Panama: Selected Issues

Abstract

Panama: Selected Issues

Assessing Risks in the Panamanian Banking Sector: Stress Testing and Contagion Analysis1

A. Executive Summary

Risks Covered

1. This paper takes a close look at the recent evolution of the banking sector, and reports results from bank-level stress testing and contagion analysis. Tapping on publicly available balance-sheet data, we assess the resilience of individual banks to adverse shocks interest-rate and growth shocks. Using data on interbank deposits, the paper then assesses the potential for propagation of shocks through these exposures by tracking the effect of simulated defaults.

Key Results

2. The analysis suggests that Panama’s banking system seems able to withstand reasonably severe shocks, while contagion risks stem primarily from foreign banks. Ample starting capital buffers and bank profitability prevent translation of higher loan defaults under stress into materially impair capital adequacy ratios. If we reverse engineer the exercise to gauge what it would take to erase one-fourth of system capital, we find that the shock would need to be not only unprecedented, but also extremely large. In terms of contagion, while failures of both domestic and foreign banks would result in significant capital losses for Panamanian banks, the risk of contagion propagation is much higher in the case of the latter.

B. Introduction

3. Panama has a deep and well-integrated banking sector, and its adequate monitoring is key to macroeconomic stability. Panama’s financial depth, as measured by credit-to-GDP, is relatively high (Chart 1)2. With its role as regional trade and financial hub, and in the absence of a lender of last resort, periodic monitoring of the banking sector risks is essential to ensure the country continues to enjoy macroeconomic stability and maintains its sustainable growth path.

Chart 1.
Chart 1.

Private Sector Credit \1

as percent of GDP

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

\1 For country groups, the chart shows the median.Source: WEO and Staff calculations.

4. This paper takes a close look at the recent evolution of the banking sector, and reports results from bank-level stress testing. Panama’s economy has grown rapidly in recent years, and the banking system has accompanied this process. As growth slows down towards its medium-term potential, it is important to monitor how banks are adjusting to this new scenario. We therefore start our analysis by discussing recent trends in banking sector performance, highlighting potential sources of vulnerabilities. Tapping on publicly available balance-sheet data, we assess the resilience of individual banks to adverse shocks. The focus is on two key factors that may put pressure on banks’ balance sheets. First, and ahead of the expected U.S. monetary policy normalization, an increase in interest rates that may compress banks’ earnings, while raising loan defaults. Second, adverse growth shocks that also affects the quality of banks’ portfolios.

5. After assessing banks’ resilience at the individual level, the paper evaluates potential contagion risks stemming from the network of domestic as well as international interbank exposures. An important lesson from the financial crisis is that focusing the analysis only on individual financial institutions may lead to overlooking the buildup of significant risks arising from the system’s interconnectedness. To this end, Section D assesses the potential for propagation of shocks through interbank deposit exposures, by tracking the effect of simulated defaults.

6. The analysis suggests that Panama’s banking system seems able to withstand reasonably severe shocks. Ample starting capital buffers and bank profitability prevent translation of higher loan defaults under stress into materially impairing capital adequacy ratios. Only a number of smaller banks would fail the stress test featuring a slowdown and moderately higher interest rates, while only extreme conditions (recession in combination with sharply rising interest rates) would lead to widespread undercapitalization. If we reverse engineer the exercise to gauge what it would take to erase one-fourth of system capital, we find that the shock would need to be not only unprecedented, but also extremely large. Under less extreme assumptions the projected decline in system capitalization would appear manageable.

7. Contagion risks stem primarily from foreign banks. While failures of both domestic and foreign banks would result in significant capital losses for Panamanian banks, the risk of contagion propagation is much higher in the case of the latter. In fact, the systemic importance of foreign banks would be further magnified when funding risk through foreign banks’ exposure to Panama is taken into account.

C. Banking Sector Performance and Vulnerabilities

8. The banking sector is generally healthy. Most banks pursue a traditional business model with limited wholesale funding, are well-capitalized, profitable, and liquid. Offshore banks’ exposure to the domestic economy is limited to interbank exposures to onshore banks, and more recently purchases of certain domestic debt. Liquidity holdings are ample, since banks need to self-insure against funding risks in the absence of a lender of last resort.

9. Panama’s financial depth, as measured by credit-to-GDP, is relatively high. This partly reflects Panama’s role of a regional trade and financial hub. For example, domestic credit includes exposures to firms in the Colon Free Zone trading mainly internationally. After overshooting in the mid-2000s, bank credit has recently grown more in line with domestic activity, reflected in a broadly flat credit-to-GDP ratio since 2010 (Chart 2).

Chart 2.
Chart 2.

Panama: Credit-to-GDP Ratio

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

Source: Countrv authorities.

10. Non-performing loans (NPLs) are low and provisions coverage appears adequate in general, though there is heterogeneity across lending segments. Loan delinquencies hardly increased during the global financial crisis given Panama’s robust growth in those years. The overall NPL ratio currently stands at 1½ percent but the household sector displays greater loan defaults, which may be partly due to a shrinking payment capacity on the back of increasing household indebtedness (up by 15 percent y-o-y). The buffer of specific loan loss provisions has grown in recent years as the SBP ordered additional provisions for doubtful exposures to the Colon Free Trade Zone. Provisions coverage is also aided by the phasing-in of countercyclical provisions from 2014 (Chart 3).

Chart 3.
Chart 3.

Portfolio Quality by Recipient Sector of Bank Credit

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

Source: Superintendency of Banks Panama (SBP).

11. However, some recent trends in key bank performance indicators need to be monitored closely (Figure 1). In particular, the system’s capital adequacy ratio (CAR) and leverage ratio, albeit high, have worsened slightly over the past years. Foreign banks operate with somewhat less capital than their domestic competitors. After falling in the aftermath of the global financial crisis, the average liquidity ratio has recovered since 2013 on account of a rebound in foreign banks, while the negative trend persists at domestic banks. Meanwhile, profitability has improved markedly in the last few years, after slumping since the financial crisis. The difference between the performance of domestic/foreign banks and the system on average is explained by a lower profitability of public institutions.

