Republic of Korea
Financial Sector Assessment Program-Technical Note-Systemic Risk Analysis, Financial Sector Stress Testing, and an Assessment of Demographic Shift in Korea

This note presents the systemic risk analysis conducted for the Republic of Korea in the course of the 2019 Korea FSAP. It comprises a forward-looking solvency analysis for banks, insurers, and pension funds, a liquidity stress test for banks, and an assessment of network and interconnectedness for a wide range of financial sector entities and their ties to the real economy. Various structural characteristics of Korea’s economy and its financial system informed the features and focus for its forward-looking risk analysis. They include Korea’s strong export orientation, limited diversification, and its key role as a node in regional and international supply chains. Korea’s financial system has grown by 40 percentage points of GDP since 2013, enhancing the importance of a deep financial sector analysis as conducted through the FSAP. Mortgage insurance schemes are widely used—which was reflected in the way the risk assessment for banks was conducted. Korea’s life and non-life insurance sector is large, highly concentrated and saturated. Fintech developments keep accelerating, in terms of its Open Banking system and e-money providers. Demographic developments in Korea are among the most adverse world-wide, implying a continuous drag on demand, downward pressure on interest rates, financial firms’ income, and hence their capitalization unless they will be altering their business models.

Abstract

This note presents the systemic risk analysis conducted for the Republic of Korea in the course of the 2019 Korea FSAP. It comprises a forward-looking solvency analysis for banks, insurers, and pension funds, a liquidity stress test for banks, and an assessment of network and interconnectedness for a wide range of financial sector entities and their ties to the real economy. Various structural characteristics of Korea’s economy and its financial system informed the features and focus for its forward-looking risk analysis. They include Korea’s strong export orientation, limited diversification, and its key role as a node in regional and international supply chains. Korea’s financial system has grown by 40 percentage points of GDP since 2013, enhancing the importance of a deep financial sector analysis as conducted through the FSAP. Mortgage insurance schemes are widely used—which was reflected in the way the risk assessment for banks was conducted. Korea’s life and non-life insurance sector is large, highly concentrated and saturated. Fintech developments keep accelerating, in terms of its Open Banking system and e-money providers. Demographic developments in Korea are among the most adverse world-wide, implying a continuous drag on demand, downward pressure on interest rates, financial firms’ income, and hence their capitalization unless they will be altering their business models.

Executive Summary 1

This note presents the systemic risk analysis conducted for the Republic of Korea in the course of the 2019 Korea FSAP. It comprises a forward-looking solvency analysis for banks, insurers, and pension funds, a liquidity stress test for banks, and an assessment of network and interconnectedness for a wide range of financial sector entities and their ties to the real economy.

Various structural characteristics of Korea’s economy and its financial system informed the features and focus for its forward-looking risk analysis. They include Korea’s strong export orientation, limited diversification, and its key role as a node in regional and international supply chains. Korea’s financial system has grown by 40 percentage points of GDP since 2013, enhancing the importance of a deep financial sector analysis as conducted through the FSAP. Mortgage insurance schemes are widely used—which was reflected in the way the risk assessment for banks was conducted. Korea’s life and non-life insurance sector is large, highly concentrated and saturated. Fintech developments keep accelerating, in terms of its Open Banking system and e-money providers. Demographic developments in Korea are among the most adverse world-wide, implying a continuous drag on demand, downward pressure on interest rates, financial firms’ income, and hence their capitalization unless they will be altering their business models.

The bank solvency risk analysis concludes that the Korean banking system is broadly resilient to a severe economic downturn scenario. The performance of banks under a macro-financial downturn scenario would be to an extent heterogeneous, with the drop in specialized banks’ capital ratios being most pronounced, and those of ODIs, including Mutual Savings Banks and Credit Cooperatives, being most dispersed. None of the nation-wide, regional, and specialized banks are found to fall short of regulatory capital minima. Fintech-implied rising competition in the banking system can have notable implications in terms of systemic liquidity, implied solvency as well as system-wide operational risks.

The FSAP team’s assessment is that the adverse FSAP macro-financial scenario is severe enough to encapsulate a COVID-19 implied fallout on economic activity. This assessment stands both in terms of the expected depth and duration of the shock (the FSAP’s scenario’s deep downturn spans over two full years before normalizing).

Measures have been taken to strengthen the liquidity of the Korean banking system and banks have ample room to withstand liquidity shocks. Overall, the banking system is likely to maintain adequate liquidity, both in terms of domestic and foreign currency, following hypothetical asset price falls, retail funding and wholesale funding shocks. In general, this is the assessment for all types of banks in Korea. However, state-owned banks’ reliance on unsecured wholesale funding would lead to a fall in FX liquidity coverage from 112 percent to 85 percent.

Insurance firms’ capital would decline substantially under the stress scenario, while all firms would stay above regulatory thresholds. The top-down solvency stress test for life and non-life insurance firms covered about 75 percent of the market. The same narrative and severity of the scenario from the banking sector solvency analysis were adopted. Under the current accounting regime, held-to-maturity designation is still allowed—and widely used among life insurers—hence a significant portion of insurers’ investments were shielded from market price changes. Life insurers specifically would experience a significant decline in income, pointing to the need for them to restructure their business to restore underwriting profitability, and shifting further from guaranteed savings products into lower guarantees and protection business.

The interconnectedness analysis for the Korean financial system suggests that specialized banks are as systemic as nation-wide commercial banks. Nation-wide banks are net lenders to specialized banks, and hence are more vulnerable to specialized banks than the other way around. Insurers are not as systemic but vulnerable to stress, should it arise, in the banking system; insurers are net lenders to all other sub-segments of the financial system in Korea.

The recommendations derived from the systemic risk analysis pertain in many respects to enhancing the Korean authorities’ quantitative assessment frameworks (Table 1). They include the urgent need for a detailed quantitative impact analysis of Fintech developments in terms of systemic liquidity and system-wide operational risks which the Open Banking and e-money developments may imply—despite their positive intent and expected benefits. Real estate market structures in Korea, including the Jeonse leasehold system, ought to be better reflected and the related risks be assessed, amid a slowing economy due to adverse demographic developments.

Table 1.

Korea: Main Recommendations

article image
Source: IMF staff. * I = immediate, ST = short-term, MT = medium-term.

Macro-Financial Environ Structure, and Scope of the Systemic Risk Analysis

A. Financial System Structure and Trends

1. As of 2018Q4, total assets of financial institutions in Korea reached about 300 percent of GDP. Since the previous 2013 FSAP, Korea’s financial system has grown by about 40 percentage points of GDP. Real estate is the central asset class where leverage is high. The asset management industry and second-pillar pension funds are less developed than other segments of the financial system but are becoming increasingly important amid the material demographic transition observed and further expected in Korea. The rapid growth of Fintech is also adding to such changing dynamics. Growth in the onshore FX derivatives market lags cross-border investment flows while similarly with other emerging currency markets, the offshore NDF market in Korean won remains large and growing, most likely reflecting electronification in NDF markets and global legal and regulatory reforms for derivatives markets which incentivized greater central clearing of NDFs. Selected indicators of the Korean financial system structure are presented in Figure 1.

Figure 1.
Figure 1.

Korea: Financial System Structure

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Bloomberg, IMF Financial Development Index, IMF World Economic Outlook, FSB Global Monitoring Report on Nonbank Financial Intermediation 2018, FSS.1/ For more details about the financial development index and its financial institutions and markets subcomponents see IMF SDN/15/08 and IMF WP/16/5.

Financial Institutions’ Assets

(in percent of total)

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: FSS/FISIS and IMF staff calculations.

2. The banking sector comprises a vast spectrum of entities. It is composed of nation-wide and regional commercial banks, specialized (policy) banks and Other Depository Institutions (ODIs); see Table 2. The state-owned financial firms are a prominent part of this landscape. The KHFC provides mortgages insurance and issues fully guaranteed mortgage-backed securities composed of “conforming” loans2 for which it sets the maximum amount, maturity and interest rates in advance.3 KAMCO has played an instrumental role after the Asian crisis when it acquired about USD 100bn of bad loans from the banking system and has also bought substantial amounts of NPLs in the wake of the credit card crisis (2003), global financial crisis (2008), and household debt crisis (2013). The National Pension Service (NPS) is the third largest in the world with USD 600bn in assets and is the largest investor in Korea. The Korean Investment Corporation (KIC) is a sovereign wealth fund with about USD 130bn assets under management.

