KDI International Conference on Household Debt from an International Perspective: Issues and Policy Directions Consumer Over-indebtedness and Financial Vulnerability: Evidence from Credit Bureau Data By Young-il Kim (KDI), Hyoung-Chan Kim (NICE), and Joo-Hee Yoo (NICE) 10 July 2015 Kim, Young Il
1. Introduction: Concerns about Household Debt 2. Data and Methods 3. Over-indebtedness and the Likelihood of Default 4. Types of the Over-indebted 5. Stress Test and Debt Vulnerability 6. Concluding Remarks
1. Introduction: Concerns about Household Debt
1. Introduction: Concerns about Household Debt 4 The amount of the household debt (as a percent of GDP) is relatively high compared to other OECD countries. Household Debt as a Percent of GDP 160 (%) 140 120 100 80 OECD 평균 average 약 72% 60 40 20 0 Denmark Netherlands Switzerland Australia Ireland New Zealand United Kingdom Canada Portugal Korea Norway Sweden United States Spain Finland Greece Japan Germany France Luxembourg Belgium Austria Italy Estonia Czech Republic Hungary Slovenia Slovak Republic Turkey Source: OECD
1. Introduction: Concerns about Household Debt 5 The growth rates of household debt have been higher than that of the disposable income. Concerns about sustainability and possible deleveraging process Growth of Household Debt and Disposable Income 12 (year-to-year, %) 10 8 6 4 2 0 2008 2009 2010 2011 2012 2013 Disposable Income Household Debt
1. Introduction: Concerns about Household Debt 6 Composition of Household(HH) debt: banks vs. non-banks The rising share of HH debt from non-bank financial inst. raises concerns about the quality deterioration of HH debt due to its high borrowing costs. Household Debt by Financial Sector: Bank vs. Non-bank 1,200,000 (unit: billion w on) 1,000,000 800,000 600,000 400,000 200,000 0 2003 2005 2007 2009 2011 2013 2015 Bank Non-bank Source: ECOS from BOK.
1. Introduction: Concerns about Household Debt 7 Concerns about the vulnerability of HH debt structure A relatively large share of balloon-payment loans with short maturity - Exposed to refinancing risks at maturity - Refinancing risks may be magnified if financial markets were hit by negative shocks A large share of floating mortgage rates - Exposed to interest rate risks
1. Introduction: Concerns about Household Debt 8 Downside risks to the macro-economy that could trigger widespread household defaults For example, Downside risks to the domestic economy - GDP growth could slowdown. - Housing markets could sluggish. Downside risks at the global level - Interest rate increases in the US - Delayed recovery of the world economy
2. Data and Methods
2. Data and Methods 10 Credit Bureau (CB) data are used for analysis in this study. The CB data cover most individuals participating in credit markets. - Record actual credit activities and transactions in most financial institutions - Have information about individual characteristics, debt contracts, delinquencies, types of lenders, and more Different from household survey data used by most previous studies - Survey data may have some limitations due to errors and psychological biases in survey and sampling. - However, the survey data may have information about households that are not circulated in credit markets.
2. Data and Methods 11 Concept and measures of over-indebtedness The concept of over-indebtedness may be ambiguous. - Different studies have used over-indebtedness in different meanings; hence, suggesting different measures. The concept of over-indebtedness is related to the likelihood of default in this study. - Indicators of over-indebtedness are evaluated for their ability to predict defaults within the following one year. - Common indicators (credit score, DSR, DSR of unsecured debt holder, LTI, number of credit commitments) are assessed for their performance as predictors of near-term defaults.
2. Data and Methods 12 About 70% of borrowers are rated above the 5 th credit score. Distribution of Borrowers by Credit Score
2. Data and Methods 13 About 70% of borrowers have DSR lower than 40%. Distribution of Borrowers by DSR
2. Data and Methods 14 About 80% of borrowers have LTI lower than 200%. Distribution of Borrowers by LTI
2. Data and Methods 15 About 80% of borrowers have loans from less than or equal to two financial institutions. Distribution of Borrowers by the Number of Credit Commitments
2. Data and Methods 16 Performance of the over-indebtedness indicators is assessed. The performance of over-indebtedness indicators are assessed for their ability to predict defaults within the following one year. - The predictability is assessed in a single dimension as well as in a multi-dimension of the over-indebtedness. - The predictability and suitability of the over-indebtedness indicators are assessed based on logistic regressions. Types of of the over-indebted The overly indebted are characterized in comparison with the average borrowers over several metrics such as - age, income, debt, geographical area, lending institutions, other measures of over-indebtedness.
