Costa Rica: Selected Issues and Analytical Notes

Selected Issues and Analytical Notes

Abstract

Selected Issues and Analytical Notes

Selected Financial Sector Issues1

A. Credit Cycle and Countercyclical Capital Buffers in Costa Rica 2

This note investigates the financial stabilization properties of the Basel III Countercyclical Capital Buffers and the early warning power of the credit gap in Costa Rica. We find that Countercyclical Capital Buffers activated based on a normalized version of the credit gap would have an effective cushioning effect on the financial system, largely due to the robust early warning properties of the normalized credit gap. However, implementation of Countercyclical Capital Buffers as part of a gradual adoption of Basel III standards, is likely to involve non-negligible challenges.

Introduction

1. Costa Rica is gradually adopting Basel III standards. The Basel III Liquidity Coverage Ratio (LCR) has been introduced in 2015 and the authorities are planning to issue a regulation to implement the Net Stable Funding Ratio (NSFR) by the beginning of 2018. They also expect to finalize regulations to introduce the Basel III definition of regulatory capital and leverage ratio by end-2017, and implemented in 2016 countercyclical provisions based on the growth of bank-specific loan portfolios. In general, banks are well placed to comply with Basel III capital requirements: the minimum capital requirement is 10 percent, and the Capital Adequacy Ratio (CAR) of banks and cooperatives satisfies Basel III requirements when adjusted for the new capital guidelines, although it declines due to increases in Risk Weighted Assets (RWA). Supervisory authorities have also advanced in implementing risk-based supervision (Pillar 2) and adopting a cross-border consolidated scheme that allows for the identification of risks taken by financial conglomerates.

2. The Countercyclical Capital Buffer (CCyB) is one of the major conceptual innovations of the Basel III framework. Along with the Capital Conservation Buffer (CCoB), the CCyB is a key macroprudential policy (MaPP) tool to promote the conservation of capital and the build-up of adequate buffers above the minimum that can be drawn down in periods of stress. Their aim is to prevent dysfunctions in the banking system and to avoid strains to the credit supply. According to Basel III guidelines, the CCoB is to be set at 2.5 percent of RWA and it can be drawn down when losses are incurred, to avoid the breach of minimum capital requirements. The CCyB varies between zero and a ‘non-binding’ cap of 2.5 percent of RWA: authorities should increase the buffer rate when risks associated with excessive credit growth build-up, and should lower it when risks materialize to sustain the flow of credit to households and corporations, and contain the risk of systemic deleverage.

3. The CCyB introduces a countercyclical dimension in banks’ capital requirements, contributing to both macroeconomic and financial stability. This tool adds an additional layer of capital during credit booms to create loss absorbing capacity to use in case of a banking crisis. Thus, its main purpose is to enhance the stability of the banking system, with any reduction in the volatility of the financial and business cycles being a welcome side effect, contributing to both macroeconomic and financial stability. However, given its methodological implications and the call for policymakers’ judgement on how buffers are built up and released, the CCyB is likely to entail nontrivial implementation challenges, hence the focus of this note.

4. The concept of credit-to-GDP gap plays a central role in guiding CCyB policies. Basel III uses the gap between the credit-to-GDP ratio and its long-term trend as a guide for setting the CCyB. This should be activated when there is excessive credit growth and the financial cycle is in its expansion phase, or equivalently when the “credit gap” is significantly positive. Similarly, CCyB policies should be loosened when signs of financial distress in the banking system appear and the financial cycle enters its downturn period, that is when the credit gap declines or turns negative. However, estimation of the credit gap crucially depends on the method that is used to back out the trend in the credit-to-GDP ratio, which could be a nontrivial task, especially when sufficiently long time series are unavailable. Hence, the Basel Committee on Banking Supervision (BCBS, 2011) recommends using this measure in conjunction with other financial indicators of excessive credit growth to guide CCyB policies, and in general assigns a significant role to the judgment of policy makers.

Methodology

5. Following the Basel III framework, we estimate the credit gap as a guide for calibrating the CCyB. The Basel Committee (BCBS 2010, 2011) selected the credit gap as the main indicator for setting the CCyB mainly because of its out-of-sample forecasting power in predicting banking crises in a large sample of countries. Subsequent research by various authors has largely confirmed its properties for advanced countries while some have questioned its performance for samples of developing countries. In this methodology, credit is defined as aggregate credit extended to households and non-financial private businesses from banks and non-banks. Importantly, this measure of credit includes all credit that is extended to the private nonfinancial sector through nonbanks, corporate bond markets, and foreign intermediaries or investors, as such indicator proves to over-perform narrower definitions of credit (such as credit extended by banks only) in signaling future banking crises. Normally, this measure is constructed at quarterly frequency.

6. We use a standard one-sided backward looking HP filter to estimate the long-run trend in the credit-to-GDP ratio. This is consistent with BCBS (2010) and BCBS (2011) guidelines, although some studies have suggested that a two-sided filter might perform better. The one-sided HP filter estimates the trend as the solution to the following minimization problem:

minTrendtΣTt=1(CTGtTrendt)2+λΣTt=3(Trendt2Trendt1+Trendt2)2)

which balances the trade-off between the size of the estimated cycles and the variation in trend growth rate, with the smoothing parameter λ establishing the relative weight of each of the two terms: the larger the smoothing parameter, the more importance is assigned to the second term, and therefore the “smoother” the trend series would be. Since financial cycles are thought to operate at very low frequencies, for them the smoothing parameter is usually set at large values. BCBS (2010), based on Borio and Lowe (2002), suggests a smoothing parameter of 400000.3

7. Since the usual credit gap measure is inconsistent with sustained fast credit growth, we consider a normalized indicator to reflect Costa Rica’s still ongoing financial deepening. The standard credit-to-GDP gap proposed by the BCBS is based on studies of countries with high levels of financial deepening, mostly with credit-to-GDP ratios above 100 percent. For countries with lower level of financial deepening such as Costa Rica, which had credit-to-GDP ratio at 60 percent as of January 2017, the standard credit gap might not be the most appropriate measure to inform decisions on CCyB. As an alternative, a percentage deviation of the credit gap relative to the trend of the credit-to-GDP ratio is proposed.4 Hence the normalized credit gap is defined as:

(CreditGap^)t=CTGtTrendtTrendt

8. Drawbacks related to this methodology apply to Costa Rica as well as other emerging countries. Chief among them are issues regarding the length of the available data and the presence of structural breaks in the series.