Table 2.

Korea: Financial System Structure

article image
Sources: FSS/FISIS; IMF, World Economic Outlook database; and IMF staff calculations.Note: 2018 as of 2018Q3. Excludes community credit cooperatives.

3. Financial holding companies (FHCs) play a systemically important role. As of end-2018, nine FHCs held about 40 percent of total financial institutions’ assets (114 percent of GDP) through complex networks of subsidiaries with operations across all segments of the financial system; four FHCs have been identified as D-SIBs. The financial firms belonging to the bank holding companies operate across a broad cross-section of the financial sector, with subsidiaries being engaged in insurance, capital markets and asset management business.

4. The business models of Korea’s banks are broadly conventional (Figure 2). Banks, including ODIs, are primarily funded by retail deposits. Assets are concentrated in loans, mostly related to real estate. Bank loans are split about equally across lending to households and firms, with 80 percent of the stock of corporate bank loans outstanding to SMEs. Most SME loans are collateralized against real estate. Commercial banks have a relatively diversified loan portfolio and securities’ holdings. State-owned banks are focused on lending to SMEs, which is reflective of Korea’s economic policy priorities.

Figure 2.
Figure 2.

Korea: Depository Institutions’ Business Models

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Source: FSS/FISIS and IMF staff calculations.

5. The insurance sector is large, highly concentrated, and saturated (Figure 3). The sector comprises 24 life insurance firms and 30 non-life insurers, managing KRW 1170 trillion in total. Insurance penetration (premiums to GDP) is one of the highest in the world, exceeded only by Taiwan Province of China and Hong Kong SAR. This reflects, in part, the central role that life insurance has played as a conduit of savings in Korea, where life insurance assets account for more than 20 percent of household financial assets, well above the OECD average. Cross-sectoral linkages exist as insurers typically also offer financial consulting services and are involved in asset management. The bancassurance market is well developed and 50 percent of new business in life is sold via banks. The market is dominated by large firms owned by FHCs and Korea’s largest conglomerates (chaebols).

Figure 3.
Figure 3.

Korea: Structure of the Insurance Sector

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: FSS, IMF World Economic Outlook Database, and IMF staff calculations.

6. The Korean capital market is one of the most active markets in Asia. The equity market has a market capitalization of around USD 1.8 trillion with foreigners holding around 35 percent of the listed Korean stocks. The largest five conglomerates account for over 50 percent of market capitalization. Compared with major advanced and emerging market economies, the price-to-book and price-to-earnings ratios of Korean companies are notably lower, which has been coined the “Korea discount” and attributed to various factors, such as in relation to North Korea, and the complicated ownership and governance structure of many Korean corporations and especially large firm conglomerates.

7. The Korean bond market is dominated by government and other public debt, while ‘green’ or ESG bonds’ importance is rising. Korea’s asset management industry has experienced robust growth over the past few years reflecting changing saving patterns but also a search for yield by households and other investors. Assets under the management of privately placed funds, derivatives-linked products such as equity-linked securities, etc., all grew at double digit rates over the past year and now amount to about KRW 500 trillion; roughly 30 percent of GDP.

8. Fintech innovations are rapidly taking root in the retail payments service sector and altering the market structure. The ongoing push to foster Fintech innovation by the Government is facilitating this. Defining features of the Korean retail banking sector include great financial institution depth and a saturated credit card market, high smartphone penetration, and customer eagerness to embrace cutting-edge technologies—providing a fertile environment for mobile payment services to thrive. The entry of large technology companies in the payment services market (as providers of stored value payment products and of payments-initiation interfaces) is beginning to alter the market structure. These developments affect the degree of concentration and competition, catalyzed by the authorities’ work plan to facilitate innovation, specifically the introduction of an “Open Banking” system, in parallel with loosening of legal restrictions for electronic financial transactions and use of personal data.

9. Korean banks have improved their capital position and asset quality since the 2013 FSAP (Table 3, Figures 4 and 5). Banks’ aggregate capital ratio at about 16 percent of risk-weighted assets does not stand out compared to other banks in the region. NPL ratios are structurally low but must be interpreted cautiously because banks sell NPLs swiftly into an active market for distressed assets4. Korean banks underperform their regional peers in terms of profitability. The share of foreign business in total assets remains relatively small at about 6 percent, up from 4 percent in 20135. Nation-wide banks’ asset exposures are diversified, regional banks’ and mutual savings banks’ exposures are concentrated in SME lending, and ODIs’ lending is concentrated in consumer credit (Figure 6).

Table 3.

Korea: Core Financial Soundness Indicators, 2012–17

article image
Sources: IMF Financial Soundness Indicators.
Figure 4.
Figure 4.

Korea: Financial System Performance

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: FSS, Haver, and IMF staff calculations.
Figure 5.
Figure 5.

Korea: Bank Profitability (Significant Institutions):1 Key Trends

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Haver, FISIS, IFS.1 Includes commercial and specialized banks that are currently active.2 Includes internet-only banks.
Figure 6.
Figure 6.

Korean Banks’ Asset Portfolio Structure, Liability Structure, and P&L Components

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Supervisory data from the FSS, publicly available data for banks, and IMF staff calculations.Notes: Based on data for underlying 8 nation-wide banks, 5 regional banks, 6 specialized banks, and 2,316 ODIs (888 Credit Unions, 79 Mutual Savings Banks, 1,122 Agricultural Cooperatives, 90 Fisheries Cooperatives, 137 Forestry Cooperatives).

10. Insurers are currently well capitalized but low interest rates are weighing on profitability (Figure 7). The median life insurers recorded a coverage of its risk-based capital (RBC) at end-2018 of 239 percent, well above the regulatory minimum of 100 percent and the recommended level of 150 percent. In the non-life sector, the RBC coverage was even higher, at 263 which is the highest level since 2012. Increasing difficulties to match higher interest rate guarantees with dwindling returns have led most insurers to extend asset durations (such as through foreign investments) and to offer more ‘protection’ instead of savings products. Still, profitability is low; the sector’s return on assets stood at 0.3 percent in 2018, reflecting decreasing investment yields which declined continuously from 5.1 percent in 2012 to 3.4 percent in 2018. With unfavorable demographics and competition from other financial institutions, growth prospects are limited, triggering foreign expansion and cooperation with fintech providers. Non-life companies, too, have material exposure to long-term saving and protection business. Profitability is impacted by underwriting losses, especially in health insurance where the government aims for lower cost of private coverage. For the non-life sector, the combined ratio hovered slightly above 100 percent in recent years, indicating that claims and operating costs exceed premium income. Nevertheless, the return on assets is still higher than in the life sector, with 1.3 percent for the median company in 2018.

Figure 7.
Figure 7.

Korea: Solvency and Profitability of the Insurance Sector

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Source: IMF staff calculations based on FSS data.

11. While well capitalized under the current regime, additional capital needs are expected under the new regime, which requires close supervisory attention. The new solvency regime, which is going to be aligned with IFRS 17, will introduce a market-consistent valuation of assets and liabilities by 2022. An initial quantitative impact study in 2018, based on the IAIS’ Insurance Capital Standards and Solvency II, showed a notable capital shortfall in the life insurance sector. A more recent study in 2019 implies substantially smaller capital needs as several parameters were adjusted to local market conditions6. The transition to the new regime has already triggered an increased issuance of subordinated debt. An increased re-classification of fixed-income assets from available-for-sale to held-to-maturity, observed recently, would need to be reverted when transitioning to a fully market-consistent asset valuation.

B. Macro-Financial Environment

12. Macro-financial stability in Korea remains vulnerable to external shocks. Korea’s economy is very open and closely integrated with international supply chains. The export-oriented manufacturing sector, accounting for about a quarter of GDP, comprises Korea’s steel, telecommunications equipment, electronics, auto, and shipping industries. Production is concentrated, with the top five conglomerates dominating the Korean stock index.