Figure 1.
Figure 1.

Panama: Selected Financial Soundness Indicators, 2005–14

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

Source: IMF Staff calculations based on SBP data.

12. Panama’s financial soundness indicators point to overall higher buffers and performance in comparison to other financial hubs (Figure 2). While Panama’s average capital adequacy ratio is lower, its capital-to-total asset ratio is higher compared to Hong Kong, Luxembourg and Singapore. Panamanian banks’ return-on-assets (ROA) is also higher, implying a higher earnings buffer for absorption of shocks. While the NPL ratio is only marginally higher than in the other financial hubs, the coverage of nonperforming assets by provisions in Panama is better than in Luxembourg and Singapore, although the value of collateral is deducted from the loan value before provisioning. Finally, at about 100 percent, the loan-to-deposit ratio (not shown below) is quite high, but not excessively so in comparison with some other emerging market economies.

Figure 2.
Figure 2.

Panama: Financial Soundness Indicators in Peer Comparison

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

Source: IMF Staff calculations based on SBP data.

13. Rising interest rates may reduce banks’ earnings, while raising loan defaults. The prospect of an end to quantitative easing in the U.S. is a risk to profitability, as banks may not be able to pass on rising funding costs to borrowers, at least in the short run. The findings in the Selected Issues Paper Interest Rates in Panama: U.S. Pass-through and its Effects on Local Economic Activity suggest that the pass-through from U.S. interest rates has been lower for lending rates relative to deposit rates, implying a negative relationship between U.S. interest rates and Panamanian banks’ lending-deposit spreads. However, major banks hope to avoid a strong spread compression, as prevailing variable-rate loans should allow re-setting interest rates in due course. While margins may be preserved to some extent, rising loan rates would affect borrowers’ payment capacity and could lead to defaults. The estimated credit risk models (reported in Section C) provide evidence that, historically, the NPL ratio indeed tended to move in lockstep with domestic loan rates.

14. Real estate prices have been growing rapidly, but there is still little evidence of price growth being out of line with fundamentals. Index data for residential real estate prices, to which the SBP recently gained access, confirm a rapid increase over the last 10 years. Even so, house price developments do not seem to be out of line with the strong economic growth over the same period. Going forward, a continued pick-up in house prices is to be expected, as medium-to-low income earners increasingly gain access to mortgages, in large part at preferential rates. While there are no data on commercial real estate prices, there is anecdotal evidence of excess capacity in some segments (lower occupancy rates at hotels and office buildings). Banks claim that their exposures to these segments are limited.

15. At over 40 percent of GDP, household debt is lower than in advanced countries but relatively high in regional comparison. After declining in line with buoyant economic growth, the household debt-to-GDP ratio has only slightly outpaced nominal GDP recently (Chart 4). The debt-service-to-income ratio cannot be calculated yet, but the SBP plans to obtain the necessary income data directly from banks.

Chart 4.
Chart 4.

Household Indebtedness

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

D. Solvency and Liquidity Risk

Solvency Risk

16. A solvency stress test for onshore banks was performed by the mission team. This section describes the properties of the credit risk model that was implemented at the SBP with the help of IMF Technical Assistance in 2011 and updated in early 2015. The following paragraphs explain the model structure, the variable selection procedure, stress test scenarios as well as the method of projecting NPLs, provisions and capital adequacy ratios. For confidentiality reasons, stress test results are reported at the system level.

17. The type of credit risk model used in this exercise takes a balance sheet approach to projecting loan defaults. This approach aims at explaining variations in impaired loans by way of changes in key macroeconomic and financial variables. Specifically, it uses a dynamic panel data model with bank-specific fixed effects. In this panel model, NPL ratios for up to 51 Panamanian onshore banks during 2003Q1-2014Q4 were regressed on the lagged dependent NPL ratios, an array of explanatory variables at the economy and/or sector level and bank fixed effects depicting time-invariant bank-specific factors. A total of seven sectoral panel models were estimated for the primary sector, construction, manufacturing, commerce, services as well as for consumer and mortgage loans. Bank-specific data (loan interest rates) were tested but resulted significant only in the panel for the construction sector. The model uses a wide definition of NPLs that includes also “special mention” loans that in Panama require provisioning as well.3

18. In view of Panama’s openness, both domestic and external explanatory variables were tested, but in the end only domestic factors resulted significant. Specifically, the sectoral models include as explanatory variables real Panamanian GDP growth and/or the average interest rate on loans charged to the respective economic sector. Other variables such as U.S. or Chinese economic growth, exports, oil prices and inflation did not result significant in any of the newly-run regressions.4 This finding may be viewed as suggesting that it is mainly domestic factors, perhaps influenced by external factors, that explain variations in NPLs. As some of these additional variables entered in the 2011 version of the model, it stands to reason that there is an increasing decoupling of loan quality from external factors. In some sectoral models the level of significance was lower than in others, particularly in the services panel.

19. Text Table 1 provides an overview of the explanatory variables used in each panel. Real GDP growth entered in all panels but mortgage lending, while the average loan interest rate for a given sector turned out significant for the primary sector, manufacturing, commerce and for mortgage loans. Explanatory variables entered mainly with their contemporaneous levels, although some were lagged by one or two quarters. The lagged dependent variable always resulted highly significant, with coefficients ranging from 0.68 to 0.75, indicating moderate inertia in the evolution of NPL ratios. The regression fit can be considered good, with values of R-squared ranging between 0.74 and 0.80.

Text Table 1.

Explanatory Variables used in Sector-Specific Credit Risk Models

article image
Note: ***, **. * denote significance at the 1, 5 and 10 percent level, respectively.