  • Length of the time series. Edge and Meisenzahl (2011) argue that the backward-looking nature of the one-sided HP filter puts too much emphasis on the end points and therefore the estimated trend depends heavily on the most recent observations. 5 As noted by Seidler and Gersl (2012), the estimated trend changes very significantly with the starting point of the series, Drehmann and Tsatsaronis (2014) and Alessandri and others (2015) shows that this problem gets much worse as the length of available time series shrinks. Both drawbacks suggest that the longest available time series should be used to attain the best possible estimation of the credit gap. Borio and Lowe (2002) suggest that this calculation should not be done for time series with length of less than 10 years at least, while many others suggest 20 years of data is the minimum requirement.

  • Structural breaks. The use of such credit gap measures may hinder the process of beneficial financial deepening undergone by many emerging market economies (Reserve Bank of India, 2013), because this measure essentially penalizes fast growth rates of credit-to-GDP, which could be a structurally desirable and positive outcome. This also induces a reverse problem: a sustained period of high growth in credit-to-GDP ratio translates to a faster trend growth estimate which could turn CCyB measures never binding. BCBS (2010) and BCBS (2011) recognize these issues and the limited information contained in the one-sided HP filtered credit gap and recommend using additional variables to guide the decision regarding CCyB rates.

For our estimates, we consider aggregate quarterly data on credit extended by banks—including foreign—cooperatives, and other financial institutions, but excluding corporate bonds (due to the shorter length of this time series). However, such aggregate credit data is only available since 1997, compared to 1987 for bank credit only, exposing a tradeoff between length and coverage of the credit series. It should also be noted that, besides the ongoing financial deepening process in Costa Rica, the shifts in exchange rate policy during the period in consideration also introduce structural breaks in the credit series given the high level of credit dollarization in the economy.

A03ufig1

Exchage Rate

(Colones/USD)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: BCCR; SUGEF; and staff calculations.

9. The additional signals of financial risk build-up recommended by the BCBS include measures of excessive credit growth, lenient credit risk pricing, and measures of high leverage. In contrast to credit gap, however, there are no specific definitional or numerical recommendations for these measures by BCBS.

  • Excessive credit growth: credit growth rates at sectoral level, such as household and construction, and persistently large current account deficits;

  • Credit mispricing: low risk premia for risky assets (low credit spreads), high equity valuations, high housing price growth rates or price to rent ratios;

  • High leverage/risk buildup: high leverage ratio in the banking sector, high leverage ratios in the corporate sector, high loan to value ratios (LTV) or debt service to income ratio (DSTI) in the household sector.

Credit gap and CCyB setting

10. Estimation results show that the absolute and normalized measures of the credit gap are highly correlated. The two measures are scaled versions of one another, and the correlation between the two series is 0.99. This means that the early warning properties of credit gap do not depend on whether one uses the measure proposed by BCBS or the normalized one proposed here.

A03ufig2

Credit Gap Measures

(Percent)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: SUGEF; BCCR;and staff calculations.

11. CCyBs activated based on the normalized credit gap would have started to build up steadily one year before the global financial crisis, peaking at 2.5 percent in 2008Q2-Q4. Using the standard credit-to-GDP gap and the BCBS thresholds to set CCyB rates, on the other hand, would have implied a less timely activation of the buffer, which would have never reached its maximum of 2.5 percent before the period of financial distress. The normalized measure proposed here entails a prompter use of the buffer: it would have been activated one year ahead of the global financial crisis, reached its maximum two quarters before the start of the great recession and released after Lehman’s failure. It is worth noting that Costa Rica has not experienced a banking crisis in recent decades, although signs of financial distress were clear around the time of the financial crisis.

A03ufig3

CCyB Simulated Rates

(Percent)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: SUGEF; BCCR;and staff calculations.

12. However, the banks were well capitalized, hence a putative capital requirement of 10 percent plus the CCyB would not have been binding for the system as whole. This does not suggest that CCyB policies are ineffective when they are not binding for the whole banking system. First, higher regulatory minimum requirements restrict the amount of capital subject to banks’ voluntary decisions. Second, even when the system wide CAR does not breach the total capital requirement, some individual banks might fail to meet this requirement, and therefore would need to increase their capital levels.

A03ufig4

System-wide CAR and Minimum Requirements

(Percent)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: SUGEF; BCCR;and staff calculations.

Early Warning Power of the Credit Gap

13. An assessment of the early warning power of the credit gap is encumbered by the absence of recent systemic banking crises. Costa Rica has not experienced systemic banking crises toward which to assess the credit-to-GDP gap. Hence, we carried out a preliminary assessment of the signaling properties of the credit gap by adopting NPL rates as an alternative measure of financial distress.

14. Regression results show that he normalized gap is a powerful predictor of systemic vulnerabilities for the Costa Rican banking system. Our preliminary assessment shows that credit gap has significant predictive power for NPL rates at 1, 2 and 3-year horizons, with the maximum predictive power for the 2-years horizon (Table 1). This finding is consistent with the existing literature showing the early warning power of the credit gap for many countries.

Table 1.

Costa Rica: Early Warning Properties of Credit Gap and Robustness Check

article image
pval in parentheses*** p<0.01, ** p<0.05, * p<0.1

15. Robustness checks suggest that the credit gap retains considerable predictive power even in presence of other early warning variables for financial distress. As a preliminary robustness check, we add other indicators which are among those that have been recommended by the BCBS and IMF guidelines as relevant early warning indicators of financial distress, such as: capital to RWA (as a measure of leverage); money growth; and the credit to GDP (in level). Results show that, even after adding these measures, the normalized credit gap retains its predictive power.