13. The macroeconomy has performed well since the last FSAP but faces headwinds. GDP growth was robust at around 3 percent in 2017 and 2018, reflecting strong export growth and business and construction investment. The semiconductor industry, a stable driver of South Korea’s growth, has experienced a cyclical slowdown. Growth is expected to remain sluggish (Table 4) due to ongoing trade tensions and low business confidence, notwithstanding a boost from monetary and fiscal policy easing. Cross-border financial flows have been volatile, while the Korean won has depreciated partly on the back of trade tensions between the United States and China.

Table 4.

Korea: Selected Economic Indicators, 2017–24

article image
Sources: Korean authorities; and IMF staff estimates and projections.

Contribution to GDP growth.

Excludes gold.

Debt service on medium- and long-term debt in percent of exports of goods and services.

14. The financial cycle has reached an advanced phase with household debt among the highest for OECD countries (Figure 8). Financial conditions in mid-2019 were close to historical average after several years in easy territory, while the recent monetary policy loosening should counter the impact of rising market risks somewhat. The ratio of total non-financial private sector debt to GDP has reached an elevated level—standing close to 200 percent—and core debt (debt of the non-financial sector owed to banks) amounts to about 130 percent of GDP, which is high in international comparison. Although the trend has been upward, an accelerating trend is not evident. The credit-to-GDP gap is close to zero. After some deleveraging, corporate credit growth has picked up again, particularly to the SME sector driven by sole proprietors and often secured by lending against real estate.

Figure 8.
Figure 8.

Korea: Macrofinancial Environment

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Bloomberg, IIF, BIS, Haver, IMF staff calculations.

15. The upward trend in house prices has moderated, but household leverage in real estate remains high (Figure 9). Housing supply has expanded over recent years and large sections of the financial sector are exposed to the housing market. Household lending growth, primarily related to housing, has slowed but remains above nominal GDP growth while household debt as a ratio of disposable income stands at about 180 percent, which is among the highest across the OECD countries.

Figure 9.
Figure 9.

Korea: Housing Market

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Haver, KOSIS, OECD, CEIC, and IMF staff calculations.Notes: The disposable income series in the denominator of the median house price and apartment price to income ratios in sub-figure 3 are trailing 4-quarter sums of quarterly disposable income flows.

16. Demographic shifts are posing a long-term challenge for the financial sector (Figure 10). By 2050 almost 40 percent of the population is expected to be older than 65, up from 13 percent now, while the working-age population will shrink significantly given Korea’s low birth rate. One implication is that the proportion of debt held by older households will increase, also reflecting reverse mortgages promoted by the KFHC, and the DSTI may rise significantly. Adverse demographic developments are having an impact on capital flows as a rising share of growing retirement savings is invested abroad by pension funds, insurance firms, and asset management companies. This demographic shift, combined with competitive pressures from China, has raised concerns that Korea might be destined for a prolonged period of low growth and inflation with an erosion of its financial buffers. Structural changes and long-term implications that these demographic shifts are expected to exert on banks, insurers and other economic agents will be discussed later in this note.

Figure 10.
Figure 10.

Korea: Demographics Shift and Household Debt

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Haver, KOSIS, and IMF staff calculations.

17. The rapid rise in COVID-19 virus cases in South Korea since February 2020 likely implies a significant drag on economic activity, which is being tempered by a proactive policy response of Korea’s government. The Korean authorities have taken bold steps to contain the COVID-19 outbreak and mitigate its impact on public health and the economy, including through large-scale testing of the population to rapidly identify, isolate, and treat infected patients. The Korean government is also using its fiscal space to mitigate the macroeconomic impact of the outbreak, including through a proposed KRW 11.7trn (0.6 percent of GDP) supplementary budget that provides resources to step up disease control efforts, support small merchants and SMEs, provide transfers to sustain consumption and employment, and support local economies hit hardest by the spread of COVID-197.

C. Korean Housing Market Structure and Dynamics

18. House prices appear to have developed in line with fundamentals at the national level, but regional heterogeneity can be observed (Figure 9). Korean nominal house price growth has moderated since the GFC. Nation-wide median house price and apartment price to disposable income ratios have trended upward to an extent since 2014 (sub-figure 3 in Figure 9). Debt-to-income ratios have been rising steadily since 2010 for younger age cohorts (up to age 50), and also for people at 60 and above. A model-based valuation, based on regional data, points to overvaluations in the Seoul market and the larger capital region of 10 percent and 5 percent respectively. More rural areas have seen house prices increase less than fundamentals in relation to income and mortgage interest rates would have predicted8.

19. The leasehold deposit market (Jeonse) is a unique feature of the Korean financial system and implies potential vulnerabilities for the Korean real estate market. Jeonse tenants give landlords a deposit equivalent to on average about 50–70 percent of the house price that is to be paid back to the tenant at the end of the two-year contract, if the lease contract is not extended9. To finance these deposits, tenants often turn to banks for Jeonse loans which doubled to KRW 72.2trn since 2014, while loan-to-deposit ratios have increased to 42 percent in 2018. In lieu of rental income, Jeonse landlords then rely on house price increases and return on any financial investments made using the Jeonse deposit. Given that landlords do not accrue a yield on Jeonse properties, in the cases where the Jeonse deposit is used to fund the property purchase, the total return on such an investment is entirely dependent on house price increases. Consequently, the investment in Jeonse properties is not sustainable in an environment in which house prices would fall over an extended period.

20. The rollover risk of Jeonse contracts could amplify negative shocks to house prices and the prices of financial assets. In a severe adverse scenario in which Jeonse deposit prices would fall, landlords may have to cover the resulting capital shortfall using their own funds. If this is done by selling financial assets, this would put pressure on asset prices. In an extreme case, landlords may need to sell their properties, which would further amplify the fall in house prices. An assessment conducted by the BOK shows that following a 20 percent fall in the Jeonse deposit price, about 78 percent of landlords could cover total leasehold deposits with their own financial assets, while a remaining 22 percent would need an additional loan10.

21. The authorities need to be vigilant on the following aspects of the Jeonse market structure: (i) who the ultimate holder of risk is and potential claims on the public exchequer; (ii) a rise in rollover risk of Jeonse contracts; and (iii) negative shocks to house prices and the economy more broadly could be amplified through adverse conditions in the Jeonse market. The potential vulnerability associated with household leverage with Jeonse leasing of their properties, possibly in addition to bank debt, could be mitigated by a policy tool that can consider and capture various factors, including the combined value of mortgage and Jeonse deposits, dynamics between house prices and Jeonse prices, and Jeonse loans held by tenants.

22. The public sector has a large footprint in the housing market. The Korean Housing-Finance Corporation (HF) offers a number of products with both a social function and a risk-sharing objective. The HF engages in direct funding of mortgages with funding from the Ministry of Land, Infrastructure and Transportation. It also offers subsidized mortgage loans to lower income households via banks, which are then securitized by the HF11. The HF securitizes non-subsidized long-term mortgages and total MBSs issued and guaranteed by the HF amounts to 116tr KRW, a large portion of which are held by insurance companies. The HF also guarantees other housing-related loans, including loans using Jeonse deposits as collateral and loans by landlords needed to cover falls in Jeonse deposit prices12. The HF offers reverse mortgages to the rapidly growing group of low-income pensioners. There is lack of transparency regarding the public sector’s total contingent liabilities related to the housing market and whether the capital buffers held are adequate. While housing market subsidies and guarantees may have a social function, it is not clear to what extent these distort the market price of risk.

D. Scope of the Systemic Risk Analysis

23. The FSAP risk analysis covers the banking and insurance sector. The analysis as presented in this note was based on public and supervisory data for 24 banks, covering 95 percent of banking system assets at end-2018. Five banks of the 24 are sub-aggregates, forming the ODI category (Credit Unions, Mutual Savings Banks, credit cooperative banks). Table 5 lists the banks in-scope of the banking system stress test, including information about their D-SIB status.

Table 5.

Korean Banks In-Scope of the Systemic Risk Analysis and Stress Test

article image
Source: FSS and IMF staff.Note: See Glossary for abbreviations used in this table. The last column (loan to deposit reg.) denotes whether the loan to deposit ratio-based regulatory constraints are relevant or not for a given bank (see FSS Handbook for details about the regulation). The FSS/STARS and BOK/SAMP models are the FSS’ and BOK’s top-down stress test model frameworks. All information provided in the table refers to end-2018.