20. Two stress test scenarios were considered. The baseline scenario applies the mission team’s projections for real GDP and the evolution of interest rates as well as credit growth during 2015-16, which does not necessarily imply stress but rather aims to project loan quality and capitalization under expected conditions (real GDP to rise slightly to 6.4 percent in 2016, and lending rates to increase by 60 basis points (bp) until the end of the projection horizon, Text Table 2). The adverse scenario contemplates a gradual slowdown of real GDP to 2 percent y-o-y by end-2016 and a likewise gradual increase in domestic loan rates by 113 basis points. Regarding lending rates, the shock originates from a rise in funding costs of 150 bp with an assumed 75 percent pass-through to loan rates (full pass-through in the baseline scenario). A direct determinant of risk-weighted assets under Basel I, credit growth is assumed to recede by -0.4 pp and -1.0 pp quarter-on-quarter in the baseline and adverse scenario, respectively.5

Text Table 2.

Stress Test Scenarios

article image

21. In addition, a reverse stress test6 was run to evaluate what kind of shock would be necessary to deliver a loss of 25 percent of system capital (“extreme scenario”). By backward induction it was ascertained that doubling the severity of the GDP shock and further increasing lending rates by 37 bp would erase one-fourth of capital. The additional rate shock is predicated on a deposit rate increase by 300 bp, half of which the bank is assumed to pass on to borrowers, implying both higher credit risk and lower net interest income.

22. NPL ratios were then projected on the basis of the seven sectoral credit risk models and using the aforementioned three scenarios. The NPL ratio is projected to rise from currently 1.3 percent to 2.4 and 3.9 percent under the adverse and extreme scenario, respectively (Text Table 3), while increasing only slightly under the baseline. Using the wider definition of NPLs that includes “special mention” loans, which also underlies the credit risk models, the delinquency ratio increases by 1.3 and 3.4 percentage points in the stress scenarios, reaching 6.7 percent in the extreme one.

Text Table 3.

Current and Projected NPL ratios

article image

23. To obtain the impact of the loan quality shock on banks’ income, the projected variation in the NPL ratio was then translated into credit losses. Projecting the increase in loan loss provisions required a forecast of the provisions-to-total-loans ratio that predicted both components separately. The effective provisioning rates for each bank and risk category were used to forecast the increase in the provisioning flow under each scenario.8 In order to project the share of impaired loans in each of the loan classification categories, a similar satellite model was estimated.9

24. The post-provision income was joined with a forecast of risk-weighted assets (RWAs) to obtain projected capital adequacy ratios. The final step of the exercise consisted in forecasting banks’ CARs by setting the projected change in capital on account of adjusted provisioning costs and other net income in relation to projected risk-weighted assets. In accordance with the Basel I accord applied in Panama, the projected change in commercial and consumer loans carries a risk weight of 100 percent and that in mortgage loans a weight of 50 percent. Note that higher provisioning does not necessarily imply falling capital, if banks’ pre-provision income is sufficiently high to absorb the additional cost. Chart 5 summarizes schematically the impact of stress conditions via higher loan impairment and lower financial margins onto net income and capital adequacy.

Chart 5.
Chart 5.

Transmission Channel from Macroeconomic Shocks to Capital Adequacy Ratio

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

25. The stress test results illustrate that the banking system appears resilient to adverse conditions. In both the baseline and the adverse scenario, the system’s capital adequacy ratio remains above 15 percent of RWA, whereas in the extreme scenario it drops to 12 percent (Text Table 4). While in the adverse scenario three small banks representing less than 1 percent of system capital would fall below the hurdle rate of 8 percent (in large part due to continued fast credit growth, which by assumption is not accommodated by capital injections), in the extreme scenario more than one-third of the banks accounting for 20 percent of overall capital would experience a decline in the CAR to below the 8 percent threshold.

Text Table 4.

Current and Projected Capital Adequacy Ratios, and Banks Failing Tests

article image

26. In interpreting the stress test results, a number of factors have to be taken into account. First, the drop in banks’ CAR in the baseline scenario despite still positive profits is somewhat fictitious, as it is explained by a softer but still reasonably strong credit growth, thus automatically raising RWAs virtually one-to-one, and also flat capital positions by assumption. The reason why such a drop in capital adequacy is seldom observed in tranquil times is that banks are periodically recapitalized by owners to retain the capacity to expand operations. As this possibility is not foreseen in the stress test, the results under the baseline clearly represent a worst-case outcome. Second, the relatively small drop in CAR under the adverse scenario relative to the baseline is owed to a partial offset of diminishing capital buffers due to rising loan impairment and lower pre-provision income by slower credit expansion and, hence, RWA growth in the denominator of the ratio. Finally, the assumptions underlying the reverse stress tests are to be qualified extreme but plausible even though Panama has never experienced the combination of such a GDP plus funding rate shock. The purpose of the reserve stress test is to show that it would take an unprecedented event combining most severe domestic and external shocks to cause a systemic event.

Liquidity Risk

27. Notwithstanding ample liquidity holdings of Pananamian banks, a liquidity stress test by way of approximating the Liquidity Coverage Ratio (LCR) was run. Required by regulation and also due to the lack of a lender of last resort, banks have historically held a relatively large share of assets in cash and highly liquid instruments. The SBP requires banks to maintain a ratio of liquid assets, as defined by regulation, to total short-term liabilities of 30 percent. The liquidity index is calculated at the 6-month horizon, and includes repayments of principal and interest on interbank and customer loans at that horizon. In the first months of 2015, the index has fluctuated in a range of 59 to 62 percent. This liquidity concept does not correspond well to the liquidity standards established as part of the Basel III framework (BCBS, 2013), which require banks to maintain adequate coverage of liabilities at the 30-day horizon (the LCR) and ensure sufficiently stable funding at the 12-month horizon (the Net Stable Funding Ratio, NSFR). Given data limitations, the mission team decided to calculate only the LCR at this point.