Country Experiences

16. The BCBS establishes that CCyB rate decisions should consider country specific conditions beside other measures of systemic risk buildup. In general, it is understood that judgment is an integral part of CCyB policies and therefore no strict international rules exist. Below we overview a range of approaches in advanced and emerging economies with the interpretation and application of CCyBs.

Advanced Economies

17. Since the announcement of the BCBS guidance, most major advanced economies have adopted and activated CCyB. While most countries have not set positive CCyB rates, they have established procedures for calibrating the CCyB rates so that markets can form well-anchored expectations regarding their future setting. Since advanced countries usually have long time series data on credit and output, the credit gap can be estimated with more certainty, and therefore it serves as a good signal. In most of these countries, the credit gap was large and positive in the few years before the global financial crisis, which gives this measure more credibility. Nonetheless, authorities tend to refer to other financial indicators as well as inputs in their decision making.

  • The Czech Republic, Hong Kong, Norway, Sweden, and the UK have actively used CCyB. The buffer rate is currently at 0.5 percent in the Czech Republic, 1.25 percent in Hong Kong, 1.5 percent in Norway, and 2 percent in Sweden. 6

  • The UK announced in 2016 a 0.5 percent CCyB rate—calibrated also by means of stress testing—to be implemented the following year. However, the Bank of England promptly discontinued the buffer to raise banks’ capacity to lend to households and firms in response to the outcome of the referendum to leave the European Union.

  • Switzerland is the first country to have used a sectoral CCyB, to enhance to banking system resilience to building vulnerabilities in the real estate sector. The buffer rate is currently set a 2 percent.7 While not yet recognized by the BCBS international standards, the Swiss experience provides an interesting case for a more tailored use of the CCyB, which in this design could replace other sectoral tools as a more transparent and easier to communicate measure.

  • In Finland, albeit the credit gap would have suggested to set a positive buffer rate, the authorities have discretionally kept the buffer at zero based on the information of a set of additional indicators and of the comprehensive MaPP stance adopted.

  • To inform its CCyB decisions, Italy has introduced a real time two-sided filtered credit-to-GDP gap as an additional measure of the credit-to-GDP gap, based on the findings by Alessandri and others (2015). It provides a better representation of the credit cycle - with less volatile results -than the standard one-sided BCBS approach. The policy stance on the CCyB is communicated not only through the quarterly policy decision statements, but also through the Financial Stability Report to shape market expectations and enhance the predictability of future policy decisions.

Emerging Economies

18. Many emerging countries have also started introducing CCyBs but setting rates at zero. 8 While these countries have largely adopted the credit gap as a guide to CCyB rate setting as in advanced economies, the exact procedure tends to be less articulated, and policy announcements are not followed or preceded by background papers detailing the underlying methodology. In this context, the experiences of Slovakia and Peru are substantially different from other developing countries.

  • In Slovakia, the credit gap does not play the same central role as it does for most other countries; rather information is pulled across a range of financial indicators, much in line with the complementary variables described above (Rychtárik, 2014). Three major categories of indicators are covered: cycles (both economic and financial), banks, and customers. For each category, a set of core and supplementary variables are defined. Core variables are meant to capture excessive credit growth and the state of the financial and the business cycles in the whole economy, while supplementary variables are to capture the excessive credit growth in specific sectors. The current values of such indicators are compared to their historical distributions and given a value between 1 to 9 (9 meaning the highest percentile). These scores are then averaged to produce an indicator of systemic risk buildup. Both the single indicators and their average are considered in decisions regarding the policy rate, with very high values being indicative of excessive credit growth.

  • In Peru instead of the credit gap as the main guiding tool for setting countercyclical capital buffers, the focus is on the economic cycle (Galindo, Rojas-Suarez, and del Valle, 2013). After running a stress scenario, the amount of RWA for each bank under stress is computed, and banks are asked to increase their capital to cover a specific fraction of the difference between the RWA under stress and the current RWA. Under this rule, the CCyB could potentially go above 2.5 percent. This procedure resembles the one in the UK in using stress test results to guide MaPP, but the procedure is not formally detailed in a publicly available paper and, as noted earlier, CCyB rates are currently at zero.

Conclusions and Policy Implications

19. CCyBs would exhibit effective countercyclical dynamics in Costa Rica, reflecting strong early warning properties of the normalized credit gap. Our estimates show that CCyBs activated by a normalized credit gap measure—to take into account desirable financial deepening—would have reached its maximum of 2.5 percent two quarters ahead of the global financial crisis, providing cushioning room during the downturn. The early warning properties of the credit gap are difficult to assess in Costa Rica due to the absence of a banking crisis during the available time span. However, using NPLs as a proxy for financial distress, the normalized credit gap proves to be a strong and robust early warning indicator, even in presence of other signals of financial risk build-up. This suggest that adopting CCyBs as part of a broader transition to Basel III would improve the stability of the Costa Rican financial system

20. However, CCyB implementation would entail non-trivial challenges, in particular, caution should be exercised in using the credit gap for determining CCyB rates. Estimating the credit gap in Costa Rica is subject to caveats, given the short time span of the available data and its incompleteness in covering all sources of credit. Hence, in Costa Rica, and more generally in those cases where the length of available time series is short, credible complementary indicators should be considered in decision making to help better assess the state of the financial cycle.

21. Other challenges to the implementation of CCyBs can be wide-ranging. First and foremost, the activation of CCyBs requires amendments to the regulation on bank capital requirements to define how the buffer would work in practice (e.g., timing of decisions and implementations, quality of capital instruments to be used to satisfy the requirement, and how it should be communicated). In addition, proactive policy making will be needed for the implementation of the CCyB, including by conducting regular analysis based on which accountable and transparent decisions should be taken on a quarterly basis. Constrained discretion should be used to forestall inaction bias, allowing policy making to be discretional but still systematic, transparent, and accountable reflecting predetermined constraints.