24. Seven life and six non-life insurers were included in the insurance sector stress test. In terms of end-2018 balance sheet assets, the analysis covers 73 percent and 76 percent of the life and non-life sector, respectively. Table 6 presents the main characteristics of the insurance undertakings included in the analysis.

Table 6.

Korea: Insurance Firms Included in the Stress Test

article image
Source: FSS and company data, and IMF staff calculations.

Risks, Vulnerabilities, and Macro-Financial Scenarios

A. Risks, Vulnerabilities, and Macro-Financial Scenarios13

25. The risks to consider for the Korean economy stem primarily from escalating trade tensions. Adverse macro-financial developments may be triggered by escalating global trade tensions, involving the US, China, the EU, and Japan. Geopolitical risks could, moreover, unfold despite North Korea’s stated commitment to denuclearization, with the related diplomatic process appearing to be repeatedly prone to breakdown; albeit the news related to North Korea have recently abated.

26. The structural macro-financial conditions against which such broad-based risks are judged to have an impact upon materialization include Korea’s export orientation and notable Fintech developments. Relevant structural features include Korea’s strong reliance on macroeconomic terms is to be material for the Korean economy, for the risk (trigger) to be considered as a basis for a scenario narrative. The probability (of a trigger) will remain subjective.

27. The adverse macro-financial scenario narrative is rooted in the escalation of global trade tensions. Upon it being triggered, the scenario is assumed to entail a broad-based, worldwide sell-off in equity markets, reflecting a general fall in investors’ risk appetite. Korea would face significant capital outflows, coupled with strong currency depreciation, due to its close ties with China in terms of trade, and hence being close to the epicenter of the escalating trade tensions and very directly hit by a global slow-down in trade through falling export and import flows. 14,15

28. The scenario would imply pressure on export-oriented Korean firms and adverse spillover effects to domestic demand through various channels (higher unemployment, drop in income, depressed confidence, etc.). The activities of large firm conglomerates in Korea would be significantly disrupted due to their cross-border supply-chain interconnectedness16. Firms’ and households’ expectations would turn adversely and imply for investment activity and private consumption to drop in anticipation of a slow-down of activity in the medium-term future, thereby further pressuring real activity in the short-term (expectation channel). Interest rate-based monetary policy would have less room to stimulate demand (through the credit channel) since interest rates are structurally lower at the outset of the scenario horizon, compared to the monetary policy stance ahead of the GFC.

29. Table 7 presents the baseline and adverse scenario calibration for a set of core macro-financial variables17. The baseline scenario is aligned with the IMF’s October 2019 WEO. The adverse scenario was assumed to start unfolding from the first quarter of 2019 onward. The scenario horizon covers the 20-quarter period from 2019Q1–2023Q418. A chart collection corresponding to the variables in Table 7 is presented in Figure 11. The adverse scenario has been calibrated such that the first year’s real GDP growth matches an estimate implied by the Growth-at-Risk (GaR) framework that the IMF FSAP team has implemented for Korea. The GaR estimate corresponding to a 5 percent tail probability equals about -3.1 percent year-on-year for the year 2019. The real GDP shock severity corresponds to an approximate 2.1x standard deviation multiple (2-year cumulative) in historical perspective.

Table 7.

Korea: Baseline and Adverse Macro-Financial Scenario—Main Features

article image
Sources: IMF WEO and IMF staff calculations.Note: URX = unemployment rate, STN = short-term money market rate (call rate), LTN = long-term interest rate (long-term sovereign bond yield), TS = term spread (LTN-STN), ESX = nominal stock prices, USDKRW = 1 USD in KRW (up means depreciation of KRW), RPP = residential property prices. MSCI World is the MSCI World Index. NFC credit = nonfinancial corporate credit stock. HH credit = household credit, including mortgages, Jeonse loans and consumer credit.
Figure 11.
Figure 11.

Korea: Baseline and Adverse Macro-Financial Scenario

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF WEO and IMF staff calculations.Note: The scenario is developed at quarterly frequency. The scenario horizon covers the 20-quarter period from 2019Q1–2023Q4.

30. The IMF WEO as of April 2020 which reflects the expected impact of COVID-19 is less adverse than the adverse scenario profile considered for the FSAP. The April 2020 IMF WEO’s baseline outlook for 2020 expects real GDP in 2020 to grow at -1.2 percent year-on-year, compared to a drop of -3.1 percent in the FSAP’s adverse scenario. The unemployment rate is expected to move to 4.5 percent according to the April 2020 WEO, comparing to 4.8 percent in the FSAP’s adverse scenario. Hence, the judgment is that the FSAP adverse scenario is sufficiently conservative conditional on the current information set as of end-April 2020 to encompass the adverse economic consequences of the COVID-19 pandemic.

31. Measures of inflation (deflator- and consumer price index-based) move up visibly in the first year of the adverse scenario. This reflects the impact of the depreciating currency through the implied rise in the costs of imports. After the value of the currency would be expected to normalize from 2020 onward, the inflation measures level off, moving temporarily into negative territory, before they recover throughout the second half of the scenario horizon.

32. Asset prices in Korea, including for real estate, would drop markedly. Asset prices would drop due to the slowdown in private sector demand for housing, inter alia for the assumed reason that loan interest rates would not fall sufficiently (to stimulate demand) because of banks being under solvency and liquidity stress and thereby implying for their funding costs to stretch to higher levels. Nation-wide nominal house prices would drop by -16 percent in cumulative terms between 2018Q4 and 2021Q2. Equity prices in Korea would drop by -39 percent from 2018Q4 to 2020Q2 (trough), before they start to recover.

33. Figure 12 shows the real GDP flow trajectories under the scenarios, along with their historical evolution during the AFC, the GFC, and the FSAP 2013 adverse scenarios. The baseline chart—involving the estimated pre-crisis trends for the AFC and GFC profiles in the figure— suggests that the FSAP 2019 baseline scenario is the weakest in comparison to the AFC, GFC pre-crisis trends as well as the FSAP 2013’s baseline. The FSAP scenarios—both in 2013 and 2019—imply a more protracted path compared to the shape of the GDP flows as observed during the AFC and GFC (more frontloading and faster recovery). This shape has been considered for the sake of conservatism.

Figure 12.
Figure 12.

Korea: Baseline and Adverse Scenario Paths for Real GDP in Historical Perspective

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF WEO and IMF staff calculations.Note: AFC = Asian Financial Crisis (1997/38). GFC = Global Financial Crisis (2007–09). The origin, “t0”, in the charts denote the years [1997Q4, 2008Q3, 2012Q4, 2018Q4] for the [AFC, GFC, FSAP 2013, FSAP 2019].

34. Figure 13 shows the cumulative real GDP deviations along different horizons, in historical perspective. The FSAP scenario’s GDP losses relative to the baseline are less negative compared to the FSAP 2013 deviations. This deviation profile is judged to be justified against the background of the weaker baseline (resulting in comparable level trajectories under the adverse scenario, see Figure 12) as well as from the perspective of the GaR benchmark estimates that take account of the cyclical position of the Korean economy and its downside risks at a 5 percent probability level.

Figure 13.
Figure 13.

Korea: Cumulative Real GDP Deviations (Adverse Relative to Baseline)

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF WEO and staff calculations.Note: T1:X denotes the sum of quarterly GDP flows (real) over X=[4,8,12] quarters. The deviation measures are computed as the sum of adverse scenario’s GDP flows minus the sum of baseline GDP flows, divided by the sum of baseline GDP flows; along 4, 8, and 12 quarters. For the AFC and GFC, the respective pre-crisis trends (average growth) over the preceding 24 quarters of the two recession episodes were used to project a ‘baseline’ that is assumed to have ensued if the AFC and GFC would not have happened. For the FSAP 2013 and FSAP 2019 calculations, the respective baseline scenarios were involved.

35. Real estate is ‘pervasive’ in the Korean economy, for the financial system’s resilience to real estate price fluctuations therefore deserving emphasis. The assessment about the valuation of housing prices and their endogenous response in the scenario is surrounded by significant uncertainty, hence implying that sensitivity analyses around the adverse scenario with a special focus on house price shocks will be considered and presented later in this note.

