments denominated in foreign currencies.

Figure 13.
Figure 13.

Claims of Brazilian Banks on the Rest of the World

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Bank for International Settlements; and IMF staff calculations.

33. Foreign financial claims on Brazil have been growing rapidly for most of the past 15 years. Indeed, foreign claims have more than quadrupled since 2005 (Figure 13), and stand at about US$442 billion (roughly 18 percent of GDP). Spain has the highest foreign claims, representing close to 7.5 percent of GDP, reflecting the significant presence of Spanish banks, most notably Santander. In the last couple of years, however, the total amount of foreign claims has stabilized. This is consistent with the slowdown in the domestic economy. Furthermore, Brazilian financial institutions have a relatively low ratio of foreign liabilities to credit to the economy (around 10 percent). This suggests a relatively low reliance on foreign funds as a source of funding, limiting the effects of any potential global liquid squeeze.

34. Banking sector flows appear geographically related to real sector activity. There are likely a large number of drivers behind Brazilian cross-border financial flows. There is some evidence that cross-border banking sector flows in Brazil tend to be associated with trade linkages as well as FDI (Figure 15). Furthermore, there has been a noticeable increase in bank as well as non-bank issuance abroad by Brazilian corporations (Figure 16), through 2014. Issuance abroad has contracted as risk appetite for Brazilian securities has subsided.

Figure 14.
Figure 14.

Claims on Brazil by BIS Reporting Banks

Citation: Policy Papers 2016, 023; 10.5089/9781498345897.007.A002

Sources: Bank for International Settlements; and IMF staff calculations.
Figure 15.