22. Effective microprudential supervision is also a prerequisite for a consistent MaPP framework. Strengthening the microprudential foundations of MaPP is key to ensure the fundamental resilience of individual institutions by dealing with issues that cannot be addressed by MaPP. Strong supervision—on a solo and consolidated basis—is essential both to ensure that macroprudential policymakers can draw on supervisory information in their risk assessment and to ensure that the MaPP stance adopted is effectively enforced across institutions. Improved regulatory frameworks and increased supervisory scrutiny are often useful prior steps to the tightening of macroprudential rules.

23. Interaction with other policies can be synergic, but at times conflicts can emerge. While in a benign macroeconomic environment both micro- and macroprudential policies act in the same direction, in a downturn phase the two policies could conflict with each other. Interactions between monetary policy, credit growth, and MaPPs and their implications for inflation and/or growth also need to be carefully considered. In this respect, a clear attribution of responsibilities and a ranking of the different policy objectives, with a preference for systemic stability, can smooth out institutional conflicts, while effective coordination mechanisms can contribute to exploit synergies among different policies.

B. Banking Sector Vulnerabilities and Stress Testing 9

This note analyzes the financial stability risks of the Costa Rican banking sector based on stress tests assessing both the solvency and liquidity stance of banks in the face of potential shocks. Results suggest that the financial sector is particularly exposed to direct and indirect (through FX risk) credit risk, but well prepared to absorb a range of shocks with available buffers. However, under an extreme scenario featuring a combination of these shocks, six banks—accounting altogether for about 15 percent of market share—would need to be recapitalized, although the total capital shortfall would only amount to less than ½ percent of GDP.

Soundness of the Banking Sector and Risks

24. The Costa Rican Financial sector is centered on banking intermediation. As of December 2016, total assets of the financial sector accounted for 75 percent of GDP, of which 82 percent (or 61 percent of GDP) from deposit institutions (including the Central Bank, commercial banks, cooperatives, and other non-bank deposit institutions) and 18 percent (or 13 percent of GDP) from other financial institutions (including insurance companies, pension funds, and other financial intermediaries). Commercial bank assets account for 56 percent of the total financial sector assets and 42 percent of GDP.

25. The banking sector remains concentrated and highly segmented. Out of the total 16 commercial banks in the system, the 4 state-owned banks control 61 percent of market share by assets, compared to 36 percent of the 10 foreign banks, and 3 percent of the two domestic banks. At the same time, foreign banks attract about half of total deposits in foreign currency, while state-owned banks account for 77 percent of deposits in domestic currency, mainly due to the explicit unlimited state guarantee on all public banks deposits in national currency.

26. Banks are well capitalized. The CAR for the banking system was 19 percent as of December 2016, well above the 10 percent minimum regulatory requirements. State-owned banks are on average better capitalized than domestic private and foreign ones, largely because of a lower share of credit in FX (48 percent of total, compared to 92 percent in private and foreign banks), which carry higher risk weigh. The latter are a result of recent increases aimed at integrating currency risk into banks regulation, and mostly targeted at FX loans to unhedged borrowers.

27. Asset quality and liquidity indicators are robust but low profitability provides limited cushion. On average, only less than 2 percent of total loans are non-performing, although this share is marginally higher for state-owned than private banks, including foreign banks. The system is liquid, with little variation across all banks. Profitability levels are low despite large intermediation margins, mostly due to quasi-fiscal operations of state-owned banks, and private banks’ requirement to finance with 17 percent of their deposits the credit fund for development (Sistema de Banca para el Desarrollo).

Costa Rica: Selected Banking Sector Soundness Indicators as of December 2016

article image
Source: SUGEF; and IMF staff calculations.

28. Deposits are the main source of funding but the share of external financing is significant. Deposits account for almost 90 percent of total liabilities across the system, with little difference among state-owned, private and foreign banks. The deposit to loans ratio is also well above 100 for all banks. However, at 30 percent of total, the share of foreign liabilities is significant and has been on an increasing path over the last years—from about 10 percent in 2010—exposing the system to rollover risks.

29. Banks are mainly oriented towards lending activities. On average, loans represent about two thirds of banks assets, the rest being comprised of cash and T-bills (22 percent), long-term bonds (8 percent) and other assets (4 percent). Given the composition of banks’ balance sheets, most potential losses and risks to solvency are likely to come from direct and indirect credit risks in the loan book.

30. FX induced risk remains an important vulnerability. The high dollarization of both bank assets and liabilities remain the main source of risk to the system. About 70 percent of banks loans are denominated in foreign currency, and 60 percent of them are extended to un-hedged borrowers, exposing the system to FX risk through credit risk. At the same time, about half of bank deposits (74 percent on average for private and foreign banks) are denominated in foreign currency, exposing banks to FX risk in case of a large currency depreciation, although the system overall has positive net FX exposure, thus mitigating such risk.

Stress Testing

31. The stress test covers the main risks to solvency and liquidity faced by the banking sector. The top-down solvency stress test includes: (i) credit risk, through an aggregate NPL shock as well as differentiated sectoral shocks; (ii) market risk, though interest and exchange rate risk; (iii) contagion risk through interbank exposure, and (iv) a set of reverse tests. The liquidity stress test models a simple liquidity drain that affects all banks in the system proportionally (the next section focuses on a more granular and longer term analysis of liquidity buffers in the system).

32. The impact of single shocks to solvency is moderate and could be absorbed by existing buffers:

  • The Credit risk shock is modeled as a system-wide proportional increase in NPLs to 8.8 percent of currently performing loans; and sectoral shocks of 7 and 11 percent of performing loans in the construction and trade sectors respectively—which account together for 67 percent of total loans. Both shocks are calibrated to the mean value plus three standard deviations of NPLs over the last decade. Consistent with the composition of banks’ balance-sheet data, results suggest that credit risk losses from a credit shock, both system-wide and sectoral, would materially affect banks capital adequacy, but losses would be limited: on average the CAR for the system would remain above 16 percent, although it would fall to 12 for private banks under the system-wide shock—still well above minimum requirements, and the CAR of two banks would fall below 10 percent.