B. Fintech Developments in Korea

36. The possible implications of Fintech developments as observed in Korea can be seen in three dimensions. Both the Open Banking initiative and the e-money service provider developments imply changing liquidity dynamics in the retail payments and money market in Korea19. These have implications for solvency of financial institutions and system-wide operational risks (Figure 14).

Figure 14.
Figure 14.

Korea: Implications of Open Banking and E-Money Developments in Korea

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF staff.

37. Notwithstanding its various expected positive implications, the Open Banking initiative could unfold its impact through rising competition and imply potential risks; in three dimensions. First, bank retail customers would be more inclined to swiftly transfer deposits across banks to where deposit rates are most competitive; using mobile phone applications that conveniently display them the pricing conditions across banks and allowing the transfer in the same place. This in turn may pressure banks’ interest expense. Second, bank customers may learn to more actively seek the most competitive offers in terms of loan interest rates, as a result of being more open-minded to where to deposit their funds, which would pressure banks’ interest income; third, rising financial literacy as a result of significant Fintech developments such as through Open Banking may enhance customers’ willingness and ability to behave as outlined above.

38. The result of such changing customer behavior may be that banks with currently lower retail deposit rates may raise them, and those with currently higher loan interest rates may lower them. Such behavior is captured in the Fintech Overlay analysis which will be presented in the solvency analysis section of this note.

39. The growing popularity of e-money providers could in the future exacerbate such rising competition through at least two channels. First, they may compete with commercial bank deposits by offering remuneration resembling interest to attract customers20, enabled by the Korean authorities’ gradual de-regulation initiative; second, banks might offer more sizable deposit rates to the e-money providers to attract their—possibly sizable in the future—re-deposited customer funds. Both aspects could imply further upward pressure on banks’ interest expense.

40. Open Banking and the growing popularity of e-money providers can in the future imply risks in terms of higher-frequency liquidity dynamics. First, the Open Banking system would imply that “stickiness” of bank deposits will be reduced. Bank customers could more likely react to ‘bad news’—whether substantiated or not—by swiftly and sizably moving funds to other banks. This would represent an ‘electronic deposit run’ and could exacerbate a liquidity squeeze for banks beyond what would be justified by banks’ fundamental solvency risk perception by the market21. Second, with e-money providers re-depositing customer funds in one or at most a few banks can imply concentration risk, i) from a system-wide operational risk viewpoint, and ii) in terms of a structural shortage of liquid funds (alongside reserves) for those banks with whom the service providers do not hold their funds.

41. Systemic liquidity risks may arise from e-money providers if their activities are not subject to adequate and formal regulation. Certain baseline requirements for e-money providers have been stipulated by EFTA (e.g., obligations to redeem the e-money they issue and to maintain dedicated accounts segregated for each line of business), but redemption risks may arise. E-money providers mitigate this risk in current practice by holding funds underlying their e-money on a 1:1 basis. Given the growing popularity of e-money, risk controls and mitigants would be strengthened by enshrining in law or regulation clear requirements as to a service providers’ treatment of funds underlying their e-money, e.g., to avoid that e-money providers would in the future deviate from their 1:1 holding practice.

42. The FSAP recommends that the Korean oversight institutions conduct a detailed Fintech and competition-related impact assessment for the banking sector and other components of the financial system. As argued above, such an assessment would not only consider medium-term solvency implications through changing competition but focus as well (primarily) on systemic liquidity and system-wide operational risk implications (Figure 14).

Forward-Looking Bank Solvency Analysis

A. Methodology

43. The schematic in Figure 15 illustrates selected elements of the modular Solvency Analysis Tool Suite for banks. It consists of several connected modules that have been set up for the Korea FSAP. The modules’ inputs, functioning and outputs are described in this section.

Figure 15.
Figure 15.

Korea: Dependencies Captured in Top-Down Solvency Model Suite for Korean Banks

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF staff.

44. Four defining features of the tool suite deserve emphasis. They include i) an account for IFRS 9 loan loss provision principles; ii) a nonlinear solvency-funding cost feedback mechanism; iii) dynamic balance sheets; and iv) a “Fintech Overlay”. Figure 15 indicates which elements the Fintech Overlay relates to from a methodological perspective. Details will be explained later in this section. Figure 16 shows the P&L structure adopted for the Korea bank solvency analysis along with a summary of the model approach adopted for each line.

Figure 16.
Figure 16.

Korea: P&L Structure and Model Approaches Adopted for the Korea Solvency Analysis

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF staff.

Credit Risk (CR) Module

45. Figure 17 shows the structure of the CR module22. It summarizes the “hybrid” PD-satellite-Z-factor methodology developed for Korea, related to item 2 in Figure 16. Appendix IV shows a table which maps the line items from the banks’ loan and trading books to their treatment under either the credit risk or market risk methodology.

Figure 17.
Figure 17.

Korea: Credit Risk Module Structure

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF staff. see Glossary for abbreviations.

46. Default rate satellite models were estimated based on confidential bank data provided by the FSS (Figure 17, Block A). They were structured and estimated based on a Bayesian Model Averaging (BMA) Methodology subject to sign constraints on the long-run multipliers of the equations23. The default rate models were developed for four portfolio segments: large nonfinancial corporates, SMEs, household mortgage loans (including Jeonse leasehold loans), and consumer credit. The default rate models were used to derive forecasts conditional on the baseline and the adverse scenario for Korea (Block B in Figure 17)24,25. The PDs at the outset of the scenario horizon and the maxima under the adverse scenario are shown in Figure 18. The model structure is sketched in Appendix V. Household mortgage portfolio PDs were found to be very reactive to the adverse scenario in terms of multiples to their starting points; the latter being very small in comparison to the other portfolios’ PDs as of 2018. Their levels under the adverse scenario were deemed sufficiently conservative in historical perspective.

Figure 18.
Figure 18.

Korea: Baseline and Adverse Scenario-Conditional PDs

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Korean banks’ and FSS data and IMF staff calculations. Annual horizon PDs. The NFC category comprises both large nonfinancial corporate exposures as well as SME exposures. The upper and lower ends of the boxes denote the 90th and 10th percentiles of the underlying distribution across banks. The error bars extending to the up- and downside denote the maxima and minima. White lines denote the median. Under the baseline (adverse) scenario, the distribution reflects the average (maximum) of the PDs along the 20-quarter scenario horizon.

47. Historical transition data in accordance with IFRS 9 classification criteria have been compiled by the FSS from the Korean banks, under FSAP guidance. The Korean regulatory asset classification scheme was mapped into the Stage 1–2-3 classification under IFRS 9: the prudential categories “normal” and “precautionary” were mapped to the accounting Stages 1 and 2, respectively. The prudential asset classes “substandard”, “doubtful” and “loss” were jointly mapped to Stage 3. The historical flows across the resulting proxy accounting stages were sourced from the banks, including the additional flow categories for new business, repayment, write-offs and asset sales. They were collected at bank-level for three portfolio segments: nonfinancial corporate (large firms and SMEs combined), household mortgages, and consumer credit. Historical transition flow data with quarterly frequency were available from 2010Q1–2018Q4 for the majority of banks and portfolios.

48. Historical Z-factors and additional related parameters were estimated based on the historical bank-portfolio-specific transition flow data (Figure 17, Block C). The Z-factor methodology is summarized in Box 1.

The Z-Score Methodology

The Z-score methodology (Belkin et al., 1998) aims to reduce the information contained in a time series of transition matrices down to one number per point in time. It was originally developed for rating transition matrices; involving letter-based grids of ratings spanning up to 10–15 classes (AAA-D). The methodology is applicable to matrices of any size and irrespective of the criteria set behind the classes in the matrix, and hence applicable to a time series of IFRS 9 transition matrices. Figure B1.1 combines Figure 1 from Belkin et al. (1998) (on the left, for an initial BBB exposure) with a modified version thereof that is based on an IFRS 9 staging structure (on the right, for an initial S2 exposure in this example). Both refer to a long-term average transition matrix.