28. The LCR resembles a traditional deposit run-off stress test measuring the degree to which outflows can be met by liquid assets, yet it reinforces the importance of maintaining high-quality liquid assets (HQLA). The quality of liquidity assets is reflected in haircuts to the balance sheet value of liquidity positions, with the highest quality assets (cash, central bank reserves, and certain domestic debt) not requiring value adjustments and those of lesser quality (“Level 2 assets”) commanding haircuts of between 15 and 50 percent according to Basel III. The total adjusted value of HQLA is required to exceed the net cash outflows in the denominator of the ratio, comprising outflows of deposits and other liabilities weighted by run-off factors net of similarly-weighted cash inflows from asset positions, including customer loans. Two important restrictions apply under the Basel III LCR framework: the share of level 2 assets may not exceed 40 percent of total HQLA, and cash inflows to be subtracted are the lower of calculated inflows and 75 percent of gross outflows, thus preventing the denominator from becoming very small or even negative.

29. A number of assumptions needed to be made to approximate the LCR for Panama. Available data for the cash outflows on maturing wholesale liabilities and repo agreements is reported only at a residual maturity of 12 months. Based on evidence shared by the SBP, it was assumed that 1/12th of wholesale liabilities and all of the repos mature within the next month. Additionally, deposits at Banco Nacional de Panamá, a state-owned bank entrusted with certain functions typically carried out by the central bank, were given the HQLA-haircut of 0 percent foreseen for reserves at central banks. Text Table 5 summarizes the components of the LCR taken into account in the present case.

Text Table 5.

Approximation of LCR for Panamanian Banking Sector

article image

30. The results of the LCR approximation confirm that the Panamanian banking system is highly liquid on average, yet some banks would not meet the new standard. Keeping in mind the margin of error introduced by the aforementioned assumptions, the median LCR is calculated to be 284 percent as at December 2014, well above the Basel III requirement of 100 percent. However, the tails of the distribution are thick, with some banks around 1000 percent and some clearly below the 100 percent mark – the lower quartile lies at 125 percent. Specifically, nine banks are assessed not to hold a sufficient amount of high-quality assets given their liability structures, although they are compliant according to the Panamanian liquidity requirement (Text Table 6). Upon closer inspection, the limiting factor is generally the volume of cash or cash-like positions that is often too low to allow inclusion of the entire volume of available Level 2 assets.

Text Table 6.

Aggregate Results of LCR Approximation

article image

31. The outcome of the liquidity test also illustrates that upgrading the liquidity regulation and data compilation would be needed to become compliant with Basel III. The current liquidity requirement ought to be brought in line with the LCR standard, while computation of the NSFR would bolster the quality of longer-term funding. In this context, efforts to compile data at the required 30-day horizon would be necessary to move from an approximation to the calculation of a Basel III-compliant LCR.

E. Interbank Contagion Risk

A Simple Model of Interbank Contagion

32. This section studies the resilience of the Panamanian banking system by tracking the effect of simulated defaults. We apply the balance sheet-based analysis proposed by Espinosa-Vega and Sole (2010) to identify banks’ contagiousness and vulnerability. A single bank is assumed to fail, spreading contagion to other banks (both creditors and borrowers) through two different channels. First, institutions which are creditors to a failing bank are affected through the credit channel, losing a fraction λ of their deposits. Second, borrowers from the failing bank are affected through the funding channel. In particular, these borrowers can only replace a fraction 1-ρ of the funding they were getting from the failed bank and need to restore their balance-sheet identity. They achieve this by selling assets at a (fire-sale) price of $1/(1+δ) on the dollar. We adopt the baseline calibration in Espinosa-Vega and Sole (2010) and assume λ=1, ρ=0.35, and δ=1. While it may be considered an extreme case, the 100-percent loss given default assumption (λ = 1) reflects our conservative approach to account for potentially high uncertainty related to bankruptcy resolutions.

33. We consider three scenarios for the threshold of capital under which banks would become insolvent. As contagion travels through the network, a key question is how low a bank’s capital adequacy ratio needs to be before the bank files for bankruptcy. We take an agnostic stance on this issue, and run the simulations with three different thresholds. A high-sensitivity scenario assumes that banks default when their CAR falls below 8 percent. In a medium-sensitivity scenario, banks are assumed to default when their CAR is below 4 percent. Finally, a low-sensitivity scenario is the most benign, assuming that banks default on their obligations (and remove their liquidity from the system) only when their capital is fully depleted. In practice, the relevant threshold will depend on country- and market-specific factors such as regulation and the ex-ante stability of the system.

34. To understand the magnitude of necessary interventions upon an exogenous default, capital losses are classified in three categories: buffer losses, injection required to restore CAR, and excess losses. After each simulation, capital losses (excluding those of the original defaulting bank) are classified as follows. Buffer losses are those over and above 8 percent of each bank’s risk-weighted assets, whereas the injection required to restore CAR represents the difference between 8 percent of risk-weighted assets and actual capital after contagion brought capital below the 8 percent threshold. For defaulting banks in scenarios 1 and 2 (i.e. those banks whose capital is below 8 and 4 percent of CAR, respectively), any remaining is classified as excess loss.

Data

35. The data include December 2014 interbank deposits among 77 Panamanian banks, and the deposits these banks have in 151 foreign banks. Of the Panamanian banks, 27 have international license (offshore), while general-license (onshore) banks include 20 owned by residents and the remaining 30 owned by foreigners. The dataset is constructed by SBP based on banks’ weekly liquidity report. Each bank reports on a weekly basis its deposits in other depository institutions, both in Panama and abroad. Since banks report the deposits they make, and not the ones they receive, the data do not include information on foreign banks’ deposits in Panama.