  • The interest rate risk shock assumes a nominal interest rate increase of 3½ percentage point, as the cumulative cut in the monetary policy rate in 2015. The analysis includes: (i) the flow impact from the gap between interest sensitive assets and liabilities; and (ii) the stock impact from bond repricing. While the simulated increase in interest rates would result in valuation losses on fixed income instruments, these would be largely offset by an increase in interest income as most banks have long cumulative interest sensitive positions across maturity buckets. Overall, the system-wide CAR would remain virtually unchanged, with a marginally negative variation for state-owned banks, and positive overall changes for private banks.

  • The FX risk shock assumes an 18 percent nominal depreciation of the bilateral exchange rate with the US dollar, like the overall depreciation occurred in 2008-09, and looks at: (i) the direct exchange rate risk effect on FX exposures; and (ii) the indirect effect through credit risk assuming 11 percent of credit in FX (18 percent of the share of total FX loans extended to unhedged borrowers) becoming non-performing. Results indicate that the simulated FX depreciation would result in both direct and indirect losses, with significant indirect losses through credit quality. Overall, the CAR for the whole system would decrease by about 3 percent and fall to about 11 percent for private banks—still above the regulatory minimum.

33. A combined solvency shock would require recapitalization of some banks, although the system would satisfy minimum CAR requirements. The combined shock includes: (i) the proportional increase in NPLs, (ii) the interest rate shock, and (iii) the FX risk shock. It should be noted that such combined shock represents a very extreme scenario with low probability of occurrence. Results show that, even subject to such extreme shock, on aggregate the system would only fall to 14 percent, thus remaining above the regulatory minimum. However, six individual banks would fall short of the minimum regulatory CAR thus requiring some recapitalization. However, even after such a severe combined solvency shock, the affected banks would only account for 15 percent of the market share, and the identified capital shortfalls would only amount to 0.3 percent of GDP.

34. Contagion risks stemming from domestic interbank exposures are limited and there is no second-round effect following the combined macro-shock. Contagion risks are assessed using a matrix of interbank exposures containing, for each bank, the net credit to every other bank in the system. The exercise illustrates what happens to other banks when one bank fails to repay its obligations in the interbank market as a result of the combined shock. Results show that there is no contagion stemming from domestic interbank exposures through second-round effects. This is because interbank lending is very thin in Costa Rica, and mainly constituted by deposits of private banks into two of the state-owned banks, equal to 8½ percent of their total deposit respectively, to finance the Sistema de Banca para el Desarrollo.

35. The reverse test indicates that system-wide NPLs would need to increase enormously for the system-wide CAR to fall below minimum requirements. The reverse stress test answers what would have to be the NPL increase necessary for: (i) the system-wide CAR, (ii) the CAR of at least 8 banks (half of total), and (iii) the CAR for 50 percent of the total market share, to fall below the regulatory minimum of 10 percent. While an increase to 39 percent would be necessary for the system-wide CAR to fall below 10 percent, NPLs would need to increase to 27 and 30 percent of currently performing loans for at least 8 banks or half of the total market share respectively to fall below the 10 percent regulatory minimum.

36. The liquidity stress test suggests that liquidity shortfalls following a short-lived drain on deposits would be manageable. The liquidity stress test assumes a widespread liquidity drain on all banks of 10 and 8 percent per day of demand deposits in domestic and foreign currency respectively; and a 5 and 3 percent per day withdrawal of time deposits in domestic and foreign currency respectively, affecting all banks. Results indicates that, although the share of liquid to total assets would tumble, all banks would remain liquid after 5 days, mainly because of maturing assets being rolled off and converted into new cash inflows.

Figure 1.
Figure 1.

Costa Rica: Stress Test Results

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Source: SUGEF, and IMF staff estimates.Notes: The sample includes a total of 16 banks, of which 4 state owned, 2 private and 10 foreign. The System-Wide Credit Risk Shock assumes 9 percent (3 s.d. over average NPLs during the last 10 years) of outstanding performing loans to become non-performing; and the Sectoral Shock assumes 7 and 11 percent (3 s.d. over average sectoral NPLs during the last 10 years) of outstanding loans in the construction and trade sectors (which account together for about 62 percent of total loans) becoming non-performing. The Interest Rate Shock assumes a 3.5 percentage points nominal interest rate increase. The FX Shock assumesa 18 percent depreciation of the FXrate, leadingto 11 percent of FX loans becoming NPL. The Liquidity Shock assumes a 10 and 8 percentper day withdrawal of demand depositsin domestic and foreign currency respectively; anda 5 and 3 percent per day withdrawal of time deposits in domestic and foreign currency respectively.
Table 2.

Costa Rica: Stress Test Results

article image
Sources: SUGEF, and IMF staff estimates.Notes: The sample includes a total of 16 banks, of which 4 state owned, 2 private and 10 foreign.

The proportional increase in NPLs assumes 9 percent (average NPLs over the last 10 years plus 3 s.d.) of outstanding performing loans to become non-performing; and the Sectoral Shock assumes 7 and 11 percent (average sectoral NPLs over the last 10 years plus 3 s.d.) of outstanding loans in the construction and trade sectors (which account together for about 62 percent of total loans) becoming non-performing.

The Interest Rate Shock assumes a 3.5 percentage points nominal interest rate increase.

The FX Shock assumes a 18 percent depreciation of the FX rate, leading to 11 percent of FX loans becoming non-performing.

The Liquidity Shock assumes a 10 and 8 percent per day withdrawal of demand deposits in domestic and foreign currency respectively; and a 5 and 3 percent per day withdrawal of time deposits in domestic and foreign currency respectively.