Figure B1.1:
Figure B1.1:

Probability Density for BBB Rating (Left) vs. S2 Exposure Under IFRS 9 (Right)

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

The Z-score method builds on the assumption that the probability density X is a function of an idiosyncratic driver Y and a systematic economy-wide driver Z. Both are independent unit normal random variables by assumption. The parameter ρ captures the correlation between Z and X, with Z explaining a fraction ρ of the variance of X.

(1)Xt=1ρYt+ρZt

The fitted transition probabilities, ∆t, are expressed as:

(2)Δt(xg+1G,xgG,Zt,ρ)=Φ(xg+1GρZt1ρ)Φ(xgGρZt1ρ)

Φ is a standard normal cumulative distribution function. The terms xgG are the “bin boundaries” (referring to the vertical lines in Figure B1.1) which are computed based on an inverse of a standard normal cumulative distribution function, with reference to a long-term average transition matrix. The historical deviation between an observed and fitted transition matrix can be computed as:

(3)minZtΣGΣgwtg(Pt(G,g)obs,transitionratesΔ(xg+1G,xgG,Zt,ρ)fittedtransitionrates)2

where the two sums in this equation indicate a summation over all elements in a transition matrix. Conditional on a ρ and the bin boundaries from the long-term average transition matrix, a Zt can be found for each point in time that minimizes eq. (3). The constant ρ and the time series Zt were estimated jointly subject to the constraint that the variance of the resulting Zt be equal one, for each bank-portfolio. A +1/-1 value for Z denotes a “1-standard deviation from normal (long-run average)” conditions. The estimated parameters ρ (bank-portfolio specific) for the Korean banks are depicted in Figure B1.2.

Figure B1.2:
Figure B1.2:

Estimated ρ for Korean Banks’ Loan Portfolios

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Note: The upper and lower ends of the boxes denote the 90th and 10th percentiles of the underlying distribution of estimated ρ across banks. The error bars extending to the up- and downside denote the maxima and minima. White lines denote the median. NFC abbreviates Nonfinancial corporate.

49. A “hybrid” PD-satellite-Z-factor methodology has been developed for Korea (Figure 17, Blocks A-G). In principle, one could use the estimated historical Z-factor times series at portfolio level—and either at bank- or aggregated banking-system level—to develop satellite models, to derive scenario-conditional paths of all Z’s into the future, and in this case not requiring the separate estimation of default rate satellite models. Default rates would be part of the transition matrix and implied by the projected Z factor paths26. This approach was not chosen for Korea because the historical transition matrix data—albeit rich from a cross-bank-portfolio perspective— was still deemed too short from a time series perspective, not capturing a full economic cycle (as starting in 2010). Default rates were longer, starting at the end of the 1990s for most portfolios. The “hybrid” methodology is summarized in Box 2.

“Hybrid” PD-Satellite-Z-factor Methodology and Dynamic Balance Sheets

Figure B2.1 depicts how the previous step (Box 1, first row in Figure B2.1) compares to the translation of a Z-score path back to a transition matrix (this box, second row in Figure B2.1).

Figure B2.1:
Figure B2.1:

From Transition Matrices to the Z-Score and Vice-versa

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

To accomplish what is indicated in the second row of this figure, a transition matrix forecast can be obtained from a Z-score forecast path in the same way as historical fit is produced at the Z-estimation stage (Box 1). The formula required to that end has the same structure as eq. (2) in Box 1:

(4)Δt(xg+1G,xgG,Zt,ρ)=Φ(xg+1GρZt˜1ρ)Φ(xgGρZt˜1ρ)

The parameter ρ and the bin boundaries are given at this stage (previously estimated, Box 1) and do not change. Only Z is an input that varies conditional on different scenarios, now denoted as Z˜t implying the transition probabilities across the transition matrix and along the scenario horizon.

The formulas that were used to imply the S1, S2, and S3 stocks are the following:

(5)S2t=S2t1+TRt12S1t1+TRt32S3t1InflowstoS2TRt12S2t1+TRt23S2t1MtS2S2t1OutflowsawayfromS2S3t=S3t1+TRt13S1t1+TRt23S2t1InflowstoS3TRt31S3t1+TRt32S3t1WROtS3t1OutflowsawayfromS3St=S1t+S2t+S3t=(1+gt)St1S1t=max(0,StS2tS3t)

The portfolio-specific gross loan growth is under explicit control (gt), pre-determined by the scenario. The repayment percentage for S2 exposures as well as the write-off/asset-sale parameter (MtS2 and WROt) were held constant at observed end-sample position for all portfolios and banks.

Based on the above set of equations, the “hybrid Z-score-PD-Satellite” methodology was designed as follows: the bank-portfolio-specific Z path for a given bank-portfolio along the 20-quarter forward horizon was set such that the exposure-weighted PD would match a path implied by the PD satellite models. The PD paths from the PD satellite models were attached to bank-portfolio-specific PD starting points using a distance-to-default transformation. The exposure-weighted PDs based on the transition matrices were computed, period by period, as follows:

(6)PDt+1|tPiT=S1t×TR13t+1|t+S2t×TR23t+1|tS1t+S2t

50. A “perfect foresight” assumption was employed (Figure 17, Block E). “Perfect foresight” means that the multiple probability-weighted scenario requirement as stipulated under IFRS 9 for accounting provision purposes is ignored for what concerns the solvency stress test analysis. Two concrete scenarios are instead considered, with 100 percent weight set for each of them separately (baseline, adverse).

51. ECL provisioning under IFRS 9 requires forming an expectation as to how the underlying risk parameters behave until the end of the lifetime of the relevant financial assets. This horizon extends beyond the 5-year horizon of the FSAP scenarios for portfolios with residual maturities larger than 5 years; that is, in particular for mortgage portfolios. The baseline risk parameters (PDs, LGDs) as of the last quarter of year 5 (2023Q4) were held constant; the adverse scenario’s parameters at 2023Q4 were assumed to decay back to baseline over an 8-year period (Figure 19)27.

Figure 19.
Figure 19.

Korea: Perfect Foresight and Risk Parameter Behavior after Initial 5-Year Horizon

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF staff. X is 5 years. Z is 8 years.

52. The solvency analysis was conducted based on a dynamic bank balance sheet assumption (Figure 17 Blocks E/F and Box 3). Gross credit growth paths for the banks’ corporate and household loan portfolios were set as part of the scenario design process, consistent with the assumed macro-financial developments in the baseline and adverse scenario, at economy-wide level. The portfolio-specific growth paths were assumed to be equal across banks. Write-offs and asset sales were allowed to be positive, which is important in Korea since banks make active use of NPL management firms for them to take care of the workout, seizure of collateral, etc. The relevant equations to account for write-offs and asset sales were shown in Box 2 in relation to the hybrid Z-factor methodologies.

53. A simple structural LGD model was employed to link housing-collateralized portfolios’ LGDs to the house price paths in the scenarios (Figure 17, Block H). The structural LGD model captures the dependence of the LGDs on the underlying housing collateral value28,29. For all other portfolios, LGDs were kept constant in the baseline and scaled by a factor of 1.15 under the adverse scenario. On top, for all portfolios in-scope of the CR Module, a conservative overlay has been applied, by taking the maximum of the model-implied (structural model or simple factor) LGDs and the bank-reported regulatory downturn LGDs30. Figure 20 shows the resulting distribution of LGDs for different portfolios under the baseline and adverse scenario.

Figure 20.
Figure 20.

Korea: Baseline and Adverse Scenario-Conditional LGDs

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Korean Banks’ and FSS data and IMF staff calculations. The upper and lower ends of the boxes denote the 90th and 10th percentiles of the underlying LGD distribution across banks. The error bars extending to the up- and downside denote the maxima and minima. White lines denote the median.

54. Effective interest rate assumptions are required for discounting purposes when computing ECLs (Figure 17, Block I). The effective interest rates as of end-2018 were sourced from the banks at bank-portfolio level (Figure 21) and held constant for the purpose of discounting (Box 3 documents how).

Figure 21.
Figure 21.

Korea: Effective Interest Rates for ECL Discounting Purposes

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: FSS and IMF staff calculations. The upper and lower ends of the boxes denote the 90th and 10th percentiles of the underlying distribution across banks. The error bars extending to the up- and downside denote the maxima and minima. White lines denote the median.