2. Data and Methods 17 Stress test of consumer debt at the aggregate level Stress scenarios are juxtaposed against the baseline scenario. - Stress scenarios: 1) interest rate changes 2) historical scenarios of past crises (the Asian financial crisis, the global financial crisis) Macro-economic distress would shift the distribution of borrowers in the dimension of over-indebtedness, credit score in this study - The distribution of borrowers would shift toward higher levels of over-indebtedness in response to macroeconomic distress. As a result the share of borrowers at risk, or the average probability of default (PD) would increase. - The bigger the macro-economic distress, the larger the impact on the share of borrowers at risk (average PD)
3. Over-indebtedness and the Likelihood 3. Over-indebtedness and the Likelihood of Default
3. Over-indebtedness and the Likelihood of Default 19 Borrowers rated in the 6 th ~10 th credit ranking show higher-thanaverage default rates. Default Rates by Credit Score
3. Over-indebtedness and the Likelihood of Default 20 Borrowers with DSR higher than 40% show higher-than-average default rates. Default Rates by DSR
3. Over-indebtedness and the Likelihood of Default 21 Borrowers with LTI higher than 600~700% show higher-than-average default rates. Default Rates by LTI
3. Over-indebtedness and the Likelihood of Default 22 Borrowers with loans from more than or equal to 3 financial institutions show higher-than-average default rates. Default Rates by the Number of Credit Commitments
3. Over-indebtedness and the Likelihood of Default 23 Credit scores are a dominant predictor of forthcoming defaults. - Including additional over-indebtedness indicators such as the no. of credit commit, DSR, and LTI in the regression model improves predictability but only slightly. Logistic Regression Results (2013Q4) (1) (2) (3) (4) (5) Constant -6.26990-3.57770-3.53780-4.34520-6.18250 (0.02250) (0.00987) (0.00991) (0.01460) (0.03020) Credit score 0.43730 (0.00332) 0.40560 (0.00453) DSR 0.00059 (0.00005) 0.00183 (0.00021) LTI 0.00011 (0.00002) -0.00062 (0.00009) No. of credit commitments 0.35760 (0.00393) 0.21290 (0.00452) C-statistic 0.811 0.539 0.505 0.683 0.846 * Numbers in parentheses are standard errors.
4. Types of the Over-indebted
25 4. Types of the Over-indebted Borrowers with low credit scores tend to have lower income, to have more no. of loans, and to depend more on non-banks than the average. - However, they tend to have lower DSR and LTI than the average due to smaller amount of debt holdings. Types of the Over-indebted with Low Credit Scores (2013Q4) 1st 6th 7th 10th Borrower types TOTAL grade grade Share of borrowers (%) 81 19 100 Average income relative to the total (%) 104 82 100 Average debt relative to the total (%) 108 65 100 Income quintile (by median income) 4 3 4 Share of borrowers by area Share of borrowers by lenders Capital (%) 53 51 52 Non-capital (%) 47 49 48 Only banks (%) 49 12 42 Only non-banks (%) 33 66 39 Both banks and non-banks (%) 19 22 19 Median age 46 45 46 Average number of loans 1.8 2.6 1.8 Median credit score 3 8 4 Median DSR (%) 21.9 16.9 21.1 Median LTI (%) 77. 5 46.1 70.7
26 4. Types of the Over-indebted Borrowers with multiple credit commitments tend to have larger amount of debt and much higher DSR & LTI, and to depend more on non-banks than the average borrowers. Types of the Over-indebted with Multiple Credit Commitments (2013Q4) Borrower types No. of loans No. of loans < 3 >=3 TOTAL Share of borrowers (%) 81 19 100 Average income relative to the total (%) 99 104 100 Average debt relative to the total (%) 86 161 100 Income quintile (by median income) 4 4 4 Share of borrowers by area Capital (%) 52 55 52 Non-capital (%) 48 45 48 Only banks (%) 50 8 42 Share of borrowers by lenders Only non-banks (%) 40 34 39 Both banks and non-banks (%) 10 59 19 Median age 46 44 46 Average number of loans 1.3 4.1 1.8 Median credit score 3 6 4 Median DSR (%) 17.5 43.7 21.1 Median LTI (%) 58.2 134.2 70.7
27 4. Types of the Over-indebted Borrowers with high DSR tend to have much higher DSR & LTI with a large amount of debt, to have more no. of credit commitments, and to depend more on non-banks than the average borrowers. Types of the Over-indebted with High DSR (2013Q4) Borrower types DSR < 60% DSR >= 60% TOTAL Share of borrowers (%) 84 16 100 Average income relative to the total (%) 99 107 100 Average debt relative to the total (%) 54 341 100 Income quintile (by median income) 4 4 4 Share of borrowers by area Share of borrowers by lenders Capital (%) 53 52 52 Non-capital (%) 47 48 48 Only banks (%) 45 24 42 Only non-banks (%) 39 39 39 Both banks and non-banks (%) 16 37 19 Median age 45 49 46 Average number of loans 1.7 2.5 1.8 Median credit score 4 4 4 Median DSR (%) 16.4 94.4 21.1 Median LTI (%) 52.7 386.3 70.7
28 4. Types of the Over-indebted Vulnerable borrowers can be identified and characterized based on over-indebtedness indicators. Over-indebted borrowers show heavy reliance on non-bank financial institutions as sources of their loans according to all the above indicators. Those classified as overly indebted in terms of one indicator tend to be classified as overly indebted in terms of other indicators. Meanwhile, different indicators elucidate idiosyncratic characteristics of the overly indebted.