C. Assessing Liquidity Buffers 10

This note evaluates the Costa Rican banking sector liquidity stance by assessing banks’ buffers visa-vis a short-term adverse funding shock as well as their resilience to a medium-term substantial loss of foreign funding. The latter is comparable in its effect to a potential withdrawal of corresponding banking relationships. Results show that the system as a whole is liquid and potential short-term liquidity gaps would be manageable, although there are pockets of vulnerability in FX liquidity. However, most banks would withstand a severe and protracted drainage of foreign funding in the medium-term by using liquid assets and available layers of counterbalancing capacity.

Short-term Liquidity Buffers

37. We use the Basel III Liquidity Coverage Ratio (LCR) to evaluate banks’ short-term liquidity buffers. The LCR measures banks’ potential net outflows over the next 30 days, and their counterbalancing capacity—i.e. the amount of available liquid assets to cover these potential outflows. Accordingly, the metric is defined as the value of liquid assets after haircuts as a share of net cash outflows. Basel III establishes that banks should maintain an LCR above 100 percent in normal times, but can use their stock of liquid assets—thereby falling below the minimum—during periods of stress. To minimize material disruption to the orderly strengthening of the banking system, and the ongoing financing of economic activity, Basel III provides for a phased introduction of the LCR, starting with 60 percent at the time of its introduction in 2015 and reaching its statutory value of 100 percent in 2019. This rule would place the LCR at 80 percent in 2017, but our analysis assumes the standard 100 percent threshold. To emulate stress conditions based on bank-by-bank balance sheet data, we use standard Basel III assumptions for deposit run-off rates, liquidity eligibility, and asset haircuts (see assumptions table below).

38. The short-term funding structure and asset composition of the banking system point to limited vulnerabilities. Retail deposits account, on average for the whole system, for about 40 percent of total funding up to 30 days, and funding sources secured by high quality collateral bring the ratio of stable funding to about 45 percent of total. Although unsecured wholesale funding accounts for a non-trivial share of the total, the bulk of it is represented by deposits from small businesses, which are typically subject to low volatility. The quality of assets is also high, with central bank reserves and domestic sovereign bonds accounting for about ¾ of liquid assets, and a 95 percent ratio of level 1 to total liquid assets11 for the system. At 14 percent, the share of liquid assets to total assets is within conventional metrics, considering the typically long maturity structure of bank assets and the short time horizon of this analysis. Overall, the composition of assets and liabilities places the share of cash inflows to cash outflows at 18 percent, although there is considerable variation across banks.

A03ufig5

Composition of Short-term Assets and Liabilities

(Percent, over 30 days)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: SUGEF.

39. However, the breakdown of funding and assets by currency points to pockets of vulnerability in foreign currency. Unsecured and wholesale funding in FX accounts for a higher share of the total, at the expense of retail deposits and secured sources of funding. Wholesale funding in FX is also likely to be more volatile than deposits from small business in domestic currency, thus raising vulnerabilities. At 9 percent of total assets, the stock of liquid assets in FX is also lower and of inferior quality compared to the overall stock of liquid assets, and 95 percent of potential cash inflows would be subject to haircuts of between 50 and 100 percent according to Basel III assumptions. This suggests that bank balance sheets would be more vulnerable to short-term liquidity stress in FX.

40. Results show that the system as a whole is liquid and potential liquidity shortfalls appear to be manageable. At 142 percent of net cash outflows, the system-wide LCR is well above the 100 percent threshold, suggesting that the banking sector overall has enough liquidity. However, 5 banks are not fully compliant with the LCR 100 percent threshold, suggesting that injection of some extra liquidity could be necessary in case of a severe drain. Nevertheless, the cumulative liquidity shortfall would amount to less than 1 percent of 2016 GDP, or 1.8 percent of broad money as of December 2016. Banks’ liquidity position in FX appear more vulnerable to a short-term adverse funding shock, with an overall shortfall under the same assumptions of 1.2 percent of GDP, or about 9 percent of Net International Reserves as of December 2016.

Summary of Short-term Liquidity Stress Test Results

(Billions of colones, unless otherwise noted)

article image
Sources: SUGEF; and IMF staff estimates.Notes: the sample includes a total of 16 banks.

LCR Basel III Assumption

article image
article image

Medium-term Loss of Foreign Funding

41. We assess the resilience of the system to a severe and protracted loss of foreign funding based on a conventional medium-term liquidity stress. This test uses bank-by-bank data by maturity buckets in FX of up to 6 months and assumes strong run-off rates that extend beyond 30 days although declining over time (see table of assumptions below). Such scenario can be used also to simulate the effect of a protracted withdrawal of Corresponding Banking Relationships (CBR) which could in principle extend beyond the LCR 30-days horizon. For this purpose, we only consider funding is foreign currency, since the Central Bank could technically provide unlimited LOLR assistance to cover shortages in local currency. FX balances are converted in domestic currency at the bilateral exchange rate of 548,18 COL/USD. It should be noted that the assumed runoff of funding is arguably severe, but in line with international experience during crisis.

42. The test assumes three layers of liquidity to meet projected cash outflows:

  • Inflows from maturing investments and lending operations, with roll-off rates (or share of usable inflows) decreasing with maturity;

  • The stock of cash and interbank loans, although the latter is very thin in Costa Rica overall, and virtually non-existent is foreign currency; and

  • Sale of securities with longer residual maturity, subject to fire-sale haircuts.

The first item represents banks’ direct liquidity buffer, and the last two their counterbalancing capacity, i.e., the set of instruments with high liquidity generation capacity by which banks can cope with funding gaps. These three layers of liquidity are assumed to be accessed in a sequential way to cover shortfall left by the previous layer. In case of any remaining shortfall after the sale of securities, the bank is considered illiquid. Offset of liquidity shortfalls by excess liquidity at longer horizons is not permissible, hence outflows in each bucket must be met with the available liquidity in the same bucket. This could result in banks being consider illiquid even if its position would theoretically turn liquid again at longer horizons.