55. An IFRS 9-compatible ECL scheme was employed (Figure 17, Block J). Box 3 summarizes the methodology for computing 12-month ECLs for Stage 1 exposures, lifetime ECLs for Stage 2 and 3 exposures, as well as for implying the provision stocks and flows.

56. RWs for credit exposures were treated differently for STA and IRB portfolios. RWs for STA exposures were kept constant at their end-2018 levels. RWs for IRB exposures were allowed to move dynamically as a function of the underlying changes in regulatory TTC-PDs. Downturn LGDs were kept constant as deemed sufficiently conservative. The TTC PDs at bank-portfolio level were made a smooth function of PiT-PDs as used for accounting provisioning, using the formula PDt+1|tTTC=φ1(φ(PDt|t1TTC)+α×Δφ(PDt+1|tPiT)). The smoothing parameter α was set to 0.5 for the base set of results. A sensitivity analysis related to α will be presented in the result section. Under both the STA and IRB approach, the RWA evolution in volume-terms also reflects the allowance for dynamic balance sheets31.

Interest Income and Expense, incl. Nonlinear Solvency-funding Cost Feedback, Feedback from Credit Risk to Interest Income, and Fintech Overlay

57. The interest income and expense module—capturing interest rate risk in the banking book—is based on cross-bank panel econometric models. The banks’ historical effective interest income and interest expense rates form the basis for these models (Figure 22). The corresponding panel models were estimated for the groups of nation-wide banks, regional banks, and specialized banks separately. The regional banks’ models were employed for the group of ODIs (credit unions, credit cooperatives and mutual savings banks). The interest income and expense panel model equations are documented in Appendix V.

Figure 22.
Figure 22.

Korea: Banks’ Historical Effective Interest Rate Evolution for Korean Banks

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: IMF staff.Notes: The bars show the distribution of the underlying individual-bank samples at a quarterly frequency since the beginning of the 2000s.

58. The interest expense model contains a feedback link from banks’ solvency position to their funding cost (Appendix V). From an economic perspective, this estimated feedback was deemed material enough to warrant its inclusion in the model for the Korean banks. From a methodological perspective, including that link meant that a sequential, iterative link from the end-period (quarter) capital ratios resulting from the overall stress test model suite to banks’ funding cost in any subsequent quarter was built in.

Lifetime (LT-) ECL and Provision Stock and Flow Calculations

The LT-ECL formula employed for the Korea solvency analysis has been structured as follows:

(6)ECLtLT,S2=Σs=t+1MTRs23,*×LGDs×S2s1(1+r)s

with M denoting the average residual maturity of a portfolio. The lifetime ECL is here measured in monetary units. The right hand-side includes a point-in-time LGD and the relevant exposure, which is the S2 stock. The denominator of the formula involves an effective loan interest rate for discounting the ECL along the lifetime of a loan portfolio. The formula for the incremental PD in the numerator, denoted TRs23,*, is:

(7)TRs23,*=TRs23×Πt=1s1(1TRt23)PiTsurvivalprobabilitycumulativesurvivalprobability

The term TRs23 is the unconditional transition probability for S2 stocks which links to the outcome of the transition matrix forecast path (Box 2). While this unconditional PD (TR2–3) moves over the lifetime of a loan portfolio in an ‘unrestricted’ manner and in relation to macro-financial conditions, the incremental PD measures the probability of default in period s conditional on not having defaulted up to period s-1 and approaches zero over time1. The exposure term in eq. (6), i.e. the S2 stock, is projected for simplicity using a linear principal repayment scheme2.

For S1 exposures under IFRS 9, the provision stocks are to equal the 12-month expected credit loss, i.e.:

(8)PROVt,S1=ECLt,S1=TRt+1|t13×LGDt+H|t×S1t

Any change in the underlying risk parameters implies a provision flow and hence an impact on capital through the P&L The term TRt+1|t13 is the expected default rate for S1 exposures conditional on end of period-t information for the following year. The LGDt+H|t term has a t+H to denote the fact that the LGD is meant to be forward-looking beyond a 1-year horizon if the expected time until collateral can be sold is more than 1 year. For S2, exposures, the lifetime ECL formula becomes relevant (eq. 1), with its provision stock is supposed to equal, that is, PROVt,S2=ECLtLT,S2. For S3 exposures, provision stocks are to cover the portion of the defaulted loan exposures that will likely not be recoverable, that is:

(9)PROVt,S3=ECLt,S3=LGDt+H|t×S3t

The total provision stock is to the sum of the stage-specific provision stocks:

(10)PROVt=PROVt,S1+PROVt,S2+PROVt,S3

The loan loss provision flow is then the change in the stock, adjusted for write-offs and asset sales:

(11)PROVFLOWt=ΔPROVt+WROt×LGDt×S3t1

The adjustment term related to the write-offs in eq. (11) accounts for the fact that exposures that are written off or sold to asset managers, whose provision stock is falling for that reason, and which should be residual net-equity neutral. The way the adjustment is designed involves the assumption that the LGD estimate based on which a provision stock had been set just before the write-off or asset sales equals the realized LGD upon the collateral sale for the exposures that are written off or sold.

1 If the maturity parameter M in eq. (6) would be set to one, then the lifetime ECL would be a 12-month ECL. In this case, the incremental PD would yet equal the unconditional PD according to eq. (7). If one may ignore in this case the discount factor in the denominator of eq. (6), then the ECL formula would attain the standard “PDxLGDxEAD” structure.2 Alternatively, it could be projected into the future based on a nonlinear repayment schedule of fixed or variable rate loans. If a nonlinear repayment schedule was employed, for variable rate loan portfolios specifically, then an expectation about the loan interest rate would have to be considered as well. For portfolios that are “mixed”, i.e. contain fixed and variable loans, two corresponding repayment schedules could be considered which could then be added together for the two portfolio components. Modelling prepayments explicitly is an option as well. All of these model options were not deemed necessary for the Korea FSAP, as they would have added undue complexity that was not warranted given the target level of detail for the analysis.

59. In addition to being included as such, the solvency-funding cost feedback was allowed to be nonlinear. A notable nonlinearity has been found (Figure 23). The shape of the nonlinear relation implies that at low initial capital ratio levels of a bank, an increase in that ratio would compress its funding cost, while at higher initial capital ratios, a further rising capital ratio may increase its funding cost32. Loan interest rates in turn depend on banks’ funding cost, for the solvency feedback to funding costs to effectively also influence lending rates at the bank level (Appendix V) 33.

Figure 23.
Figure 23.

Korea: Nonlinear Solvency-to-Funding Cost Feedback for Korean Banks

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Supervisory data from the FSS, publicly available data for banks, and IMF staff calculations. The estimated feedback from solvency to the banks’ cost of debt is significantly negative on average over the capital ratio range observed for the banks. The light grey (yellow) shaded area indicate the 5th-95th percentile range of the banks’ shareholder equity/TA ratios under the baseline (adverse) scenario across banks and the scenario horizon. Solid lines (dashed lines) depict the mean (10th and 90th percentiles) of the derivative of the cost of debt to the initial capital ratio. The interpretation for example for the regional banks: a 1p.p. shift to the capital ratio starting at 1 percent (4 percent) would induce a -0.25 p.p. (-0.1 p.p.) response of their cost of debt.

60. An additional feedback mechanism from materializing credit risk to interest income is captured. It is the structural link from the CR Module—as a result of which initially performing exposures (Stage 1 and 2) move to nonperforming status (Stage 3)—which implies no interest income by assumption in Stage 3. This is depicted by the connecting line between credit risk and interest income in Figure 15.

61. The interest income and expense models were used to implement a “Fintech Overlay” analysis. The methodology (Box 4) is embedded in the stress test model suite, implying that it can be used in “scenario-conditional” mode, and hence be set as an overlay to both the baseline and the adverse scenario. The Fintech Overlay concerns only one of the three dimensions (Fig. 14)—the medium-term solvency impact on banks—under the assumption of “no business model change”.