36. Interbank deposit data were matched with bank-level data on capital and risk-weighted assets. SBP is the supervisor of origin for 52 of the 77 Panamanian banks included in the dataset. While balance-sheet data are available for all 77 banks, only those supervised by SBP are required to report regulatory capital and risk-weighted assets data to SBP.10 Risk-weighted assets of the remaining 25 banks were estimated by multiplying their unweighted assets by the average ratio of risk-weighted assets to unweighted assets of the banks under SBP’s supervision. A similar approach was taken with regulatory capital. However, this left 11 banks below the required capital adequacy ratio (CAR) of 8 percent. For those banks we set the capital so that the CAR is equal to the median CAR of the other banks.

Text Table 7.

Interbank Deposits by License Type 1/

article image

GL stands for “General license, local”, GF for “General license, foreign”, IL for “International license”.

Results

37. Capital losses following an exogenous default are only weakly related to the number of contagious defaults. Following the default of an institution, both the capital loss of the entire banking system as well as the number of contagious defaults are important indicators of the system’s resilience to shocks for several reasons. For example, to assess capital needs and potential fiscal impact, information on capital losses is crucial. For monitoring the risks of a banking crisis, the number of defaults is also relevant. Figure 3 shows, for each scenario, the link between total capital losses and number of contagious defaults. While there is a natural positive relationship (to the extent that capital losses are necessary for contagion), similar levels of capital impairment can be associated with widely different outcomes in terms of defaults. For example, in Scenario 1, seven simulations generate losses greater than U$S 1.5bn (or 1.4 percent of total assets, 13.2 percent of equity, and 3.2 percent of GDP). While up to 9 contagious defaults take place in the worst of these simulations, some feature only one or zero contagious defaults. On the one hand, the failure of a bank that receives many but small deposits (relative to depositors’ capital) in the interbank market is likely to generate substantial capital losses but few instances of contagion. On the other hand, the failure of banks that receive large deposits from comparatively fewer banks is likely to be highly contagious.

Figure 3.
Figure 3.

Panama: Total Capital Losses and Number of Defaults 1/

(Based on three scenarios related to sensitivity of bank failures to CAR)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

38. Contagion risks stem mainly from foreign banks, while exposures to domestic general-license banks, albeit substantial, are much less contagious. Figure 4 shows, for each scenario, total capital losses and number of defaults caused by the 30 banks that inflict the largest losses on the system. The license type of the original defaulting bank is displayed in lieu of its name. The largest capital loss is generated by the simulated default of a foreign institution, and amounts to U$S 2.8bn, or 24 percent of the system’s capital and 6 percent of GDP. The largest capital injection required amounts to U$S 1.4bn, or 12 percent of capital and 3 percent of GDP. While both foreign and general-license banks feature prominently in all three scenarios, the losses generated by the latter are mostly absorbed by the system’s buffers. Furthermore, even in cases where defaults take place, the injection required to restore adequacy ratios is typically much smaller following the default of a Panamanian bank than that of a foreign bank. For instance, in Scenario 1, the largest injection required by the default of a general-license bank amounts to U$S180mn (Figure 4, top panel, seventh bank starting from the right). In comparison, the average injection required by the default of one of the foreign banks displayed in Figure 4 under this scenario is U$S330mn.

Figure 4.
Figure 4.

Panama: Capital Losses as Function of Original Defaulting Bank 1/2/

(US$ millions, 30 most systemically important institutions)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

1\ Each column displays total capital losses in the banking system (excluding those of the original defaulting bank) following the exogenous default of one bank (the license type of the original defaulting bank is shown in lieu of its name). Buffer losses are those over and above 8 percent of RWA, whereas injection to restore CAR represents the difference between 8 percent of RWA and actual capital after contagion. For defaulting banks (i.e. those whose capital is below 8, 4, and 0 percent of CAR in scenarios 1,2, and 3, respectively), if remaining capital is positive then that amount is classified as excess loss.2\ Total number of failures excludes original defaulting institution.

39. The systemic importance of foreign banks would be even larger if the data included foreign exposures to Panama. The dataset used here does not include foreign banks’ deposits in Panama and is therefore insufficient to assess foreign funding risks. Appendix C provides an assessment of this source of risk based on cross-border exposures reported to the Bank for International Settlements (BIS). The results suggest that Panama could face a significant contraction in foreign credit availability if a bank deleveraging cycle were triggered in Europe or North America.

40. Capital losses are fairly similar across scenarios, partly reflecting the fact that vulnerable banks tend to be less contagious (Figure 3). Total capital losses across all simulations amount to $39.6bn in Scenario 1, $37.6bn in Scenario 2, and $36.4bn under Scenario 3. Differences in capital losses across scenarios will only arise if they feature different sets of failing banks, and this only happens to a limited extent in our results. The reasons are that exposures to banks that are prone to contagious defaults are small, and exposure chains are in general short. In this regard, it is worth noting for example that, out of the 50 contagious simulations in Scenario 1, 41 do not cause defaults beyond the first contagion round, 6 take two rounds, and the remaining ones take three rounds.

41. In fact, banks’ contagiousness is negatively correlated with their vulnerability, implying that banks whose failures would have knock-on effects are typically more resilient to other banks’ failures. We define a bank’s contagiousness as the number of other institutions that would fail as a result of the latter’s failure. Conversely, a bank’s vulnerability is defined as the number of simulations in which the bank failed as a result of another bank’s failure.11 Text Table 8 presents estimates from regressing Panamanian banks’ contagiousness on their vulnerability, excluding from the sample those banks that are neither contagious nor vulnerable.12

Text Table 8.

Contagiousness Vulnerability Regressions 1/

article image

For banks that can both originate and be affected by cascades, the estimates correspond to regressions of “contagiousness” on “vulnerability”. t statistics in parentheses

* p<0.05, ** p<0.01, *** p<0.001

42. Offshore banks have little systemic importance in terms of deposit exposures, and spillovers to the onshore system are negligible. Text Table 9 aggregates the results of all the simulations distinguishing by the license type of the originating bank and of the banks affected by the cascade.13 It is evident that the relationship between the offshore banking system and onshore banks is weak and asymmetric. While the failure of some onshore banks may cause problems in parts of the offshore center, the reverse does not hold. Furthermore, only two banks would have their CAR brought down below 8 percent by failures of offshore banks. No offshore bank’s failure would impair another bank’s capital to the extent of bringing it below 4 percent.