43. Results show that funding gaps could generally be met using first and second layers. While system-wide shortfalls are considerable in the shorter maturity buckets, they tend to decrease with maturity, to the point that cash inflows more than offset simulated outflows at maturities beyond 6 months, resulting in no gap. Even at shorter maturities where liquidity shortfalls are larger, funding needs can generally be met after utilizing inflows from operations, and the gap generally disappears when accessing the available stock of cash. Only one bank fails to pass the test and remains illiquid after selling available securities, with the unfunded gap emerging at maturities between 16 to 90 days, and returning in a long liquid position afterwards. At about 4½ billion colones, however, the cumulative funding gap for the bank only amounts to 0.1 percent of the Central Bank’s international reserves as of December 2016.

44. There is also very limited risk of a simultaneous deterioration in both the liquidity and solvency position. Reassuringly, the bank left illiquid after the sale of available securities displays a robust solvency position, and its CAR remains above the regulatory minimum even after a severe combined solvency shock. Hence, there is little risk that liquidity problems in the aftermath of a severe loss of foreign funding, including from the withdrawal of CBR, may spill over to solvency in the medium-term.

Figure 2.
Figure 2.

System-Wide Cash Flows in Medium-Term Liquidity Stress Scenario

(Billions of Colones)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: SUGEF; and IMF staff calculations.
Table 3.

Medium-Term Liquidity Stress Scenario Assumptions

(Foreign Currency Flows)

article image

D. Spillovers from Stress in International Banks12

This note uses the IMF Bank Contagion Module to estimate spillovers to Costa Rica from stress in international banks. We find that the upstream exposure of Costa Rica to BIS—reporting banks is higher than in other countries in the region. Hence, foreign credit availability in Costa Rica would be materially affected by spillovers from a severe shock originated abroad, with financial integration through Panama playing an important role in the transmission of shocks.

45. We assess the impact of financial spillovers to Costa Rica from stress in international banks based on the IMF Bank Contagion Module. The model estimates potential rollover risks for Costa Rica stemming from both foreign banks’ affiliates operating in Costa Rica that are funded by their parent system, and direct cross-border lending by foreign banks to Costa Rica borrowers. The analysis is based on BIS banking statistics and bank-level data as of June 2016.13 In the scenario analyzed here, rollover risks are triggered by assuming substantial bank losses in the value of private and public sector assets in selected countries, resulting in uniform deleveraging across domestic and external claims. If the banks do not have sufficient capital buffers to cover the triggered losses, they have to deleverage by reducing both their foreign and domestic assets to restore their capital-to-asset ratios,14 thus squeezing credit lines to Costa Rica and other countries. Our analysis includes the transmission of shocks through Panama, a non-BIS-reporting jurisdiction, given the country’s importance as regional financial center.

46. The upstream exposure of Costa Rica to BIS-reporting banks is higher than in other countries in the region. The country’s upstream exposure to BIS-reporting banks15 captures the upper bound of rollover risks for Costa Rica in a worst-case scenario without any replacement, either domestic or external, of the loss of credit by BIS reporting banks to Costa Rican borrowers. This was about 10½ percent of GDP in June 2016, or 14 percent of total domestic and foreign credit to the non-bank private and public sector. 16 This value is marginally lower than the corresponding exposure as of October 2015 (11½ and 14½ respectively), even though the decrease may reflect changes in the number of banks reporting to the BIS disclosing bilateral positions. Although limited, the upstream exposure for Costa Rica is higher than for other countries in the region, with the exclusion of Panama and El Salvador. The lion share of Costa Rica’s total upstream exposure is accounted by foreign claims by Canadian and, to a lesser extent, U.S. bank lenders, although claims by U.S. banks in Costa Rica have been decreasing over the last two years (foreign claims by Canadian banks are not publicly disclosed).

A03ufig10

Foreign Claims of BIS Reporting Banks

(Percent of GDP, June 2016)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: BIS, and IMF, WEO.
A03ufig11

Foreign Claims of U.S. BIS-Reporting Banks to Costa Rica

(Amount Outstanding, USD Billion)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Sources: BIS,

47. Foreign credit availability in Costa Rica would be materially affected by spillovers from a severe shock originated abroad. For example, a 10 percent loss on assets of BIS-reporting banks in U.S. and Canada would reduce credit in Costa Rica by almost 8 percent of GDP. In contrast, a similar shock would reduce credit in the Dominican Republic and Guatemala by only about 3 percent of GDP, and about one percent of GDP in Honduras and Nicaragua. The most sizeable impact on foreign credit availability for Costa Rican borrowers would stem from losses in Canadian assets, which would result in a credit squeeze in Costa Rica of more than 7½ percent of GDP, compared to 6½ percent as of October 2015. These calculations do not take into account the amount of local stable funding for Canadian banks from deposits in Costa Rica which would cushion banks’ need to deleverage in the country.

A03ufig12

Spillovers from International Banks’ Exposures

(Effect on domestic credit of 10% loss in assets of BIS reporting banks, percent of GDP)

Citation: IMF Staff Country Reports 2017, 157; 10.5089/9781484304563.002.A003

Source: IMF/RES Bank Contagion Module, based on BIS data as of June 2016. Note: Includes the transmission of shocks through Panama. In Panama, the loss of credit includes credit by banks in the off-shore center with minimal links to the domestic economy.

48. Financial regional integration is important in the transmission of shocks. Almost three quarters and two thirds of the estimated credit losses in Costa Rica resulting from a shock originating in Japan and Europe would be transmitted through cross-border lending from Panama, which is more dependent on Japanese and European banks’ funding. However, given the limited upstream exposures of Costa Rican borrowers to those countries, the overall reduction in foreign credit from stress in these countries would be rather limited (less than one percent of GDP) in Costa Rica.

Spillovers to Costa Rica from International Banks’ Exposure as of June 2016

article image
Source: IMF, Research Department Macro-Financial Division Bank Contagion Module based on BIS, ECB, IFS, and Bankscope data.Notes: Estimates include the transmission of shocks through Panama.

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 2016Q2 data.