Market Risk (MR) Module

62. The methodologies applied to exposures subject to MtM revaluation depend on their exposure type (bonds, equity). For sovereign, nonfinancial and financial bond exposures under the FVOCI and FVPL category, a modified-duration formula was employed to revalue the exposures as a function of their reported residual duration and the relevant bond yield assumptions under the scenarios34.

63. Domestic and foreign equity exposures were revalued using a direct link to the assumed equity index paths under the scenarios. The KOSPI and the MSCI World index were used to revalue Korean banks’ domestic and foreign equity exposures under the FVPL category.

64. Korean banks’ domestic NFC exposures in the FVOCI category were subject to both the MR and CR Module. In line with the accounting standard, the CR revaluation component was reflected in the P&L (tax non-neutral), while the FVOCI differential beyond the CR impact was reflected through the OCI account (tax neutral).

Fintech Overlay – Methodology

The Fintech Overlay methodology is anchored in the bank-level panel econometric models for interest income and expenses (Appendix V) and focused on the medium-term solvency impact of rising competition. The panel models have bank-specific intercepts. These intercepts will be adjusted based on the following methodology.

A Fintech Overlay “strength” parameter is defined as α which can range between zero and one. The share of household loans in total loans of a bank is denoted as hbL; the share of household deposits in total liabilities as hbD. The initial bank-specific intercepts for the interest income rate (IIR) and interest expense rate (IER) models are denoted as cbLandcbD. The cross-bank percentiles over the initial intercepts are defined as:

(1)c^L=perc((cb=1L,...,cb=BL),1α)andc^D=perc((cb=1D,...,cb=BD),α)

where α determines the percentile. The adjustment to the IIR models’ intercepts, denoted as c˜bL, is then done as follows:

(2)c˜bL=min(cbL,c^L(α)+(1hbL)(cbLc^L(α)))

For the IER models, the intercepts are adjusted as:

(3)c˜bD=min(cbD,c^D(α)+(1hbD)(cbDc^D(α)))

The underlying idea is that an adjustment of an intercept in, say, the IIR model, downward to a lower cross-bank percentile would be scaled by the share of household loans in total loans; likewise, the upward adjustment of an expense rate model’s intercept by the share of household deposits in total liabilities. Banks that do not have any household deposits (as for example some selected specialized banks) would not face any change in their funding cost, by assumption. The max and min operators in eqs. (2) and (3) imply that banks whose intercept would already stand below/above the required target intercept would not change.

Two additional assumptions are embedded in the methodology: first, a “no business model change” assumption, as fees and other sources of income are not modified; second, there is no consideration and account of caps on transactions fees which the Open Banking initiative in Korea implies, due to insufficient data in this respect.

The methodology is simplistic, yet useful for the purpose of deriving an approximate impact of heightened competition, in a scenario-conditional mode of the overall integrated stress test model suite. It can be further developed and refined in numerous dimensions; for example, by the Korean oversight institutions.

Regulatory Capital Ratio Thresholds

65. Regulatory capital thresholds are employed for nation-wide, regional, and specialized banks. Inclusive of the CCB (2.5 percent), the threshold for non-DSIBs amounts to 7 percent, 8.5 percent and 10.5 percent for CET1/RWA, Tier1/RWA and total capital (Tier 1 + Tier 2)/RWA ratios, respectively. For DSIBs, these three thresholds are 1 percentage point higher35.

B. Results

Headline Bank Solvency Stress Test Results

66. The Korean banking system appears resilient under the FSAP adverse macro-financial scenario. A few regional, specialized banks and ODIs face the most sizable capital losses across banks. Credit risk losses and diminished net interest income would be the most pronounced source of pressure on banks’ capital ratios (Figure 24). Specialized banks would experience a notable change in their capital ratios in both absolute and relative terms (Figure 25). The cross-bank heterogeneity in terms of capital responses is most visible in the ODI category.

Figure 24.
Figure 24.

Korea: Solvency Stress Test Results: CET1 Ratios, Sub-Sector Aggregates

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Supervisory data from the FSS, publicly available data for banks, and IMF staff calculations.Notes: The forecasts conditional on the baseline scenario are displayed in cumulative terms up until Year 5 (end-2023). The adverse scenario results are reported in cumulative terms up to the low point, as indicated in the title of the Figures (year 1–5 correspond to 2019–23). The capital ratios are defined as CET1/RWA for nation-wide, regional and specialized banks. For ODIs, the ratios are defined as accounting equity (net of loan loss provision stocks) over total assets. OCI carries an asterisk to indicate that its impact is not measured before tax but after tax, despite in the figures being positioned to the left of the tax impact.
Figure 25.
Figure 25.

Korea: Changes in CET1 Capital Ratios from Starting Point to Low Point Under the Adverse Scenario

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Supervisory data from the FSS, publicly available data for banks, and IMF staff calculations.Notes: The upper and lower ends of the boxes denote the 90th and 10th percentiles of the underlying distribution across banks. The error bars extending to the up- and downside denote the maxima and minima. White lines denote the median. “Factors” are defined as the capital ratio at the banks’ respective low points divided by their initial capital ratios at end-2018.

67. The system-wide capital depletion from the starting point amounts to about 22 percent of initial capital levels (Figure 26). This represents about 2.9 percent of GDP as of 2018.36 All banks’ capital ratios would stay above regulatory minima when allowing the consumption of the capital conservation buffer (CCB).37 A subset of the specialized banks would consume between 0.3 and 1.4 percentage points of their CCB under the adverse scenario.

Figure 26.
Figure 26.

Korea: Capital Depletion Under the Adverse Scenario

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Supervisory data from the FSS, publicly available data for banks, and IMF staff calculations.

68. The banks’ asset exposure shares and their portfolio-loss-in-total-loss shares are not fully proportional, i.e. some portfolios are more sensitive under the adverse scenario than others. Consumer credit losses are about aligned with their portfolio shares while nonfinancial corporate loss shares are higher and household mortgage-related loss shares—including from Jeonse loans—lower relative to their portfolio shares (Figure 27). The latter finding reflects the widespread use of mortgage insurance schemes in Korea which protect the banks’ capital and provide relief in terms of risk weight densities for such portfolios under both the baseline and the adverse scenario38.

Figure 27.
Figure 27.
Figure 27.

Korea: Asset Exposure Shares vs. Credit and Market Loss Shares Under the Adverse Scenario

Citation: IMF Staff Country Reports 2020, 279; 10.5089/9781513557052.002.A001

Sources: Supervisory data from the FSS, publicly available data for banks, and IMF staff calculations.Notes: Each dot (24) is a bank (including five ODI aggregates). The “loss share” is based on the maximum cumulative loan loss per bank-portfolio from along the adverse scenario horizon (2019–23). The “exposure share” is measured as of end-2018.

Fintech Overlay

69. The pressure on capital ratios may grow if Fintech-induced competition would intensify. The “Fintech Overlay” has been used to gauge the effect of intensifying competition due to the “Open Banking” initiative as well as Fintech developments in the retail payment sector, suggesting potentially notable impacts on regional banks, ODIs and selected specialized banks. The analysis assumes that Open Banking and e-money may put upward pressure on retail deposit rates, coupled with downward pressure on loan interest rates, and not allowing banks to adjust their business models, e.g. by raising fees. The aim was to thereby examine how Open Banking may impact the solvency conditions of banks under some conservative assumptions (related system-wide liquidity aspects are discussed later in this note).

70. Under the baseline scenario, regional banks’ capital ratios may fall by 0.6 to 1.3 percentage points by the end of the five-year horizon under the Fintech Overlay. For nationwide and specialized banks, the effects are “nonlinear” in the sense that only once the strength of the overlay is set to high levels, the impact on the banks’ capital ratios would become notable. When activating the Fintech Overlay, the system-wide capital depletion would rise to 4 percent, from 3 percent without the overlay. In this case, several regional and specialized banks’ as well as ODIs’ capital ratios would be further stretched. In aggregate, nation-wide, regional, and specialized banks would consume about 0.4, 0.7, and 2.2 percentage points of their CCBs. Figure 28 shows the impact in terms of capital ratios and capital depletion from Year 0. Figure 29 shows the impact on NIMs, which are the main underlying channel through which stronger competition would influence the capital ratios. This effect is not a baseline forecast, as banks will likely change their business model to counteract such possible losses in income.