Text Table 9.

Number of Failures and Capital Losses by License Type 1/

article image

GL stands for “General license, local”, GF for “General license, foreign”, IL for “International license”. Each cell aggregates the effects of simulations by the type of license of the affected institutions. The results are colored according to deciles (e.g. the darkest red corresponds to highest contagion levels):

Color code (deciles)

F. A Combined Shock Scenario Based on Stress Testing Results

43. This section considers the systemic implications of a scenario where banks that do not perform well in the individual stress tests simultaneously fail, thereby affecting the rest of the banking system. In other words, we put together the results of the individual stress-testing exercise (Section C) with the interbank contagion analysis (Section D). We do so by tracking the potential effects of the simultaneous failure of those banks that fall below the regulatory CAR in the severe stress-testing scenario.

44. The simultaneous shock would have modest implications for the rest of the banking system. In all contagion scenarios (high-, medium-, and low-sensitivity), total capital losses (excluding those of the originally defaulting institutions) would amount to $147mn.14 Even though almost all of these losses (97 percent) are concentrated in the onshore system, the shock would not cause the default of any onshore bank in any of the scenarios (Chart 7). In terms of the offshore system, one bank would require a capital injection of less than $1mn in order to restore its CAR to 8 percent. Moreover, this offshore bank would only fail in the high-sensitivity scenario, as the post-shock capital would remain above 4 percent of risk-weighted assets.

Chart 7.
Chart 7.

Capital Losses with Combined Shock Scenario

(US$ millions)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

45. Further assessments of systemic-risk should consider the response to common shocks, and incorporate additional channels for contagion. The methodology employed in this section is well-suited for studying the systemic implications of idiosyncratic bank failures. Such failures are not necessarily related to underlying developments in the real economy, as is the case, for example, when bankruptcy is filed upon fraud or corruption cases. However, from a regulatory standpoint, it is also important to evaluate how the system would cope should a shock affect several institutions simultaneously. In this sense, it is particularly important to take into account and understand the potential implications of common exposures in the real economy. These common exposures may also open additional avenues for propagation that are not taken into account in the analysis of this section. More specifically, institutions under stress may be forced to deleverage through fire sales, thereby indirectly affecting the balance sheet of all institutions holding similar asset classes.

G. Concluding Remarks

46. Panama’s banking system has performed well in recent years and appears resilient to a variety of shocks. Key financial soundness indicators remain at levels indicating low risk, although some of them have lately worsened somewhat. Pockets of vulnerabilities exist in the dependence of local interest rates on U.S. monetary policy action and in excess capacity in commercial real estate. Another weakness is the relatively high debt of households that, going forward, may impair their payment capacity in times of distress.

47. The banking system seems able to withstand reasonably severe shocks. Ample initial capital buffers and bank profitability prevent translation of higher loan defaults under stress into materially impair capital adequacy ratios. Only a number of smaller banks would fail the stress test featuring a slowdown and a moderate increase in rates interest rates, while only extreme conditions (recession in combination with sharply rising interest rates) would lead to widespread undercapitalization. However, it would take an unprecedented degree of shocks to cause a systemic event erasing one-fourth of system capital, whereas under less extreme assumptions the projected decline in system capitalization would appear manageable.

48. On average liquidity holdings are more than adequate, although pockets of vulnerability exist. As was to be expected from the large share of highly liquid instruments that Panamanian banks maintain, the median LCR is about three times the minimum required by Basel III. At the same time, a number of banks would not meet the LCR requirement at this point, owing to insufficient cash positions limiting the inclusion of lesser-quality liquid assets in the overall stock of high-quality liquid assets. Updating the liquidity regulation and corresponding data collection to ensure compliance with Basel III liquidity norms should therefore be a policy priority.

49. Contagion risks stem mainly from foreign banks, while exposures to domestic general-license banks, albeit substantial, are much less contagious. While failures of both domestic and foreign banks would result in significant capital losses for Panamanian banks, the risk of contagion propagation is much higher in the case of the latter. In fact, the systemic importance of foreign banks would be further magnified when funding risk through foreign banks’ exposure to Panama is taken into account. As a final cautionary note, and as noted in the previous section, a more comprehensive systemic-risk analysis should consider the response to common shocks, and incorporate additional channels for contagion.

References

  • Cerutti, E. (2013), Banks’ Foreign Credit Exposures and Borrowers’ Rollover Risks: Measurement, Evolution and Determinants, IMF Working Paper WP/13/09.

    • Search Google Scholar
    • Export Citation
  • Espinosa-Vega, M. and J. Sole (2010), Cross-Border Financial Surveillance: A Network Perspective, IMF Working Paper WP/10/105.

  • Basel Committee on Banking Supervision (2013), Basel III: The Liquidity Coverage Ratio and liquidity risk monitoring tools, January 2013.

    • Search Google Scholar
    • Export Citation
  • Beck, N. and J.N. Katz (2009), Modeling Dynamics in Time-Series-Cross-Section Political Economy Data, California Institute of Technology Social Science Working Paper 1304.

    • Search Google Scholar
    • Export Citation
  • Henry, J. and C. Kok (2013), A Macro Stress Testing Framework for Assessing Systemic Risks in the Banking Sector, ECB Occasional Paper No. 152.

    • Search Google Scholar
    • Export Citation
  • Wawro, G.J., C. Samii, and I.P. Kristensen (2011), On the Use of Fixed Effects Estimators for Time-Series Cross-Section Data, mimeo. Available at: https://files.nyu.edu/cds2083/public/docs/kristensen_etal.pdf.