Greece, Ireland, Portugal, Italy, Spain, France, Germany, Netherlands, and the UK.

These lender countries stopped disclosing bilateral positions with Costa Rica, estimates results only include the transmission of shocks through Panama.

References

  • Alessandri, P., Pierluigi Bologna, Roberta Fiori, and Enrico Sette, 2015, A note on the implementation of a Countercyclical Capital Buffer in Italy, Occasional Paper No. 278 (Rome: Banca d’Italia), https://ideas.repec.org/p/bdi/opques/qef_278_15.html.

    • Search Google Scholar
    • Export Citation
  • Basel Committee on Banking Supervision, 2010, Guidance for National Authorities Operating the Countercyclical Capital Buffer (Switzerland: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation
  • Basel Committee on Banking Supervision, 2011, Basel III: A global Regulatory Framework for More Resilient Banks and Banking Systems (Switzerland: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation
  • Borio, C., and Philip Lowe, 2002, “Assessing the Risk of Banking Crises,Bank for International Settlements Quaterly Review, (December).

    • Search Google Scholar
    • Export Citation
  • Drehmann, M., and Kostas Tsatsatonis, 2014, “The Credit-to-GDP Gap and Countercyclical Capital Buffers: Questions and Answers,Bank for Interntional Settlements (March). http://www.bis.org/publ/qtrpdf/r_qt1403g.htm.

    • Search Google Scholar
    • Export Citation
  • Edge, R., and Ralf R.Meisenzahl, 2011, “The Unreliability of Credit-to-GDP Ratio Gaps in Real Time: Implications for Countercyclical Capital Buffers,International Journal of Central Banking, Vol. 7 (December), pp.26198.

    • Search Google Scholar
    • Export Citation
  • Galindo, A., Liliana Rojas-Suarez, and Marielle del Valle, 2013, Macro-prudential Regulations in Andean Countries (Washington: Center for Global Development).

    • Search Google Scholar
    • Export Citation
  • Orphanides, A., and Simon van Norden, 2002, “The Unreliability of Output Gap Estimates in Real Time,The Review of Economics and Statistics, Vol. 84(4) (November), pp. 56983.

    • Search Google Scholar
    • Export Citation
  • Reserve Bank of India, 2013, Implementation of Counter-cyclical Capital Buffer, (December).

  • Rychtárik, Š, 2014, “Analytical Background for the Countercyclical Capital Buffer Decisions in SlovakiaBank of Slovakia BIATEC - Banking Journal, (April), pp. 1015.

    • Search Google Scholar
    • Export Citation
  • Gersl, A., and Jakub Seidler, 2012, “Excessive Credit Growth and Countercyclical Capital Buffers in Basel III: An Empirical Evidence from Central and East European Countries,Univerty of Finance and Administartion’s ACTA VSFS Journal, Vol. 6 (July), pp. 91107.

    • Search Google Scholar
    • Export Citation
1

The authors thank Oscar Mario Morales Berrocal, Jairo Dávila Castañeda, Jaime Odio Chinchilla, Grace Hernández Loría, and Guillermo Olivier Cruz Mendez for providing the data needed for these analyses.

2

Prepared by Pierluigi Bologna, Valentina Flamini, and Rasool Zandvakil.

3

This value should be compared to 1600, which is the value of the smoothing parameter used for estimating business cycles. The difference between the two reflects the idea that financial cycles are thought to be 4-5 times longer than business cycles.

4

The motivation for the alternative measure can be better understood by going through an example. Consider two countries with different levels of financial deepening: country A with a credit-to-GDP ratio of 30 percent and a trend value of 20 percent, and country B with a credit-to-GDP ratio of 110 percent and a trend value of 100 percent. Per the BCBS standard approach, both countries have a credit gap of 10 percentage points. However, for country A to reach a credit-to-GDP ratio of 30 percent credit must grow about 50 percent faster than GDP, while in country B credit has to grow only about 10 percent faster than GDP. If one thinks that a constant fraction of the new credit extended above the trend value (10 percentage points in the case of country B) is excessive and increases systemic vulnerability, the increase in capital and loss absorbing capacity in country A should be much larger than that of country B, i.e., around 5 times larger. Using the percentage deviation of credit gap from the trend averts this issue.

8

For a list of emerging countries that are members of the BIS refer to https://www.bis.org/bcbs/ccyb/

9

Prepared by Valentina Flamini.

10

Prepared by Valentina Flamini.

11

Level 1 assets generally include cash, central bank reserves, and certain marketable securities backed by sovereigns and central banks, among others. These assets are typically of the highest quality and the most liquid, and there is no limit on the extent to which a bank can hold these assets to meet the LCR. Level 2 assets include certain government securities, covered bonds and corporate debt securities, lower rated corporate bonds, residential mortgage backed securities and equities that meet certain conditions. Level 2 assets may not in aggregate account for more than 40 percent of a bank’s stock of HQLA and are subject to lower eligibility criteria. See assumption table below for more information.

12

Prepared by Valentina Flamini with model estimates provided by Paola Ganum (RES).

13

For methodological details see Cerutti, Eugenio, Stijn Claessens, and Patrick McGuire, 2012, “Systemic Risks in Global Banking: What can Available Data Tell Us and What More Dare are Needed?” BIS Working Paper 376, Bank for International Settlements.

14

Bank recapitalizations as well as other remedial policy actions (e.g., ring fencing, monetary policy, etc.) at the host and/or home country level are not assumed.

15

Based on consolidated claims on Costa Rica of BIS reporting banks—excluding domestic deposits of subsidiaries of these banks in Costa Rica.

16

Total credit to the non-bank sectors in Costa Rica is calculated by adding IFS local (both domestic and foreign owned) banks’ claims on non-bank borrowers and BIS reporting banks’ direct cross-border claims on non-bank sectors (BIS Locational Banking Statistics Table 6B).

Costa Rica: Selected Issues and Analytical Notes
Author: International Monetary Fund. Western Hemisphere Dept.