5. Stress Test and Debt Vulnerability
5. Stress Test and Debt Vulnerability 30 Macroeconomic distress would increase the number of borrowers at risk and default rates. Household debt vulnerability can be assessed by analyzing how sensitive the number of borrowers (or the amount of debt) at risk would be subject to changes in a macroeconomic environment. Stress test procedure in this paper: (1) Stress scenarios are juxtaposed against the baseline scenario. (2) Stress scenarios shift the distribution of borrowers toward regions of worse credit scores, which are associated with higher probability of defaults. (3) As more borrowers are associated with higher probability of default, the share of borrowers at risk, the average PD increases.
5. Stress Test and Debt Vulnerability 31 Stress scenarios reflect macroeconomic distress (1) Scenarios drawn as distribution of macroeconomic variables from the Bayesian VAR with t-copula à Scenarios of Interest rate risks (2) Historical scenarios - Asian Financial Crisis (AFC in 1998~99) - Global Financial Crisis (GFC 2008~09) Stress scenarios shift the distribution of borrowers toward worse credit scores, which are more strongly associated with higher PDs. The change of the distribution of borrowers in the credit score dimension can be represented by the change in the relationship between credit scores and the probability of default. The change in the relationship between credit scores and the probability of default can be explained by changes in a macroeconomic environment. Thus, we can model how the distribution of borrowers in the credit score dimension can be affected by changes in a macroeconomic environment in this study.
5. Stress Test and Debt Vulnerability 32 If the relationship between credit score and the risk amount were expressed visually by a straight line, the change in the relationship could be translated into changes in the mean and the slope. The mean and the slope can be expressed as functions of macroeconomic variables. Macroeconomic Distress and a change in the Relationship between Credit Score and Risk Amount Risk Amount Macro-economic distress Score Risk Amount Mean change Risk Amount Slope Change Score Score
5. Stress Test and Debt Vulnerability 33 Stress scenarios for the interest rate risks Interest rates in Korea and the U.S. increase - by 0.5 ~ 1.0%p - by 1.0 ~ 1.5%p - by 1.5 ~ 2.0%p - by 2.0 ~ 3.0%p - by 3.0 ~ 4.0%p - by 4.0 ~ 5.0%p - by 5.0%p ~ For each stress scenario, we can show how the distribution of borrowers changes in the credit score dimension. We can also show how the share of borrowers at risk, or the average PD changes.
5. Stress Test and Debt Vulnerability 34 Large increases in interest rates shift the distribution of borrowers toward worse credit scores. - Small increases of less than 1.5%p in interest rates do not appear to have much impact on the distribution of borrowers in credit score dimension. Interest Rate Scenarios and the Distribution of Borrowers in the Credit Score Dimension
5. Stress Test and Debt Vulnerability 35 Large increases of more than 5%p in interest rates increase the share of borrowers at risk by more than 1%p. - Small increases of less than 1.5%p in interest rates do not appear to have much impact on the share of borrowers at risk, or the average PD. Interest Rate Scenarios and the Share of Borrowers at Risk
5. Stress Test and Debt Vulnerability 36 Stress scenarios reflecting past economic crises Historical scenarios of past economic crises - The Asian financial crisis (AFC) in 1997~99 - The global financial crisis (GFC) in 2007~09 For each stress scenario, we can show how the distribution of borrowers changes in the credit score dimension. We can also show how the share of borrowers at risk, or the average PD changes.
5. Stress Test and Debt Vulnerability 37 Macroeconomic distress redistributes borrowers to worse credit score groups, and its impact is much bigger for the Asian financial crisis, the worse macroeconomic condition. Historical Scenarios and the Distribution of Borrowers in the Credit Score Dimension
5. Stress Test and Debt Vulnerability 38 The harder the macroeconomic distress, the bigger the impact on the default rates. - Historical scenarios of the Asian financial crisis (AFC) and the recent global financial crisis (GFC) result in large increases in the share of borrowers at risk, or the average PD. Historical Scenarios and the Share of Borrowers at Risk
6. Concluding Remarks
6. Concluding Remarks 40 Analyzed the Credit Bureau data to shed light on over-indebtedness and debt vulnerability of household sector. Assessed the performance of over-indebtedness indicators for their ability to predict near-term defaults. Identified and characterized the overly indebted based on the over- indebtedness indicators Stress-tested consumer debt to assess their vulnerability - Severe macroeconomic distress would increase the share of borrowers at risk, the average PD dramatically. - Moderate changes in a macroeconomic environment did not have much impact on the aggregate default rates. Reducing the share of overly indebted with high credit risk is important to make the economy and the financial system more resilient.
Thank you