    • Search Google Scholar
    • Export Citation

Appendix I. Contagion Analysis - Capital Losses by License Type

The top-left panel in Appendix I Figures 13 correspond to the results shown in Figure 3 in the main text, whereas the remaining panels present the breakdown by the license type of the affected institution.

Appendix I Figure 1.
Appendix I Figure 1.

Capital Losses as Function of Bank Originating Cascade 1/2/High-sensitivity scenario Breakdown by Type of Bank Being Affected

(US$ millions, 30 most systemically important institutions)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

1\ Each column displays total capital losses in the banking system (excluding those of the original defaulting bank) following the exogenous default of one bank (the license type of the original defaulting bank is shown in lieu of its name). Buffer losses are those over and above 8 percent of RWA, whereas injection to restore CAR represents the difference between 8 percent of RWA and actual capital. For defaulting banks (i.e. those whose capital is below 8, 4, and 0 percent of CAR in scenarios 1,2, and 3, respectively), if remaining capital is positive then that amount is classified as excess loss.2\ Total number of failures excludes original defaulting institution.
Appendix I Figure 2.
Appendix I Figure 2.

Capital Losses as Function of Bank Originating Cascade 1/2/Medium-sensitivity scenario Breakdown by Type of Bank Being Affected

(US$ millions, 30 most systemically important institutions)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

1\ Each column displays total capital losses in the banking system (excluding those of the original defaulting bank) following the exogenous default of one bank (the license type of the original defaulting bank is shown in lieu of its name). Buffer losses are those over and above 8 percent of RWA, whereas injection to restore CAR represents the difference between 8 percent of RWA and actual capital. For defaulting banks (i.e. those whose capital is below 8, 4, and 0 percent of CAR in scenarios 1,2, and 3, respectively), if remaining capital is positive then that amount is classified as excess loss.2\ Total number of failures excludes original defaulting institution.
Appendix I Figure 3.
Appendix I Figure 3.

Capital Losses as Function of Bank Originating Cascade 1/2/Low-sensitivity scenario Breakdown by Type of Bank Being Affected

(US$ millions, 30 most systemically important institutions)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

1\ Each column displays total capital losses in the banking system (excluding those of the original defaulting bank) following the exogenous default of one bank (the license type of the original defaulting bank is shown in lieu of its name). Buffer losses are those over and above 8 percent of RWA, whereas injection to restore CAR represents the difference between 8 percent of RWA and actual capital. For defaulting banks (i.e. those whose capital is below 8, 4, and 0 percent of CAR in scenarios 1,2, and 3, respectively), if remaining capital is positive then that amount is classified as excess loss.2\ Total number of failures excludes original defaulting institution.

Appendix II. Foreign Downstream Exposures

Panama’s claims on foreigners—its “downstream exposures”—are broadly distributed internationally, with a significant share corresponding to LA5 economies. The data reported to the BIS can also be used to study Panama’s exposure by borrower country. The Panamanian banking system’s global downstream exposure amounts to almost 40 percent of GDP as of 2014Q3, significantly higher than the corresponding figures for other BIS-reporting LA5 economies.

The high bilateral downstream exposures through lending to foreign clients imply significant potential credit losses. Estimated using inputs from the IMF’s Vulnerability Exercises for Advanced and Emerging Economies and employing standard loss-given-default ratios (75 percent for the sovereign and 60 percent for the private sector), potential downstream losses amount to almost 5 percent of GDP, with Brazil accounting for about a quarter of this total. Compared to other BIS-reporting LA5 countries, potential losses of Panamanian banks are several times larger. Overall, the results confirm the findings of the main text in terms of the potential credit risks stemming from foreign exposures of Panamanian internationally-active banks.

Downstream exposure of Panama’s banking system

article image
Source: IMF staff calculations based on BIS, Bankscope, ECB, and IFS data.

Total refers to the downstream exposure (cross-border claims plus foreign claims) of the Panamanian banking system vis-à-vis all borrowers in the rest of the world. For the other countries indicated as column heading, these are the bilateral claims of the Panamaian banking system vis-à-vis borrowers in each country.

A01ufig1

Panama: downstream exposure potential losses (% of GDP)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

Source: IMF staff calculations based on BIS, Bankscope, ECB, and IFS data, using the IMF’s Research Department Bank Contagion Module.

Appendix III. Assessing Foreign Funding Risk

This Appendix assesses the funding risk from a foreign systemic event based on BIS bilateral banking statistics. To complement the analysis of the main text, here we examine the risks posed by the exposures of foreign banks to Panamanian residents (that is, the “upstream exposures” of Panama). BIS-reporting banks’ claims on Panama amount to about 73 percent of GDP. This is significantly higher than the upstream exposures of other LA countries and gives rise to considerable funding risk.

Upstream exposure of Panama to foreign credit availability

article image
Source: IMF staff calculations based on BIS, Bankscope, ECB, and IFS data.

Foreign credit availability in Panama may be significantly affected in case of adverse shocks to foreign banks’ balance sheets. This approach measures the potential impact of a shock to the banking system in a foreign country on the availability of foreign credit in Panama (see Cerutti (2013) for details on the methodology). It is assumed that the shock leads to defaults on a given percentage of bank assets, which forces the banks to deleverage in order to restore capital at the prescribed adequacy level. The exercise covers two shocks: a uniform 3 percent default rate for banking systems in each creditor country considered (scenario A), and a country-specific default rate based on the change in their NPL indicators in recent years (scenario B).1 Intuitively, higher exposure and weaker capital position translate into stronger impact from deleveraging. The results in Appendix III Figure 1 indicate that shocks in Europe and North America lead to significant declines in credit availability to Panama, with a joint shock (to bank balance sheets in European countries, the U.S., and Canada) exerting credit contractions in Panama of 7.3 percent of GDP and 16.6 percent of GDP in scenarios A and B, respectively.2

Appendix III Figure 1.
Appendix III Figure 1.

Impact on Credit Availability from shocks to BIS-reporting Banks of Selected Countries (% of GDP)

Citation: IMF Staff Country Reports 2015, 238; 10.5089/9781513550862.002.A001

Source: IMF staff calculations based on BIS, Bankscope, ECB, and IFS data, using the IMF’s Research Department Bank Contagion Module.1/ Country group “EUR” includes Greece, Ireland, Portugal, Italy, Spain, France, Germany, The Netherlands, and UK.
1

Prepared by D. Cerdeiro, M. Hadzi-Vaskov and T. Wezel. We would like to thank Paola Ganum, Etibar Jafarov, Camelia Minoiu, and Wei Shi for their contributions. We are also very grateful to the Panamanian authorities (Superintendencia de Bancos de Panama) for excellent cooperation and sharing the data.

2

In addition, and consistent with its role as a financial center, total assets of the banking sector represent about 250% of GDP at end-2014.

3

The NPL variable underwent a logistic transformation, as is standard in regression-based credit risk models. Since by construction the dependent variable is bounded between zero and one, in the regression the logit-transformed value was used to create an unrestricted variable and thus avoid non-normality of the error term.

4

External factors nonetheless may have significant impact on the quality of Panamanian banks’ credit exposures. Based on BIS data, Appendix B suggests that high bilateral downstream exposures through lending to foreign clients imply significant potential credit losses.

5

Other necessary assumptions comprise a tax rate on bank profits of 28 percent; a dividend payout rate of 25 percent; and quarterly growth rates of investment and other income as well as operating costs of 2, 1¾ and 1½ percent in the baseline, adverse and extreme scenario, respectively (the decline of operating costs assumes cost-saving measures by banks under adverse conditions). The drop in income is assumed to be more moderate than may be called for under stressful conditions since rising interest rates, particularly on the large holdings of liquid assets, would have a mitigating effect.

6

Details on the properties of a reverse stress test can be found in Henry and Kok (2013).

7

Including loans classified as in the special mention category.

8

The effective provisioning rates rather than the statutory ones were taken because loans in the system were amply collateralized. To arrive at effective rates, the statutory rates of 20, 50, 80 and 100 percent for special mention, substandard, doubtful and loss loans, respectively, were adjusted by considering the average degree of collateralization in each of the seven sectors.

9

To obtain the share of each of the categories of impaired loans for the projection, a fixed effects panel data model was estimated with non-performing loans (as previously estimated for each bank and economic sector) as explanatory variable.

10

For some banks, the regulatory capital is significantly larger than the capital that shows up in the balance sheet, as regulatory capital corresponds to a larger holding.

11

Espinosa-Vega and Sole (2010) refer to this measure as the bank’s hazard.

12

Several banks are neither contagious nor vulnerable. If these were included in the regression, the estimated slope would be positive. The regressions in Table 2 are thus conditional on the bank having meaningful exposures in the interbank deposit network.

13

Contrary to Figure 4, Text Table 9 is based on the results of all simulations, and not just the ones corresponding to the 30 most important institutions.

14

For conciseness, here we show the results for the high-sensitivity scenario. However, total capital losses differ by less than U$S 0.1 million across scenarios. In the medium- and low-sensitivity scenarios no contagious default takes place, and thus no capital injection is required.

1

Scenario A: default rate equals 3 percent on all on-balance sheet claims (all borrowing sectors/all countries). Scenario B: losses amounting to 2.5 percent, 5 percent, and 10 percent, respectively, of bank balance sheet claims in countries whose change in non-performing loans (NPLs) from 2007 to 2012 was between 0 and 2.5 percent (Canada, Germany, France, and The Netherlands); between 2.5 and 5 percent (US and UK); and above 5 percent (Greece, Ireland, Italy, Portugal, and Spain).

2

The reduction in foreign banks’ credit due to the impact of the shock to their balance sheet assumes a uniform deleveraging across domestic and external claims. All simulations are based on BIS bilateral banking statistics as of 2014Q3.

Panama: Selected Issues
Author: International Monetary Fund. Western Hemisphere Dept.
  • View in gallery

    Private Sector Credit \1

    as percent of GDP

  • View in gallery

    Panama: Credit-to-GDP Ratio

  • View in gallery

    Portfolio Quality by Recipient Sector of Bank Credit

  • View in gallery

    Panama: Selected Financial Soundness Indicators, 2005–14

  • View in gallery

    Panama: Financial Soundness Indicators in Peer Comparison

  • View in gallery

    Household Indebtedness

  • View in gallery

    Transmission Channel from Macroeconomic Shocks to Capital Adequacy Ratio

  • View in gallery

    Panama: Total Capital Losses and Number of Defaults 1/

    (Based on three scenarios related to sensitivity of bank failures to CAR)

  • View in gallery

    Panama: Capital Losses as Function of Original Defaulting Bank 1/2/

    (US$ millions, 30 most systemically important institutions)

  • View in gallery
  • View in gallery

    Capital Losses with Combined Shock Scenario

    (US$ millions)

  • View in gallery

    Capital Losses as Function of Bank Originating Cascade 1/2/High-sensitivity scenario Breakdown by Type of Bank Being Affected

    (US$ millions, 30 most systemically important institutions)

  • View in gallery

    Capital Losses as Function of Bank Originating Cascade 1/2/Medium-sensitivity scenario Breakdown by Type of Bank Being Affected

    (US$ millions, 30 most systemically important institutions)

  • View in gallery

    Capital Losses as Function of Bank Originating Cascade 1/2/Low-sensitivity scenario Breakdown by Type of Bank Being Affected

    (US$ millions, 30 most systemically important institutions)

  • View in gallery

    Panama: downstream exposure potential losses (% of GDP)

  • View in gallery

    Impact on Credit Availability from shocks to BIS-reporting Banks of Selected Countries (% of GDP)